diff options
Diffstat (limited to 'src/gallium/drivers')
183 files changed, 39998 insertions, 2305 deletions
diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index cb0631baf5..98554d7f52 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -64,9 +64,13 @@ #define ROUNDUP16(k) (((k) + 0xf) & ~0xf) -#define CELL_MAX_SPUS 6 +#define CELL_MAX_SPUS 8 #define CELL_MAX_SAMPLERS 4 +#define CELL_MAX_TEXTURE_LEVELS 12 /* 2k x 2k */ +#define CELL_MAX_CONSTANTS 32 /**< number of float[4] constants */ +#define CELL_MAX_WIDTH 1024 /**< max framebuffer width */ +#define CELL_MAX_HEIGHT 1024 /**< max framebuffer width */ #define TILE_SIZE 32 @@ -94,43 +98,94 @@ #define CELL_CMD_STATE_BIND_VS 18 #define CELL_CMD_STATE_FRAGMENT_PROGRAM 19 #define CELL_CMD_STATE_ATTRIB_FETCH 20 -#define CELL_CMD_VS_EXECUTE 22 -#define CELL_CMD_FLUSH_BUFFER_RANGE 23 +#define CELL_CMD_STATE_FS_CONSTANTS 21 +#define CELL_CMD_STATE_RASTERIZER 22 +#define CELL_CMD_VS_EXECUTE 23 +#define CELL_CMD_FLUSH_BUFFER_RANGE 24 +#define CELL_CMD_FENCE 25 +/** Command/batch buffers */ #define CELL_NUM_BUFFERS 4 #define CELL_BUFFER_SIZE (4*1024) /**< 16KB would be the max */ #define CELL_BUFFER_STATUS_FREE 10 #define CELL_BUFFER_STATUS_USED 20 +/** Debug flags */ +#define CELL_DEBUG_CHECKER (1 << 0) +#define CELL_DEBUG_ASM (1 << 1) +#define CELL_DEBUG_SYNC (1 << 2) +#define CELL_DEBUG_FRAGMENT_OPS (1 << 3) +#define CELL_DEBUG_FRAGMENT_OP_FALLBACK (1 << 4) +#define CELL_DEBUG_CMD (1 << 5) +#define CELL_DEBUG_CACHE (1 << 6) -#define CELL_DEBUG_CHECKER (1 << 0) -#define CELL_DEBUG_SYNC (1 << 1) +#define CELL_FENCE_IDLE 0 +#define CELL_FENCE_EMITTED 1 +#define CELL_FENCE_SIGNALLED 2 +#define CELL_FACING_FRONT 0 +#define CELL_FACING_BACK 1 + +struct cell_fence +{ + /** There's a 16-byte status qword per SPU */ + volatile uint status[CELL_MAX_SPUS][4]; +}; -/** Max instructions for doing per-fragment operations */ -#define SPU_MAX_FRAGMENT_OPS_INSTS 64 +/** + * Fence command sent to SPUs. In response, the SPUs will write + * CELL_FENCE_STATUS_SIGNALLED back to the fence status word in main memory. + */ +struct cell_command_fence +{ + uint64_t opcode; /**< CELL_CMD_FENCE */ + struct cell_fence *fence; +}; /** * Command to specify per-fragment operations state and generated code. + * Note that this is a variant-length structure, allocated with as + * much memory as needed to hold the generated code; the "code" + * field *must* be the last field in the structure. Also, the entire + * length of the structure (including the variant code field) must be + * a multiple of 8 bytes; we require that this structure itself be + * a multiple of 8 bytes, and that the generated code also be a multiple + * of 8 bytes. + * + * Also note that the dsa, blend, blend_color fields are really only needed + * for the fallback/C per-pixel code. They're not used when we generate + * dynamic SPU fragment code (which is the normal case), and will eventually + * be removed from this structure. */ struct cell_command_fragment_ops { uint64_t opcode; /**< CELL_CMD_STATE_FRAGMENT_OPS */ + + /* Fields for the fallback case */ struct pipe_depth_stencil_alpha_state dsa; struct pipe_blend_state blend; - unsigned code[SPU_MAX_FRAGMENT_OPS_INSTS]; + struct pipe_blend_color blend_color; + + /* Fields for the generated SPU code */ + unsigned total_code_size; + unsigned front_code_index; + unsigned back_code_index; + /* this field has variant length, and must be the last field in + * the structure + */ + unsigned code[0]; }; /** Max instructions for fragment programs */ -#define SPU_MAX_FRAGMENT_PROGRAM_INSTS 128 +#define SPU_MAX_FRAGMENT_PROGRAM_INSTS 512 /** - * Command to send a fragment progra to SPUs. + * Command to send a fragment program to SPUs. */ struct cell_command_fragment_program { @@ -145,7 +200,7 @@ struct cell_command_fragment_program */ struct cell_command_framebuffer { - uint64_t opcode; /**< CELL_CMD_FRAMEBUFFER */ + uint64_t opcode; /**< CELL_CMD_STATE_FRAMEBUFFER */ int width, height; void *color_start, *depth_start; enum pipe_format color_format, depth_format; @@ -153,6 +208,16 @@ struct cell_command_framebuffer /** + * Tell SPUs about rasterizer state. + */ +struct cell_command_rasterizer +{ + uint64_t opcode; /**< CELL_CMD_STATE_RASTERIZER */ + struct pipe_rasterizer_state rasterizer; +}; + + +/** * Clear framebuffer to the given value/color. */ struct cell_command_clear_surface @@ -248,23 +313,27 @@ struct cell_command_sampler struct cell_command_texture { uint64_t opcode; /**< CELL_CMD_STATE_TEXTURE */ + uint target; /**< PIPE_TEXTURE_x */ uint unit; - void *start; /**< Address in main memory */ - ushort width, height; + void *start[CELL_MAX_TEXTURE_LEVELS]; /**< Address in main memory */ + ushort width[CELL_MAX_TEXTURE_LEVELS]; + ushort height[CELL_MAX_TEXTURE_LEVELS]; + ushort depth[CELL_MAX_TEXTURE_LEVELS]; }; -/** XXX unions don't seem to work */ -/* XXX this should go away; all commands should be placed in batch buffers */ -struct cell_command +#define MAX_SPU_FUNCTIONS 12 +/** + * Used to tell the PPU about the address of particular functions in the + * SPU's address space. + */ +struct cell_spu_function_info { -#if 0 - struct cell_command_framebuffer fb; - struct cell_command_clear_surface clear; - struct cell_command_render render; -#endif - struct cell_command_vs vs; -} ALIGN16_ATTRIB; + uint num; + char names[MAX_SPU_FUNCTIONS][16]; + uint addrs[MAX_SPU_FUNCTIONS]; + char pad[12]; /**< Pad struct to multiple of 16 bytes (256 currently) */ +}; /** This is the object passed to spe_create_thread() */ @@ -273,11 +342,13 @@ struct cell_init_info unsigned id; unsigned num_spus; unsigned debug_flags; /**< mask of CELL_DEBUG_x flags */ - struct cell_command *cmd; + float inv_timebase; /**< 1.0/timebase, for perf measurement */ /** Buffers for command batches, vertex/index data */ ubyte *buffers[CELL_NUM_BUFFERS]; uint *buffer_status; /**< points at cell_context->buffer_status */ + + struct cell_spu_function_info *spu_functions; } ALIGN16_ATTRIB; diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile index b28f4c5c31..9358a47284 100644 --- a/src/gallium/drivers/cell/ppu/Makefile +++ b/src/gallium/drivers/cell/ppu/Makefile @@ -24,6 +24,7 @@ SOURCES = \ cell_clear.c \ cell_context.c \ cell_draw_arrays.c \ + cell_fence.c \ cell_flush.c \ cell_gen_fragment.c \ cell_gen_fp.c \ diff --git a/src/gallium/drivers/cell/ppu/cell_batch.c b/src/gallium/drivers/cell/ppu/cell_batch.c index 16882c0129..962775cd33 100644 --- a/src/gallium/drivers/cell/ppu/cell_batch.c +++ b/src/gallium/drivers/cell/ppu/cell_batch.c @@ -28,6 +28,7 @@ #include "cell_context.h" #include "cell_batch.h" +#include "cell_fence.h" #include "cell_spu.h" @@ -42,7 +43,9 @@ uint cell_get_empty_buffer(struct cell_context *cell) { - uint buf = 0, tries = 0; + static uint prev_buffer = 0; + uint buf = (prev_buffer + 1) % CELL_NUM_BUFFERS; + uint tries = 0; /* Find a buffer that's marked as free by all SPUs */ while (1) { @@ -58,8 +61,13 @@ cell_get_empty_buffer(struct cell_context *cell) cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_USED; } /* - printf("PPU: ALLOC BUFFER %u\n", buf); + printf("PPU: ALLOC BUFFER %u, %u tries\n", buf, tries); */ + prev_buffer = buf; + + /* release tex buffer associated w/ prev use of this batch buf */ + cell_free_fenced_buffers(cell, &cell->fenced_buffers[buf]); + return buf; } } @@ -82,6 +90,37 @@ cell_get_empty_buffer(struct cell_context *cell) /** + * Append a fence command to the current batch buffer. + * Note that we're sure there's always room for this because of the + * adjusted size check in cell_batch_free_space(). + */ +static void +emit_fence(struct cell_context *cell) +{ + const uint batch = cell->cur_batch; + const uint size = cell->buffer_size[batch]; + struct cell_command_fence *fence_cmd; + struct cell_fence *fence = &cell->fenced_buffers[batch].fence; + uint i; + + /* set fence status to emitted, not yet signalled */ + for (i = 0; i < cell->num_spus; i++) { + fence->status[i][0] = CELL_FENCE_EMITTED; + } + + ASSERT(size + sizeof(struct cell_command_fence) <= CELL_BUFFER_SIZE); + + fence_cmd = (struct cell_command_fence *) (cell->buffer[batch] + size); + fence_cmd->opcode = CELL_CMD_FENCE; + fence_cmd->fence = fence; + + /* update batch buffer size */ + cell->buffer_size[batch] = size + sizeof(struct cell_command_fence); + assert(sizeof(struct cell_command_fence) % 8 == 0); +} + + +/** * Flush the current batch buffer to the SPUs. * An empty buffer will be found and set as the new current batch buffer * for subsequent commands/data. @@ -91,7 +130,7 @@ cell_batch_flush(struct cell_context *cell) { static boolean flushing = FALSE; uint batch = cell->cur_batch; - const uint size = cell->buffer_size[batch]; + uint size = cell->buffer_size[batch]; uint spu, cmd_word; assert(!flushing); @@ -99,6 +138,14 @@ cell_batch_flush(struct cell_context *cell) if (size == 0) return; + /* Before we use this batch buffer, make sure any fenced texture buffers + * are released. + */ + if (cell->fenced_buffers[batch].head) { + emit_fence(cell); + size = cell->buffer_size[batch]; + } + flushing = TRUE; assert(batch < CELL_NUM_BUFFERS); @@ -139,6 +186,7 @@ uint cell_batch_free_space(const struct cell_context *cell) { uint free = CELL_BUFFER_SIZE - cell->buffer_size[cell->cur_batch]; + free -= sizeof(struct cell_command_fence); return free; } @@ -169,7 +217,7 @@ cell_batch_append(struct cell_context *cell, const void *data, uint bytes) size = cell->buffer_size[cell->cur_batch]; - if (size + bytes > CELL_BUFFER_SIZE) { + if (bytes > cell_batch_free_space(cell)) { cell_batch_flush(cell); size = 0; } @@ -223,7 +271,7 @@ cell_batch_alloc_aligned(struct cell_context *cell, uint bytes, padbytes = (alignment - (size % alignment)) % alignment; - if (padbytes + size + bytes > CELL_BUFFER_SIZE) { + if (padbytes + bytes > cell_batch_free_space(cell)) { cell_batch_flush(cell); size = 0; } diff --git a/src/gallium/drivers/cell/ppu/cell_clear.c b/src/gallium/drivers/cell/ppu/cell_clear.c index c9c0c721bb..037635e466 100644 --- a/src/gallium/drivers/cell/ppu/cell_clear.c +++ b/src/gallium/drivers/cell/ppu/cell_clear.c @@ -106,4 +106,17 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, clr->surface = surfIndex; clr->value = clearValue; } + + /* Technically, the surface's contents are now known and cleared, + * so we could set the status to PIPE_SURFACE_STATUS_CLEAR. But + * it turns out it's quite painful to recognize when any particular + * surface goes from PIPE_SURFACE_STATUS_CLEAR to + * PIPE_SURFACE_STATUS_DEFINED (i.e. with known contents), because + * the drawing commands could be operating on numerous draw buffers, + * which we'd have to iterate through to set all their stati... + * For now, we cheat a bit and set the surface's status to DEFINED + * right here. Later we should revisit this and set the status to + * CLEAR here, and find a better place to set the status to DEFINED. + */ + ps->status = PIPE_SURFACE_STATUS_DEFINED; } diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index 71f1a3049d..22d552d8e3 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -47,6 +47,7 @@ #include "cell_clear.h" #include "cell_context.h" #include "cell_draw_arrays.h" +#include "cell_fence.h" #include "cell_flush.h" #include "cell_state.h" #include "cell_surface.h" @@ -62,6 +63,8 @@ cell_destroy_context( struct pipe_context *pipe ) { struct cell_context *cell = cell_context(pipe); + util_delete_keymap(cell->fragment_ops_cache, NULL); + cell_spu_exit(cell); align_free(cell); @@ -85,13 +88,16 @@ cell_draw_create(struct cell_context *cell) } -#ifdef DEBUG static const struct debug_named_value cell_debug_flags[] = { {"checker", CELL_DEBUG_CHECKER},/**< modulate tile clear color by SPU ID */ + {"asm", CELL_DEBUG_ASM}, /**< dump SPU asm code */ {"sync", CELL_DEBUG_SYNC}, /**< SPUs do synchronous DMA */ + {"fragops", CELL_DEBUG_FRAGMENT_OPS}, /**< SPUs emit fragment ops debug messages*/ + {"fragopfallback", CELL_DEBUG_FRAGMENT_OP_FALLBACK}, /**< SPUs use reference implementation for fragment ops*/ + {"cmd", CELL_DEBUG_CMD}, /**< SPUs dump command buffer info */ + {"cache", CELL_DEBUG_CACHE}, /**< report texture cache stats on exit */ {NULL, 0} }; -#endif struct pipe_context * @@ -99,6 +105,7 @@ cell_create_context(struct pipe_screen *screen, struct cell_winsys *cws) { struct cell_context *cell; + uint i; /* some fields need to be 16-byte aligned, so align the whole object */ cell = (struct cell_context*) align_malloc(sizeof(struct cell_context), 16); @@ -125,11 +132,14 @@ cell_create_context(struct pipe_screen *screen, cell_init_state_functions(cell); cell_init_shader_functions(cell); cell_init_surface_functions(cell); - cell_init_texture_functions(cell); cell_init_vertex_functions(cell); cell->draw = cell_draw_create(cell); + /* Create cache of fragment ops generated code */ + cell->fragment_ops_cache = + util_new_keymap(sizeof(struct cell_fragment_ops_key), ~0, NULL); + cell_init_vbuf(cell); draw_set_rasterize_stage(cell->draw, cell->vbuf); @@ -143,17 +153,31 @@ cell_create_context(struct pipe_screen *screen, cell_debug_flags, 0 ); + for (i = 0; i < CELL_NUM_BUFFERS; i++) + cell_fence_init(&cell->fenced_buffers[i].fence); + + /* * SPU stuff */ - cell->num_spus = 6; - /* XXX is this in SDK 3.0 only? - cell->num_spus = spe_cpu_info_get(SPE_COUNT_PHYSICAL_SPES, -1); - */ + /* This call only works with SDK 3.0. Anyone still using 2.1??? */ + cell->num_cells = spe_cpu_info_get(SPE_COUNT_PHYSICAL_CPU_NODES, -1); + cell->num_spus = spe_cpu_info_get(SPE_COUNT_USABLE_SPES, 0); + if (cell->debug_flags) { + printf("Cell: found %d Cell(s) with %u SPUs\n", + cell->num_cells, cell->num_spus); + } + if (getenv("CELL_NUM_SPUS")) { + cell->num_spus = atoi(getenv("CELL_NUM_SPUS")); + assert(cell->num_spus > 0); + } cell_start_spus(cell); cell_init_batch_buffers(cell); + /* make sure SPU initializations are done before proceeding */ + cell_flush_int(cell, CELL_FLUSH_WAIT); + return &cell->pipe; } diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 14914b9c6f..eb1397bb3f 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -38,6 +38,7 @@ #include "cell/common.h" #include "rtasm/rtasm_ppc_spe.h" #include "tgsi/tgsi_scan.h" +#include "util/u_keymap.h" struct cell_vbuf_render; @@ -67,31 +68,29 @@ struct cell_fragment_shader_state /** - * Cell blend state atom, subclass of pipe_blend_state. + * Key for mapping per-fragment state to cached SPU machine code. + * keymap(cell_fragment_ops_key) => cell_command_fragment_ops */ -struct cell_blend_state +struct cell_fragment_ops_key { - struct pipe_blend_state base; - - /** - * Generated code to perform alpha blending - */ - struct spe_function code; + struct pipe_blend_state blend; + struct pipe_blend_color blend_color; + struct pipe_depth_stencil_alpha_state dsa; + enum pipe_format color_format; + enum pipe_format zs_format; }; +struct cell_buffer_node; + /** - * Cell depth/stencil/alpha state atom, subclass of - * pipe_depth_stencil_alpha_state. + * Fenced buffer list. List of buffers which can be unreferenced after + * the fence has been executed/signalled. */ -struct cell_depth_stencil_alpha_state +struct cell_buffer_list { - struct pipe_depth_stencil_alpha_state base; - - /** - * Generated code to perform alpha, stencil, and depth testing on the SPE - */ - struct spe_function code; + struct cell_fence fence ALIGN16_ATTRIB; + struct cell_buffer_node *head; }; @@ -104,10 +103,10 @@ struct cell_context struct cell_winsys *winsys; - const struct cell_blend_state *blend; + const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; uint num_samplers; - const struct cell_depth_stencil_alpha_state *depth_stencil; + const struct pipe_depth_stencil_alpha_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; const struct cell_vertex_shader_state *vs; const struct cell_fragment_shader_state *fs; @@ -135,6 +134,11 @@ struct cell_context uint *tex_map; uint dirty; + uint dirty_textures; /* bitmask of texture units */ + uint dirty_samplers; /* bitmask of sampler units */ + + /** Cache of code generated for per-fragment ops */ + struct keymap *fragment_ops_cache; /** The primitive drawing context */ struct draw_context *draw; @@ -149,8 +153,9 @@ struct cell_context /** Mapped constant buffers */ void *mapped_constants[PIPE_SHADER_TYPES]; + struct cell_spu_function_info spu_functions ALIGN16_ATTRIB; - uint num_spus; + uint num_cells, num_spus; /** Buffers for command batches, vertex/index data */ uint buffer_size[CELL_NUM_BUFFERS]; @@ -162,6 +167,14 @@ struct cell_context uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BUFFERS][4] ALIGN16_ATTRIB; + /** Associated with each command/batch buffer is a list of pipe_buffers + * that are fenced. When the last command in a buffer is executed, the + * fence will be signalled, indicating that any pipe_buffers preceeding + * that fence can be unreferenced (and probably freed). + */ + struct cell_buffer_list fenced_buffers[CELL_NUM_BUFFERS]; + + struct spe_function attrib_fetch; unsigned attrib_fetch_offsets[PIPE_MAX_ATTRIBS]; diff --git a/src/gallium/drivers/cell/ppu/cell_fence.c b/src/gallium/drivers/cell/ppu/cell_fence.c new file mode 100644 index 0000000000..867b5dcaa0 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_fence.c @@ -0,0 +1,168 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <unistd.h> +#include "util/u_memory.h" +#include "pipe/p_inlines.h" +#include "cell_context.h" +#include "cell_batch.h" +#include "cell_fence.h" +#include "cell_texture.h" + + +void +cell_fence_init(struct cell_fence *fence) +{ + uint i; + ASSERT_ALIGN16(fence->status); + for (i = 0; i < CELL_MAX_SPUS; i++) { + fence->status[i][0] = CELL_FENCE_IDLE; + } +} + + +boolean +cell_fence_signalled(const struct cell_context *cell, + const struct cell_fence *fence) +{ + uint i; + for (i = 0; i < cell->num_spus; i++) { + if (fence->status[i][0] != CELL_FENCE_SIGNALLED) + return FALSE; + /*assert(fence->status[i][0] == CELL_FENCE_EMITTED);*/ + } + return TRUE; +} + + +void +cell_fence_finish(const struct cell_context *cell, + const struct cell_fence *fence) +{ + while (!cell_fence_signalled(cell, fence)) { + usleep(10); + } + +#ifdef DEBUG + { + uint i; + for (i = 0; i < cell->num_spus; i++) { + assert(fence->status[i][0] == CELL_FENCE_SIGNALLED); + } + } +#endif +} + + + + +struct cell_buffer_node +{ + struct pipe_buffer *buffer; + struct cell_buffer_node *next; +}; + + +static void +cell_add_buffer_to_list(struct cell_context *cell, + struct cell_buffer_list *list, + struct pipe_buffer *buffer) +{ + struct pipe_screen *ps = cell->pipe.screen; + struct cell_buffer_node *node = CALLOC_STRUCT(cell_buffer_node); + /* create new list node which references the buffer, insert at head */ + if (node) { + pipe_buffer_reference(ps, &node->buffer, buffer); + node->next = list->head; + list->head = node; + } +} + + +/** + * Wait for completion of the given fence, then unreference any buffers + * on the list. + * This typically unrefs/frees texture buffers after any rendering which uses + * them has completed. + */ +void +cell_free_fenced_buffers(struct cell_context *cell, + struct cell_buffer_list *list) +{ + if (list->head) { + struct pipe_screen *ps = cell->pipe.screen; + struct cell_buffer_node *node; + + cell_fence_finish(cell, &list->fence); + + /* traverse the list, unreferencing buffers, freeing nodes */ + node = list->head; + while (node) { + struct cell_buffer_node *next = node->next; + assert(node->buffer); + pipe_buffer_unmap(ps, node->buffer); +#if 0 + printf("Unref buffer %p\n", node->buffer); + if (node->buffer->refcount == 1) + printf(" Delete!\n"); +#endif + pipe_buffer_reference(ps, &node->buffer, NULL); + FREE(node); + node = next; + } + list->head = NULL; + } +} + + +/** + * This should be called for each render command. + * Any texture buffers that are current bound will be added to a fenced + * list to be freed later when the fence is executed/signalled. + */ +void +cell_add_fenced_textures(struct cell_context *cell) +{ + struct cell_buffer_list *list = &cell->fenced_buffers[cell->cur_batch]; + uint i; + + for (i = 0; i < cell->num_textures; i++) { + struct cell_texture *ct = cell->texture[i]; + if (ct) { + uint level; + for (level = 0; level < CELL_MAX_TEXTURE_LEVELS; level++) { + if (ct->tiled_buffer[level]) { +#if 0 + printf("Adding texture %p buffer %p to list\n", + ct, ct->tiled_buffer[level]); +#endif + cell_add_buffer_to_list(cell, list, ct->tiled_buffer[level]); + } + } + } + } +} diff --git a/src/gallium/drivers/cell/ppu/cell_fence.h b/src/gallium/drivers/cell/ppu/cell_fence.h new file mode 100644 index 0000000000..536b4ba411 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_fence.h @@ -0,0 +1,57 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_FENCE_H +#define CELL_FENCE_H + + +extern void +cell_fence_init(struct cell_fence *fence); + + +extern boolean +cell_fence_signalled(const struct cell_context *cell, + const struct cell_fence *fence); + + +extern void +cell_fence_finish(const struct cell_context *cell, + const struct cell_fence *fence); + + + +extern void +cell_free_fenced_buffers(struct cell_context *cell, + struct cell_buffer_list *list); + + +extern void +cell_add_fenced_textures(struct cell_context *cell); + + +#endif /* CELL_FENCE_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_flush.c b/src/gallium/drivers/cell/ppu/cell_flush.c index 6596b72010..a64967b4b9 100644 --- a/src/gallium/drivers/cell/ppu/cell_flush.c +++ b/src/gallium/drivers/cell/ppu/cell_flush.c @@ -49,7 +49,7 @@ cell_flush(struct pipe_context *pipe, unsigned flags, flags |= CELL_FLUSH_WAIT; } - if (flags & PIPE_FLUSH_SWAPBUFFERS) + if (flags & (PIPE_FLUSH_SWAPBUFFERS | PIPE_FLUSH_RENDER_CACHE)) flags |= CELL_FLUSH_WAIT; draw_flush( cell->draw ); diff --git a/src/gallium/drivers/cell/ppu/cell_gen_fp.c b/src/gallium/drivers/cell/ppu/cell_gen_fp.c index 6ffe94eb14..96a1743fc1 100644 --- a/src/gallium/drivers/cell/ppu/cell_gen_fp.c +++ b/src/gallium/drivers/cell/ppu/cell_gen_fp.c @@ -37,7 +37,7 @@ * \author Brian Paul */ - +#include <math.h> #include "pipe/p_defines.h" #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" @@ -51,25 +51,43 @@ #include "cell_gen_fp.h" -/** Set to 1 to enable debug/disassembly printfs */ -#define DISASSEM 01 +#define MAX_TEMPS 16 +#define MAX_IMMED 8 +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 /** * Context needed during code generation. */ struct codegen { + struct cell_context *cell; int inputs_reg; /**< 1st function parameter */ int outputs_reg; /**< 2nd function parameter */ int constants_reg; /**< 3rd function parameter */ - int temp_regs[8][4]; /**< maps TGSI temps to SPE registers */ + int temp_regs[MAX_TEMPS][4]; /**< maps TGSI temps to SPE registers */ + int imm_regs[MAX_IMMED][4]; /**< maps TGSI immediates to SPE registers */ + + int num_imm; /**< number of immediates */ int one_reg; /**< register containing {1.0, 1.0, 1.0, 1.0} */ /** Per-instruction temps / intermediate temps */ int num_itemps; - int itemps[3]; + int itemps[12]; + + /** Current IF/ELSE/ENDIF nesting level */ + int if_nesting; + /** Index of execution mask register */ + int exec_mask_reg; + + /** KIL mask: indicates which fragments have been killed */ + int kill_mask_reg; + + int frame_size; /**< Stack frame size, in words */ struct spe_function *f; boolean error; @@ -112,19 +130,78 @@ get_const_one_reg(struct codegen *gen) { if (gen->one_reg <= 0) { gen->one_reg = spe_allocate_available_register(gen->f); - } - /* one = {1.0, 1.0, 1.0, 1.0} */ - spe_load_float(gen->f, gen->one_reg, 1.0f); -#if DISASSEM - printf("il\tr%d, 1.0f\n", gen->one_reg); -#endif + spe_indent(gen->f, 4); + spe_comment(gen->f, -4, "INIT CONSTANT 1.0:"); + + /* one = {1.0, 1.0, 1.0, 1.0} */ + spe_load_float(gen->f, gen->one_reg, 1.0f); + + spe_indent(gen->f, -4); + } return gen->one_reg; } /** + * Return index of the pixel execution mask. + * The register is allocated an initialized upon the first call. + * + * The pixel execution mask controls which pixels in a quad are + * modified, according to surrounding conditionals, loops, etc. + */ +static int +get_exec_mask_reg(struct codegen *gen) +{ + if (gen->exec_mask_reg <= 0) { + gen->exec_mask_reg = spe_allocate_available_register(gen->f); + + spe_indent(gen->f, 4); + spe_comment(gen->f, -4, "INIT EXEC MASK = ~0:"); + + /* exec_mask = {~0, ~0, ~0, ~0} */ + spe_load_int(gen->f, gen->exec_mask_reg, ~0); + + spe_indent(gen->f, -4); + } + + return gen->exec_mask_reg; +} + + +static boolean +is_register_src(struct codegen *gen, int channel, + const struct tgsi_full_src_register *src) +{ + int swizzle = tgsi_util_get_full_src_register_extswizzle(src, channel); + int sign_op = tgsi_util_get_full_src_register_sign_mode(src, channel); + + if (swizzle > TGSI_SWIZZLE_W || sign_op != TGSI_UTIL_SIGN_KEEP) { + return FALSE; + } + if (src->SrcRegister.File == TGSI_FILE_TEMPORARY || + src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + return TRUE; + } + return FALSE; +} + + +static boolean +is_memory_dst(struct codegen *gen, int channel, + const struct tgsi_full_dst_register *dst) +{ + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + return TRUE; + } + else { + return FALSE; + } +} + + +/** * Return the index of the SPU temporary containing the named TGSI * source register. If the TGSI register is a TGSI_FILE_TEMPORARY we * just return the corresponding SPE register. If the TGIS register @@ -136,35 +213,93 @@ get_src_reg(struct codegen *gen, int channel, const struct tgsi_full_src_register *src) { - int reg; + int reg = -1; + int swizzle = tgsi_util_get_full_src_register_extswizzle(src, channel); + boolean reg_is_itemp = FALSE; + uint sign_op; + + assert(swizzle >= TGSI_SWIZZLE_X); + assert(swizzle <= TGSI_EXTSWIZZLE_ONE); + + if (swizzle == TGSI_EXTSWIZZLE_ONE) { + /* Load const one float and early out */ + reg = get_const_one_reg(gen); + } + else if (swizzle == TGSI_EXTSWIZZLE_ZERO) { + /* Load const zero float and early out */ + reg = get_itemp(gen); + spe_xor(gen->f, reg, reg, reg); + } + else { + assert(swizzle < 4); - /* XXX need to examine src swizzle info here. - * That will involve changing the channel var... + switch (src->SrcRegister.File) { + case TGSI_FILE_TEMPORARY: + reg = gen->temp_regs[src->SrcRegister.Index][swizzle]; + break; + case TGSI_FILE_INPUT: + { + /* offset is measured in quadwords, not bytes */ + int offset = src->SrcRegister.Index * 4 + swizzle; + reg = get_itemp(gen); + reg_is_itemp = TRUE; + /* Load: reg = memory[(machine_reg) + offset] */ + spe_lqd(gen->f, reg, gen->inputs_reg, offset * 16); + } + break; + case TGSI_FILE_IMMEDIATE: + reg = gen->imm_regs[src->SrcRegister.Index][swizzle]; + break; + case TGSI_FILE_CONSTANT: + { + /* offset is measured in quadwords, not bytes */ + int offset = src->SrcRegister.Index * 4 + swizzle; + reg = get_itemp(gen); + reg_is_itemp = TRUE; + /* Load: reg = memory[(machine_reg) + offset] */ + spe_lqd(gen->f, reg, gen->constants_reg, offset * 16); + } + break; + default: + assert(0); + } + } + + /* + * Handle absolute value, negate or set-negative of src register. */ + sign_op = tgsi_util_get_full_src_register_sign_mode(src, channel); + if (sign_op != TGSI_UTIL_SIGN_KEEP) { + /* + * All sign ops are done by manipulating bit 31, the IEEE float sign bit. + */ + const int bit31mask_reg = get_itemp(gen); + int result_reg; + + if (reg_is_itemp) { + /* re-use 'reg' for the result */ + result_reg = reg; + } + else { + /* alloc a new reg for the result */ + result_reg = get_itemp(gen); + } + /* mask with bit 31 set, the rest cleared */ + spe_load_uint(gen->f, bit31mask_reg, (1 << 31)); - switch (src->SrcRegister.File) { - case TGSI_FILE_TEMPORARY: - reg = gen->temp_regs[src->SrcRegister.Index][channel]; - break; - case TGSI_FILE_INPUT: - { - /* offset is measured in quadwords, not bytes */ - int offset = src->SrcRegister.Index * 4 + channel; - reg = get_itemp(gen); - /* Load: reg = memory[(machine_reg) + offset] */ - spe_lqd(gen->f, reg, gen->inputs_reg, offset); -#if DISASSEM - printf("lqd\tr%d, r%d + %d\n", reg, gen->inputs_reg, offset); -#endif + if (sign_op == TGSI_UTIL_SIGN_CLEAR) { + spe_andc(gen->f, result_reg, reg, bit31mask_reg); } - break; - case TGSI_FILE_IMMEDIATE: - /* xxx fall-through for now / fix */ - case TGSI_FILE_CONSTANT: - /* xxx fall-through for now / fix */ - default: - assert(0); + else if (sign_op == TGSI_UTIL_SIGN_SET) { + spe_and(gen->f, result_reg, reg, bit31mask_reg); + } + else { + assert(sign_op == TGSI_UTIL_SIGN_TOGGLE); + spe_xor(gen->f, result_reg, reg, bit31mask_reg); + } + + reg = result_reg; } return reg; @@ -183,11 +318,14 @@ get_dst_reg(struct codegen *gen, int channel, const struct tgsi_full_dst_register *dest) { - int reg; + int reg = -1; switch (dest->DstRegister.File) { case TGSI_FILE_TEMPORARY: - reg = gen->temp_regs[dest->DstRegister.Index][channel]; + if (gen->if_nesting > 0) + reg = get_itemp(gen); + else + reg = gen->temp_regs[dest->DstRegister.Index][channel]; break; case TGSI_FILE_OUTPUT: reg = get_itemp(gen); @@ -211,19 +349,59 @@ store_dest_reg(struct codegen *gen, int value_reg, int channel, const struct tgsi_full_dst_register *dest) { + /* + * XXX need to implement dst reg clamping/saturation + */ +#if 0 + switch (inst->Instruction.Saturate) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + break; + case TGSI_SAT_MINUS_PLUS_ONE: + break; + default: + assert( 0 ); + } +#endif + switch (dest->DstRegister.File) { case TGSI_FILE_TEMPORARY: - /* no-op */ + if (gen->if_nesting > 0) { + int d_reg = gen->temp_regs[dest->DstRegister.Index][channel]; + int exec_reg = get_exec_mask_reg(gen); + /* Mix d with new value according to exec mask: + * d[i] = mask_reg[i] ? value_reg : d_reg + */ + spe_selb(gen->f, d_reg, d_reg, value_reg, exec_reg); + } + else { + /* we're not inside a condition or loop: do nothing special */ + + } break; case TGSI_FILE_OUTPUT: { /* offset is measured in quadwords, not bytes */ int offset = dest->DstRegister.Index * 4 + channel; - /* Store: memory[(machine_reg) + offset] = reg */ - spe_stqd(gen->f, value_reg, gen->outputs_reg, offset); -#if DISASSEM - printf("stqd\tr%d, r%d + %d\n", value_reg, gen->outputs_reg, offset); -#endif + if (gen->if_nesting > 0) { + int exec_reg = get_exec_mask_reg(gen); + int curval_reg = get_itemp(gen); + /* First read the current value from memory: + * Load: curval = memory[(machine_reg) + offset] + */ + spe_lqd(gen->f, curval_reg, gen->outputs_reg, offset * 16); + /* Mix curval with newvalue according to exec mask: + * d[i] = mask_reg[i] ? value_reg : d_reg + */ + spe_selb(gen->f, curval_reg, curval_reg, value_reg, exec_reg); + /* Store: memory[(machine_reg) + offset] = curval */ + spe_stqd(gen->f, curval_reg, gen->outputs_reg, offset * 16); + } + else { + /* Store: memory[(machine_reg) + offset] = reg */ + spe_stqd(gen->f, value_reg, gen->outputs_reg, offset * 16); + } } break; default: @@ -232,27 +410,114 @@ store_dest_reg(struct codegen *gen, } + +static void +emit_prologue(struct codegen *gen) +{ + gen->frame_size = 1024; /* XXX temporary, should be dynamic */ + + spe_comment(gen->f, -4, "Function prologue:"); + + /* save $lr on stack # stqd $lr,16($sp) */ + spe_stqd(gen->f, SPE_REG_RA, SPE_REG_SP, 16); + + if (gen->frame_size >= 512) { + /* offset is too large for ai instruction */ + int offset_reg = spe_allocate_available_register(gen->f); + int sp_reg = spe_allocate_available_register(gen->f); + /* offset = -framesize */ + spe_load_int(gen->f, offset_reg, -gen->frame_size); + /* sp = $sp */ + spe_move(gen->f, sp_reg, SPE_REG_SP); + /* $sp = $sp + offset_reg */ + spe_a(gen->f, SPE_REG_SP, SPE_REG_SP, offset_reg); + /* save $sp in stack frame */ + spe_stqd(gen->f, sp_reg, SPE_REG_SP, 0); + /* clean up */ + spe_release_register(gen->f, offset_reg); + spe_release_register(gen->f, sp_reg); + } + else { + /* save stack pointer # stqd $sp,-frameSize($sp) */ + spe_stqd(gen->f, SPE_REG_SP, SPE_REG_SP, -gen->frame_size); + + /* adjust stack pointer # ai $sp,$sp,-frameSize */ + spe_ai(gen->f, SPE_REG_SP, SPE_REG_SP, -gen->frame_size); + } +} + + +static void +emit_epilogue(struct codegen *gen) +{ + const int return_reg = 3; + + spe_comment(gen->f, -4, "Function epilogue:"); + + spe_comment(gen->f, 0, "return the killed mask"); + if (gen->kill_mask_reg > 0) { + /* shader called KIL, return the "alive" mask */ + spe_move(gen->f, return_reg, gen->kill_mask_reg); + } + else { + /* return {0,0,0,0} */ + spe_load_uint(gen->f, return_reg, 0); + } + + spe_comment(gen->f, 0, "restore stack and return"); + if (gen->frame_size >= 512) { + /* offset is too large for ai instruction */ + int offset_reg = spe_allocate_available_register(gen->f); + /* offset = framesize */ + spe_load_int(gen->f, offset_reg, gen->frame_size); + /* $sp = $sp + offset */ + spe_a(gen->f, SPE_REG_SP, SPE_REG_SP, offset_reg); + /* clean up */ + spe_release_register(gen->f, offset_reg); + } + else { + /* restore stack pointer # ai $sp,$sp,frameSize */ + spe_ai(gen->f, SPE_REG_SP, SPE_REG_SP, gen->frame_size); + } + + /* restore $lr # lqd $lr,16($sp) */ + spe_lqd(gen->f, SPE_REG_RA, SPE_REG_SP, 16); + + /* return from function call */ + spe_bi(gen->f, SPE_REG_RA, 0, 0); +} + + static boolean emit_MOV(struct codegen *gen, const struct tgsi_full_instruction *inst) { - int ch; + int ch, src_reg[4], dst_reg[4]; + + spe_comment(gen->f, -4, "MOV:"); for (ch = 0; ch < 4; ch++) { if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { - int src_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); - int dst_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); - /* XXX we don't always need to actually emit a mov instruction here */ - spe_move(gen->f, dst_reg, src_reg); -#if DISASSEM - printf("mov\tr%d, r%d\n", dst_reg, src_reg); -#endif - store_dest_reg(gen, dst_reg, ch, &inst->FullDstRegisters[0]); + src_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + dst_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + } + } + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + if (is_register_src(gen, ch, &inst->FullSrcRegisters[0]) && + is_memory_dst(gen, ch, &inst->FullDstRegisters[0])) { + /* special-case: register to memory store */ + store_dest_reg(gen, src_reg[ch], ch, &inst->FullDstRegisters[0]); + } + else { + spe_move(gen->f, dst_reg[ch], src_reg[ch]); + store_dest_reg(gen, dst_reg[ch], ch, &inst->FullDstRegisters[0]); + } free_itemps(gen); } } return true; } - /** * Emit addition instructions. Recall that a single TGSI_OPCODE_ADD * becomes (up to) four SPU "fa" instructions because we're doing SOA @@ -261,24 +526,25 @@ emit_MOV(struct codegen *gen, const struct tgsi_full_instruction *inst) static boolean emit_ADD(struct codegen *gen, const struct tgsi_full_instruction *inst) { - int ch; - /* Loop over Red/Green/Blue/Alpha channels */ + int ch, s1_reg[4], s2_reg[4], d_reg[4]; + + spe_comment(gen->f, -4, "ADD:"); + /* Loop over Red/Green/Blue/Alpha channels, fetch src operands */ for (ch = 0; ch < 4; ch++) { /* If the dest R, G, B or A writemask is enabled... */ if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { - /* get indexes of the two src, one dest SPE registers */ - int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); - int s2_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); - int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); - + s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + s2_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + } + } + /* Loop over Red/Green/Blue/Alpha channels, do the add, store results */ + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { /* Emit actual SPE instruction: d = s1 + s2 */ - spe_fa(gen->f, d_reg, s1_reg, s2_reg); -#if DISASSEM - printf("fa\tr%d, r%d, r%d\n", d_reg, s1_reg, s2_reg); -#endif - + spe_fa(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch]); /* Store the result (a no-op for TGSI_FILE_TEMPORARY dests) */ - store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]); /* Free any intermediate temps we allocated */ free_itemps(gen); } @@ -286,6 +552,99 @@ emit_ADD(struct codegen *gen, const struct tgsi_full_instruction *inst) return true; } +/** + * Emit subtract. See emit_ADD for comments. + */ +static boolean +emit_SUB(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch, s1_reg[4], s2_reg[4], d_reg[4]; + spe_comment(gen->f, -4, "SUB:"); + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + s2_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + } + } + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + /* d = s1 - s2 */ + spe_fs(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch]); + store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + return true; +} + +/** + * Emit multiply add. See emit_ADD for comments. + */ +static boolean +emit_MAD(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch, s1_reg[4], s2_reg[4], s3_reg[4], d_reg[4]; + spe_comment(gen->f, -4, "MAD:"); + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + s2_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + s3_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[2]); + d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + } + } + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + /* d = s1 * s2 + s3 */ + spe_fma(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch], s3_reg[ch]); + store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + return true; +} + + +/** + * Emit linear interpolate. See emit_ADD for comments. + */ +static boolean +emit_LERP(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch, s1_reg[4], s2_reg[4], s3_reg[4], d_reg[4], tmp_reg[4]; + spe_comment(gen->f, -4, "LERP:"); + /* setup/get src/dst/temp regs */ + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + s2_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + s3_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[2]); + d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + tmp_reg[ch] = get_itemp(gen); + } + } + + /* d = s3 + s1(s2 - s3) */ + /* do all subtracts, then all fma, then all stores to better pipeline */ + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + spe_fs(gen->f, tmp_reg[ch], s2_reg[ch], s3_reg[ch]); + } + } + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + spe_fma(gen->f, d_reg[ch], tmp_reg[ch], s1_reg[ch], s3_reg[ch]); + } + } + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]); + } + } + free_itemps(gen); + return true; +} /** * Emit multiply. See emit_ADD for comments. @@ -293,17 +652,63 @@ emit_ADD(struct codegen *gen, const struct tgsi_full_instruction *inst) static boolean emit_MUL(struct codegen *gen, const struct tgsi_full_instruction *inst) { + int ch, s1_reg[4], s2_reg[4], d_reg[4]; + spe_comment(gen->f, -4, "MUL:"); + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + s2_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + } + } + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + /* d = s1 * s2 */ + spe_fm(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch]); + store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + return true; +} + +/** + * Emit reciprocal. See emit_ADD for comments. + */ +static boolean +emit_RCP(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ int ch; + spe_comment(gen->f, -4, "RCP:"); for (ch = 0; ch < 4; ch++) { if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); - int s2_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); - /* d = s1 * s2 */ - spe_fm(gen->f, d_reg, s1_reg, s2_reg); -#if DISASSEM - printf("fm\tr%d, r%d, r%d\n", d_reg, s1_reg, s2_reg); -#endif + /* d = 1/s1 */ + spe_frest(gen->f, d_reg, s1_reg); + spe_fi(gen->f, d_reg, s1_reg, d_reg); + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + return true; +} + +/** + * Emit reciprocal sqrt. See emit_ADD for comments. + */ +static boolean +emit_RSQ(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + spe_comment(gen->f, -4, "RSQ:"); + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + /* d = 1/s1 */ + spe_frsqest(gen->f, d_reg, s1_reg); + spe_fi(gen->f, d_reg, s1_reg, d_reg); store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); free_itemps(gen); } @@ -311,6 +716,270 @@ emit_MUL(struct codegen *gen, const struct tgsi_full_instruction *inst) return true; } +/** + * Emit absolute value. See emit_ADD for comments. + */ +static boolean +emit_ABS(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + spe_comment(gen->f, -4, "ABS:"); + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + const int bit31mask_reg = get_itemp(gen); + + /* mask with bit 31 set, the rest cleared */ + spe_load_uint(gen->f, bit31mask_reg, (1 << 31)); + + /* d = sign bit cleared in s1 */ + spe_andc(gen->f, d_reg, s1_reg, bit31mask_reg); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + return true; +} + +/** + * Emit 3 component dot product. See emit_ADD for comments. + */ +static boolean +emit_DP3(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + int s1x_reg, s1y_reg, s1z_reg; + int s2x_reg, s2y_reg, s2z_reg; + int t0_reg = get_itemp(gen), t1_reg = get_itemp(gen); + + spe_comment(gen->f, -4, "DP3:"); + + s1x_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]); + s2x_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[1]); + s1y_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]); + s2y_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[1]); + s1z_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]); + s2z_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[1]); + + /* t0 = x0 * x1 */ + spe_fm(gen->f, t0_reg, s1x_reg, s2x_reg); + + /* t1 = y0 * y1 */ + spe_fm(gen->f, t1_reg, s1y_reg, s2y_reg); + + /* t0 = z0 * z1 + t0 */ + spe_fma(gen->f, t0_reg, s1z_reg, s2z_reg, t0_reg); + + /* t0 = t0 + t1 */ + spe_fa(gen->f, t0_reg, t0_reg, t1_reg); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + spe_move(gen->f, d_reg, t0_reg); + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + } + } + + free_itemps(gen); + return true; +} + +/** + * Emit 4 component dot product. See emit_ADD for comments. + */ +static boolean +emit_DP4(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + int s0x_reg, s0y_reg, s0z_reg, s0w_reg; + int s1x_reg, s1y_reg, s1z_reg, s1w_reg; + int t0_reg = get_itemp(gen), t1_reg = get_itemp(gen); + + spe_comment(gen->f, -4, "DP4:"); + + s0x_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]); + s1x_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[1]); + s0y_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]); + s1y_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[1]); + s0z_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]); + s1z_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[1]); + s0w_reg = get_src_reg(gen, CHAN_W, &inst->FullSrcRegisters[0]); + s1w_reg = get_src_reg(gen, CHAN_W, &inst->FullSrcRegisters[1]); + + /* t0 = x0 * x1 */ + spe_fm(gen->f, t0_reg, s0x_reg, s1x_reg); + + /* t1 = y0 * y1 */ + spe_fm(gen->f, t1_reg, s0y_reg, s1y_reg); + + /* t0 = z0 * z1 + t0 */ + spe_fma(gen->f, t0_reg, s0z_reg, s1z_reg, t0_reg); + + /* t1 = w0 * w1 + t1 */ + spe_fma(gen->f, t1_reg, s0w_reg, s1w_reg, t1_reg); + + /* t0 = t0 + t1 */ + spe_fa(gen->f, t0_reg, t0_reg, t1_reg); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + spe_move(gen->f, d_reg, t0_reg); + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + } + } + + free_itemps(gen); + return true; +} + +/** + * Emit homogeneous dot product. See emit_ADD for comments. + */ +static boolean +emit_DPH(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + /* XXX rewrite this function to look more like DP3/DP4 */ + int ch; + spe_comment(gen->f, -4, "DPH:"); + + int s1_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]); + int s2_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[1]); + int tmp_reg = get_itemp(gen); + + /* t = x0 * x1 */ + spe_fm(gen->f, tmp_reg, s1_reg, s2_reg); + + s1_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]); + s2_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[1]); + /* t = y0 * y1 + t */ + spe_fma(gen->f, tmp_reg, s1_reg, s2_reg, tmp_reg); + + s1_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]); + s2_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[1]); + /* t = z0 * z1 + t */ + spe_fma(gen->f, tmp_reg, s1_reg, s2_reg, tmp_reg); + + s2_reg = get_src_reg(gen, CHAN_W, &inst->FullSrcRegisters[1]); + /* t = w1 + t */ + spe_fa(gen->f, tmp_reg, s2_reg, tmp_reg); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + spe_move(gen->f, d_reg, tmp_reg); + store_dest_reg(gen, tmp_reg, ch, &inst->FullDstRegisters[0]); + } + } + + free_itemps(gen); + return true; +} + +/** + * Emit 3-component vector normalize. + */ +static boolean +emit_NRM3(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + int src_reg[3]; + int t0_reg = get_itemp(gen), t1_reg = get_itemp(gen); + + spe_comment(gen->f, -4, "NRM3:"); + + src_reg[0] = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]); + src_reg[1] = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]); + src_reg[2] = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]); + + /* t0 = x * x */ + spe_fm(gen->f, t0_reg, src_reg[0], src_reg[0]); + + /* t1 = y * y */ + spe_fm(gen->f, t1_reg, src_reg[1], src_reg[1]); + + /* t0 = z * z + t0 */ + spe_fma(gen->f, t0_reg, src_reg[2], src_reg[2], t0_reg); + + /* t0 = t0 + t1 */ + spe_fa(gen->f, t0_reg, t0_reg, t1_reg); + + /* t1 = 1.0 / sqrt(t0) */ + spe_frsqest(gen->f, t1_reg, t0_reg); + spe_fi(gen->f, t1_reg, t0_reg, t1_reg); + + for (ch = 0; ch < 3; ch++) { /* NOTE: omit W channel */ + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + /* dst = src[ch] * t1 */ + spe_fm(gen->f, d_reg, src_reg[ch], t1_reg); + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + } + } + + free_itemps(gen); + return true; +} + + +/** + * Emit cross product. See emit_ADD for comments. + */ +static boolean +emit_XPD(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + spe_comment(gen->f, -4, "XPD:"); + + int s1_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]); + int s2_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[1]); + int tmp_reg = get_itemp(gen); + + /* t = z0 * y1 */ + spe_fm(gen->f, tmp_reg, s1_reg, s2_reg); + + s1_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]); + s2_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[1]); + /* t = y0 * z1 - t */ + spe_fms(gen->f, tmp_reg, s1_reg, s2_reg, tmp_reg); + + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << CHAN_X)) { + store_dest_reg(gen, tmp_reg, CHAN_X, &inst->FullDstRegisters[0]); + } + + s1_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]); + s2_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[1]); + /* t = x0 * z1 */ + spe_fm(gen->f, tmp_reg, s1_reg, s2_reg); + + s1_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]); + s2_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[1]); + /* t = z0 * x1 - t */ + spe_fms(gen->f, tmp_reg, s1_reg, s2_reg, tmp_reg); + + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << CHAN_Y)) { + store_dest_reg(gen, tmp_reg, CHAN_Y, &inst->FullDstRegisters[0]); + } + + s1_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]); + s2_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[1]); + /* t = y0 * x1 */ + spe_fm(gen->f, tmp_reg, s1_reg, s2_reg); + + s1_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]); + s2_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[1]); + /* t = x0 * y1 - t */ + spe_fms(gen->f, tmp_reg, s1_reg, s2_reg, tmp_reg); + + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << CHAN_Z)) { + store_dest_reg(gen, tmp_reg, CHAN_Z, &inst->FullDstRegisters[0]); + } + + free_itemps(gen); + return true; +} /** * Emit set-if-greater-than. @@ -323,6 +992,8 @@ emit_SGT(struct codegen *gen, const struct tgsi_full_instruction *inst) { int ch; + spe_comment(gen->f, -4, "SGT:"); + for (ch = 0; ch < 4; ch++) { if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); @@ -331,26 +1002,776 @@ emit_SGT(struct codegen *gen, const struct tgsi_full_instruction *inst) /* d = (s1 > s2) */ spe_fcgt(gen->f, d_reg, s1_reg, s2_reg); -#if DISASSEM - printf("fcgt\tr%d, r%d, r%d\n", d_reg, s1_reg, s2_reg); -#endif /* convert d from 0x0/0xffffffff to 0.0/1.0 */ /* d = d & one_reg */ spe_and(gen->f, d_reg, d_reg, get_const_one_reg(gen)); -#if DISASSEM - printf("and\tr%d, r%d, r%d\n", d_reg, d_reg, get_const_one_reg(gen)); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return true; +} + +/** + * Emit set-if_less-then. See emit_SGT for comments. + */ +static boolean +emit_SLT(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + + spe_comment(gen->f, -4, "SLT:"); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int s2_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + + /* d = (s1 < s2) */ + spe_fcgt(gen->f, d_reg, s2_reg, s1_reg); + + /* convert d from 0x0/0xffffffff to 0.0/1.0 */ + /* d = d & one_reg */ + spe_and(gen->f, d_reg, d_reg, get_const_one_reg(gen)); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return true; +} + +/** + * Emit set-if_greater-then-or-equal. See emit_SGT for comments. + */ +static boolean +emit_SGE(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + + spe_comment(gen->f, -4, "SGE:"); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int s2_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + + /* d = (s1 >= s2) */ + spe_fcgt(gen->f, d_reg, s2_reg, s1_reg); + + /* convert d from 0x0/0xffffffff to 0.0/1.0 */ + /* d = ~d & one_reg */ + spe_andc(gen->f, d_reg, get_const_one_reg(gen), d_reg); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return true; +} + +/** + * Emit set-if_less-then-or-equal. See emit_SGT for comments. + */ +static boolean +emit_SLE(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + + spe_comment(gen->f, -4, "SLE:"); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int s2_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + + /* d = (s1 <= s2) */ + spe_fcgt(gen->f, d_reg, s1_reg, s2_reg); + + /* convert d from 0x0/0xffffffff to 0.0/1.0 */ + /* d = ~d & one_reg */ + spe_andc(gen->f, d_reg, get_const_one_reg(gen), d_reg); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return true; +} + +/** + * Emit set-if_equal. See emit_SGT for comments. + */ +static boolean +emit_SEQ(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + + spe_comment(gen->f, -4, "SEQ:"); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int s2_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + + /* d = (s1 == s2) */ + spe_fceq(gen->f, d_reg, s1_reg, s2_reg); + + /* convert d from 0x0/0xffffffff to 0.0/1.0 */ + /* d = d & one_reg */ + spe_and(gen->f, d_reg, d_reg, get_const_one_reg(gen)); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return true; +} + +/** + * Emit set-if_not_equal. See emit_SGT for comments. + */ +static boolean +emit_SNE(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + + spe_comment(gen->f, -4, "SNE:"); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int s2_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + + /* d = (s1 != s2) */ + spe_fceq(gen->f, d_reg, s1_reg, s2_reg); + spe_nor(gen->f, d_reg, d_reg, d_reg); + + /* convert d from 0x0/0xffffffff to 0.0/1.0 */ + /* d = d & one_reg */ + spe_and(gen->f, d_reg, d_reg, get_const_one_reg(gen)); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return true; +} + +/** + * Emit compare. See emit_SGT for comments. + */ +static boolean +emit_CMP(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + + spe_comment(gen->f, -4, "CMP:"); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int s2_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + int s3_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[2]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + int zero_reg = get_itemp(gen); + + spe_xor(gen->f, zero_reg, zero_reg, zero_reg); + + /* d = (s1 < 0) ? s2 : s3 */ + spe_fcgt(gen->f, d_reg, zero_reg, s1_reg); + spe_selb(gen->f, d_reg, s3_reg, s2_reg, d_reg); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return true; +} + +/** + * Emit trunc. + * Convert float to signed int + * Convert signed int to float + */ +static boolean +emit_TRUNC(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + + spe_comment(gen->f, -4, "TRUNC:"); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + + /* Convert float to int */ + spe_cflts(gen->f, d_reg, s1_reg, 0); + + /* Convert int to float */ + spe_csflt(gen->f, d_reg, d_reg, 0); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return true; +} + +/** + * Emit floor. + * If negative int subtract one + * Convert float to signed int + * Convert signed int to float + */ +static boolean +emit_FLR(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + + spe_comment(gen->f, -4, "FLR:"); + + int zero_reg = get_itemp(gen); + spe_xor(gen->f, zero_reg, zero_reg, zero_reg); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + int tmp_reg = get_itemp(gen); + + /* If negative, subtract 1.0 */ + spe_fcgt(gen->f, tmp_reg, zero_reg, s1_reg); + spe_selb(gen->f, tmp_reg, zero_reg, get_const_one_reg(gen), tmp_reg); + spe_fs(gen->f, tmp_reg, s1_reg, tmp_reg); + + /* Convert float to int */ + spe_cflts(gen->f, tmp_reg, tmp_reg, 0); + + /* Convert int to float */ + spe_csflt(gen->f, d_reg, tmp_reg, 0); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return true; +} + +/** + * Compute frac = Input - FLR(Input) + */ +static boolean +emit_FRC(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + + spe_comment(gen->f, -4, "FRC:"); + + int zero_reg = get_itemp(gen); + spe_xor(gen->f, zero_reg, zero_reg, zero_reg); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + int tmp_reg = get_itemp(gen); + + /* If negative, subtract 1.0 */ + spe_fcgt(gen->f, tmp_reg, zero_reg, s1_reg); + spe_selb(gen->f, tmp_reg, zero_reg, get_const_one_reg(gen), tmp_reg); + spe_fs(gen->f, tmp_reg, s1_reg, tmp_reg); + + /* Convert float to int */ + spe_cflts(gen->f, tmp_reg, tmp_reg, 0); + + /* Convert int to float */ + spe_csflt(gen->f, tmp_reg, tmp_reg, 0); + + /* d = s1 - FLR(s1) */ + spe_fs(gen->f, d_reg, s1_reg, tmp_reg); + + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return true; +} + + +#if 0 +static void +print_functions(struct cell_context *cell) +{ + struct cell_spu_function_info *funcs = &cell->spu_functions; + uint i; + for (i = 0; i < funcs->num; i++) { + printf("SPU func %u: %s at %u\n", + i, funcs->names[i], funcs->addrs[i]); + } +} #endif + +static uint +lookup_function(struct cell_context *cell, const char *funcname) +{ + const struct cell_spu_function_info *funcs = &cell->spu_functions; + uint i, addr = 0; + for (i = 0; i < funcs->num; i++) { + if (strcmp(funcs->names[i], funcname) == 0) { + addr = funcs->addrs[i]; + } + } + assert(addr && "spu function not found"); + return addr / 4; /* discard 2 least significant bits */ +} + + +/** + * Emit code to call a SPU function. + * Used to implement instructions like SIN/COS/POW/TEX/etc. + * If scalar, only the X components of the src regs are used, and the + * result is replicated across the dest register's XYZW components. + */ +static boolean +emit_function_call(struct codegen *gen, + const struct tgsi_full_instruction *inst, + char *funcname, uint num_args, boolean scalar) +{ + const uint addr = lookup_function(gen->cell, funcname); + char comment[100]; + int s_regs[3]; + int func_called = FALSE; + uint a, ch; + int retval_reg = -1; + + assert(num_args <= 3); + + snprintf(comment, sizeof(comment), "CALL %s:", funcname); + spe_comment(gen->f, -4, comment); + + if (scalar) { + for (a = 0; a < num_args; a++) { + s_regs[a] = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[a]); + } + /* we'll call the function, put the return value in this register, + * then replicate it across all write-enabled components in d_reg. + */ + retval_reg = spe_allocate_available_register(gen->f); + } + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int d_reg; + ubyte usedRegs[SPE_NUM_REGS]; + uint i, numUsed; + + if (!scalar) { + for (a = 0; a < num_args; a++) { + s_regs[a] = get_src_reg(gen, ch, &inst->FullSrcRegisters[a]); + } + } + + d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + + if (!scalar || !func_called) { + /* for a scalar function, we'll really only call the function once */ + + numUsed = spe_get_registers_used(gen->f, usedRegs); + assert(numUsed < gen->frame_size / 16 - 2); + + /* save registers to stack */ + for (i = 0; i < numUsed; i++) { + uint reg = usedRegs[i]; + int offset = 2 + i; + spe_stqd(gen->f, reg, SPE_REG_SP, 16 * offset); + } + + /* setup function arguments */ + for (a = 0; a < num_args; a++) { + spe_move(gen->f, 3 + a, s_regs[a]); + } + + /* branch to function, save return addr */ + spe_brasl(gen->f, SPE_REG_RA, addr); + + /* save function's return value */ + if (scalar) + spe_move(gen->f, retval_reg, 3); + else + spe_move(gen->f, d_reg, 3); + + /* restore registers from stack */ + for (i = 0; i < numUsed; i++) { + uint reg = usedRegs[i]; + if (reg != d_reg && reg != retval_reg) { + int offset = 2 + i; + spe_lqd(gen->f, reg, SPE_REG_SP, 16 * offset); + } + } + + func_called = TRUE; + } + + if (scalar) { + spe_move(gen->f, d_reg, retval_reg); + } + store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]); free_itemps(gen); } } + if (scalar) { + spe_release_register(gen->f, retval_reg); + } + return true; } +static boolean +emit_TEX(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + const uint target = inst->InstructionExtTexture.Texture; + const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + uint addr; + int ch; + int coord_regs[4], d_regs[4]; + + switch (target) { + case TGSI_TEXTURE_1D: + case TGSI_TEXTURE_2D: + addr = lookup_function(gen->cell, "spu_tex_2d"); + break; + case TGSI_TEXTURE_3D: + addr = lookup_function(gen->cell, "spu_tex_3d"); + break; + case TGSI_TEXTURE_CUBE: + addr = lookup_function(gen->cell, "spu_tex_cube"); + break; + default: + ASSERT(0 && "unsupported texture target"); + return FALSE; + } + + assert(inst->FullSrcRegisters[1].SrcRegister.File == TGSI_FILE_SAMPLER); + + spe_comment(gen->f, -4, "CALL tex:"); + + /* get src/dst reg info */ + for (ch = 0; ch < 4; ch++) { + coord_regs[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + d_regs[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + } + + { + ubyte usedRegs[SPE_NUM_REGS]; + uint i, numUsed; + + numUsed = spe_get_registers_used(gen->f, usedRegs); + assert(numUsed < gen->frame_size / 16 - 2); + + /* save registers to stack */ + for (i = 0; i < numUsed; i++) { + uint reg = usedRegs[i]; + int offset = 2 + i; + spe_stqd(gen->f, reg, SPE_REG_SP, 16 * offset); + } + + /* setup function arguments (XXX depends on target) */ + for (i = 0; i < 4; i++) { + spe_move(gen->f, 3 + i, coord_regs[i]); + } + spe_load_uint(gen->f, 7, unit); /* sampler unit */ + + /* branch to function, save return addr */ + spe_brasl(gen->f, SPE_REG_RA, addr); + + /* save function's return values (four pixel's colors) */ + for (i = 0; i < 4; i++) { + spe_move(gen->f, d_regs[i], 3 + i); + } + + /* restore registers from stack */ + for (i = 0; i < numUsed; i++) { + uint reg = usedRegs[i]; + if (reg != d_regs[0] && + reg != d_regs[1] && + reg != d_regs[2] && + reg != d_regs[3]) { + int offset = 2 + i; + spe_lqd(gen->f, reg, SPE_REG_SP, 16 * offset); + } + } + } + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + store_dest_reg(gen, d_regs[ch], ch, &inst->FullDstRegisters[0]); + free_itemps(gen); + } + } + + return TRUE; +} + + +/** + * KILL if any of src reg values are less than zero. + */ +static boolean +emit_KIL(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch; + int s_regs[4], kil_reg = -1, cmp_reg, zero_reg; + + spe_comment(gen->f, -4, "CALL kil:"); + + /* zero = {0,0,0,0} */ + zero_reg = get_itemp(gen); + spe_load_uint(gen->f, zero_reg, 0); + + cmp_reg = get_itemp(gen); + + /* get src regs */ + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + s_regs[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + } + } + + /* test if any src regs are < 0 */ + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + if (kil_reg >= 0) { + /* cmp = 0 > src ? : ~0 : 0 */ + spe_fcgt(gen->f, cmp_reg, zero_reg, s_regs[ch]); + /* kil = kil | cmp */ + spe_or(gen->f, kil_reg, kil_reg, cmp_reg); + } + else { + kil_reg = get_itemp(gen); + /* kil = 0 > src ? : ~0 : 0 */ + spe_fcgt(gen->f, kil_reg, zero_reg, s_regs[ch]); + } + } + } + + if (gen->if_nesting) { + /* may have been a conditional kil */ + spe_and(gen->f, kil_reg, kil_reg, gen->exec_mask_reg); + } + + /* allocate the kill mask reg if needed */ + if (gen->kill_mask_reg <= 0) { + gen->kill_mask_reg = spe_allocate_available_register(gen->f); + spe_move(gen->f, gen->kill_mask_reg, kil_reg); + } + else { + spe_or(gen->f, gen->kill_mask_reg, gen->kill_mask_reg, kil_reg); + } + + free_itemps(gen); + + return TRUE; +} + + + +/** + * Emit max. See emit_SGT for comments. + */ +static boolean +emit_MAX(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch, s0_reg[4], s1_reg[4], d_reg[4], tmp_reg[4]; + + spe_comment(gen->f, -4, "MAX:"); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + s0_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + tmp_reg[ch] = get_itemp(gen); + } + } + + /* d = (s0 > s1) ? s0 : s1 */ + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + spe_fcgt(gen->f, tmp_reg[ch], s0_reg[ch], s1_reg[ch]); + } + } + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + spe_selb(gen->f, d_reg[ch], s1_reg[ch], s0_reg[ch], tmp_reg[ch]); + } + } + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]); + } + } + + free_itemps(gen); + return true; +} + +/** + * Emit max. See emit_SGT for comments. + */ +static boolean +emit_MIN(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + int ch, s0_reg[4], s1_reg[4], d_reg[4], tmp_reg[4]; + + spe_comment(gen->f, -4, "MIN:"); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + s0_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]); + d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + tmp_reg[ch] = get_itemp(gen); + } + } + + /* d = (s1 > s0) ? s0 : s1 */ + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + spe_fcgt(gen->f, tmp_reg[ch], s1_reg[ch], s0_reg[ch]); + } + } + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + spe_selb(gen->f, d_reg[ch], s1_reg[ch], s0_reg[ch], tmp_reg[ch]); + } + } + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]); + } + } + + free_itemps(gen); + return true; +} + +static boolean +emit_IF(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + const int channel = 0; + const int exec_reg = get_exec_mask_reg(gen); + + spe_comment(gen->f, -4, "IF:"); + + /* update execution mask with the predicate register */ + int tmp_reg = get_itemp(gen); + int s1_reg = get_src_reg(gen, channel, &inst->FullSrcRegisters[0]); + + /* tmp = (s1_reg == 0) */ + spe_ceqi(gen->f, tmp_reg, s1_reg, 0); + /* tmp = !tmp */ + spe_complement(gen->f, tmp_reg, tmp_reg); + /* exec_mask = exec_mask & tmp */ + spe_and(gen->f, exec_reg, exec_reg, tmp_reg); + + gen->if_nesting++; + + free_itemps(gen); + + return true; +} + + +static boolean +emit_ELSE(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + const int exec_reg = get_exec_mask_reg(gen); + + spe_comment(gen->f, -4, "ELSE:"); + + /* exec_mask = !exec_mask */ + spe_complement(gen->f, exec_reg, exec_reg); + + return true; +} + + +static boolean +emit_ENDIF(struct codegen *gen, const struct tgsi_full_instruction *inst) +{ + const int exec_reg = get_exec_mask_reg(gen); + + spe_comment(gen->f, -4, "ENDIF:"); + + /* XXX todo: pop execution mask */ + + spe_load_int(gen->f, exec_reg, ~0x0); + + gen->if_nesting--; + return true; +} + + +static boolean +emit_DDX_DDY(struct codegen *gen, const struct tgsi_full_instruction *inst, + boolean ddx) +{ + int ch; + + spe_comment(gen->f, -4, ddx ? "DDX:" : "DDY:"); + + for (ch = 0; ch < 4; ch++) { + if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch)) { + int s_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]); + int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]); + + int t1_reg = get_itemp(gen); + int t2_reg = get_itemp(gen); + + spe_splat_word(gen->f, t1_reg, s_reg, 0); /* upper-left pixel */ + if (ddx) { + spe_splat_word(gen->f, t2_reg, s_reg, 1); /* upper-right pixel */ + } + else { + spe_splat_word(gen->f, t2_reg, s_reg, 2); /* lower-left pixel */ + } + spe_fs(gen->f, d_reg, t2_reg, t1_reg); + + free_itemps(gen); + } + } + + return true; +} + + + + /** * Emit END instruction. * We just return from the shader function at this point. @@ -361,11 +1782,8 @@ emit_SGT(struct codegen *gen, const struct tgsi_full_instruction *inst) static boolean emit_END(struct codegen *gen) { - /* return from function call */ - spe_bi(gen->f, SPE_REG_RA, 0, 0); -#if DISASSEM - printf("bi\trRA\n"); -#endif + spe_comment(gen->f, -4, "END:"); + emit_epilogue(gen); return true; } @@ -379,19 +1797,101 @@ emit_instruction(struct codegen *gen, { switch (inst->Instruction.Opcode) { case TGSI_OPCODE_MOV: + case TGSI_OPCODE_SWZ: return emit_MOV(gen, inst); case TGSI_OPCODE_MUL: return emit_MUL(gen, inst); case TGSI_OPCODE_ADD: return emit_ADD(gen, inst); + case TGSI_OPCODE_SUB: + return emit_SUB(gen, inst); + case TGSI_OPCODE_MAD: + return emit_MAD(gen, inst); + case TGSI_OPCODE_LERP: + return emit_LERP(gen, inst); + case TGSI_OPCODE_DP3: + return emit_DP3(gen, inst); + case TGSI_OPCODE_DP4: + return emit_DP4(gen, inst); + case TGSI_OPCODE_DPH: + return emit_DPH(gen, inst); + case TGSI_OPCODE_NRM: + return emit_NRM3(gen, inst); + case TGSI_OPCODE_XPD: + return emit_XPD(gen, inst); + case TGSI_OPCODE_RCP: + return emit_RCP(gen, inst); + case TGSI_OPCODE_RSQ: + return emit_RSQ(gen, inst); + case TGSI_OPCODE_ABS: + return emit_ABS(gen, inst); case TGSI_OPCODE_SGT: return emit_SGT(gen, inst); + case TGSI_OPCODE_SLT: + return emit_SLT(gen, inst); + case TGSI_OPCODE_SGE: + return emit_SGE(gen, inst); + case TGSI_OPCODE_SLE: + return emit_SLE(gen, inst); + case TGSI_OPCODE_SEQ: + return emit_SEQ(gen, inst); + case TGSI_OPCODE_SNE: + return emit_SNE(gen, inst); + case TGSI_OPCODE_CMP: + return emit_CMP(gen, inst); + case TGSI_OPCODE_MAX: + return emit_MAX(gen, inst); + case TGSI_OPCODE_MIN: + return emit_MIN(gen, inst); + case TGSI_OPCODE_TRUNC: + return emit_TRUNC(gen, inst); + case TGSI_OPCODE_FLR: + return emit_FLR(gen, inst); + case TGSI_OPCODE_FRC: + return emit_FRC(gen, inst); case TGSI_OPCODE_END: return emit_END(gen); + case TGSI_OPCODE_COS: + return emit_function_call(gen, inst, "spu_cos", 1, TRUE); + case TGSI_OPCODE_SIN: + return emit_function_call(gen, inst, "spu_sin", 1, TRUE); + case TGSI_OPCODE_POW: + return emit_function_call(gen, inst, "spu_pow", 2, TRUE); + case TGSI_OPCODE_EXPBASE2: + return emit_function_call(gen, inst, "spu_exp2", 1, TRUE); + case TGSI_OPCODE_LOGBASE2: + return emit_function_call(gen, inst, "spu_log2", 1, TRUE); + case TGSI_OPCODE_TEX: + /* fall-through for now */ + case TGSI_OPCODE_TXD: + /* fall-through for now */ + case TGSI_OPCODE_TXB: + /* fall-through for now */ + case TGSI_OPCODE_TXL: + /* fall-through for now */ + case TGSI_OPCODE_TXP: + return emit_TEX(gen, inst); + case TGSI_OPCODE_KIL: + return emit_KIL(gen, inst); + + case TGSI_OPCODE_IF: + return emit_IF(gen, inst); + case TGSI_OPCODE_ELSE: + return emit_ELSE(gen, inst); + case TGSI_OPCODE_ENDIF: + return emit_ENDIF(gen, inst); + + case TGSI_OPCODE_DDX: + return emit_DDX_DDY(gen, inst, true); + case TGSI_OPCODE_DDY: + return emit_DDX_DDY(gen, inst, false); + /* XXX lots more cases to do... */ default: + fprintf(stderr, "Cell: unimplemented TGSI instruction %d!\n", + inst->Instruction.Opcode); return false; } @@ -401,48 +1901,93 @@ emit_instruction(struct codegen *gen, /** + * Emit code for a TGSI immediate value (vector of four floats). + * This involves register allocation and initialization. + * XXX the initialization should be done by a "prepare" stage, not + * per quad execution! + */ +static boolean +emit_immediate(struct codegen *gen, const struct tgsi_full_immediate *immed) +{ + int ch; + + assert(gen->num_imm < MAX_TEMPS); + + spe_comment(gen->f, -4, "IMMEDIATE:"); + + for (ch = 0; ch < 4; ch++) { + float val = immed->u.ImmediateFloat32[ch].Float; + + if (ch > 0 && val == immed->u.ImmediateFloat32[ch - 1].Float) { + /* re-use previous register */ + gen->imm_regs[gen->num_imm][ch] = gen->imm_regs[gen->num_imm][ch - 1]; + } + else { + int reg = spe_allocate_available_register(gen->f); + + if (reg < 0) + return false; + + /* update immediate map */ + gen->imm_regs[gen->num_imm][ch] = reg; + + /* emit initializer instruction */ + spe_load_float(gen->f, reg, val); + } + } + + gen->num_imm++; + + return true; +} + + + +/** * Emit "code" for a TGSI declaration. * We only care about TGSI TEMPORARY register declarations at this time. * For each TGSI TEMPORARY we allocate four SPE registers. */ -static void -emit_declaration(struct codegen *gen, const struct tgsi_full_declaration *decl) +static boolean +emit_declaration(struct cell_context *cell, + struct codegen *gen, const struct tgsi_full_declaration *decl) { int i, ch; switch (decl->Declaration.File) { case TGSI_FILE_TEMPORARY: -#if DISASSEM - printf("Declare temp reg %d .. %d\n", - decl->DeclarationRange.First, - decl->DeclarationRange.Last); -#endif for (i = decl->DeclarationRange.First; i <= decl->DeclarationRange.Last; i++) { + assert(i < MAX_TEMPS); for (ch = 0; ch < 4; ch++) { gen->temp_regs[i][ch] = spe_allocate_available_register(gen->f); + if (gen->temp_regs[i][ch] < 0) + return false; /* out of regs */ } /* XXX if we run out of SPE registers, we need to spill * to SPU memory. someday... */ -#if DISASSEM - printf(" SPE regs: %d %d %d %d\n", - gen->temp_regs[i][0], - gen->temp_regs[i][1], - gen->temp_regs[i][2], - gen->temp_regs[i][3]); -#endif + { + char buf[100]; + sprintf(buf, "TGSI temp[%d] maps to SPU regs [$%d $%d $%d $%d]", i, + gen->temp_regs[i][0], gen->temp_regs[i][1], + gen->temp_regs[i][2], gen->temp_regs[i][3]); + spe_comment(gen->f, -4, buf); + } } break; default: ; /* ignore */ } + + return true; } + /** * Translate TGSI shader code to SPE instructions. This is done when * the state tracker gives us a new shader (via pipe->create_fs_state()). @@ -460,6 +2005,7 @@ cell_gen_fragment_program(struct cell_context *cell, struct codegen gen; memset(&gen, 0, sizeof(gen)); + gen.cell = cell; gen.f = f; /* For SPE function calls: reg $3 = first param, $4 = second param, etc. */ @@ -472,50 +2018,50 @@ cell_gen_fragment_program(struct cell_context *cell, spe_allocate_register(f, gen.outputs_reg); spe_allocate_register(f, gen.constants_reg); -#if DISASSEM - printf("Begin %s\n", __FUNCTION__); - tgsi_dump(tokens, 0); -#endif + if (cell->debug_flags & CELL_DEBUG_ASM) { + spe_print_code(f, true); + spe_indent(f, 8); + printf("Begin %s\n", __FUNCTION__); + tgsi_dump(tokens, 0); + } tgsi_parse_init(&parse, tokens); + emit_prologue(&gen); + while (!tgsi_parse_end_of_tokens(&parse) && !gen.error) { tgsi_parse_token(&parse); switch (parse.FullToken.Token.Type) { case TGSI_TOKEN_TYPE_IMMEDIATE: -#if 0 - if (!note_immediate(&gen, &parse.FullToken.FullImmediate )) - goto fail; -#endif + if (!emit_immediate(&gen, &parse.FullToken.FullImmediate)) + gen.error = true; break; case TGSI_TOKEN_TYPE_DECLARATION: - emit_declaration(&gen, &parse.FullToken.FullDeclaration); + if (!emit_declaration(cell, &gen, &parse.FullToken.FullDeclaration)) + gen.error = true; break; case TGSI_TOKEN_TYPE_INSTRUCTION: - if (!emit_instruction(&gen, &parse.FullToken.FullInstruction )) { + if (!emit_instruction(&gen, &parse.FullToken.FullInstruction)) gen.error = true; - } break; default: assert(0); - } } - if (gen.error) { /* terminate the SPE code */ return emit_END(&gen); } -#if DISASSEM - printf("cell_gen_fragment_program nr instructions: %d\n", f->num_inst); - printf("End %s\n", __FUNCTION__); -#endif + if (cell->debug_flags & CELL_DEBUG_ASM) { + printf("cell_gen_fragment_program nr instructions: %d\n", f->num_inst); + printf("End %s\n", __FUNCTION__); + } tgsi_parse_free( &parse ); diff --git a/src/gallium/drivers/cell/ppu/cell_gen_fragment.c b/src/gallium/drivers/cell/ppu/cell_gen_fragment.c index 06219d4e98..2c64eb1bcc 100644 --- a/src/gallium/drivers/cell/ppu/cell_gen_fragment.c +++ b/src/gallium/drivers/cell/ppu/cell_gen_fragment.c @@ -54,12 +54,17 @@ * \param ifragZ_reg register containing integer fragment Z values (in) * \param ifbZ_reg register containing integer frame buffer Z values (in/out) * \param zmask_reg register containing result of Z test/comparison (out) + * + * Returns true if the Z-buffer needs to be updated. */ -static void -gen_depth_test(const struct pipe_depth_stencil_alpha_state *dsa, - struct spe_function *f, +static boolean +gen_depth_test(struct spe_function *f, + const struct pipe_depth_stencil_alpha_state *dsa, int mask_reg, int ifragZ_reg, int ifbZ_reg, int zmask_reg) { + /* NOTE: we use clgt below, not cgt, because we want to compare _unsigned_ + * quantities. This only makes a difference for 32-bit Z values though. + */ ASSERT(dsa->depth.enabled); switch (dsa->depth.func) { @@ -79,28 +84,28 @@ gen_depth_test(const struct pipe_depth_stencil_alpha_state *dsa, case PIPE_FUNC_GREATER: /* zmask = (ifragZ > ref) */ - spe_cgt(f, zmask_reg, ifragZ_reg, ifbZ_reg); + spe_clgt(f, zmask_reg, ifragZ_reg, ifbZ_reg); /* mask = (mask & zmask) */ spe_and(f, mask_reg, mask_reg, zmask_reg); break; case PIPE_FUNC_LESS: /* zmask = (ref > ifragZ) */ - spe_cgt(f, zmask_reg, ifbZ_reg, ifragZ_reg); + spe_clgt(f, zmask_reg, ifbZ_reg, ifragZ_reg); /* mask = (mask & zmask) */ spe_and(f, mask_reg, mask_reg, zmask_reg); break; case PIPE_FUNC_LEQUAL: /* zmask = (ifragZ > ref) */ - spe_cgt(f, zmask_reg, ifragZ_reg, ifbZ_reg); + spe_clgt(f, zmask_reg, ifragZ_reg, ifbZ_reg); /* mask = (mask & ~zmask) */ spe_andc(f, mask_reg, mask_reg, zmask_reg); break; case PIPE_FUNC_GEQUAL: /* zmask = (ref > ifragZ) */ - spe_cgt(f, zmask_reg, ifbZ_reg, ifragZ_reg); + spe_clgt(f, zmask_reg, ifbZ_reg, ifragZ_reg); /* mask = (mask & ~zmask) */ spe_andc(f, mask_reg, mask_reg, zmask_reg); break; @@ -129,7 +134,10 @@ gen_depth_test(const struct pipe_depth_stencil_alpha_state *dsa, * framebufferZ = (ztest_passed ? fragmentZ : framebufferZ; */ spe_selb(f, ifbZ_reg, ifbZ_reg, ifragZ_reg, mask_reg); + return true; } + + return false; } @@ -229,7 +237,40 @@ gen_alpha_test(const struct pipe_depth_stencil_alpha_state *dsa, spe_release_register(f, amask_reg); } +/* This pair of functions is used inline to allocate and deallocate + * optional constant registers. Once a constant is discovered to be + * needed, we will likely need it again, so we don't want to deallocate + * it and have to allocate and load it again unnecessarily. + */ +static inline void +setup_optional_register(struct spe_function *f, boolean *is_already_set, unsigned int *r) +{ + if (*is_already_set) return; + *r = spe_allocate_available_register(f); + *is_already_set = true; +} + +static inline void +release_optional_register(struct spe_function *f, boolean *is_already_set, unsigned int r) +{ + if (!*is_already_set) return; + spe_release_register(f, r); + *is_already_set = false; +} + +static inline void +setup_const_register(struct spe_function *f, boolean *is_already_set, unsigned int *r, float value) +{ + if (*is_already_set) return; + setup_optional_register(f, is_already_set, r); + spe_load_float(f, *r, value); +} +static inline void +release_const_register(struct spe_function *f, boolean *is_already_set, unsigned int r) +{ + release_optional_register(f, is_already_set, r); +} /** * Generate SPE code to implement the given blend mode for a quad of pixels. @@ -242,6 +283,7 @@ gen_alpha_test(const struct pipe_depth_stencil_alpha_state *dsa, */ static void gen_blend(const struct pipe_blend_state *blend, + const struct pipe_blend_color *blend_color, struct spe_function *f, enum pipe_format color_format, int fragR_reg, int fragG_reg, int fragB_reg, int fragA_reg, @@ -262,10 +304,17 @@ gen_blend(const struct pipe_blend_state *blend, int fbB_reg = spe_allocate_available_register(f); int fbA_reg = spe_allocate_available_register(f); - int one_reg = spe_allocate_available_register(f); int tmp_reg = spe_allocate_available_register(f); - boolean one_reg_set = false; /* avoid setting one_reg more than once */ + /* Optional constant registers we might or might not end up using; + * if we do use them, make sure we only allocate them once by + * keeping a flag on each one. + */ + boolean one_reg_set = false; + unsigned int one_reg; + boolean constR_reg_set = false, constG_reg_set = false, + constB_reg_set = false, constA_reg_set = false; + unsigned int constR_reg, constG_reg, constB_reg, constA_reg; ASSERT(blend->blend_enable); @@ -346,127 +395,449 @@ gen_blend(const struct pipe_blend_state *blend, spe_release_register(f, mask_reg); } - /* - * Compute Src RGB terms + * Compute Src RGB terms. We're actually looking for the value + * of (the appropriate RGB factors) * (the incoming source RGB color), + * because in some cases (like PIPE_BLENDFACTOR_ONE and + * PIPE_BLENDFACTOR_ZERO) we can avoid doing unnecessary math. */ switch (blend->rgb_src_factor) { case PIPE_BLENDFACTOR_ONE: + /* factors = (1,1,1), so term = (R,G,B) */ spe_move(f, term1R_reg, fragR_reg); spe_move(f, term1G_reg, fragG_reg); spe_move(f, term1B_reg, fragB_reg); break; case PIPE_BLENDFACTOR_ZERO: - spe_zero(f, term1R_reg); - spe_zero(f, term1G_reg); - spe_zero(f, term1B_reg); + /* factors = (0,0,0), so term = (0,0,0) */ + spe_load_float(f, term1R_reg, 0.0f); + spe_load_float(f, term1G_reg, 0.0f); + spe_load_float(f, term1B_reg, 0.0f); break; case PIPE_BLENDFACTOR_SRC_COLOR: + /* factors = (R,G,B), so term = (R*R, G*G, B*B) */ spe_fm(f, term1R_reg, fragR_reg, fragR_reg); spe_fm(f, term1G_reg, fragG_reg, fragG_reg); spe_fm(f, term1B_reg, fragB_reg, fragB_reg); break; case PIPE_BLENDFACTOR_SRC_ALPHA: + /* factors = (A,A,A), so term = (R*A, G*A, B*A) */ spe_fm(f, term1R_reg, fragR_reg, fragA_reg); spe_fm(f, term1G_reg, fragG_reg, fragA_reg); spe_fm(f, term1B_reg, fragB_reg, fragA_reg); break; - /* XXX more cases */ + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* factors = (1-R,1-G,1-B), so term = (R*(1-R), G*(1-G), B*(1-B)) + * or in other words term = (R-R*R, G-G*G, B-B*B) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term1R_reg, fragR_reg, fragR_reg, fragR_reg); + spe_fnms(f, term1G_reg, fragG_reg, fragG_reg, fragG_reg); + spe_fnms(f, term1B_reg, fragB_reg, fragB_reg, fragB_reg); + break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* factors = (Rfb,Gfb,Bfb), so term = (R*Rfb, G*Gfb, B*Bfb) */ + spe_fm(f, term1R_reg, fragR_reg, fbR_reg); + spe_fm(f, term1G_reg, fragG_reg, fbG_reg); + spe_fm(f, term1B_reg, fragB_reg, fbB_reg); + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* factors = (1-Rfb,1-Gfb,1-Bfb), so term = (R*(1-Rfb),G*(1-Gfb),B*(1-Bfb)) + * or term = (R-R*Rfb, G-G*Gfb, B-B*Bfb) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term1R_reg, fragR_reg, fbR_reg, fragR_reg); + spe_fnms(f, term1G_reg, fragG_reg, fbG_reg, fragG_reg); + spe_fnms(f, term1B_reg, fragB_reg, fbB_reg, fragB_reg); + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + /* factors = (1-A,1-A,1-A), so term = (R*(1-A),G*(1-A),B*(1-A)) + * or term = (R-R*A,G-G*A,B-B*A) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term1R_reg, fragR_reg, fragA_reg, fragR_reg); + spe_fnms(f, term1G_reg, fragG_reg, fragA_reg, fragG_reg); + spe_fnms(f, term1B_reg, fragB_reg, fragA_reg, fragB_reg); + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + /* factors = (Afb, Afb, Afb), so term = (R*Afb, G*Afb, B*Afb) */ + spe_fm(f, term1R_reg, fragR_reg, fbA_reg); + spe_fm(f, term1G_reg, fragG_reg, fbA_reg); + spe_fm(f, term1B_reg, fragB_reg, fbA_reg); + break; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + /* factors = (1-Afb, 1-Afb, 1-Afb), so term = (R*(1-Afb),G*(1-Afb),B*(1-Afb)) + * or term = (R-R*Afb,G-G*Afb,b-B*Afb) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term1R_reg, fragR_reg, fbA_reg, fragR_reg); + spe_fnms(f, term1G_reg, fragG_reg, fbA_reg, fragG_reg); + spe_fnms(f, term1B_reg, fragB_reg, fbA_reg, fragB_reg); + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* We need the optional constant color registers */ + setup_const_register(f, &constR_reg_set, &constR_reg, blend_color->color[0]); + setup_const_register(f, &constG_reg_set, &constG_reg, blend_color->color[1]); + setup_const_register(f, &constB_reg_set, &constB_reg, blend_color->color[2]); + /* now, factor = (Rc,Gc,Bc), so term = (R*Rc,G*Gc,B*Bc) */ + spe_fm(f, term1R_reg, fragR_reg, constR_reg); + spe_fm(f, term1G_reg, fragG_reg, constG_reg); + spe_fm(f, term1B_reg, fragB_reg, constB_reg); + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + /* we'll need the optional constant alpha register */ + setup_const_register(f, &constA_reg_set, &constA_reg, blend_color->color[3]); + /* factor = (Ac,Ac,Ac), so term = (R*Ac,G*Ac,B*Ac) */ + spe_fm(f, term1R_reg, fragR_reg, constA_reg); + spe_fm(f, term1G_reg, fragG_reg, constA_reg); + spe_fm(f, term1B_reg, fragB_reg, constA_reg); + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* We need the optional constant color registers */ + setup_const_register(f, &constR_reg_set, &constR_reg, blend_color->color[0]); + setup_const_register(f, &constG_reg_set, &constG_reg, blend_color->color[1]); + setup_const_register(f, &constB_reg_set, &constB_reg, blend_color->color[2]); + /* factor = (1-Rc,1-Gc,1-Bc), so term = (R*(1-Rc),G*(1-Gc),B*(1-Bc)) + * or term = (R-R*Rc, G-G*Gc, B-B*Bc) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term1R_reg, fragR_reg, constR_reg, fragR_reg); + spe_fnms(f, term1G_reg, fragG_reg, constG_reg, fragG_reg); + spe_fnms(f, term1B_reg, fragB_reg, constB_reg, fragB_reg); + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + /* We need the optional constant color registers */ + setup_const_register(f, &constR_reg_set, &constR_reg, blend_color->color[0]); + setup_const_register(f, &constG_reg_set, &constG_reg, blend_color->color[1]); + setup_const_register(f, &constB_reg_set, &constB_reg, blend_color->color[2]); + /* factor = (1-Ac,1-Ac,1-Ac), so term = (R*(1-Ac),G*(1-Ac),B*(1-Ac)) + * or term = (R-R*Ac,G-G*Ac,B-B*Ac) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term1R_reg, fragR_reg, constA_reg, fragR_reg); + spe_fnms(f, term1G_reg, fragG_reg, constA_reg, fragG_reg); + spe_fnms(f, term1B_reg, fragB_reg, constA_reg, fragB_reg); + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + /* We'll need the optional {1,1,1,1} register */ + setup_const_register(f, &one_reg_set, &one_reg, 1.0f); + /* factor = (min(A,1-Afb),min(A,1-Afb),min(A,1-Afb)), so + * term = (R*min(A,1-Afb), G*min(A,1-Afb), B*min(A,1-Afb)) + * We could expand the term (as a*min(b,c) == min(a*b,a*c) + * as long as a is positive), but then we'd have to do three + * spe_float_min() functions instead of one, so this is simpler. + */ + /* tmp = 1 - Afb */ + spe_fs(f, tmp_reg, one_reg, fbA_reg); + /* tmp = min(A,tmp) */ + spe_float_min(f, tmp_reg, fragA_reg, tmp_reg); + /* term = R*tmp */ + spe_fm(f, term1R_reg, fragR_reg, tmp_reg); + spe_fm(f, term1G_reg, fragG_reg, tmp_reg); + spe_fm(f, term1B_reg, fragB_reg, tmp_reg); + break; + + /* These are special D3D cases involving a second color output + * from the fragment shader. I'm not sure we can support them + * yet... XXX + */ + case PIPE_BLENDFACTOR_SRC1_COLOR: + case PIPE_BLENDFACTOR_SRC1_ALPHA: + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + default: ASSERT(0); } /* - * Compute Src Alpha term + * Compute Src Alpha term. Like the above, we're looking for + * the full term A*factor, not just the factor itself, because + * in many cases we can avoid doing unnecessary multiplies. */ switch (blend->alpha_src_factor) { + case PIPE_BLENDFACTOR_ZERO: + /* factor = 0, so term = 0 */ + spe_load_float(f, term1A_reg, 0.0f); + break; + + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: /* fall through */ case PIPE_BLENDFACTOR_ONE: + /* factor = 1, so term = A */ spe_move(f, term1A_reg, fragA_reg); break; + case PIPE_BLENDFACTOR_SRC_COLOR: + /* factor = A, so term = A*A */ spe_fm(f, term1A_reg, fragA_reg, fragA_reg); break; case PIPE_BLENDFACTOR_SRC_ALPHA: spe_fm(f, term1A_reg, fragA_reg, fragA_reg); break; - /* XXX more cases */ + + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* factor = 1-A, so term = A*(1-A) = A-A*A */ + /* fnms(a,b,c,d) computes a = d - b*c */ + spe_fnms(f, term1A_reg, fragA_reg, fragA_reg, fragA_reg); + break; + + case PIPE_BLENDFACTOR_DST_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_DST_COLOR: + /* factor = Afb, so term = A*Afb */ + spe_fm(f, term1A_reg, fragA_reg, fbA_reg); + break; + + case PIPE_BLENDFACTOR_INV_DST_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* factor = 1-Afb, so term = A*(1-Afb) = A - A*Afb */ + /* fnms(a,b,c,d) computes a = d - b*c */ + spe_fnms(f, term1A_reg, fragA_reg, fbA_reg, fragA_reg); + break; + + case PIPE_BLENDFACTOR_CONST_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_CONST_COLOR: + /* We need the optional constA_reg register */ + setup_const_register(f, &constA_reg_set, &constA_reg, blend_color->color[3]); + /* factor = Ac, so term = A*Ac */ + spe_fm(f, term1A_reg, fragA_reg, constA_reg); + break; + + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* We need the optional constA_reg register */ + setup_const_register(f, &constA_reg_set, &constA_reg, blend_color->color[3]); + /* factor = 1-Ac, so term = A*(1-Ac) = A-A*Ac */ + /* fnms(a,b,c,d) computes a = d - b*c */ + spe_fnms(f, term1A_reg, fragA_reg, constA_reg, fragA_reg); + break; + + /* These are special D3D cases involving a second color output + * from the fragment shader. I'm not sure we can support them + * yet... XXX + */ + case PIPE_BLENDFACTOR_SRC1_COLOR: + case PIPE_BLENDFACTOR_SRC1_ALPHA: + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: default: ASSERT(0); } /* - * Compute Dest RGB terms + * Compute Dest RGB term. Like the above, we're looking for + * the full term (Rfb,Gfb,Bfb)*(factor), not just the factor itself, because + * in many cases we can avoid doing unnecessary multiplies. */ switch (blend->rgb_dst_factor) { case PIPE_BLENDFACTOR_ONE: + /* factors = (1,1,1), so term = (Rfb,Gfb,Bfb) */ spe_move(f, term2R_reg, fbR_reg); spe_move(f, term2G_reg, fbG_reg); spe_move(f, term2B_reg, fbB_reg); break; case PIPE_BLENDFACTOR_ZERO: - spe_zero(f, term2R_reg); - spe_zero(f, term2G_reg); - spe_zero(f, term2B_reg); + /* factor s= (0,0,0), so term = (0,0,0) */ + spe_load_float(f, term2R_reg, 0.0f); + spe_load_float(f, term2G_reg, 0.0f); + spe_load_float(f, term2B_reg, 0.0f); break; case PIPE_BLENDFACTOR_SRC_COLOR: + /* factors = (R,G,B), so term = (R*Rfb, G*Gfb, B*Bfb) */ spe_fm(f, term2R_reg, fbR_reg, fragR_reg); spe_fm(f, term2G_reg, fbG_reg, fragG_reg); spe_fm(f, term2B_reg, fbB_reg, fragB_reg); break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* factors = (1-R,1-G,1-B), so term = (Rfb*(1-R), Gfb*(1-G), Bfb*(1-B)) + * or in other words term = (Rfb-Rfb*R, Gfb-Gfb*G, Bfb-Bfb*B) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term2R_reg, fragR_reg, fbR_reg, fbR_reg); + spe_fnms(f, term2G_reg, fragG_reg, fbG_reg, fbG_reg); + spe_fnms(f, term2B_reg, fragB_reg, fbB_reg, fbB_reg); + break; case PIPE_BLENDFACTOR_SRC_ALPHA: + /* factors = (A,A,A), so term = (Rfb*A, Gfb*A, Bfb*A) */ spe_fm(f, term2R_reg, fbR_reg, fragA_reg); spe_fm(f, term2G_reg, fbG_reg, fragA_reg); spe_fm(f, term2B_reg, fbB_reg, fragA_reg); break; case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - /* one = {1.0, 1.0, 1.0, 1.0} */ - if (!one_reg_set) { - spe_load_float(f, one_reg, 1.0f); - one_reg_set = true; - } - /* tmp = one - fragA */ - spe_fs(f, tmp_reg, one_reg, fragA_reg); - /* term = fb * tmp */ - spe_fm(f, term2R_reg, fbR_reg, tmp_reg); - spe_fm(f, term2G_reg, fbG_reg, tmp_reg); - spe_fm(f, term2B_reg, fbB_reg, tmp_reg); - break; - /* XXX more cases */ + /* factors = (1-A,1-A,1-A) so term = (Rfb-Rfb*A,Gfb-Gfb*A,Bfb-Bfb*A) */ + /* fnms(a,b,c,d) computes a = d - b*c */ + spe_fnms(f, term2R_reg, fbR_reg, fragA_reg, fbR_reg); + spe_fnms(f, term2G_reg, fbG_reg, fragA_reg, fbG_reg); + spe_fnms(f, term2B_reg, fbB_reg, fragA_reg, fbB_reg); + break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* factors = (Rfb,Gfb,Bfb), so term = (Rfb*Rfb, Gfb*Gfb, Bfb*Bfb) */ + spe_fm(f, term2R_reg, fbR_reg, fbR_reg); + spe_fm(f, term2G_reg, fbG_reg, fbG_reg); + spe_fm(f, term2B_reg, fbB_reg, fbB_reg); + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* factors = (1-Rfb,1-Gfb,1-Bfb), so term = (Rfb*(1-Rfb),Gfb*(1-Gfb),Bfb*(1-Bfb)) + * or term = (Rfb-Rfb*Rfb, Gfb-Gfb*Gfb, Bfb-Bfb*Bfb) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term2R_reg, fbR_reg, fbR_reg, fbR_reg); + spe_fnms(f, term2G_reg, fbG_reg, fbG_reg, fbG_reg); + spe_fnms(f, term2B_reg, fbB_reg, fbB_reg, fbB_reg); + break; + + case PIPE_BLENDFACTOR_DST_ALPHA: + /* factors = (Afb, Afb, Afb), so term = (Rfb*Afb, Gfb*Afb, Bfb*Afb) */ + spe_fm(f, term2R_reg, fbR_reg, fbA_reg); + spe_fm(f, term2G_reg, fbG_reg, fbA_reg); + spe_fm(f, term2B_reg, fbB_reg, fbA_reg); + break; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + /* factors = (1-Afb, 1-Afb, 1-Afb), so term = (Rfb*(1-Afb),Gfb*(1-Afb),Bfb*(1-Afb)) + * or term = (Rfb-Rfb*Afb,Gfb-Gfb*Afb,Bfb-Bfb*Afb) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term2R_reg, fbR_reg, fbA_reg, fbR_reg); + spe_fnms(f, term2G_reg, fbG_reg, fbA_reg, fbG_reg); + spe_fnms(f, term2B_reg, fbB_reg, fbA_reg, fbB_reg); + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* We need the optional constant color registers */ + setup_const_register(f, &constR_reg_set, &constR_reg, blend_color->color[0]); + setup_const_register(f, &constG_reg_set, &constG_reg, blend_color->color[1]); + setup_const_register(f, &constB_reg_set, &constB_reg, blend_color->color[2]); + /* now, factor = (Rc,Gc,Bc), so term = (Rfb*Rc,Gfb*Gc,Bfb*Bc) */ + spe_fm(f, term2R_reg, fbR_reg, constR_reg); + spe_fm(f, term2G_reg, fbG_reg, constG_reg); + spe_fm(f, term2B_reg, fbB_reg, constB_reg); + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + /* we'll need the optional constant alpha register */ + setup_const_register(f, &constA_reg_set, &constA_reg, blend_color->color[3]); + /* factor = (Ac,Ac,Ac), so term = (Rfb*Ac,Gfb*Ac,Bfb*Ac) */ + spe_fm(f, term2R_reg, fbR_reg, constA_reg); + spe_fm(f, term2G_reg, fbG_reg, constA_reg); + spe_fm(f, term2B_reg, fbB_reg, constA_reg); + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* We need the optional constant color registers */ + setup_const_register(f, &constR_reg_set, &constR_reg, blend_color->color[0]); + setup_const_register(f, &constG_reg_set, &constG_reg, blend_color->color[1]); + setup_const_register(f, &constB_reg_set, &constB_reg, blend_color->color[2]); + /* factor = (1-Rc,1-Gc,1-Bc), so term = (Rfb*(1-Rc),Gfb*(1-Gc),Bfb*(1-Bc)) + * or term = (Rfb-Rfb*Rc, Gfb-Gfb*Gc, Bfb-Bfb*Bc) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term2R_reg, fbR_reg, constR_reg, fbR_reg); + spe_fnms(f, term2G_reg, fbG_reg, constG_reg, fbG_reg); + spe_fnms(f, term2B_reg, fbB_reg, constB_reg, fbB_reg); + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + /* We need the optional constant color registers */ + setup_const_register(f, &constR_reg_set, &constR_reg, blend_color->color[0]); + setup_const_register(f, &constG_reg_set, &constG_reg, blend_color->color[1]); + setup_const_register(f, &constB_reg_set, &constB_reg, blend_color->color[2]); + /* factor = (1-Ac,1-Ac,1-Ac), so term = (Rfb*(1-Ac),Gfb*(1-Ac),Bfb*(1-Ac)) + * or term = (Rfb-Rfb*Ac,Gfb-Gfb*Ac,Bfb-Bfb*Ac) + * fnms(a,b,c,d) computes a = d - b*c + */ + spe_fnms(f, term2R_reg, fbR_reg, constA_reg, fbR_reg); + spe_fnms(f, term2G_reg, fbG_reg, constA_reg, fbG_reg); + spe_fnms(f, term2B_reg, fbB_reg, constA_reg, fbB_reg); + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: /* not supported for dest RGB */ + ASSERT(0); + break; + + /* These are special D3D cases involving a second color output + * from the fragment shader. I'm not sure we can support them + * yet... XXX + */ + case PIPE_BLENDFACTOR_SRC1_COLOR: + case PIPE_BLENDFACTOR_SRC1_ALPHA: + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + default: ASSERT(0); } /* - * Compute Dest Alpha term + * Compute Dest Alpha term. Like the above, we're looking for + * the full term Afb*factor, not just the factor itself, because + * in many cases we can avoid doing unnecessary multiplies. */ switch (blend->alpha_dst_factor) { case PIPE_BLENDFACTOR_ONE: + /* factor = 1, so term = Afb */ spe_move(f, term2A_reg, fbA_reg); break; case PIPE_BLENDFACTOR_ZERO: - spe_zero(f, term2A_reg); + /* factor = 0, so term = 0 */ + spe_load_float(f, term2A_reg, 0.0f); break; - case PIPE_BLENDFACTOR_SRC_ALPHA: + + case PIPE_BLENDFACTOR_SRC_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_SRC_COLOR: + /* factor = A, so term = Afb*A */ spe_fm(f, term2A_reg, fbA_reg, fragA_reg); break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - /* one = {1.0, 1.0, 1.0, 1.0} */ - if (!one_reg_set) { - spe_load_float(f, one_reg, 1.0f); - one_reg_set = true; - } - /* tmp = one - fragA */ - spe_fs(f, tmp_reg, one_reg, fragA_reg); - /* termA = fbA * tmp */ - spe_fm(f, term2A_reg, fbA_reg, tmp_reg); + + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* factor = 1-A, so term = Afb*(1-A) = Afb-Afb*A */ + /* fnms(a,b,c,d) computes a = d - b*c */ + spe_fnms(f, term2A_reg, fbA_reg, fragA_reg, fbA_reg); + break; + + case PIPE_BLENDFACTOR_DST_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_DST_COLOR: + /* factor = Afb, so term = Afb*Afb */ + spe_fm(f, term2A_reg, fbA_reg, fbA_reg); + break; + + case PIPE_BLENDFACTOR_INV_DST_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* factor = 1-Afb, so term = Afb*(1-Afb) = Afb - Afb*Afb */ + /* fnms(a,b,c,d) computes a = d - b*c */ + spe_fnms(f, term2A_reg, fbA_reg, fbA_reg, fbA_reg); + break; + + case PIPE_BLENDFACTOR_CONST_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_CONST_COLOR: + /* We need the optional constA_reg register */ + setup_const_register(f, &constA_reg_set, &constA_reg, blend_color->color[3]); + /* factor = Ac, so term = Afb*Ac */ + spe_fm(f, term2A_reg, fbA_reg, constA_reg); + break; + + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: /* fall through */ + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* We need the optional constA_reg register */ + setup_const_register(f, &constA_reg_set, &constA_reg, blend_color->color[3]); + /* factor = 1-Ac, so term = Afb*(1-Ac) = Afb-Afb*Ac */ + /* fnms(a,b,c,d) computes a = d - b*c */ + spe_fnms(f, term2A_reg, fbA_reg, constA_reg, fbA_reg); + break; + + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: /* not supported for dest alpha */ + ASSERT(0); break; - /* XXX more cases */ + + /* These are special D3D cases involving a second color output + * from the fragment shader. I'm not sure we can support them + * yet... XXX + */ + case PIPE_BLENDFACTOR_SRC1_COLOR: + case PIPE_BLENDFACTOR_SRC1_ALPHA: + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: default: ASSERT(0); } /* - * Combine Src/Dest RGB terms + * Combine Src/Dest RGB terms as per the blend equation. */ switch (blend->rgb_func) { case PIPE_BLEND_ADD: @@ -479,7 +850,21 @@ gen_blend(const struct pipe_blend_state *blend, spe_fs(f, fragG_reg, term1G_reg, term2G_reg); spe_fs(f, fragB_reg, term1B_reg, term2B_reg); break; - /* XXX more cases */ + case PIPE_BLEND_REVERSE_SUBTRACT: + spe_fs(f, fragR_reg, term2R_reg, term1R_reg); + spe_fs(f, fragG_reg, term2G_reg, term1G_reg); + spe_fs(f, fragB_reg, term2B_reg, term1B_reg); + break; + case PIPE_BLEND_MIN: + spe_float_min(f, fragR_reg, term1R_reg, term2R_reg); + spe_float_min(f, fragG_reg, term1G_reg, term2G_reg); + spe_float_min(f, fragB_reg, term1B_reg, term2B_reg); + break; + case PIPE_BLEND_MAX: + spe_float_max(f, fragR_reg, term1R_reg, term2R_reg); + spe_float_max(f, fragG_reg, term1G_reg, term2G_reg); + spe_float_max(f, fragB_reg, term1B_reg, term2B_reg); + break; default: ASSERT(0); } @@ -494,7 +879,15 @@ gen_blend(const struct pipe_blend_state *blend, case PIPE_BLEND_SUBTRACT: spe_fs(f, fragA_reg, term1A_reg, term2A_reg); break; - /* XXX more cases */ + case PIPE_BLEND_REVERSE_SUBTRACT: + spe_fs(f, fragA_reg, term2A_reg, term1A_reg); + break; + case PIPE_BLEND_MIN: + spe_float_min(f, fragA_reg, term1A_reg, term2A_reg); + break; + case PIPE_BLEND_MAX: + spe_float_max(f, fragA_reg, term1A_reg, term2A_reg); + break; default: ASSERT(0); } @@ -514,8 +907,14 @@ gen_blend(const struct pipe_blend_state *blend, spe_release_register(f, fbB_reg); spe_release_register(f, fbA_reg); - spe_release_register(f, one_reg); spe_release_register(f, tmp_reg); + + /* Free any optional registers that actually got used */ + release_const_register(f, &one_reg_set, one_reg); + release_const_register(f, &constR_reg_set, constR_reg); + release_const_register(f, &constG_reg_set, constG_reg); + release_const_register(f, &constB_reg_set, constB_reg); + release_const_register(f, &constA_reg_set, constA_reg); } @@ -524,24 +923,74 @@ gen_logicop(const struct pipe_blend_state *blend, struct spe_function *f, int fragRGBA_reg, int fbRGBA_reg) { - /* XXX to-do */ - /* operate on 32-bit packed pixels, not float colors */ -} - - -static void -gen_colormask(uint colormask, - struct spe_function *f, - int fragRGBA_reg, int fbRGBA_reg) -{ - /* XXX to-do */ - /* operate on 32-bit packed pixels, not float colors */ + /* We've got four 32-bit RGBA packed pixels in each of + * fragRGBA_reg and fbRGBA_reg, not sets of floating-point + * reds, greens, blues, and alphas. + * */ + ASSERT(blend->logicop_enable); + + switch(blend->logicop_func) { + case PIPE_LOGICOP_CLEAR: /* 0 */ + spe_zero(f, fragRGBA_reg); + break; + case PIPE_LOGICOP_NOR: /* ~(s | d) */ + spe_nor(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg); + break; + case PIPE_LOGICOP_AND_INVERTED: /* ~s & d */ + /* andc R, A, B computes R = A & ~B */ + spe_andc(f, fragRGBA_reg, fbRGBA_reg, fragRGBA_reg); + break; + case PIPE_LOGICOP_COPY_INVERTED: /* ~s */ + spe_complement(f, fragRGBA_reg, fragRGBA_reg); + break; + case PIPE_LOGICOP_AND_REVERSE: /* s & ~d */ + /* andc R, A, B computes R = A & ~B */ + spe_andc(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg); + break; + case PIPE_LOGICOP_INVERT: /* ~d */ + /* Note that (A nor A) == ~(A|A) == ~A */ + spe_nor(f, fragRGBA_reg, fbRGBA_reg, fbRGBA_reg); + break; + case PIPE_LOGICOP_XOR: /* s ^ d */ + spe_xor(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg); + break; + case PIPE_LOGICOP_NAND: /* ~(s & d) */ + spe_nand(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg); + break; + case PIPE_LOGICOP_AND: /* s & d */ + spe_and(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg); + break; + case PIPE_LOGICOP_EQUIV: /* ~(s ^ d) */ + spe_xor(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg); + spe_complement(f, fragRGBA_reg, fragRGBA_reg); + break; + case PIPE_LOGICOP_NOOP: /* d */ + spe_move(f, fragRGBA_reg, fbRGBA_reg); + break; + case PIPE_LOGICOP_OR_INVERTED: /* ~s | d */ + /* orc R, A, B computes R = A | ~B */ + spe_orc(f, fragRGBA_reg, fbRGBA_reg, fragRGBA_reg); + break; + case PIPE_LOGICOP_COPY: /* s */ + break; + case PIPE_LOGICOP_OR_REVERSE: /* s | ~d */ + /* orc R, A, B computes R = A | ~B */ + spe_orc(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg); + break; + case PIPE_LOGICOP_OR: /* s | d */ + spe_or(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg); + break; + case PIPE_LOGICOP_SET: /* 1 */ + spe_load_int(f, fragRGBA_reg, 0xffffffff); + break; + default: + ASSERT(0); + } } - /** - * Generate code to pack a quad of float colors into a four 32-bit integers. + * Generate code to pack a quad of float colors into four 32-bit integers. * * \param f SPE function to append instruction onto. * \param color_format the dest color packing format @@ -557,13 +1006,16 @@ gen_pack_colors(struct spe_function *f, int r_reg, int g_reg, int b_reg, int a_reg, int rgba_reg) { + int rg_reg = spe_allocate_available_register(f); + int ba_reg = spe_allocate_available_register(f); + /* Convert float[4] in [0.0,1.0] to int[4] in [0,~0], with clamping */ spe_cfltu(f, r_reg, r_reg, 32); spe_cfltu(f, g_reg, g_reg, 32); spe_cfltu(f, b_reg, b_reg, 32); spe_cfltu(f, a_reg, a_reg, 32); - /* Shift the most significant bytes to least the significant positions. + /* Shift the most significant bytes to the least significant positions. * I.e.: reg = reg >> 24 */ spe_rotmi(f, r_reg, r_reg, -24); @@ -595,12 +1047,713 @@ gen_pack_colors(struct spe_function *f, * OR-ing all those together gives us four packed colors: * RGBA = {0xffffffff, 0xaa114477, 0xbb225588, 0xcc336699} */ - spe_or(f, rgba_reg, r_reg, g_reg); - spe_or(f, rgba_reg, rgba_reg, b_reg); - spe_or(f, rgba_reg, rgba_reg, a_reg); + spe_or(f, rg_reg, r_reg, g_reg); + spe_or(f, ba_reg, a_reg, b_reg); + spe_or(f, rgba_reg, rg_reg, ba_reg); + + spe_release_register(f, rg_reg); + spe_release_register(f, ba_reg); +} + +static void +gen_colormask(struct spe_function *f, + uint colormask, + enum pipe_format color_format, + int fragRGBA_reg, int fbRGBA_reg) +{ + /* We've got four 32-bit RGBA packed pixels in each of + * fragRGBA_reg and fbRGBA_reg, not sets of floating-point + * reds, greens, blues, and alphas. Further, the pixels + * are packed according to the given color format, not + * necessarily RGBA... + */ + unsigned int r_mask; + unsigned int g_mask; + unsigned int b_mask; + unsigned int a_mask; + + /* Calculate exactly where the bits for any particular color + * end up, so we can mask them correctly. + */ + switch(color_format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + /* ARGB */ + a_mask = 0xff000000; + r_mask = 0x00ff0000; + g_mask = 0x0000ff00; + b_mask = 0x000000ff; + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + /* BGRA */ + b_mask = 0xff000000; + g_mask = 0x00ff0000; + r_mask = 0x0000ff00; + a_mask = 0x000000ff; + break; + default: + ASSERT(0); + } + + /* For each R, G, B, and A component we're supposed to mask out, + * clear its bits. Then our mask operation later will work + * as expected. + */ + if (!(colormask & PIPE_MASK_R)) { + r_mask = 0; + } + if (!(colormask & PIPE_MASK_G)) { + g_mask = 0; + } + if (!(colormask & PIPE_MASK_B)) { + b_mask = 0; + } + if (!(colormask & PIPE_MASK_A)) { + a_mask = 0; + } + + /* Get a temporary register to hold the mask that will be applied to the fragment */ + int colormask_reg = spe_allocate_available_register(f); + + /* The actual mask we're going to use is an OR of the remaining R, G, B, and A + * masks. Load the result value into our temporary register. + */ + spe_load_uint(f, colormask_reg, r_mask | g_mask | b_mask | a_mask); + + /* Use the mask register to select between the fragment color + * values and the frame buffer color values. Wherever the + * mask has a 0 bit, the current frame buffer color should override + * the fragment color. Wherever the mask has a 1 bit, the + * fragment color should persevere. The Select Bits (selb rt, rA, rB, rM) + * instruction will select bits from its first operand rA wherever the + * the mask bits rM are 0, and from its second operand rB wherever the + * mask bits rM are 1. That means that the frame buffer color is the + * first operand, and the fragment color the second. + */ + spe_selb(f, fragRGBA_reg, fbRGBA_reg, fragRGBA_reg, colormask_reg); + + /* Release the temporary register and we're done */ + spe_release_register(f, colormask_reg); } +/* This function is annoyingly similar to gen_depth_test(), above, except + * that instead of comparing two varying values (i.e. fragment and buffer), + * we're comparing a varying value with a static value. As such, we have + * access to the Compare Immediate instructions where we don't in + * gen_depth_test(), which is what makes us very different. + * + * There's some added complexity if there's a non-trivial state->mask + * value; then stencil and reference both must be masked + * + * The return value in the stencil_pass_reg is a bitmask of valid + * fragments that also passed the stencil test. The bitmask of valid + * fragments that failed would be found in (fragment_mask_reg & ~stencil_pass_reg). + */ +static void +gen_stencil_test(struct spe_function *f, const struct pipe_stencil_state *state, + unsigned int stencil_max_value, + unsigned int fragment_mask_reg, unsigned int fbS_reg, + unsigned int stencil_pass_reg) +{ + /* Generate code that puts the set of passing fragments into the stencil_pass_reg + * register, taking into account whether each fragment was active to begin with. + */ + switch (state->func) { + case PIPE_FUNC_EQUAL: + if (state->value_mask == stencil_max_value) { + /* stencil_pass = fragment_mask & (s == reference) */ + spe_compare_equal_uint(f, stencil_pass_reg, fbS_reg, state->ref_value); + spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + } + else { + /* stencil_pass = fragment_mask & ((s&mask) == (reference&mask)) */ + unsigned int tmp_masked_stencil = spe_allocate_available_register(f); + spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->value_mask); + spe_compare_equal_uint(f, stencil_pass_reg, tmp_masked_stencil, state->value_mask & state->ref_value); + spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + spe_release_register(f, tmp_masked_stencil); + } + break; + case PIPE_FUNC_NOTEQUAL: + if (state->value_mask == stencil_max_value) { + /* stencil_pass = fragment_mask & ~(s == reference) */ + spe_compare_equal_uint(f, stencil_pass_reg, fbS_reg, state->ref_value); + spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + } + else { + /* stencil_pass = fragment_mask & ~((s&mask) == (reference&mask)) */ + unsigned int tmp_masked_stencil = spe_allocate_available_register(f); + spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->value_mask); + spe_compare_equal_uint(f, stencil_pass_reg, tmp_masked_stencil, state->value_mask & state->ref_value); + spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + spe_release_register(f, tmp_masked_stencil); + } + break; + + case PIPE_FUNC_LESS: + if (state->value_mask == stencil_max_value) { + /* stencil_pass = fragment_mask & (reference < s) */ + spe_compare_greater_uint(f, stencil_pass_reg, fbS_reg, state->ref_value); + spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + } + else { + /* stencil_pass = fragment_mask & ((reference&mask) < (s & mask)) */ + unsigned int tmp_masked_stencil = spe_allocate_available_register(f); + spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->value_mask); + spe_compare_greater_uint(f, stencil_pass_reg, tmp_masked_stencil, state->value_mask & state->ref_value); + spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + spe_release_register(f, tmp_masked_stencil); + } + break; + + case PIPE_FUNC_GREATER: + if (state->value_mask == stencil_max_value) { + /* stencil_pass = fragment_mask & (reference > s) */ + /* There's no convenient Compare Less Than Immediate instruction, so + * we'll have to do this one the harder way, by loading a register and + * comparing directly. Compare Logical Greater Than Word (clgt) + * treats its operands as unsigned - no sign extension. + */ + unsigned int tmp_reg = spe_allocate_available_register(f); + spe_load_uint(f, tmp_reg, state->ref_value); + spe_clgt(f, stencil_pass_reg, tmp_reg, fbS_reg); + spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + spe_release_register(f, tmp_reg); + } + else { + /* stencil_pass = fragment_mask & ((reference&mask) > (s&mask)) */ + unsigned int tmp_reg = spe_allocate_available_register(f); + unsigned int tmp_masked_stencil = spe_allocate_available_register(f); + spe_load_uint(f, tmp_reg, state->value_mask & state->ref_value); + spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->value_mask); + spe_clgt(f, stencil_pass_reg, tmp_reg, tmp_masked_stencil); + spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + spe_release_register(f, tmp_reg); + spe_release_register(f, tmp_masked_stencil); + } + break; + + case PIPE_FUNC_GEQUAL: + if (state->value_mask == stencil_max_value) { + /* stencil_pass = fragment_mask & (reference >= s) + * = fragment_mask & ~(s > reference) */ + spe_compare_greater_uint(f, stencil_pass_reg, fbS_reg, state->ref_value); + spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + } + else { + /* stencil_pass = fragment_mask & ~((s&mask) > (reference&mask)) */ + unsigned int tmp_masked_stencil = spe_allocate_available_register(f); + spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->value_mask); + spe_compare_greater_uint(f, stencil_pass_reg, tmp_masked_stencil, state->value_mask & state->ref_value); + spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + spe_release_register(f, tmp_masked_stencil); + } + break; + + case PIPE_FUNC_LEQUAL: + if (state->value_mask == stencil_max_value) { + /* stencil_pass = fragment_mask & (reference <= s) ] + * = fragment_mask & ~(reference > s) */ + /* As above, we have to do this by loading a register */ + unsigned int tmp_reg = spe_allocate_available_register(f); + spe_load_uint(f, tmp_reg, state->ref_value); + spe_clgt(f, stencil_pass_reg, tmp_reg, fbS_reg); + spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + spe_release_register(f, tmp_reg); + } + else { + /* stencil_pass = fragment_mask & ~((reference&mask) > (s&mask)) */ + unsigned int tmp_reg = spe_allocate_available_register(f); + unsigned int tmp_masked_stencil = spe_allocate_available_register(f); + spe_load_uint(f, tmp_reg, state->ref_value & state->value_mask); + spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->value_mask); + spe_clgt(f, stencil_pass_reg, tmp_reg, tmp_masked_stencil); + spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg); + spe_release_register(f, tmp_reg); + spe_release_register(f, tmp_masked_stencil); + } + break; + + case PIPE_FUNC_NEVER: + /* stencil_pass = fragment_mask & 0 = 0 */ + spe_load_uint(f, stencil_pass_reg, 0); + break; + + case PIPE_FUNC_ALWAYS: + /* stencil_pass = fragment_mask & 1 = fragment_mask */ + spe_move(f, stencil_pass_reg, fragment_mask_reg); + break; + } + + /* The fragments that passed the stencil test are now in stencil_pass_reg. + * The fragments that failed would be (fragment_mask_reg & ~stencil_pass_reg). + */ +} + +/* This function generates code that calculates a set of new stencil values + * given the earlier values and the operation to apply. It does not + * apply any tests. It is intended to be called up to 3 times + * (for the stencil fail operation, for the stencil pass-z fail operation, + * and for the stencil pass-z pass operation) to collect up to three + * possible sets of values, and for the caller to combine them based + * on the result of the tests. + * + * stencil_max_value should be (2^n - 1) where n is the number of bits + * in the stencil buffer - in other words, it should be usable as a mask. + */ +static void +gen_stencil_values(struct spe_function *f, unsigned int stencil_op, + unsigned int stencil_ref_value, unsigned int stencil_max_value, + unsigned int fbS_reg, unsigned int newS_reg) +{ + /* The code below assumes that newS_reg and fbS_reg are not the same + * register; if they can be, the calculations below will have to use + * an additional temporary register. For now, mark the assumption + * with an assertion that will fail if they are the same. + */ + ASSERT(fbS_reg != newS_reg); + + /* The code also assumes the the stencil_max_value is of the form + * 2^n-1 and can therefore be used as a mask for the valid bits in + * addition to a maximum. Make sure this is the case as well. + * The clever math below exploits the fact that incrementing a + * binary number serves to flip all the bits of a number starting at + * the LSB and continuing to (and including) the first zero bit + * found. That means that a number and its increment will always + * have at least one bit in common (the high order bit, if nothing + * else) *unless* the number is zero, *or* the number is of a form + * consisting of some number of 1s in the low-order bits followed + * by nothing but 0s in the high-order bits. The latter case + * implies it's of the form 2^n-1. + */ + ASSERT(stencil_max_value > 0 && ((stencil_max_value + 1) & stencil_max_value) == 0); + + switch(stencil_op) { + case PIPE_STENCIL_OP_KEEP: + /* newS = S */ + spe_move(f, newS_reg, fbS_reg); + break; + + case PIPE_STENCIL_OP_ZERO: + /* newS = 0 */ + spe_zero(f, newS_reg); + break; + + case PIPE_STENCIL_OP_REPLACE: + /* newS = stencil reference value */ + spe_load_uint(f, newS_reg, stencil_ref_value); + break; + + case PIPE_STENCIL_OP_INCR: { + /* newS = (s == max ? max : s + 1) */ + unsigned int equals_reg = spe_allocate_available_register(f); + + spe_compare_equal_uint(f, equals_reg, fbS_reg, stencil_max_value); + /* Add Word Immediate computes rT = rA + 10-bit signed immediate */ + spe_ai(f, newS_reg, fbS_reg, 1); + /* Select from the current value or the new value based on the equality test */ + spe_selb(f, newS_reg, newS_reg, fbS_reg, equals_reg); + + spe_release_register(f, equals_reg); + break; + } + case PIPE_STENCIL_OP_DECR: { + /* newS = (s == 0 ? 0 : s - 1) */ + unsigned int equals_reg = spe_allocate_available_register(f); + + spe_compare_equal_uint(f, equals_reg, fbS_reg, 0); + /* Add Word Immediate with a (-1) value works */ + spe_ai(f, newS_reg, fbS_reg, -1); + /* Select from the current value or the new value based on the equality test */ + spe_selb(f, newS_reg, newS_reg, fbS_reg, equals_reg); + + spe_release_register(f, equals_reg); + break; + } + case PIPE_STENCIL_OP_INCR_WRAP: + /* newS = (s == max ? 0 : s + 1), but since max is 2^n-1, we can + * do a normal add and mask off the correct bits + */ + spe_ai(f, newS_reg, fbS_reg, 1); + spe_and_uint(f, newS_reg, newS_reg, stencil_max_value); + break; + + case PIPE_STENCIL_OP_DECR_WRAP: + /* newS = (s == 0 ? max : s - 1), but we'll pull the same mask trick as above */ + spe_ai(f, newS_reg, fbS_reg, -1); + spe_and_uint(f, newS_reg, newS_reg, stencil_max_value); + break; + + case PIPE_STENCIL_OP_INVERT: + /* newS = ~s. We take advantage of the mask/max value to invert only + * the valid bits for the field so we don't have to do an extra "and". + */ + spe_xor_uint(f, newS_reg, fbS_reg, stencil_max_value); + break; + + default: + ASSERT(0); + } +} + + +/* This function generates code to get all the necessary possible + * stencil values. For each of the output registers (fail_reg, + * zfail_reg, and zpass_reg), it either allocates a new register + * and calculates a new set of values based on the stencil operation, + * or it reuses a register allocation and calculation done for an + * earlier (matching) operation, or it reuses the fbS_reg register + * (if the stencil operation is KEEP, which doesn't change the + * stencil buffer). + * + * Since this function allocates a variable number of registers, + * to avoid incurring complex logic to free them, they should + * be allocated after a spe_allocate_register_set() call + * and released by the corresponding spe_release_register_set() call. + */ +static void +gen_get_stencil_values(struct spe_function *f, const struct pipe_stencil_state *stencil, + const unsigned int depth_enabled, + unsigned int fbS_reg, + unsigned int *fail_reg, unsigned int *zfail_reg, + unsigned int *zpass_reg) +{ + unsigned zfail_op; + + /* Stenciling had better be enabled here */ + ASSERT(stencil->enabled); + + /* If the depth test is not enabled, it is treated as though it always + * passes, which means that the zfail_op is not considered - a + * failing stencil test triggers the fail_op, and a passing one + * triggers the zpass_op + * + * As an optimization, override calculation of the zfail_op values + * if they aren't going to be used. By setting the value of + * the operation to PIPE_STENCIL_OP_KEEP, its value will be assumed + * to match the incoming stencil values, and no calculation will + * be done. + */ + if (depth_enabled) { + zfail_op = stencil->zfail_op; + } + else { + zfail_op = PIPE_STENCIL_OP_KEEP; + } + + /* One-sided or front-facing stencil */ + if (stencil->fail_op == PIPE_STENCIL_OP_KEEP) { + *fail_reg = fbS_reg; + } + else { + *fail_reg = spe_allocate_available_register(f); + gen_stencil_values(f, stencil->fail_op, stencil->ref_value, + 0xff, fbS_reg, *fail_reg); + } + + /* Check the possibly overridden value, not the structure value */ + if (zfail_op == PIPE_STENCIL_OP_KEEP) { + *zfail_reg = fbS_reg; + } + else if (zfail_op == stencil->fail_op) { + *zfail_reg = *fail_reg; + } + else { + *zfail_reg = spe_allocate_available_register(f); + gen_stencil_values(f, stencil->zfail_op, stencil->ref_value, + 0xff, fbS_reg, *zfail_reg); + } + + if (stencil->zpass_op == PIPE_STENCIL_OP_KEEP) { + *zpass_reg = fbS_reg; + } + else if (stencil->zpass_op == stencil->fail_op) { + *zpass_reg = *fail_reg; + } + else if (stencil->zpass_op == zfail_op) { + *zpass_reg = *zfail_reg; + } + else { + *zpass_reg = spe_allocate_available_register(f); + gen_stencil_values(f, stencil->zpass_op, stencil->ref_value, + 0xff, fbS_reg, *zpass_reg); + } +} + +/* Note that fbZ_reg may *not* be set on entry, if in fact + * the depth test is not enabled. This function must not use + * the register if depth is not enabled. + */ +static boolean +gen_stencil_depth_test(struct spe_function *f, + const struct pipe_depth_stencil_alpha_state *dsa, + const uint facing, + const int mask_reg, const int fragZ_reg, + const int fbZ_reg, const int fbS_reg) +{ + /* True if we've generated code that could require writeback to the + * depth and/or stencil buffers + */ + boolean modified_buffers = false; + + boolean need_to_calculate_stencil_values; + boolean need_to_writemask_stencil_values; + + struct pipe_stencil_state *stencil; + + /* Registers. We may or may not actually allocate these, depending + * on whether the state values indicate that we need them. + */ + unsigned int stencil_pass_reg, stencil_fail_reg; + unsigned int stencil_fail_values, stencil_pass_depth_fail_values, stencil_pass_depth_pass_values; + unsigned int stencil_writemask_reg; + unsigned int zmask_reg; + unsigned int newS_reg; + + /* Stenciling is quite complex: up to six different configurable stencil + * operations/calculations can be required (three each for front-facing + * and back-facing fragments). Many of those operations will likely + * be identical, so there's good reason to try to avoid calculating + * the same values more than once (which unfortunately makes the code less + * straightforward). + * + * To make register management easier, we start a new + * register set; we can release all the registers in the set at + * once, and avoid having to keep track of exactly which registers + * we allocate. We can still allocate and free registers as + * desired (if we know we no longer need a register), but we don't + * have to spend the complexity to track the more difficult variant + * register usage scenarios. + */ + spe_comment(f, 0, "Allocating stencil register set"); + spe_allocate_register_set(f); + + /* The facing we're given is the fragment facing; it doesn't + * exactly match the stencil facing. If stencil is enabled, + * but two-sided stencil is *not* enabled, we use the same + * stencil settings for both front- and back-facing fragments. + * We only use the "back-facing" stencil for backfacing fragments + * if two-sided stenciling is enabled. + */ + if (facing == CELL_FACING_BACK && dsa->stencil[1].enabled) { + stencil = &dsa->stencil[1]; + } + else { + stencil = &dsa->stencil[0]; + } + + /* Calculate the writemask. If the writemask is trivial (either + * all 0s, meaning that we don't need to calculate any stencil values + * because they're not going to change the stencil anyway, or all 1s, + * meaning that we have to calculate the stencil values but do not + * need to mask them), we can avoid generating code. Don't forget + * that we need to consider backfacing stencil, if enabled. + * + * Note that if the backface stencil is *not* enabled, the backface + * stencil will have the same values as the frontface stencil. + */ + if (stencil->fail_op == PIPE_STENCIL_OP_KEEP && + stencil->zfail_op == PIPE_STENCIL_OP_KEEP && + stencil->zpass_op == PIPE_STENCIL_OP_KEEP) { + need_to_calculate_stencil_values = false; + need_to_writemask_stencil_values = false; + } + else if (stencil->write_mask == 0x0) { + /* All changes are writemasked out, so no need to calculate + * what those changes might be, and no need to write anything back. + */ + need_to_calculate_stencil_values = false; + need_to_writemask_stencil_values = false; + } + else if (stencil->write_mask == 0xff) { + /* Still trivial, but a little less so. We need to write the stencil + * values, but we don't need to mask them. + */ + need_to_calculate_stencil_values = true; + need_to_writemask_stencil_values = false; + } + else { + /* The general case: calculate, mask, and write */ + need_to_calculate_stencil_values = true; + need_to_writemask_stencil_values = true; + + /* While we're here, generate code that calculates what the + * writemask should be. If backface stenciling is enabled, + * and the backface writemask is not the same as the frontface + * writemask, we'll have to generate code that merges the + * two masks into a single effective mask based on fragment facing. + */ + spe_comment(f, 0, "Computing stencil writemask"); + stencil_writemask_reg = spe_allocate_available_register(f); + spe_load_uint(f, stencil_writemask_reg, dsa->stencil[facing].write_mask); + } + + /* At least one-sided stenciling must be on. Generate code that + * runs the stencil test on the basic/front-facing stencil, leaving + * the mask of passing stencil bits in stencil_pass_reg. This mask will + * be used both to mask the set of active pixels, and also to + * determine how the stencil buffer changes. + * + * This test will *not* change the value in mask_reg (because we don't + * yet know whether to apply the two-sided stencil or one-sided stencil). + */ + spe_comment(f, 0, "Running basic stencil test"); + stencil_pass_reg = spe_allocate_available_register(f); + gen_stencil_test(f, stencil, 0xff, mask_reg, fbS_reg, stencil_pass_reg); + + /* Generate code that, given the mask of valid fragments and the + * mask of valid fragments that passed the stencil test, computes + * the mask of valid fragments that failed the stencil test. We + * have to do this before we run a depth test (because the + * depth test should not be performed on fragments that failed the + * stencil test, and because the depth test will update the + * mask of valid fragments based on the results of the depth test). + */ + spe_comment(f, 0, "Computing stencil fail mask and updating fragment mask"); + stencil_fail_reg = spe_allocate_available_register(f); + spe_andc(f, stencil_fail_reg, mask_reg, stencil_pass_reg); + /* Now remove the stenciled-out pixels from the valid fragment mask, + * so we can later use the valid fragment mask in the depth test. + */ + spe_and(f, mask_reg, mask_reg, stencil_pass_reg); + + /* We may not need to calculate stencil values, if the writemask is off */ + if (need_to_calculate_stencil_values) { + /* Generate code that calculates exactly which stencil values we need, + * without calculating the same value twice (say, if two different + * stencil ops have the same value). This code will work for one-sided + * and two-sided stenciling (so that we take into account that operations + * may match between front and back stencils), and will also take into + * account whether the depth test is enabled (if the depth test is off, + * we don't need any of the zfail results, because the depth test always + * is considered to pass if it is disabled). Any register value that + * does not need to be calculated will come back with the same value + * that's in fbS_reg. + * + * This function will allocate a variant number of registers that + * will be released as part of the register set. + */ + spe_comment(f, 0, facing == CELL_FACING_FRONT ? "Computing front-facing stencil values" : "Computing back-facing stencil values"); + gen_get_stencil_values(f, stencil, dsa->depth.enabled, fbS_reg, + &stencil_fail_values, &stencil_pass_depth_fail_values, + &stencil_pass_depth_pass_values); + } + + /* We now have all the stencil values we need. We also need + * the results of the depth test to figure out which + * stencil values will become the new stencil values. (Even if + * we aren't actually calculating stencil values, we need to apply + * the depth test if it's enabled.) + * + * The code generated by gen_depth_test() returns the results of the + * test in the given register, but also alters the mask_reg based + * on the results of the test. + */ + if (dsa->depth.enabled) { + spe_comment(f, 0, "Running stencil depth test"); + zmask_reg = spe_allocate_available_register(f); + modified_buffers |= gen_depth_test(f, dsa, mask_reg, fragZ_reg, fbZ_reg, zmask_reg); + } + + if (need_to_calculate_stencil_values) { + + /* If we need to writemask the stencil values before going into + * the stencil buffer, we'll have to use a new register to + * hold the new values. If not, we can just keep using the + * current register. + */ + if (need_to_writemask_stencil_values) { + newS_reg = spe_allocate_available_register(f); + spe_comment(f, 0, "Saving current stencil values for writemasking"); + spe_move(f, newS_reg, fbS_reg); + } + else { + newS_reg = fbS_reg; + } + + /* Merge in the selected stencil fail values */ + if (stencil_fail_values != fbS_reg) { + spe_comment(f, 0, "Loading stencil fail values"); + spe_selb(f, newS_reg, newS_reg, stencil_fail_values, stencil_fail_reg); + modified_buffers = true; + } + + /* Same for the stencil pass/depth fail values. If this calculation + * is not needed (say, if depth test is off), then the + * stencil_pass_depth_fail_values register will be equal to fbS_reg + * and we'll skip the calculation. + */ + if (stencil_pass_depth_fail_values != fbS_reg) { + /* We don't actually have a stencil pass/depth fail mask yet. + * Calculate it here from the stencil passing mask and the + * depth passing mask. Note that zmask_reg *must* have been + * set above if we're here. + */ + unsigned int stencil_pass_depth_fail_mask = spe_allocate_available_register(f); + spe_comment(f, 0, "Loading stencil pass/depth fail values"); + spe_andc(f, stencil_pass_depth_fail_mask, stencil_pass_reg, zmask_reg); + + spe_selb(f, newS_reg, newS_reg, stencil_pass_depth_fail_values, stencil_pass_depth_fail_mask); + + spe_release_register(f, stencil_pass_depth_fail_mask); + modified_buffers = true; + } + + /* Same for the stencil pass/depth pass mask. Note that we + * *can* get here with zmask_reg being unset (if the depth + * test is off but the stencil test is on). In this case, + * we assume the depth test passes, and don't need to mask + * the stencil pass mask with the Z mask. + */ + if (stencil_pass_depth_pass_values != fbS_reg) { + if (dsa->depth.enabled) { + unsigned int stencil_pass_depth_pass_mask = spe_allocate_available_register(f); + /* We'll need a separate register */ + spe_comment(f, 0, "Loading stencil pass/depth pass values"); + spe_and(f, stencil_pass_depth_pass_mask, stencil_pass_reg, zmask_reg); + spe_selb(f, newS_reg, newS_reg, stencil_pass_depth_pass_values, stencil_pass_depth_pass_mask); + spe_release_register(f, stencil_pass_depth_pass_mask); + } + else { + /* We can use the same stencil-pass register */ + spe_comment(f, 0, "Loading stencil pass values"); + spe_selb(f, newS_reg, newS_reg, stencil_pass_depth_pass_values, stencil_pass_reg); + } + modified_buffers = true; + } + + /* Almost done. If we need to writemask, do it now, leaving the + * results in the fbS_reg register passed in. If we don't need + * to writemask, then the results are *already* in the fbS_reg, + * so there's nothing more to do. + */ + + if (need_to_writemask_stencil_values && modified_buffers) { + /* The Select Bytes command makes a fine writemask. Where + * the mask is 0, the first (original) values are retained, + * effectively masking out changes. Where the mask is 1, the + * second (new) values are retained, incorporating changes. + */ + spe_comment(f, 0, "Writemasking new stencil values"); + spe_selb(f, fbS_reg, fbS_reg, newS_reg, stencil_writemask_reg); + } + + } /* done calculating stencil values */ + + /* The stencil and/or depth values have been applied, and the + * mask_reg, fbS_reg, and fbZ_reg values have been updated. + * We're all done, except that we've allocated a fair number + * of registers that we didn't bother tracking. Release all + * those registers as part of the register set, and go home. + */ + spe_comment(f, 0, "Releasing stencil register set"); + spe_release_register_set(f); + + /* Return true if we could have modified the stencil and/or + * depth buffers. + */ + return modified_buffers; +} /** @@ -621,14 +1774,19 @@ gen_pack_colors(struct spe_function *f, * should be much faster. * * \param cell the rendering context (in) - * \param f the generated function (out) + * \param facing whether the generated code is for front-facing or + * back-facing fragments + * \param f the generated function (in/out); on input, the function + * must already have been initialized. On exit, whatever + * instructions within the generated function have had + * the fragment ops appended. */ void -cell_gen_fragment_function(struct cell_context *cell, struct spe_function *f) +cell_gen_fragment_function(struct cell_context *cell, const uint facing, struct spe_function *f) { - const struct pipe_depth_stencil_alpha_state *dsa = - &cell->depth_stencil->base; - const struct pipe_blend_state *blend = &cell->blend->base; + const struct pipe_depth_stencil_alpha_state *dsa = cell->depth_stencil; + const struct pipe_blend_state *blend = cell->blend; + const struct pipe_blend_color *blend_color = &cell->blend_color; const enum pipe_format color_format = cell->framebuffer.cbufs[0]->format; /* For SPE function calls: reg $3 = first param, $4 = second param, etc. */ @@ -643,6 +1801,8 @@ cell_gen_fragment_function(struct cell_context *cell, struct spe_function *f) const int fragA_reg = 11; /* vector float */ const int mask_reg = 12; /* vector uint */ + ASSERT(facing == CELL_FACING_FRONT || facing == CELL_FACING_BACK); + /* offset of quad from start of tile * XXX assuming 4-byte pixels for color AND Z/stencil!!!! */ @@ -651,7 +1811,12 @@ cell_gen_fragment_function(struct cell_context *cell, struct spe_function *f) int fbRGBA_reg; /**< framebuffer's RGBA colors for quad */ int fbZS_reg; /**< framebuffer's combined z/stencil values for quad */ - spe_init_func(f, SPU_MAX_FRAGMENT_OPS_INSTS * SPE_INST_SIZE); + if (cell->debug_flags & CELL_DEBUG_ASM) { + spe_print_code(f, true); + spe_indent(f, 8); + spe_comment(f, -4, facing == CELL_FACING_FRONT ? "Begin front-facing per-fragment ops": "Begin back-facing per-fragment ops"); + } + spe_allocate_register(f, x_reg); spe_allocate_register(f, y_reg); spe_allocate_register(f, color_tile_reg); @@ -674,8 +1839,9 @@ cell_gen_fragment_function(struct cell_context *cell, struct spe_function *f) ASSERT(TILE_SIZE == 32); - spe_rotmi(f, x2_reg, x_reg, -1); /* x2 = x / 2 */ + spe_comment(f, 0, "Compute quad offset within tile"); spe_rotmi(f, y2_reg, y_reg, -1); /* y2 = y / 2 */ + spe_rotmi(f, x2_reg, x_reg, -1); /* x2 = x / 2 */ spe_shli(f, y2_reg, y2_reg, 4); /* y2 *= 16 */ spe_a(f, quad_offset_reg, y2_reg, x2_reg); /* offset = y2 + x2 */ spe_shli(f, quad_offset_reg, quad_offset_reg, 4); /* offset *= 16 */ @@ -684,139 +1850,196 @@ cell_gen_fragment_function(struct cell_context *cell, struct spe_function *f) spe_release_register(f, y2_reg); } - + /* Generate the alpha test, if needed. */ if (dsa->alpha.enabled) { gen_alpha_test(dsa, f, mask_reg, fragA_reg); } + /* If we need the stencil buffers (because one- or two-sided stencil is + * enabled) or the depth buffer (because the depth test is enabled), + * go grab them. Note that if either one- or two-sided stencil is + * enabled, dsa->stencil[0].enabled will be true. + */ if (dsa->depth.enabled || dsa->stencil[0].enabled) { const enum pipe_format zs_format = cell->framebuffer.zsbuf->format; boolean write_depth_stencil; - int fbZ_reg = spe_allocate_available_register(f); /* Z values */ - int fbS_reg = spe_allocate_available_register(f); /* Stencil values */ + /* We may or may not need to allocate a register for Z or stencil values */ + boolean fbS_reg_set = false, fbZ_reg_set = false; + unsigned int fbS_reg, fbZ_reg = 0; + + spe_comment(f, 0, "Fetching Z/stencil quad from tile"); /* fetch quad of depth/stencil values from tile at (x,y) */ /* Load: fbZS_reg = memory[depth_tile_reg + offset_reg] */ + /* XXX Not sure this is allowed if we've only got a 16-bit Z buffer... */ spe_lqx(f, fbZS_reg, depth_tile_reg, quad_offset_reg); - if (dsa->depth.enabled) { - /* Extract Z bits from fbZS_reg into fbZ_reg */ - if (zs_format == PIPE_FORMAT_S8Z24_UNORM || - zs_format == PIPE_FORMAT_X8Z24_UNORM) { - int mask_reg = spe_allocate_available_register(f); - spe_fsmbi(f, mask_reg, 0x7777); /* mask[0,1,2,3] = 0x00ffffff */ - spe_and(f, fbZ_reg, fbZS_reg, mask_reg); /* fbZ = fbZS & mask */ - spe_release_register(f, mask_reg); - /* OK, fbZ_reg has four 24-bit Z values now */ - } - else { - /* XXX handle other z/stencil formats */ - ASSERT(0); - } + /* From the Z/stencil buffer format, pull out the bits we need for + * Z and/or stencil. We'll also convert the incoming fragment Z + * value in fragZ_reg from a floating point value in [0.0..1.0] to + * an unsigned integer value with the appropriate resolution. + * Note that even if depth or stencil is *not* enabled, if it's + * present in the buffer, we pull it out and put it back later; + * otherwise, we can inadvertently destroy the contents of + * buffers we're not supposed to touch (e.g., if the user is + * clearing the depth buffer but not the stencil buffer, a + * quad of constant depth is drawn over the surface; the stencil + * buffer must be maintained). + */ + switch(zs_format) { - /* Convert fragZ values from float[4] to uint[4] */ - if (zs_format == PIPE_FORMAT_S8Z24_UNORM || - zs_format == PIPE_FORMAT_X8Z24_UNORM || - zs_format == PIPE_FORMAT_Z24S8_UNORM || - zs_format == PIPE_FORMAT_Z24X8_UNORM) { - /* 24-bit Z values */ - int scale_reg = spe_allocate_available_register(f); + case PIPE_FORMAT_S8Z24_UNORM: /* fall through */ + case PIPE_FORMAT_X8Z24_UNORM: + /* Pull out both Z and stencil */ + setup_optional_register(f, &fbZ_reg_set, &fbZ_reg); + setup_optional_register(f, &fbS_reg_set, &fbS_reg); - /* scale_reg[0,1,2,3] = float(2^24-1) */ - spe_load_float(f, scale_reg, (float) 0xffffff); + /* four 24-bit Z values in the low-order bits */ + spe_and_uint(f, fbZ_reg, fbZS_reg, 0x00ffffff); - /* XXX these two instructions might be combined */ - spe_fm(f, fragZ_reg, fragZ_reg, scale_reg); /* fragZ *= scale */ - spe_cfltu(f, fragZ_reg, fragZ_reg, 0); /* fragZ = (int) fragZ */ + /* Incoming fragZ_reg value is a float in 0.0...1.0; convert + * to a 24-bit unsigned integer + */ + spe_cfltu(f, fragZ_reg, fragZ_reg, 32); + spe_rotmi(f, fragZ_reg, fragZ_reg, -8); - spe_release_register(f, scale_reg); - } - else { - /* XXX handle 16-bit Z format */ - ASSERT(0); - } - } + /* four 8-bit stencil values in the high-order bits */ + spe_rotmi(f, fbS_reg, fbZS_reg, -24); + break; - if (dsa->stencil[0].enabled) { - /* Extract Stencil bit sfrom fbZS_reg into fbS_reg */ - if (zs_format == PIPE_FORMAT_S8Z24_UNORM || - zs_format == PIPE_FORMAT_X8Z24_UNORM) { - /* XXX extract with a shift */ - ASSERT(0); - } - else if (zs_format == PIPE_FORMAT_Z24S8_UNORM || - zs_format == PIPE_FORMAT_Z24X8_UNORM) { - /* XXX extract with a mask */ - ASSERT(0); - } - } + case PIPE_FORMAT_Z24S8_UNORM: /* fall through */ + case PIPE_FORMAT_Z24X8_UNORM: + setup_optional_register(f, &fbZ_reg_set, &fbZ_reg); + setup_optional_register(f, &fbS_reg_set, &fbS_reg); + + /* shift by 8 to get the upper 24-bit values */ + spe_rotmi(f, fbS_reg, fbZS_reg, -8); + + /* Incoming fragZ_reg value is a float in 0.0...1.0; convert + * to a 24-bit unsigned integer + */ + spe_cfltu(f, fragZ_reg, fragZ_reg, 32); + spe_rotmi(f, fragZ_reg, fragZ_reg, -8); + + /* 8-bit stencil in the low-order bits - mask them out */ + spe_and_uint(f, fbS_reg, fbZS_reg, 0x000000ff); + break; + + case PIPE_FORMAT_Z32_UNORM: + setup_optional_register(f, &fbZ_reg_set, &fbZ_reg); + /* Copy over 4 32-bit values */ + spe_move(f, fbZ_reg, fbZS_reg); + + /* Incoming fragZ_reg value is a float in 0.0...1.0; convert + * to a 32-bit unsigned integer + */ + spe_cfltu(f, fragZ_reg, fragZ_reg, 32); + /* No stencil, so can't do anything there */ + break; + case PIPE_FORMAT_Z16_UNORM: + /* XXX Not sure this is correct, but it was here before, so we're + * going with it for now + */ + setup_optional_register(f, &fbZ_reg_set, &fbZ_reg); + /* Copy over 4 32-bit values */ + spe_move(f, fbZ_reg, fbZS_reg); + + /* Incoming fragZ_reg value is a float in 0.0...1.0; convert + * to a 16-bit unsigned integer + */ + spe_cfltu(f, fragZ_reg, fragZ_reg, 32); + spe_rotmi(f, fragZ_reg, fragZ_reg, -16); + /* No stencil */ + + default: + ASSERT(0); /* invalid format */ + } + /* If stencil is enabled, use the stencil-specific code + * generator to generate both the stencil and depth (if needed) + * tests. Otherwise, if only depth is enabled, generate + * a quick depth test. The test generators themselves will + * report back whether the depth/stencil buffer has to be + * written back. + */ if (dsa->stencil[0].enabled) { - /* XXX this may involve depth testing too */ - // gen_stencil_test(dsa, f, ... ); - ASSERT(0); + /* This will perform the stencil and depth tests, and update + * the mask_reg, fbZ_reg, and fbS_reg as required by the + * tests. + */ + ASSERT(fbS_reg_set); + spe_comment(f, 0, "Perform stencil test"); + + /* Note that fbZ_reg may not be set on entry, if stenciling + * is enabled but there's no Z-buffer. The + * gen_stencil_depth_test() function must ignore the + * fbZ_reg register if depth is not enabled. + */ + write_depth_stencil = gen_stencil_depth_test(f, dsa, facing, mask_reg, fragZ_reg, fbZ_reg, fbS_reg); } else if (dsa->depth.enabled) { int zmask_reg = spe_allocate_available_register(f); - gen_depth_test(dsa, f, mask_reg, fragZ_reg, fbZ_reg, zmask_reg); + ASSERT(fbZ_reg_set); + spe_comment(f, 0, "Perform depth test"); + write_depth_stencil = gen_depth_test(f, dsa, mask_reg, fragZ_reg, fbZ_reg, zmask_reg); spe_release_register(f, zmask_reg); } - - /* do we need to write Z and/or Stencil back into framebuffer? */ - write_depth_stencil = (dsa->depth.writemask | - dsa->stencil[0].write_mask | - dsa->stencil[1].write_mask); + else { + write_depth_stencil = false; + } if (write_depth_stencil) { /* Merge latest Z and Stencil values into fbZS_reg. * fbZ_reg has four Z vals in bits [23..0] or bits [15..0]. * fbS_reg has four 8-bit Z values in bits [7..0]. */ + spe_comment(f, 0, "Store quad's depth/stencil values in tile"); if (zs_format == PIPE_FORMAT_S8Z24_UNORM || zs_format == PIPE_FORMAT_X8Z24_UNORM) { spe_shli(f, fbS_reg, fbS_reg, 24); /* fbS = fbS << 24 */ spe_or(f, fbZS_reg, fbS_reg, fbZ_reg); /* fbZS = fbS | fbZ */ } - else if (zs_format == PIPE_FORMAT_S8Z24_UNORM || - zs_format == PIPE_FORMAT_X8Z24_UNORM) { - /* XXX to do */ - ASSERT(0); + else if (zs_format == PIPE_FORMAT_Z24S8_UNORM || + zs_format == PIPE_FORMAT_Z24X8_UNORM) { + spe_shli(f, fbZ_reg, fbZ_reg, 8); /* fbZ = fbZ << 8 */ + spe_or(f, fbZS_reg, fbS_reg, fbZ_reg); /* fbZS = fbS | fbZ */ + } + else if (zs_format == PIPE_FORMAT_Z32_UNORM) { + spe_move(f, fbZS_reg, fbZ_reg); /* fbZS = fbZ */ } else if (zs_format == PIPE_FORMAT_Z16_UNORM) { - /* XXX to do */ - ASSERT(0); + spe_move(f, fbZS_reg, fbZ_reg); /* fbZS = fbZ */ } else if (zs_format == PIPE_FORMAT_S8_UNORM) { - /* XXX to do */ - ASSERT(0); + ASSERT(0); /* XXX to do */ } else { - /* bad zs_format */ - ASSERT(0); + ASSERT(0); /* bad zs_format */ } /* Store: memory[depth_tile_reg + quad_offset_reg] = fbZS */ spe_stqx(f, fbZS_reg, depth_tile_reg, quad_offset_reg); } - spe_release_register(f, fbZ_reg); - spe_release_register(f, fbS_reg); + /* Don't need these any more */ + release_optional_register(f, &fbZ_reg_set, fbZ_reg); + release_optional_register(f, &fbS_reg_set, fbS_reg); } - /* Get framebuffer quad/colors. We'll need these for blending, * color masking, and to obey the quad/pixel mask. * Load: fbRGBA_reg = memory[color_tile + quad_offset] * Note: if mask={~0,~0,~0,~0} and we're not blending or colormasking * we could skip this load. */ + spe_comment(f, 0, "Fetch quad colors from tile"); spe_lqx(f, fbRGBA_reg, color_tile_reg, quad_offset_reg); - if (blend->blend_enable) { - gen_blend(blend, f, color_format, + spe_comment(f, 0, "Perform blending"); + gen_blend(blend, blend_color, f, color_format, fragR_reg, fragG_reg, fragB_reg, fragA_reg, fbRGBA_reg); } @@ -829,19 +2052,21 @@ cell_gen_fragment_function(struct cell_context *cell, struct spe_function *f) int rgba_reg = spe_allocate_available_register(f); /* Pack four float colors as four 32-bit int colors */ + spe_comment(f, 0, "Convert float quad colors to packed int framebuffer colors"); gen_pack_colors(f, color_format, fragR_reg, fragG_reg, fragB_reg, fragA_reg, rgba_reg); if (blend->logicop_enable) { + spe_comment(f, 0, "Compute logic op"); gen_logicop(blend, f, rgba_reg, fbRGBA_reg); } - if (blend->colormask != 0xf) { - gen_colormask(blend->colormask, f, rgba_reg, fbRGBA_reg); + if (blend->colormask != PIPE_MASK_RGBA) { + spe_comment(f, 0, "Compute color mask"); + gen_colormask(f, blend->colormask, color_format, rgba_reg, fbRGBA_reg); } - /* Mix fragment colors with framebuffer colors using the quad/pixel mask: * if (mask[i]) * rgba[i] = rgba[i]; @@ -853,6 +2078,7 @@ cell_gen_fragment_function(struct cell_context *cell, struct spe_function *f) /* Store updated quad in tile: * memory[color_tile + quad_offset] = rgba_reg; */ + spe_comment(f, 0, "Store quad colors into color tile"); spe_stqx(f, rgba_reg, color_tile_reg, quad_offset_reg); spe_release_register(f, rgba_reg); @@ -862,9 +2088,14 @@ cell_gen_fragment_function(struct cell_context *cell, struct spe_function *f) spe_bi(f, SPE_REG_RA, 0, 0); /* return from function call */ - spe_release_register(f, fbRGBA_reg); spe_release_register(f, fbZS_reg); spe_release_register(f, quad_offset_reg); -} + if (cell->debug_flags & CELL_DEBUG_ASM) { + char buffer[1024]; + sprintf(buffer, "End %s-facing per-fragment ops: %d instructions", + facing == CELL_FACING_FRONT ? "front" : "back", f->num_inst); + spe_comment(f, -4, buffer); + } +} diff --git a/src/gallium/drivers/cell/ppu/cell_gen_fragment.h b/src/gallium/drivers/cell/ppu/cell_gen_fragment.h index b59de198dc..21b35d1faf 100644 --- a/src/gallium/drivers/cell/ppu/cell_gen_fragment.h +++ b/src/gallium/drivers/cell/ppu/cell_gen_fragment.h @@ -31,7 +31,7 @@ extern void -cell_gen_fragment_function(struct cell_context *cell, struct spe_function *f); +cell_gen_fragment_function(struct cell_context *cell, const uint facing, struct spe_function *f); #endif /* CELL_GEN_FRAGMENT_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 475c6ef0ce..81efd137c7 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -35,9 +35,9 @@ #include "draw/draw_context.h" #include "cell_context.h" #include "cell_flush.h" +#include "cell_pipe_state.h" #include "cell_state.h" #include "cell_texture.h" -#include "cell_state_per_fragment.h" @@ -45,24 +45,18 @@ static void * cell_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { - struct cell_blend_state *cb = MALLOC(sizeof(struct cell_blend_state)); - - (void) memcpy(cb, blend, sizeof(*blend)); -#if 0 - cell_generate_alpha_blend(cb); -#endif - return cb; + return mem_dup(blend, sizeof(*blend)); } static void -cell_bind_blend_state(struct pipe_context *pipe, void *state) +cell_bind_blend_state(struct pipe_context *pipe, void *blend) { struct cell_context *cell = cell_context(pipe); draw_flush(cell->draw); - cell->blend = (struct cell_blend_state *) state; + cell->blend = (struct pipe_blend_state *) blend; cell->dirty |= CELL_NEW_BLEND; } @@ -70,10 +64,7 @@ cell_bind_blend_state(struct pipe_context *pipe, void *state) static void cell_delete_blend_state(struct pipe_context *pipe, void *blend) { - struct cell_blend_state *cb = (struct cell_blend_state *) blend; - - spe_release_func(& cb->code); - FREE(cb); + FREE(blend); } @@ -95,41 +86,29 @@ cell_set_blend_color(struct pipe_context *pipe, static void * cell_create_depth_stencil_alpha_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) + const struct pipe_depth_stencil_alpha_state *dsa) { - struct cell_depth_stencil_alpha_state *cdsa = - MALLOC(sizeof(struct cell_depth_stencil_alpha_state)); - - (void) memcpy(cdsa, depth_stencil, sizeof(*depth_stencil)); -#if 0 - cell_generate_depth_stencil_test(cdsa); -#endif - return cdsa; + return mem_dup(dsa, sizeof(*dsa)); } static void cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, - void *depth_stencil) + void *dsa) { struct cell_context *cell = cell_context(pipe); draw_flush(cell->draw); - cell->depth_stencil = - (struct cell_depth_stencil_alpha_state *) depth_stencil; + cell->depth_stencil = (struct pipe_depth_stencil_alpha_state *) dsa; cell->dirty |= CELL_NEW_DEPTH_STENCIL; } static void -cell_delete_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth) +cell_delete_depth_stencil_alpha_state(struct pipe_context *pipe, void *dsa) { - struct cell_depth_stencil_alpha_state *cdsa = - (struct cell_depth_stencil_alpha_state *) depth; - - spe_release_func(& cdsa->code); - FREE(cdsa); + FREE(dsa); } @@ -191,24 +170,23 @@ cell_set_polygon_stipple( struct pipe_context *pipe, static void * cell_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup) + const struct pipe_rasterizer_state *rasterizer) { - struct pipe_rasterizer_state *state - = MALLOC(sizeof(struct pipe_rasterizer_state)); - memcpy(state, setup, sizeof(struct pipe_rasterizer_state)); - return state; + return mem_dup(rasterizer, sizeof(*rasterizer)); } static void -cell_bind_rasterizer_state(struct pipe_context *pipe, void *setup) +cell_bind_rasterizer_state(struct pipe_context *pipe, void *rast) { + struct pipe_rasterizer_state *rasterizer = + (struct pipe_rasterizer_state *) rast; struct cell_context *cell = cell_context(pipe); /* pass-through to draw module */ - draw_set_rasterizer_state(cell->draw, setup); + draw_set_rasterizer_state(cell->draw, rasterizer); - cell->rasterizer = (struct pipe_rasterizer_state *)setup; + cell->rasterizer = rasterizer; cell->dirty |= CELL_NEW_RASTERIZER; } @@ -235,17 +213,24 @@ cell_bind_sampler_states(struct pipe_context *pipe, unsigned num, void **samplers) { struct cell_context *cell = cell_context(pipe); + uint i, changed = 0x0; assert(num <= CELL_MAX_SAMPLERS); draw_flush(cell->draw); - memcpy(cell->sampler, samplers, num * sizeof(void *)); - memset(&cell->sampler[num], 0, (CELL_MAX_SAMPLERS - num) * - sizeof(void *)); - cell->num_samplers = num; + for (i = 0; i < CELL_MAX_SAMPLERS; i++) { + struct pipe_sampler_state *new_samp = i < num ? samplers[i] : NULL; + if (cell->sampler[i] != new_samp) { + cell->sampler[i] = new_samp; + changed |= (1 << i); + } + } - cell->dirty |= CELL_NEW_SAMPLER; + if (changed) { + cell->dirty |= CELL_NEW_SAMPLER; + cell->dirty_samplers |= changed; + } } @@ -263,27 +248,25 @@ cell_set_sampler_textures(struct pipe_context *pipe, unsigned num, struct pipe_texture **texture) { struct cell_context *cell = cell_context(pipe); - uint i; + uint i, changed = 0x0; assert(num <= CELL_MAX_SAMPLERS); - /* Check for no-op */ - if (num == cell->num_textures && - !memcmp(cell->texture, texture, num * sizeof(struct pipe_texture *))) - return; - - draw_flush(cell->draw); - for (i = 0; i < CELL_MAX_SAMPLERS; i++) { - struct pipe_texture *tex = i < num ? texture[i] : NULL; - - pipe_texture_reference((struct pipe_texture **) &cell->texture[i], tex); + struct pipe_texture *new_tex = i < num ? texture[i] : NULL; + if ((struct pipe_texture *) cell->texture[i] != new_tex) { + pipe_texture_reference((struct pipe_texture **) &cell->texture[i], + new_tex); + changed |= (1 << i); + } } - cell->num_textures = num; - cell_update_texture_mapping(cell); + cell->num_textures = num; - cell->dirty |= CELL_NEW_TEXTURE; + if (changed) { + cell->dirty |= CELL_NEW_TEXTURE; + cell->dirty_textures |= changed; + } } diff --git a/src/gallium/drivers/cell/ppu/cell_render.c b/src/gallium/drivers/cell/ppu/cell_render.c index dd25ae880e..79cb8df82f 100644 --- a/src/gallium/drivers/cell/ppu/cell_render.c +++ b/src/gallium/drivers/cell/ppu/cell_render.c @@ -152,6 +152,7 @@ cell_flush_prim_buffer(struct cell_context *cell) struct cell_command_render *render = &cell_global.command[i].render; render->prim_type = PIPE_PRIM_TRIANGLES; render->num_verts = cell->prim_buffer.num_verts; + render->front_winding = cell->rasterizer->front_winding; render->vertex_size = cell->vertex_info->size * 4; render->xmin = cell->prim_buffer.xmin; render->ymin = cell->prim_buffer.ymin; diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c index 139b3719b6..d223557950 100644 --- a/src/gallium/drivers/cell/ppu/cell_screen.c +++ b/src/gallium/drivers/cell/ppu/cell_screen.c @@ -58,9 +58,9 @@ cell_get_param(struct pipe_screen *screen, int param) case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: return CELL_MAX_SAMPLERS; case PIPE_CAP_NPOT_TEXTURES: - return 0; + return 1; case PIPE_CAP_TWO_SIDED_STENCIL: - return 0; + return 1; case PIPE_CAP_GLSL: return 1; case PIPE_CAP_S3TC: @@ -68,21 +68,21 @@ cell_get_param(struct pipe_screen *screen, int param) case PIPE_CAP_ANISOTROPIC_FILTER: return 0; case PIPE_CAP_POINT_SPRITE: - return 0; + return 1; case PIPE_CAP_MAX_RENDER_TARGETS: return 1; case PIPE_CAP_OCCLUSION_QUERY: - return 0; + return 1; case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 0; + return 10; case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: - return 12; /* max 2Kx2K */ + return CELL_MAX_TEXTURE_LEVELS; case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: return 8; /* max 128x128x128 */ case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: - return 12; /* max 2Kx2K */ + return CELL_MAX_TEXTURE_LEVELS; default: - return 0; + return 10; } } @@ -108,7 +108,7 @@ cell_get_paramf(struct pipe_screen *screen, int param) return 16.0; /* arbitrary */ default: - return 0; + return 10; } } diff --git a/src/gallium/drivers/cell/ppu/cell_spu.c b/src/gallium/drivers/cell/ppu/cell_spu.c index 9508227e29..28e5e6d706 100644 --- a/src/gallium/drivers/cell/ppu/cell_spu.c +++ b/src/gallium/drivers/cell/ppu/cell_spu.c @@ -36,6 +36,7 @@ #include "cell_spu.h" #include "pipe/p_format.h" #include "pipe/p_state.h" +#include "util/u_memory.h" #include "cell/common.h" @@ -52,6 +53,35 @@ struct cell_global_info cell_global; /** + * Scan /proc/cpuinfo to determine the timebase for the system. + * This is used by the SPUs to convert 'decrementer' ticks to seconds. + * There may be a better way to get this value... + */ +static unsigned +get_timebase(void) +{ + FILE *f = fopen("/proc/cpuinfo", "r"); + unsigned timebase; + + assert(f); + while (!feof(f)) { + char line[80]; + fgets(line, sizeof(line), f); + if (strncmp(line, "timebase", 8) == 0) { + char *colon = strchr(line, ':'); + if (colon) { + timebase = atoi(colon + 2); + break; + } + } + } + fclose(f); + + return timebase; +} + + +/** * Write a 1-word message to the given SPE mailbox. */ void @@ -114,6 +144,7 @@ cell_start_spus(struct cell_context *cell) { static boolean one_time_init = FALSE; uint i, j; + uint timebase = get_timebase(); if (one_time_init) { fprintf(stderr, "PPU: Multiple rendering contexts not yet supported " @@ -123,24 +154,29 @@ cell_start_spus(struct cell_context *cell) one_time_init = TRUE; - assert(cell->num_spus <= MAX_SPUS); - - ASSERT_ALIGN16(&cell_global.command[0]); - ASSERT_ALIGN16(&cell_global.command[1]); + assert(cell->num_spus <= CELL_MAX_SPUS); ASSERT_ALIGN16(&cell_global.inits[0]); ASSERT_ALIGN16(&cell_global.inits[1]); + /* + * Initialize the global 'inits' structure for each SPU. + * A pointer to the init struct will be passed to each SPU. + * The SPUs will then each grab their init info with mfc_get(). + */ for (i = 0; i < cell->num_spus; i++) { cell_global.inits[i].id = i; cell_global.inits[i].num_spus = cell->num_spus; cell_global.inits[i].debug_flags = cell->debug_flags; - cell_global.inits[i].cmd = &cell_global.command[i]; + cell_global.inits[i].inv_timebase = 1000.0f / timebase; + for (j = 0; j < CELL_NUM_BUFFERS; j++) { cell_global.inits[i].buffers[j] = cell->buffer[j]; } cell_global.inits[i].buffer_status = &cell->buffer_status[0][0][0]; + cell_global.inits[i].spu_functions = &cell->spu_functions; + cell_global.spe_contexts[i] = spe_context_create(0, NULL); if (!cell_global.spe_contexts[i]) { fprintf(stderr, "spe_context_create() failed\n"); diff --git a/src/gallium/drivers/cell/ppu/cell_spu.h b/src/gallium/drivers/cell/ppu/cell_spu.h index 137f26612e..c93958a9ed 100644 --- a/src/gallium/drivers/cell/ppu/cell_spu.h +++ b/src/gallium/drivers/cell/ppu/cell_spu.h @@ -30,14 +30,12 @@ #include <libspe2.h> -#include <libmisc.h> +#include <pthread.h> #include "cell/common.h" #include "cell_context.h" -#define MAX_SPUS 8 - /** * Global vars, for now anyway. */ @@ -46,14 +44,13 @@ struct cell_global_info /** * SPU/SPE handles, etc */ - spe_context_ptr_t spe_contexts[MAX_SPUS]; - pthread_t spe_threads[MAX_SPUS]; + spe_context_ptr_t spe_contexts[CELL_MAX_SPUS]; + pthread_t spe_threads[CELL_MAX_SPUS]; /** - * Data sent to SPUs + * Data sent to SPUs at start-up */ - struct cell_init_info inits[MAX_SPUS]; - struct cell_command command[MAX_SPUS]; + struct cell_init_info inits[CELL_MAX_SPUS]; }; diff --git a/src/gallium/drivers/cell/ppu/cell_state.h b/src/gallium/drivers/cell/ppu/cell_state.h index a7771a55a3..b193170f9c 100644 --- a/src/gallium/drivers/cell/ppu/cell_state.h +++ b/src/gallium/drivers/cell/ppu/cell_state.h @@ -44,8 +44,9 @@ #define CELL_NEW_TEXTURE 0x800 #define CELL_NEW_VERTEX 0x1000 #define CELL_NEW_VS 0x2000 -#define CELL_NEW_CONSTANTS 0x4000 -#define CELL_NEW_VERTEX_INFO 0x8000 +#define CELL_NEW_VS_CONSTANTS 0x4000 +#define CELL_NEW_FS_CONSTANTS 0x8000 +#define CELL_NEW_VERTEX_INFO 0x10000 extern void diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 2da3097983..0a0af81f53 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -25,18 +25,155 @@ * **************************************************************************/ +#include "pipe/p_inlines.h" #include "util/u_memory.h" #include "cell_context.h" #include "cell_gen_fragment.h" #include "cell_state.h" #include "cell_state_emit.h" -#include "cell_state_per_fragment.h" #include "cell_batch.h" #include "cell_texture.h" #include "draw/draw_context.h" #include "draw/draw_private.h" +/** + * Find/create a cell_command_fragment_ops object corresponding to the + * current blend/stencil/z/colormask/etc. state. + */ +static struct cell_command_fragment_ops * +lookup_fragment_ops(struct cell_context *cell) +{ + struct cell_fragment_ops_key key; + struct cell_command_fragment_ops *ops; + + /* + * Build key + */ + memset(&key, 0, sizeof(key)); + key.blend = *cell->blend; + key.blend_color = cell->blend_color; + key.dsa = *cell->depth_stencil; + + if (cell->framebuffer.cbufs[0]) + key.color_format = cell->framebuffer.cbufs[0]->format; + else + key.color_format = PIPE_FORMAT_NONE; + + if (cell->framebuffer.zsbuf) + key.zs_format = cell->framebuffer.zsbuf->format; + else + key.zs_format = PIPE_FORMAT_NONE; + + /* + * Look up key in cache. + */ + ops = (struct cell_command_fragment_ops *) + util_keymap_lookup(cell->fragment_ops_cache, &key); + + /* + * If not found, create/save new fragment ops command. + */ + if (!ops) { + struct spe_function spe_code_front, spe_code_back; + unsigned int facing_dependent, total_code_size; + + if (0) + debug_printf("**** Create New Fragment Ops\n"); + + /* Prepare the buffer that will hold the generated code. The + * "0" passed in for the size means that the SPE code will + * use a default size. + */ + spe_init_func(&spe_code_front, 0); + spe_init_func(&spe_code_back, 0); + + /* Generate new code. Always generate new code for both front-facing + * and back-facing fragments, even if it's the same code in both + * cases. + */ + cell_gen_fragment_function(cell, CELL_FACING_FRONT, &spe_code_front); + cell_gen_fragment_function(cell, CELL_FACING_BACK, &spe_code_back); + + /* Make sure the code is a multiple of 8 bytes long; this is + * required to ensure that the dual pipe instruction alignment + * is correct. It's also important for the SPU unpacking, + * which assumes 8-byte boundaries. + */ + unsigned int front_code_size = spe_code_size(&spe_code_front); + while (front_code_size % 8 != 0) { + spe_lnop(&spe_code_front); + front_code_size = spe_code_size(&spe_code_front); + } + unsigned int back_code_size = spe_code_size(&spe_code_back); + while (back_code_size % 8 != 0) { + spe_lnop(&spe_code_back); + back_code_size = spe_code_size(&spe_code_back); + } + + /* Determine whether the code we generated is facing-dependent, by + * determining whether the generated code is different for the front- + * and back-facing fragments. + */ + if (front_code_size == back_code_size && memcmp(spe_code_front.store, spe_code_back.store, front_code_size) == 0) { + /* Code is identical; only need one copy. */ + facing_dependent = 0; + total_code_size = front_code_size; + } + else { + /* Code is different for front-facing and back-facing fragments. + * Need to send both copies. + */ + facing_dependent = 1; + total_code_size = front_code_size + back_code_size; + } + + /* alloc new fragment ops command. Note that this structure + * has variant length based on the total code size required. + */ + ops = CALLOC_VARIANT_LENGTH_STRUCT(cell_command_fragment_ops, total_code_size); + /* populate the new cell_command_fragment_ops object */ + ops->opcode = CELL_CMD_STATE_FRAGMENT_OPS; + ops->total_code_size = total_code_size; + ops->front_code_index = 0; + memcpy(ops->code, spe_code_front.store, front_code_size); + if (facing_dependent) { + /* We have separate front- and back-facing code. Append the + * back-facing code to the buffer. Be careful because the code + * size is in bytes, but the buffer is of unsigned elements. + */ + ops->back_code_index = front_code_size / sizeof(spe_code_front.store[0]); + memcpy(ops->code + ops->back_code_index, spe_code_back.store, back_code_size); + } + else { + /* Use the same code for front- and back-facing fragments */ + ops->back_code_index = ops->front_code_index; + } + + /* Set the fields for the fallback case. Note that these fields + * (and the whole fallback case) will eventually go away. + */ + ops->dsa = *cell->depth_stencil; + ops->blend = *cell->blend; + ops->blend_color = cell->blend_color; + + /* insert cell_command_fragment_ops object into keymap/cache */ + util_keymap_insert(cell->fragment_ops_cache, &key, ops, NULL); + + /* release rtasm buffer */ + spe_release_func(&spe_code_front); + spe_release_func(&spe_code_back); + } + else { + if (0) + debug_printf("**** Re-use Fragment Ops\n"); + } + + return ops; +} + + + static void emit_state_cmd(struct cell_context *cell, uint cmd, const void *state, uint state_size) @@ -73,6 +210,13 @@ cell_emit_state(struct cell_context *cell) #endif } + if (cell->dirty & (CELL_NEW_RASTERIZER)) { + struct cell_command_rasterizer *rast = + cell_batch_alloc(cell, sizeof(*rast)); + rast->opcode = CELL_CMD_STATE_RASTERIZER; + rast->rasterizer = *cell->rasterizer; + } + if (cell->dirty & (CELL_NEW_FS)) { /* Send new fragment program to SPUs */ struct cell_command_fragment_program *fp @@ -90,59 +234,81 @@ cell_emit_state(struct cell_context *cell) } } + if (cell->dirty & (CELL_NEW_FS_CONSTANTS)) { + const uint shader = PIPE_SHADER_FRAGMENT; + const uint num_const = cell->constants[shader].size / sizeof(float); + uint i, j; + float *buf = cell_batch_alloc(cell, 16 + num_const * sizeof(float)); + uint64_t *ibuf = (uint64_t *) buf; + const float *constants = pipe_buffer_map(cell->pipe.screen, + cell->constants[shader].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + ibuf[0] = CELL_CMD_STATE_FS_CONSTANTS; + ibuf[1] = num_const; + j = 4; + for (i = 0; i < num_const; i++) { + buf[j++] = constants[i]; + } + pipe_buffer_unmap(cell->pipe.screen, cell->constants[shader].buffer); + } + if (cell->dirty & (CELL_NEW_FRAMEBUFFER | CELL_NEW_DEPTH_STENCIL | CELL_NEW_BLEND)) { - /* XXX we don't want to always do codegen here. We should have - * a hash/lookup table to cache previous results... - */ - struct cell_command_fragment_ops *fops - = cell_batch_alloc(cell, sizeof(*fops)); - struct spe_function spe_code; - - /* generate new code */ - cell_gen_fragment_function(cell, &spe_code); - /* put the new code into the batch buffer */ - fops->opcode = CELL_CMD_STATE_FRAGMENT_OPS; - memcpy(&fops->code, spe_code.store, - SPU_MAX_FRAGMENT_OPS_INSTS * SPE_INST_SIZE); - fops->dsa = cell->depth_stencil->base; - fops->blend = cell->blend->base; - /* free codegen buffer */ - spe_release_func(&spe_code); + struct cell_command_fragment_ops *fops, *fops_cmd; + /* Note that cell_command_fragment_ops is a variant-sized record */ + fops = lookup_fragment_ops(cell); + fops_cmd = cell_batch_alloc(cell, sizeof(*fops_cmd) + fops->total_code_size); + memcpy(fops_cmd, fops, sizeof(*fops) + fops->total_code_size); } if (cell->dirty & CELL_NEW_SAMPLER) { uint i; for (i = 0; i < CELL_MAX_SAMPLERS; i++) { - if (cell->sampler[i]) { - struct cell_command_sampler *sampler - = cell_batch_alloc(cell, sizeof(*sampler)); - sampler->opcode = CELL_CMD_STATE_SAMPLER; - sampler->unit = i; - sampler->state = *cell->sampler[i]; + if (cell->dirty_samplers & (1 << i)) { + if (cell->sampler[i]) { + struct cell_command_sampler *sampler + = cell_batch_alloc(cell, sizeof(*sampler)); + sampler->opcode = CELL_CMD_STATE_SAMPLER; + sampler->unit = i; + sampler->state = *cell->sampler[i]; + } } } + cell->dirty_samplers = 0x0; } if (cell->dirty & CELL_NEW_TEXTURE) { uint i; for (i = 0;i < CELL_MAX_SAMPLERS; i++) { - struct cell_command_texture *texture - = cell_batch_alloc(cell, sizeof(*texture)); - texture->opcode = CELL_CMD_STATE_TEXTURE; - texture->unit = i; - if (cell->texture[i]) { - texture->start = cell->texture[i]->tiled_data; - texture->width = cell->texture[i]->base.width[0]; - texture->height = cell->texture[i]->base.height[0]; - } - else { - texture->start = NULL; - texture->width = 1; - texture->height = 1; + if (cell->dirty_textures & (1 << i)) { + struct cell_command_texture *texture + = cell_batch_alloc(cell, sizeof(*texture)); + texture->opcode = CELL_CMD_STATE_TEXTURE; + texture->unit = i; + if (cell->texture[i]) { + uint level; + for (level = 0; level < CELL_MAX_TEXTURE_LEVELS; level++) { + texture->start[level] = cell->texture[i]->tiled_mapped[level]; + texture->width[level] = cell->texture[i]->base.width[level]; + texture->height[level] = cell->texture[i]->base.height[level]; + texture->depth[level] = cell->texture[i]->base.depth[level]; + } + texture->target = cell->texture[i]->base.target; + } + else { + uint level; + for (level = 0; level < CELL_MAX_TEXTURE_LEVELS; level++) { + texture->start[level] = NULL; + texture->width[level] = 0; + texture->height[level] = 0; + texture->depth[level] = 0; + } + texture->target = 0; + } } } + cell->dirty_textures = 0x0; } if (cell->dirty & CELL_NEW_VERTEX_INFO) { diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c index 3a0d066da2..cda39f8d59 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c @@ -191,13 +191,18 @@ cell_set_constant_buffer(struct pipe_context *pipe, assert(shader < PIPE_SHADER_TYPES); assert(index == 0); + draw_flush(cell->draw); + /* note: reference counting */ winsys_buffer_reference(ws, &cell->constants[shader].buffer, buf->buffer); cell->constants[shader].size = buf->size; - cell->dirty |= CELL_NEW_CONSTANTS; + if (shader == PIPE_SHADER_VERTEX) + cell->dirty |= CELL_NEW_VS_CONSTANTS; + else if (shader == PIPE_SHADER_FRAGMENT) + cell->dirty |= CELL_NEW_FS_CONSTANTS; } diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c index 732c64082e..c9203fee08 100644 --- a/src/gallium/drivers/cell/ppu/cell_surface.c +++ b/src/gallium/drivers/cell/ppu/cell_surface.c @@ -27,6 +27,7 @@ #include "util/u_rect.h" #include "cell_context.h" +#include "cell_surface.h" void diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index b6590dfb86..9f83ab8fa4 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -28,6 +28,7 @@ * Authors: * Keith Whitwell <keith@tungstengraphics.com> * Michel Dänzer <michel@tungstengraphics.com> + * Brian Paul */ #include "pipe/p_context.h" @@ -42,30 +43,31 @@ #include "cell_texture.h" -/* Simple, maximally packed layout. - */ -static unsigned minify( unsigned d ) +static unsigned +minify(unsigned d) { return MAX2(1, d>>1); } static void -cell_texture_layout(struct cell_texture * spt) +cell_texture_layout(struct cell_texture *ct) { - struct pipe_texture *pt = &spt->base; + struct pipe_texture *pt = &ct->base; unsigned level; unsigned width = pt->width[0]; unsigned height = pt->height[0]; unsigned depth = pt->depth[0]; - spt->buffer_size = 0; + ct->buffer_size = 0; for ( level = 0 ; level <= pt->last_level ; level++ ) { unsigned size; unsigned w_tile, h_tile; + assert(level < CELL_MAX_TEXTURE_LEVELS); + /* width, height, rounded up to tile size */ w_tile = align(width, TILE_SIZE); h_tile = align(height, TILE_SIZE); @@ -76,9 +78,9 @@ cell_texture_layout(struct cell_texture * spt) pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w_tile); pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h_tile); - spt->stride[level] = pt->nblocksx[level] * pt->block.size; + ct->stride[level] = pt->nblocksx[level] * pt->block.size; - spt->level_offset[level] = spt->buffer_size; + ct->level_offset[level] = ct->buffer_size; size = pt->nblocksx[level] * pt->nblocksy[level] * pt->block.size; if (pt->target == PIPE_TEXTURE_CUBE) @@ -86,7 +88,7 @@ cell_texture_layout(struct cell_texture * spt) else size *= depth; - spt->buffer_size += size; + ct->buffer_size += size; width = minify(width); height = minify(height); @@ -100,26 +102,25 @@ cell_texture_create(struct pipe_screen *screen, const struct pipe_texture *templat) { struct pipe_winsys *ws = screen->winsys; - struct cell_texture *spt = CALLOC_STRUCT(cell_texture); - if (!spt) + struct cell_texture *ct = CALLOC_STRUCT(cell_texture); + if (!ct) return NULL; - spt->base = *templat; - spt->base.refcount = 1; - spt->base.screen = screen; + ct->base = *templat; + ct->base.refcount = 1; + ct->base.screen = screen; - cell_texture_layout(spt); + cell_texture_layout(ct); - spt->buffer = ws->buffer_create(ws, 32, - PIPE_BUFFER_USAGE_PIXEL, - spt->buffer_size); + ct->buffer = ws->buffer_create(ws, 32, PIPE_BUFFER_USAGE_PIXEL, + ct->buffer_size); - if (!spt->buffer) { - FREE(spt); + if (!ct->buffer) { + FREE(ct); return NULL; } - return &spt->base; + return &ct->base; } @@ -135,177 +136,310 @@ cell_texture_release(struct pipe_screen *screen, __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); */ if (--(*pt)->refcount <= 0) { - struct cell_texture *spt = cell_texture(*pt); + /* Delete this texture now. + * But note that the underlying pipe_buffer may linger... + */ + struct cell_texture *ct = cell_texture(*pt); + uint i; /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); + DBG("%s deleting %p\n", __FUNCTION__, (void *) ct); */ - pipe_buffer_reference(screen, &spt->buffer, NULL); + pipe_buffer_reference(screen, &ct->buffer, NULL); - FREE(spt); + for (i = 0; i < CELL_MAX_TEXTURE_LEVELS; i++) { + /* Unreference the tiled image buffer. + * It may not actually be deleted until a fence is hit. + */ + if (ct->tiled_buffer[i]) { + ct->tiled_mapped[i] = NULL; + winsys_buffer_reference(screen->winsys, &ct->tiled_buffer[i], NULL); + } + } + + FREE(ct); } *pt = NULL; } -#if 0 + +/** + * Convert image from linear layout to tiled layout. 4-byte pixels. + */ static void -cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture, - uint face, uint levelsMask) +twiddle_image_uint(uint w, uint h, uint tile_size, uint *dst, + uint src_stride, const uint *src) { - /* XXX TO DO: re-tile the texture data ... */ + const uint tile_size2 = tile_size * tile_size; + const uint h_t = (h + tile_size - 1) / tile_size; + const uint w_t = (w + tile_size - 1) / tile_size; -} -#endif + uint it, jt; /* tile counters */ + uint i, j; /* intra-tile counters */ + src_stride /= 4; /* convert from bytes to pixels */ -static struct pipe_surface * -cell_get_tex_surface(struct pipe_screen *screen, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice, - unsigned usage) -{ - struct pipe_winsys *ws = screen->winsys; - struct cell_texture *spt = cell_texture(pt); - struct pipe_surface *ps; + /* loop over dest tiles */ + for (it = 0; it < h_t; it++) { + for (jt = 0; jt < w_t; jt++) { + /* start of dest tile: */ + uint *tdst = dst + (it * w_t + jt) * tile_size2; - ps = ws->surface_alloc(ws); - if (ps) { - assert(ps->refcount); - assert(ps->winsys); - winsys_buffer_reference(ws, &ps->buffer, spt->buffer); - ps->format = pt->format; - ps->block = pt->block; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->nblocksx = pt->nblocksx[level]; - ps->nblocksy = pt->nblocksy[level]; - ps->stride = spt->stride[level]; - ps->offset = spt->level_offset[level]; - ps->usage = usage; + /* compute size of this tile (may be smaller than tile_size) */ + /* XXX note: a compiler bug was found here. That's why the code + * looks as it does. + */ + uint tile_width = w - jt * tile_size; + tile_width = MIN2(tile_width, tile_size); + uint tile_height = h - it * tile_size; + tile_height = MIN2(tile_height, tile_size); - /* XXX may need to override usage flags (see sp_texture.c) */ + /* loop over texels in the tile */ + for (i = 0; i < tile_height; i++) { + for (j = 0; j < tile_width; j++) { + const uint srci = it * tile_size + i; + const uint srcj = jt * tile_size + j; + ASSERT(srci < h); + ASSERT(srcj < w); + tdst[i * tile_size + j] = src[srci * src_stride + srcj]; + } + } + } + } +} - pipe_texture_reference(&ps->texture, pt); - ps->face = face; - ps->level = level; - ps->zslice = zslice; - if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { - ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - ps->nblocksy * - ps->stride; - } - else { - assert(face == 0); - assert(zslice == 0); +/** + * For Cell. Basically, rearrange the pixels/quads from this layout: + * +--+--+--+--+ + * |p0|p1|p2|p3|.... + * +--+--+--+--+ + * + * to this layout: + * +--+--+ + * |p0|p1|.... + * +--+--+ + * |p2|p3| + * +--+--+ + */ +static void +twiddle_tile(const uint *tileIn, uint *tileOut) +{ + int y, x; + + for (y = 0; y < TILE_SIZE; y+=2) { + for (x = 0; x < TILE_SIZE; x+=2) { + int k = 4 * (y/2 * TILE_SIZE/2 + x/2); + tileOut[y * TILE_SIZE + (x + 0)] = tileIn[k]; + tileOut[y * TILE_SIZE + (x + 1)] = tileIn[k+1]; + tileOut[(y + 1) * TILE_SIZE + (x + 0)] = tileIn[k+2]; + tileOut[(y + 1) * TILE_SIZE + (x + 1)] = tileIn[k+3]; } } - return ps; } - /** - * Copy tile data from linear layout to tiled layout. - * XXX this should be rolled into the future surface-creation code. - * XXX also need "untile" code... + * Convert image from tiled layout to linear layout. 4-byte pixels. */ static void -tile_copy_data(uint w, uint h, uint tile_size, uint *dst, const uint *src) +untwiddle_image_uint(uint w, uint h, uint tile_size, uint *dst, + uint dst_stride, const uint *src) { const uint tile_size2 = tile_size * tile_size; - const uint h_t = h / tile_size, w_t = w / tile_size; - + const uint h_t = (h + tile_size - 1) / tile_size; + const uint w_t = (w + tile_size - 1) / tile_size; + uint *tile_buf; uint it, jt; /* tile counters */ uint i, j; /* intra-tile counters */ - /* loop over dest tiles */ + dst_stride /= 4; /* convert from bytes to pixels */ + + tile_buf = align_malloc(tile_size * tile_size * 4, 16); + + /* loop over src tiles */ for (it = 0; it < h_t; it++) { for (jt = 0; jt < w_t; jt++) { - /* start of dest tile: */ - uint *tdst = dst + (it * w_t + jt) * tile_size2; + /* start of src tile: */ + const uint *tsrc = src + (it * w_t + jt) * tile_size2; + + twiddle_tile(tsrc, tile_buf); + tsrc = tile_buf; + + /* compute size of this tile (may be smaller than tile_size) */ + /* XXX note: a compiler bug was found here. That's why the code + * looks as it does. + */ + uint tile_width = w - jt * tile_size; + tile_width = MIN2(tile_width, tile_size); + uint tile_height = h - it * tile_size; + tile_height = MIN2(tile_height, tile_size); + /* loop over texels in the tile */ - for (i = 0; i < tile_size; i++) { - for (j = 0; j < tile_size; j++) { - const uint srci = it * tile_size + i; - const uint srcj = jt * tile_size + j; - *tdst++ = src[srci * w + srcj]; + for (i = 0; i < tile_height; i++) { + for (j = 0; j < tile_width; j++) { + uint dsti = it * tile_size + i; + uint dstj = jt * tile_size + j; + ASSERT(dsti < h); + ASSERT(dstj < w); + dst[dsti * dst_stride + dstj] = tsrc[i * tile_size + j]; } } } } -} + align_free(tile_buf); +} /** * Convert linear texture image data to tiled format for SPU usage. - * XXX recast this in terms of pipe_surfaces (aka texture views). */ static void -cell_tile_texture(struct cell_context *cell, - struct cell_texture *texture) +cell_twiddle_texture(struct pipe_screen *screen, + struct pipe_surface *surface) { - struct pipe_screen *screen = cell->pipe.screen; - uint face = 0, level = 0, zslice = 0; - struct pipe_surface *surf; - const uint w = texture->base.width[0], h = texture->base.height[0]; - const uint *src; - - /* temporary restrictions: */ - assert(w >= TILE_SIZE); - assert(h >= TILE_SIZE); - assert(w % TILE_SIZE == 0); - assert(h % TILE_SIZE == 0); - - surf = screen->get_tex_surface(screen, &texture->base, face, level, zslice, - PIPE_BUFFER_USAGE_CPU_WRITE); - ASSERT(surf); - - src = (const uint *) pipe_surface_map(surf, PIPE_BUFFER_USAGE_CPU_WRITE); - - if (texture->tiled_data) { - align_free(texture->tiled_data); + struct cell_texture *ct = cell_texture(surface->texture); + const uint level = surface->level; + const uint texWidth = ct->base.width[level]; + const uint texHeight = ct->base.height[level]; + const uint bufWidth = align(texWidth, TILE_SIZE); + const uint bufHeight = align(texHeight, TILE_SIZE); + const void *map = pipe_buffer_map(screen, surface->buffer, + PIPE_BUFFER_USAGE_CPU_READ); + const uint *src = (const uint *) ((const ubyte *) map + surface->offset); + + switch (ct->base.format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_B8G8R8A8_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + { + int numFaces = ct->base.target == PIPE_TEXTURE_CUBE ? 6 : 1; + int offset = bufWidth * bufHeight * 4 * surface->face; + uint *dst; + + if (!ct->tiled_buffer[level]) { + /* allocate buffer for tiled data now */ + struct pipe_winsys *ws = screen->winsys; + uint bytes = bufWidth * bufHeight * 4 * numFaces; + ct->tiled_buffer[level] = ws->buffer_create(ws, 16, + PIPE_BUFFER_USAGE_PIXEL, + bytes); + /* and map it */ + ct->tiled_mapped[level] = ws->buffer_map(ws, ct->tiled_buffer[level], + PIPE_BUFFER_USAGE_GPU_READ); + } + dst = (uint *) ((ubyte *) ct->tiled_mapped[level] + offset); + + twiddle_image_uint(texWidth, texHeight, TILE_SIZE, dst, + surface->stride, src); + } + break; + default: + printf("Cell: twiddle unsupported texture format %s\n", pf_name(ct->base.format)); + ; } - texture->tiled_data = align_malloc(w * h * 4, 16); - tile_copy_data(w, h, TILE_SIZE, texture->tiled_data, src); + pipe_buffer_unmap(screen, surface->buffer); +} + + +/** + * Convert SPU tiled texture image data to linear format for app usage. + */ +static void +cell_untwiddle_texture(struct pipe_screen *screen, + struct pipe_surface *surface) +{ + struct cell_texture *ct = cell_texture(surface->texture); + const uint level = surface->level; + const uint texWidth = ct->base.width[level]; + const uint texHeight = ct->base.height[level]; + const void *map = pipe_buffer_map(screen, surface->buffer, + PIPE_BUFFER_USAGE_CPU_READ); + const uint *src = (const uint *) ((const ubyte *) map + surface->offset); + + switch (ct->base.format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_B8G8R8A8_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + { + int numFaces = ct->base.target == PIPE_TEXTURE_CUBE ? 6 : 1; + int offset = surface->stride * texHeight * 4 * surface->face; + uint *dst; + + if (!ct->untiled_data[level]) { + ct->untiled_data[level] = + align_malloc(surface->stride * texHeight * 4 * numFaces, 16); + } + + dst = (uint *) ((ubyte *) ct->untiled_data[level] + offset); - pipe_surface_unmap(surf); + untwiddle_image_uint(texWidth, texHeight, TILE_SIZE, dst, + surface->stride, src); + } + break; + default: + { + ct->untiled_data[level] = NULL; + printf("Cell: untwiddle unsupported texture format %s\n", pf_name(ct->base.format)); + } + } - pipe_surface_reference(&surf, NULL); + pipe_buffer_unmap(screen, surface->buffer); } -void -cell_update_texture_mapping(struct cell_context *cell) +static struct pipe_surface * +cell_get_tex_surface(struct pipe_screen *screen, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + unsigned usage) { -#if 0 - uint face = 0, level = 0, zslice = 0; -#endif - uint i; - - for (i = 0; i < CELL_MAX_SAMPLERS; i++) { - if (cell->texture[i]) - cell_tile_texture(cell, cell->texture[i]); - } + struct pipe_winsys *ws = screen->winsys; + struct cell_texture *ct = cell_texture(pt); + struct pipe_surface *ps; -#if 0 - if (cell->tex_surf && cell->tex_map) { - pipe_surface_unmap(cell->tex_surf); - cell->tex_map = NULL; - } + ps = ws->surface_alloc(ws); + if (ps) { + assert(ps->refcount); + assert(ps->winsys); + winsys_buffer_reference(ws, &ps->buffer, ct->buffer); + ps->format = pt->format; + ps->block = pt->block; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = ct->stride[level]; + ps->offset = ct->level_offset[level]; + ps->usage = usage; - /* XXX free old surface */ + /* XXX may need to override usage flags (see sp_texture.c) */ - cell->tex_surf = cell_get_tex_surface(&cell->pipe, - &cell->texture[0]->base, - face, level, zslice); + pipe_texture_reference(&ps->texture, pt); + ps->face = face; + ps->level = level; + ps->zslice = zslice; - cell->tex_map = pipe_surface_map(cell->tex_surf); -#endif + if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { + ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * + ps->nblocksy * + ps->stride; + } + else { + assert(face == 0); + assert(zslice == 0); + } + + if (ps->usage & PIPE_BUFFER_USAGE_CPU_READ) { + /* convert from tiled to linear layout */ + cell_untwiddle_texture(screen, ps); + } + } + return ps; } @@ -313,11 +447,17 @@ static void cell_tex_surface_release(struct pipe_screen *screen, struct pipe_surface **s) { - /* Effectively do the texture_update work here - if texture images - * needed post-processing to put them into hardware layout, this is - * where it would happen. For softpipe, nothing to do. - */ - assert ((*s)->texture); + struct cell_texture *ct = cell_texture((*s)->texture); + const uint level = (*s)->level; + + if (((*s)->usage & PIPE_BUFFER_USAGE_CPU_READ) && (ct->untiled_data[level])) + { + align_free(ct->untiled_data[level]); + ct->untiled_data[level] = NULL; + } + + /* XXX if done rendering to teximage, re-tile */ + pipe_texture_reference(&(*s)->texture, NULL); screen->winsys->surface_release(screen->winsys, s); @@ -325,11 +465,15 @@ cell_tex_surface_release(struct pipe_screen *screen, static void * -cell_surface_map( struct pipe_screen *screen, - struct pipe_surface *surface, - unsigned flags ) +cell_surface_map(struct pipe_screen *screen, + struct pipe_surface *surface, + unsigned flags) { ubyte *map; + struct cell_texture *ct = cell_texture(surface->texture); + const uint level = surface->level; + + assert(ct); if (flags & ~surface->usage) { assert(0); @@ -339,22 +483,15 @@ cell_surface_map( struct pipe_screen *screen, map = pipe_buffer_map( screen, surface->buffer, flags ); if (map == NULL) return NULL; - - /* May want to different things here depending on read/write nature - * of the map: - */ - if (surface->texture && - (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) + else { - /* Do something to notify sharing contexts of a texture change. - * In softpipe, that would mean flushing the texture cache. - */ -#if 0 - cell_screen(screen)->timestamp++; -#endif + if ((surface->usage & PIPE_BUFFER_USAGE_CPU_READ) && (ct->untiled_data[level])) { + return (void *) ((ubyte *) ct->untiled_data[level] + surface->offset); + } + else { + return (void *) (map + surface->offset); + } } - - return map + surface->offset; } @@ -362,17 +499,21 @@ static void cell_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) { - pipe_buffer_unmap( screen, surface->buffer ); -} + struct cell_texture *ct = cell_texture(surface->texture); + assert(ct); -void -cell_init_texture_functions(struct cell_context *cell) -{ - /*cell->pipe.texture_update = cell_texture_update;*/ + if ((ct->base.tex_usage & PIPE_TEXTURE_USAGE_SAMPLER) && + (surface->usage & PIPE_BUFFER_USAGE_CPU_WRITE)) { + /* convert from linear to tiled layout */ + cell_twiddle_texture(screen, surface); + } + + pipe_buffer_unmap( screen, surface->buffer ); } + void cell_init_screen_texture_funcs(struct pipe_screen *screen) { diff --git a/src/gallium/drivers/cell/ppu/cell_texture.h b/src/gallium/drivers/cell/ppu/cell_texture.h index 6d37e95ebc..7018b0c9bf 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.h +++ b/src/gallium/drivers/cell/ppu/cell_texture.h @@ -40,15 +40,19 @@ struct cell_texture { struct pipe_texture base; - unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; - unsigned long stride[PIPE_MAX_TEXTURE_LEVELS]; + unsigned long level_offset[CELL_MAX_TEXTURE_LEVELS]; + unsigned long stride[CELL_MAX_TEXTURE_LEVELS]; /* The data is held here: */ struct pipe_buffer *buffer; unsigned long buffer_size; - void *tiled_data; /* XXX this may be temporary */ /*ALIGN16*/ + /** Texture data in tiled layout is held here */ + struct pipe_buffer *tiled_buffer[CELL_MAX_TEXTURE_LEVELS]; + /** Mapped, tiled texture data */ + void *tiled_mapped[CELL_MAX_TEXTURE_LEVELS]; + void *untiled_data[CELL_MAX_TEXTURE_LEVELS]; }; @@ -62,14 +66,6 @@ cell_texture(struct pipe_texture *pt) extern void -cell_update_texture_mapping(struct cell_context *cell); - - -extern void -cell_init_texture_functions(struct cell_context *cell); - - -extern void cell_init_screen_texture_funcs(struct pipe_screen *screen); diff --git a/src/gallium/drivers/cell/ppu/cell_vbuf.c b/src/gallium/drivers/cell/ppu/cell_vbuf.c index aa63435b93..65ba51b6bb 100644 --- a/src/gallium/drivers/cell/ppu/cell_vbuf.c +++ b/src/gallium/drivers/cell/ppu/cell_vbuf.c @@ -38,6 +38,7 @@ #include "cell_batch.h" #include "cell_context.h" +#include "cell_fence.h" #include "cell_flush.h" #include "cell_spu.h" #include "cell_vbuf.h" @@ -108,6 +109,11 @@ cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, __FUNCTION__, cvbr->vertex_buf, vertices_used); */ + /* Make sure texture buffers aren't released until we're done rendering + * with them. + */ + cell_add_fenced_textures(cell); + /* Tell SPUs they can release the vert buf */ if (cvbr->vertex_buf != ~0U) { struct cell_command_release_verts *release diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c index 566df7f59e..9cba537d9e 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c @@ -73,8 +73,8 @@ emit_matrix_transpose(struct spe_function *p, int col3; - spe_lqd(p, shuf_hi, shuf_ptr, 3); - spe_lqd(p, shuf_lo, shuf_ptr, 4); + spe_lqd(p, shuf_hi, shuf_ptr, 3*16); + spe_lqd(p, shuf_lo, shuf_ptr, 4*16); spe_shufb(p, t1, row0, row2, shuf_hi); spe_shufb(p, t2, row0, row2, shuf_lo); @@ -122,13 +122,13 @@ emit_matrix_transpose(struct spe_function *p, */ switch (count) { case 4: - spe_stqd(p, col3, dest_ptr, 3); + spe_stqd(p, col3, dest_ptr, 3 * 16); case 3: - spe_stqd(p, col2, dest_ptr, 2); + spe_stqd(p, col2, dest_ptr, 2 * 16); case 2: - spe_stqd(p, col1, dest_ptr, 1); + spe_stqd(p, col1, dest_ptr, 1 * 16); case 1: - spe_stqd(p, col0, dest_ptr, 0); + spe_stqd(p, col0, dest_ptr, 0 * 16); } @@ -145,6 +145,8 @@ emit_matrix_transpose(struct spe_function *p, } +#if 0 +/* This appears to not be used currently */ static void emit_fetch(struct spe_function *p, unsigned in_ptr, unsigned *offset, @@ -166,17 +168,17 @@ emit_fetch(struct spe_function *p, float scale_signed = 0.0; float scale_unsigned = 0.0; - spe_lqd(p, v0, in_ptr, 0 + offset[0]); - spe_lqd(p, v1, in_ptr, 1 + offset[0]); - spe_lqd(p, v2, in_ptr, 2 + offset[0]); - spe_lqd(p, v3, in_ptr, 3 + offset[0]); + spe_lqd(p, v0, in_ptr, (0 + offset[0]) * 16); + spe_lqd(p, v1, in_ptr, (1 + offset[0]) * 16); + spe_lqd(p, v2, in_ptr, (2 + offset[0]) * 16); + spe_lqd(p, v3, in_ptr, (3 + offset[0]) * 16); offset[0] += 4; switch (bytes) { case 1: scale_signed = 1.0f / 127.0f; scale_unsigned = 1.0f / 255.0f; - spe_lqd(p, tmp, shuf_ptr, 1); + spe_lqd(p, tmp, shuf_ptr, 1 * 16); spe_shufb(p, v0, v0, v0, tmp); spe_shufb(p, v1, v1, v1, tmp); spe_shufb(p, v2, v2, v2, tmp); @@ -185,7 +187,7 @@ emit_fetch(struct spe_function *p, case 2: scale_signed = 1.0f / 32767.0f; scale_unsigned = 1.0f / 65535.0f; - spe_lqd(p, tmp, shuf_ptr, 2); + spe_lqd(p, tmp, shuf_ptr, 2 * 16); spe_shufb(p, v0, v0, v0, tmp); spe_shufb(p, v1, v1, v1, tmp); spe_shufb(p, v2, v2, v2, tmp); @@ -241,11 +243,11 @@ emit_fetch(struct spe_function *p, switch (count) { case 1: - spe_stqd(p, float_zero, out_ptr, 1); + spe_stqd(p, float_zero, out_ptr, 1 * 16); case 2: - spe_stqd(p, float_zero, out_ptr, 2); + spe_stqd(p, float_zero, out_ptr, 2 * 16); case 3: - spe_stqd(p, float_one, out_ptr, 3); + spe_stqd(p, float_one, out_ptr, 3 * 16); } if (float_zero != -1) { @@ -256,6 +258,7 @@ emit_fetch(struct spe_function *p, spe_release_register(p, float_one); } } +#endif void cell_update_vertex_fetch(struct draw_context *draw) diff --git a/src/gallium/drivers/cell/spu/.gitignore b/src/gallium/drivers/cell/spu/.gitignore new file mode 100644 index 0000000000..2be9a2d324 --- /dev/null +++ b/src/gallium/drivers/cell/spu/.gitignore @@ -0,0 +1 @@ +g3d_spu diff --git a/src/gallium/drivers/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile index 1ae0dfb8c1..116453b79c 100644 --- a/src/gallium/drivers/cell/spu/Makefile +++ b/src/gallium/drivers/cell/spu/Makefile @@ -16,8 +16,10 @@ PROG_SPU_EMBED_O = $(PROG)_spu-embed.o SOURCES = \ - spu_main.c \ + spu_command.c \ spu_dcache.c \ + spu_funcs.c \ + spu_main.c \ spu_per_fragment_op.c \ spu_render.c \ spu_texture.c \ diff --git a/src/gallium/drivers/cell/spu/spu_colorpack.h b/src/gallium/drivers/cell/spu/spu_colorpack.h index fd8dc6ded3..d7ce005524 100644 --- a/src/gallium/drivers/cell/spu/spu_colorpack.h +++ b/src/gallium/drivers/cell/spu/spu_colorpack.h @@ -31,6 +31,7 @@ #define SPU_COLORPACK_H +#include <transpose_matrix4x4.h> #include <spu_intrinsics.h> @@ -84,10 +85,10 @@ spu_unpack_B8G8R8A8(uint color) vector unsigned int color_u4 = spu_splats(color); color_u4 = spu_shuffle(color_u4, color_u4, ((vector unsigned char) { - 10, 10, 10, 10, - 5, 5, 5, 5, + 2, 2, 2, 2, + 1, 1, 1, 1, 0, 0, 0, 0, - 15, 15, 15, 15}) ); + 3, 3, 3, 3}) ); return spu_convtf(color_u4, 32); } @@ -98,13 +99,47 @@ spu_unpack_A8R8G8B8(uint color) vector unsigned int color_u4 = spu_splats(color); color_u4 = spu_shuffle(color_u4, color_u4, ((vector unsigned char) { - 5, 5, 5, 5, - 10, 10, 10, 10, - 15, 15, 15, 15, + 1, 1, 1, 1, + 2, 2, 2, 2, + 3, 3, 3, 3, 0, 0, 0, 0}) ); - return spu_convtf(color_u4, 32); } +/** + * \param color_in - array of 32-bit packed ARGB colors + * \param color_out - returns float colors in RRRR, GGGG, BBBB, AAAA order + */ +static INLINE void +spu_unpack_A8R8G8B8_transpose4(const vector unsigned int color_in[4], + vector float color_out[4]) +{ + vector unsigned int c0; + + c0 = spu_shuffle(color_in[0], color_in[0], + ((vector unsigned char) { + 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 0, 0, 0, 0}) ); + color_out[0] = spu_convtf(c0, 32); + + c0 = spu_shuffle(color_in[1], color_in[1], + ((vector unsigned char) { + 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 0, 0, 0, 0}) ); + color_out[1] = spu_convtf(c0, 32); + + c0 = spu_shuffle(color_in[2], color_in[2], + ((vector unsigned char) { + 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 0, 0, 0, 0}) ); + color_out[2] = spu_convtf(c0, 32); + + c0 = spu_shuffle(color_in[3], color_in[3], + ((vector unsigned char) { + 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 0, 0, 0, 0}) ); + color_out[3] = spu_convtf(c0, 32); + + _transpose_matrix4x4(color_out, color_out); +} + + + #endif /* SPU_COLORPACK_H */ diff --git a/src/gallium/drivers/cell/spu/spu_command.c b/src/gallium/drivers/cell/spu/spu_command.c new file mode 100644 index 0000000000..8500d19754 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_command.c @@ -0,0 +1,815 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * SPU command processing code + */ + + +#include <stdio.h> +#include <libmisc.h> + +#include "pipe/p_defines.h" + +#include "spu_command.h" +#include "spu_main.h" +#include "spu_render.h" +#include "spu_per_fragment_op.h" +#include "spu_texture.h" +#include "spu_tile.h" +#include "spu_vertex_shader.h" +#include "spu_dcache.h" +#include "cell/common.h" + + +struct spu_vs_context draw; + + +/** + * Buffers containing dynamically generated SPU code: + */ +static unsigned char attribute_fetch_code_buffer[136 * PIPE_MAX_ATTRIBS] + ALIGN16_ATTRIB; + + + +static INLINE int +align(int value, int alignment) +{ + return (value + alignment - 1) & ~(alignment - 1); +} + + + +/** + * Tell the PPU that this SPU has finished copying a buffer to + * local store and that it may be reused by the PPU. + * This is done by writting a 16-byte batch-buffer-status block back into + * main memory (in cell_context->buffer_status[]). + */ +static void +release_buffer(uint buffer) +{ + /* Evidently, using less than a 16-byte status doesn't work reliably */ + static const vector unsigned int status = {CELL_BUFFER_STATUS_FREE, + CELL_BUFFER_STATUS_FREE, + CELL_BUFFER_STATUS_FREE, + CELL_BUFFER_STATUS_FREE}; + const uint index = 4 * (spu.init.id * CELL_NUM_BUFFERS + buffer); + uint *dst = spu.init.buffer_status + index; + + ASSERT(buffer < CELL_NUM_BUFFERS); + + mfc_put((void *) &status, /* src in local memory */ + (unsigned int) dst, /* dst in main memory */ + sizeof(status), /* size */ + TAG_MISC, /* tag is unimportant */ + 0, /* tid */ + 0 /* rid */); +} + + +/** + * Write CELL_FENCE_SIGNALLED back to the fence status qword in main memory. + * There's a qword of status per SPU. + */ +static void +cmd_fence(struct cell_command_fence *fence_cmd) +{ + static const vector unsigned int status = {CELL_FENCE_SIGNALLED, + CELL_FENCE_SIGNALLED, + CELL_FENCE_SIGNALLED, + CELL_FENCE_SIGNALLED}; + uint *dst = (uint *) fence_cmd->fence; + dst += 4 * spu.init.id; /* main store/memory address, not local store */ + ASSERT_ALIGN16(dst); + mfc_put((void *) &status, /* src in local memory */ + (unsigned int) dst, /* dst in main memory */ + sizeof(status), /* size */ + TAG_FENCE, /* tag */ + 0, /* tid */ + 0 /* rid */); +} + + +static void +cmd_clear_surface(const struct cell_command_clear_surface *clear) +{ + D_PRINTF(CELL_DEBUG_CMD, "CLEAR SURF %u to 0x%08x\n", clear->surface, clear->value); + + if (clear->surface == 0) { + spu.fb.color_clear_value = clear->value; + if (spu.init.debug_flags & CELL_DEBUG_CHECKER) { + uint x = (spu.init.id << 4) | (spu.init.id << 12) | + (spu.init.id << 20) | (spu.init.id << 28); + spu.fb.color_clear_value ^= x; + } + } + else { + spu.fb.depth_clear_value = clear->value; + } + +#define CLEAR_OPT 1 +#if CLEAR_OPT + + /* Simply set all tiles' status to CLEAR. + * When we actually begin rendering into a tile, we'll initialize it to + * the clear value. If any tiles go untouched during the frame, + * really_clear_tiles() will set them to the clear value. + */ + if (clear->surface == 0) { + memset(spu.ctile_status, TILE_STATUS_CLEAR, sizeof(spu.ctile_status)); + } + else { + memset(spu.ztile_status, TILE_STATUS_CLEAR, sizeof(spu.ztile_status)); + } + +#else + + /* + * This path clears the whole framebuffer to the clear color right now. + */ + + /* + printf("SPU: %s num=%d w=%d h=%d\n", + __FUNCTION__, num_tiles, spu.fb.width_tiles, spu.fb.height_tiles); + */ + + /* init a single tile to the clear value */ + if (clear->surface == 0) { + clear_c_tile(&spu.ctile); + } + else { + clear_z_tile(&spu.ztile); + } + + /* walk over my tiles, writing the 'clear' tile's data */ + { + const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; + uint i; + for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { + uint tx = i % spu.fb.width_tiles; + uint ty = i / spu.fb.width_tiles; + if (clear->surface == 0) + put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0); + else + put_tile(tx, ty, &spu.ztile, TAG_SURFACE_CLEAR, 1); + } + } + + if (spu.init.debug_flags & CELL_DEBUG_SYNC) { + wait_on_mask(1 << TAG_SURFACE_CLEAR); + } + +#endif /* CLEAR_OPT */ + + D_PRINTF(CELL_DEBUG_CMD, "CLEAR SURF done\n"); +} + + +static void +cmd_release_verts(const struct cell_command_release_verts *release) +{ + D_PRINTF(CELL_DEBUG_CMD, "RELEASE VERTS %u\n", release->vertex_buf); + ASSERT(release->vertex_buf != ~0U); + release_buffer(release->vertex_buf); +} + + +/** + * Process a CELL_CMD_STATE_FRAGMENT_OPS command. + * This involves installing new fragment ops SPU code. + * If this function is never called, we'll use a regular C fallback function + * for fragment processing. + */ +static void +cmd_state_fragment_ops(const struct cell_command_fragment_ops *fops) +{ + D_PRINTF(CELL_DEBUG_CMD, "CMD_STATE_FRAGMENT_OPS\n"); + + /* Copy state info (for fallback case only - this will eventually + * go away when the fallback case goes away) + */ + memcpy(&spu.depth_stencil_alpha, &fops->dsa, sizeof(fops->dsa)); + memcpy(&spu.blend, &fops->blend, sizeof(fops->blend)); + memcpy(&spu.blend_color, &fops->blend_color, sizeof(fops->blend_color)); + + /* Make sure the SPU knows which buffers it's expected to read when + * it's told to pull tiles. + */ + spu.read_depth_stencil = (spu.depth_stencil_alpha.depth.enabled || spu.depth_stencil_alpha.stencil[0].enabled); + + /* If we're forcing the fallback code to be used (for debug purposes), + * install that. Otherwise install the incoming SPU code. + */ + if ((spu.init.debug_flags & CELL_DEBUG_FRAGMENT_OP_FALLBACK) != 0) { + static unsigned int warned = 0; + if (!warned) { + fprintf(stderr, "Cell Warning: using fallback per-fragment code\n"); + warned = 1; + } + /* The following two lines aren't really necessary if you + * know the debug flags won't change during a run, and if you + * know that the function pointers are initialized correctly. + * We set them here to allow a person to change the debug + * flags during a run (from inside a debugger). + */ + spu.fragment_ops[CELL_FACING_FRONT] = spu_fallback_fragment_ops; + spu.fragment_ops[CELL_FACING_BACK] = spu_fallback_fragment_ops; + return; + } + + /* Make sure the SPU code buffer is large enough to hold the incoming code. + * Note that we *don't* use align_malloc() and align_free(), because + * those utility functions are *not* available in SPU code. + * */ + if (spu.fragment_ops_code_size < fops->total_code_size) { + if (spu.fragment_ops_code != NULL) { + free(spu.fragment_ops_code); + } + spu.fragment_ops_code_size = fops->total_code_size; + spu.fragment_ops_code = malloc(fops->total_code_size); + if (spu.fragment_ops_code == NULL) { + /* Whoops. */ + fprintf(stderr, "CELL Warning: failed to allocate fragment ops code (%d bytes) - using fallback\n", fops->total_code_size); + spu.fragment_ops_code = NULL; + spu.fragment_ops_code_size = 0; + spu.fragment_ops[CELL_FACING_FRONT] = spu_fallback_fragment_ops; + spu.fragment_ops[CELL_FACING_BACK] = spu_fallback_fragment_ops; + return; + } + } + + /* Copy the SPU code from the command buffer to the spu buffer */ + memcpy(spu.fragment_ops_code, fops->code, fops->total_code_size); + + /* Set the pointers for the front-facing and back-facing fragments + * to the specified offsets within the code. Note that if the + * front-facing and back-facing code are the same, they'll have + * the same offset. + */ + spu.fragment_ops[CELL_FACING_FRONT] = (spu_fragment_ops_func) &spu.fragment_ops_code[fops->front_code_index]; + spu.fragment_ops[CELL_FACING_BACK] = (spu_fragment_ops_func) &spu.fragment_ops_code[fops->back_code_index]; +} + +static void +cmd_state_fragment_program(const struct cell_command_fragment_program *fp) +{ + D_PRINTF(CELL_DEBUG_CMD, "CMD_STATE_FRAGMENT_PROGRAM\n"); + /* Copy SPU code from batch buffer to spu buffer */ + memcpy(spu.fragment_program_code, fp->code, + SPU_MAX_FRAGMENT_PROGRAM_INSTS * 4); +#if 01 + /* Point function pointer at new code */ + spu.fragment_program = (spu_fragment_program_func)spu.fragment_program_code; +#endif +} + + +static uint +cmd_state_fs_constants(const uint64_t *buffer, uint pos) +{ + const uint num_const = buffer[pos + 1]; + const float *constants = (const float *) &buffer[pos + 2]; + uint i; + + D_PRINTF(CELL_DEBUG_CMD, "CMD_STATE_FS_CONSTANTS (%u)\n", num_const); + + /* Expand each float to float[4] for SOA execution */ + for (i = 0; i < num_const; i++) { + D_PRINTF(CELL_DEBUG_CMD, " const[%u] = %f\n", i, constants[i]); + spu.constants[i] = spu_splats(constants[i]); + } + + /* return new buffer pos (in 8-byte words) */ + return pos + 2 + num_const / 2; +} + + +static void +cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) +{ + D_PRINTF(CELL_DEBUG_CMD, "FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", + cmd->width, + cmd->height, + cmd->color_start, + cmd->color_format, + cmd->depth_format); + + ASSERT_ALIGN16(cmd->color_start); + ASSERT_ALIGN16(cmd->depth_start); + + spu.fb.color_start = cmd->color_start; + spu.fb.depth_start = cmd->depth_start; + spu.fb.color_format = cmd->color_format; + spu.fb.depth_format = cmd->depth_format; + spu.fb.width = cmd->width; + spu.fb.height = cmd->height; + spu.fb.width_tiles = (spu.fb.width + TILE_SIZE - 1) / TILE_SIZE; + spu.fb.height_tiles = (spu.fb.height + TILE_SIZE - 1) / TILE_SIZE; + + switch (spu.fb.depth_format) { + case PIPE_FORMAT_Z32_UNORM: + spu.fb.zsize = 4; + spu.fb.zscale = (float) 0xffffffffu; + break; + case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: + spu.fb.zsize = 4; + spu.fb.zscale = (float) 0x00ffffffu; + break; + case PIPE_FORMAT_Z16_UNORM: + spu.fb.zsize = 2; + spu.fb.zscale = (float) 0xffffu; + break; + default: + spu.fb.zsize = 0; + break; + } +} + + +/** + * Tex texture mask_s/t and scale_s/t fields depend on the texture size and + * sampler wrap modes. + */ +static void +update_tex_masks(struct spu_texture *texture, + const struct pipe_sampler_state *sampler) +{ + uint i; + + for (i = 0; i < CELL_MAX_TEXTURE_LEVELS; i++) { + int width = texture->level[i].width; + int height = texture->level[i].height; + + if (sampler->wrap_s == PIPE_TEX_WRAP_REPEAT) + texture->level[i].mask_s = spu_splats(width - 1); + else + texture->level[i].mask_s = spu_splats(~0); + + if (sampler->wrap_t == PIPE_TEX_WRAP_REPEAT) + texture->level[i].mask_t = spu_splats(height - 1); + else + texture->level[i].mask_t = spu_splats(~0); + + if (sampler->normalized_coords) { + texture->level[i].scale_s = spu_splats((float) width); + texture->level[i].scale_t = spu_splats((float) height); + } + else { + texture->level[i].scale_s = spu_splats(1.0f); + texture->level[i].scale_t = spu_splats(1.0f); + } + } +} + + +static void +cmd_state_sampler(const struct cell_command_sampler *sampler) +{ + uint unit = sampler->unit; + + D_PRINTF(CELL_DEBUG_CMD, "SAMPLER [%u]\n", unit); + + spu.sampler[unit] = sampler->state; + + switch (spu.sampler[unit].min_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + spu.min_sample_texture_2d[unit] = sample_texture_2d_bilinear; + break; + case PIPE_TEX_FILTER_ANISO: + /* fall-through, for now */ + case PIPE_TEX_FILTER_NEAREST: + spu.min_sample_texture_2d[unit] = sample_texture_2d_nearest; + break; + default: + ASSERT(0); + } + + switch (spu.sampler[sampler->unit].mag_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + spu.mag_sample_texture_2d[unit] = sample_texture_2d_bilinear; + break; + case PIPE_TEX_FILTER_ANISO: + /* fall-through, for now */ + case PIPE_TEX_FILTER_NEAREST: + spu.mag_sample_texture_2d[unit] = sample_texture_2d_nearest; + break; + default: + ASSERT(0); + } + + switch (spu.sampler[sampler->unit].min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + case PIPE_TEX_MIPFILTER_LINEAR: + spu.sample_texture_2d[unit] = sample_texture_2d_lod; + break; + case PIPE_TEX_MIPFILTER_NONE: + spu.sample_texture_2d[unit] = spu.mag_sample_texture_2d[unit]; + break; + default: + ASSERT(0); + } + + update_tex_masks(&spu.texture[unit], &spu.sampler[unit]); +} + + +static void +cmd_state_texture(const struct cell_command_texture *texture) +{ + const uint unit = texture->unit; + uint i; + + D_PRINTF(CELL_DEBUG_CMD, "TEXTURE [%u]\n", texture->unit); + + spu.texture[unit].max_level = 0; + spu.texture[unit].target = texture->target; + + for (i = 0; i < CELL_MAX_TEXTURE_LEVELS; i++) { + uint width = texture->width[i]; + uint height = texture->height[i]; + uint depth = texture->depth[i]; + + D_PRINTF(CELL_DEBUG_CMD, " LEVEL %u: at %p size[0] %u x %u\n", i, + texture->start[i], texture->width[i], texture->height[i]); + + spu.texture[unit].level[i].start = texture->start[i]; + spu.texture[unit].level[i].width = width; + spu.texture[unit].level[i].height = height; + spu.texture[unit].level[i].depth = depth; + + spu.texture[unit].level[i].tiles_per_row = + (width + TILE_SIZE - 1) / TILE_SIZE; + + spu.texture[unit].level[i].bytes_per_image = + 4 * align(width, TILE_SIZE) * align(height, TILE_SIZE) * depth; + + spu.texture[unit].level[i].max_s = spu_splats((int) width - 1); + spu.texture[unit].level[i].max_t = spu_splats((int) height - 1); + + if (texture->start[i]) + spu.texture[unit].max_level = i; + } + + update_tex_masks(&spu.texture[unit], &spu.sampler[unit]); +} + + +static void +cmd_state_vertex_info(const struct vertex_info *vinfo) +{ + D_PRINTF(CELL_DEBUG_CMD, "VERTEX_INFO num_attribs=%u\n", vinfo->num_attribs); + ASSERT(vinfo->num_attribs >= 1); + ASSERT(vinfo->num_attribs <= 8); + memcpy(&spu.vertex_info, vinfo, sizeof(*vinfo)); +} + + +static void +cmd_state_vs_array_info(const struct cell_array_info *vs_info) +{ + const unsigned attr = vs_info->attr; + + ASSERT(attr < PIPE_MAX_ATTRIBS); + draw.vertex_fetch.src_ptr[attr] = vs_info->base; + draw.vertex_fetch.pitch[attr] = vs_info->pitch; + draw.vertex_fetch.size[attr] = vs_info->size; + draw.vertex_fetch.code_offset[attr] = vs_info->function_offset; + draw.vertex_fetch.dirty = 1; +} + + +static void +cmd_state_attrib_fetch(const struct cell_attribute_fetch_code *code) +{ + mfc_get(attribute_fetch_code_buffer, + (unsigned int) code->base, /* src */ + code->size, + TAG_BATCH_BUFFER, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << TAG_BATCH_BUFFER); + + draw.vertex_fetch.code = attribute_fetch_code_buffer; +} + + +static void +cmd_finish(void) +{ + D_PRINTF(CELL_DEBUG_CMD, "FINISH\n"); + really_clear_tiles(0); + /* wait for all outstanding DMAs to finish */ + mfc_write_tag_mask(~0); + mfc_read_tag_status_all(); + /* send mbox message to PPU */ + spu_write_out_mbox(CELL_CMD_FINISH); +} + + +/** + * Execute a batch of commands which was sent to us by the PPU. + * See the cell_emit_state.c code to see where the commands come from. + * + * The opcode param encodes the location of the buffer and its size. + */ +static void +cmd_batch(uint opcode) +{ + const uint buf = (opcode >> 8) & 0xff; + uint size = (opcode >> 16); + uint64_t buffer[CELL_BUFFER_SIZE / 8] ALIGN16_ATTRIB; + const unsigned usize = size / sizeof(buffer[0]); + uint pos; + + D_PRINTF(CELL_DEBUG_CMD, "BATCH buffer %u, len %u, from %p\n", + buf, size, spu.init.buffers[buf]); + + ASSERT((opcode & CELL_CMD_OPCODE_MASK) == CELL_CMD_BATCH); + + ASSERT_ALIGN16(spu.init.buffers[buf]); + + size = ROUNDUP16(size); + + ASSERT_ALIGN16(spu.init.buffers[buf]); + + mfc_get(buffer, /* dest */ + (unsigned int) spu.init.buffers[buf], /* src */ + size, + TAG_BATCH_BUFFER, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << TAG_BATCH_BUFFER); + + /* Tell PPU we're done copying the buffer to local store */ + D_PRINTF(CELL_DEBUG_CMD, "release batch buf %u\n", buf); + release_buffer(buf); + + /* + * Loop over commands in the batch buffer + */ + for (pos = 0; pos < usize; /* no incr */) { + switch (buffer[pos]) { + /* + * rendering commands + */ + case CELL_CMD_CLEAR_SURFACE: + { + struct cell_command_clear_surface *clr + = (struct cell_command_clear_surface *) &buffer[pos]; + cmd_clear_surface(clr); + pos += sizeof(*clr) / 8; + } + break; + case CELL_CMD_RENDER: + { + struct cell_command_render *render + = (struct cell_command_render *) &buffer[pos]; + uint pos_incr; + cmd_render(render, &pos_incr); + pos += pos_incr; + } + break; + /* + * state-update commands + */ + case CELL_CMD_STATE_FRAMEBUFFER: + { + struct cell_command_framebuffer *fb + = (struct cell_command_framebuffer *) &buffer[pos]; + cmd_state_framebuffer(fb); + pos += sizeof(*fb) / 8; + } + break; + case CELL_CMD_STATE_FRAGMENT_OPS: + { + struct cell_command_fragment_ops *fops + = (struct cell_command_fragment_ops *) &buffer[pos]; + cmd_state_fragment_ops(fops); + /* This is a variant-sized command */ + pos += (sizeof(*fops) + fops->total_code_size)/ 8; + } + break; + case CELL_CMD_STATE_FRAGMENT_PROGRAM: + { + struct cell_command_fragment_program *fp + = (struct cell_command_fragment_program *) &buffer[pos]; + cmd_state_fragment_program(fp); + pos += sizeof(*fp) / 8; + } + break; + case CELL_CMD_STATE_FS_CONSTANTS: + pos = cmd_state_fs_constants(buffer, pos); + break; + case CELL_CMD_STATE_RASTERIZER: + { + struct cell_command_rasterizer *rast = + (struct cell_command_rasterizer *) &buffer[pos]; + spu.rasterizer = rast->rasterizer; + pos += sizeof(*rast) / 8; + } + break; + case CELL_CMD_STATE_SAMPLER: + { + struct cell_command_sampler *sampler + = (struct cell_command_sampler *) &buffer[pos]; + cmd_state_sampler(sampler); + pos += sizeof(*sampler) / 8; + } + break; + case CELL_CMD_STATE_TEXTURE: + { + struct cell_command_texture *texture + = (struct cell_command_texture *) &buffer[pos]; + cmd_state_texture(texture); + pos += sizeof(*texture) / 8; + } + break; + case CELL_CMD_STATE_VERTEX_INFO: + cmd_state_vertex_info((struct vertex_info *) &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct vertex_info)) / 8); + break; + case CELL_CMD_STATE_VIEWPORT: + (void) memcpy(& draw.viewport, &buffer[pos+1], + sizeof(struct pipe_viewport_state)); + pos += (1 + ROUNDUP8(sizeof(struct pipe_viewport_state)) / 8); + break; + case CELL_CMD_STATE_UNIFORMS: + draw.constants = (const float (*)[4]) (uintptr_t) buffer[pos + 1]; + pos += 2; + break; + case CELL_CMD_STATE_VS_ARRAY_INFO: + cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct cell_array_info)) / 8); + break; + case CELL_CMD_STATE_BIND_VS: +#if 0 + spu_bind_vertex_shader(&draw, + (struct cell_shader_info *) &buffer[pos+1]); +#endif + pos += (1 + ROUNDUP8(sizeof(struct cell_shader_info)) / 8); + break; + case CELL_CMD_STATE_ATTRIB_FETCH: + cmd_state_attrib_fetch((struct cell_attribute_fetch_code *) + &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct cell_attribute_fetch_code)) / 8); + break; + /* + * misc commands + */ + case CELL_CMD_FINISH: + cmd_finish(); + pos += 1; + break; + case CELL_CMD_FENCE: + { + struct cell_command_fence *fence_cmd = + (struct cell_command_fence *) &buffer[pos]; + cmd_fence(fence_cmd); + pos += sizeof(*fence_cmd) / 8; + } + break; + case CELL_CMD_RELEASE_VERTS: + { + struct cell_command_release_verts *release + = (struct cell_command_release_verts *) &buffer[pos]; + cmd_release_verts(release); + pos += sizeof(*release) / 8; + } + break; + case CELL_CMD_FLUSH_BUFFER_RANGE: { + struct cell_buffer_range *br = (struct cell_buffer_range *) + &buffer[pos+1]; + + spu_dcache_mark_dirty((unsigned) br->base, br->size); + pos += (1 + ROUNDUP8(sizeof(struct cell_buffer_range)) / 8); + break; + } + default: + printf("SPU %u: bad opcode: 0x%llx\n", spu.init.id, buffer[pos]); + ASSERT(0); + break; + } + } + + D_PRINTF(CELL_DEBUG_CMD, "BATCH complete\n"); +} + + +#define PERF 0 + + +/** + * Main loop for SPEs: Get a command, execute it, repeat. + */ +void +command_loop(void) +{ + int exitFlag = 0; + uint t0, t1; + + D_PRINTF(CELL_DEBUG_CMD, "Enter command loop\n"); + + while (!exitFlag) { + unsigned opcode; + + D_PRINTF(CELL_DEBUG_CMD, "Wait for cmd...\n"); + + if (PERF) + spu_write_decrementer(~0); + + /* read/wait from mailbox */ + opcode = (unsigned int) spu_read_in_mbox(); + D_PRINTF(CELL_DEBUG_CMD, "got cmd 0x%x\n", opcode); + + if (PERF) + t0 = spu_read_decrementer(); + + switch (opcode & CELL_CMD_OPCODE_MASK) { + case CELL_CMD_EXIT: + D_PRINTF(CELL_DEBUG_CMD, "EXIT\n"); + exitFlag = 1; + break; + case CELL_CMD_VS_EXECUTE: +#if 0 + spu_execute_vertex_shader(&draw, &cmd.vs); +#endif + break; + case CELL_CMD_BATCH: + cmd_batch(opcode); + break; + default: + printf("Bad opcode 0x%x!\n", opcode & CELL_CMD_OPCODE_MASK); + } + + if (PERF) { + t1 = spu_read_decrementer(); + printf("wait mbox time: %gms batch time: %gms\n", + (~0u - t0) * spu.init.inv_timebase, + (t0 - t1) * spu.init.inv_timebase); + } + } + + D_PRINTF(CELL_DEBUG_CMD, "Exit command loop\n"); + + if (spu.init.debug_flags & CELL_DEBUG_CACHE) + spu_dcache_report(); +} + +/* Initialize this module; we manage the fragment ops buffer here. */ +void +spu_command_init(void) +{ + /* Install default/fallback fragment processing function. + * This will normally be overriden by a code-gen'd function + * unless CELL_FORCE_FRAGMENT_OPS_FALLBACK is set. + */ + spu.fragment_ops[CELL_FACING_FRONT] = spu_fallback_fragment_ops; + spu.fragment_ops[CELL_FACING_BACK] = spu_fallback_fragment_ops; + + /* Set up the basic empty buffer for code-gen'ed fragment ops */ + spu.fragment_ops_code = NULL; + spu.fragment_ops_code_size = 0; +} + +void +spu_command_close(void) +{ + /* Deallocate the code-gen buffer for fragment ops, and reset the + * fragment ops functions to their initial setting (just to leave + * things in a good state). + */ + if (spu.fragment_ops_code != NULL) { + free(spu.fragment_ops_code); + } + spu_command_init(); +} diff --git a/src/gallium/drivers/cell/spu/spu_command.h b/src/gallium/drivers/cell/spu/spu_command.h new file mode 100644 index 0000000000..83dcdade28 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_command.h @@ -0,0 +1,35 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +extern void +command_loop(void); + +extern void +spu_command_init(void); + +extern void +spu_command_close(void); diff --git a/src/gallium/drivers/cell/spu/spu_dcache.c b/src/gallium/drivers/cell/spu/spu_dcache.c index 167404cdc5..a6d67634fd 100644 --- a/src/gallium/drivers/cell/spu/spu_dcache.c +++ b/src/gallium/drivers/cell/spu/spu_dcache.c @@ -36,7 +36,9 @@ #define CACHE_SET_TAGID(set) (((set) & 0x03) + TAG_DCACHE0) #define CACHE_LOG2NNWAY 2 #define CACHE_LOG2NSETS 6 -/*#define CACHE_STATS 1*/ +#ifdef DEBUG +#define CACHE_STATS 1 +#endif #include <cache-api.h> /* Yes folks, this is ugly. diff --git a/src/gallium/drivers/cell/spu/spu_funcs.c b/src/gallium/drivers/cell/spu/spu_funcs.c new file mode 100644 index 0000000000..ff3d609d25 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_funcs.c @@ -0,0 +1,173 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * SPU functions accessed by shaders. + * + * Authors: Brian Paul + */ + + +#include <string.h> +#include <libmisc.h> +#include <math.h> +#include <cos14_v.h> +#include <sin14_v.h> +#include <simdmath/exp2f4.h> +#include <simdmath/log2f4.h> +#include <simdmath/powf4.h> + +#include "cell/common.h" +#include "spu_main.h" +#include "spu_funcs.h" +#include "spu_texture.h" + + +/** For "return"-ing four vectors */ +struct vec_4x4 +{ + vector float v[4]; +}; + + +static vector float +spu_cos(vector float x) +{ + return _cos14_v(x); +} + +static vector float +spu_sin(vector float x) +{ + return _sin14_v(x); +} + +static vector float +spu_pow(vector float x, vector float y) +{ + return _powf4(x, y); +} + +static vector float +spu_exp2(vector float x) +{ + return _exp2f4(x); +} + +static vector float +spu_log2(vector float x) +{ + return _log2f4(x); +} + + +static struct vec_4x4 +spu_tex_2d(vector float s, vector float t, vector float r, vector float q, + unsigned unit) +{ + struct vec_4x4 colors; + (void) r; + (void) q; + spu.sample_texture_2d[unit](s, t, unit, 0, 0, colors.v); + return colors; +} + +static struct vec_4x4 +spu_tex_3d(vector float s, vector float t, vector float r, vector float q, + unsigned unit) +{ + struct vec_4x4 colors; + (void) r; + (void) q; + spu.sample_texture_2d[unit](s, t, unit, 0, 0, colors.v); + return colors; +} + +static struct vec_4x4 +spu_tex_cube(vector float s, vector float t, vector float r, vector float q, + unsigned unit) +{ + struct vec_4x4 colors; + (void) q; + sample_texture_cube(s, t, r, unit, colors.v); + return colors; +} + + +/** + * Add named function to list of "exported" functions that will be + * made available to the PPU-hosted code generator. + */ +static void +export_func(struct cell_spu_function_info *spu_functions, + const char *name, void *addr) +{ + uint n = spu_functions->num; + ASSERT(strlen(name) < 16); + strcpy(spu_functions->names[n], name); + spu_functions->addrs[n] = (uint) addr; + spu_functions->num++; + ASSERT(spu_functions->num <= 16); +} + + +/** + * Return info about the SPU's function to the PPU / main memory. + * The PPU needs to know the address of some SPU-side functions so + * that we can generate shader code with function calls. + */ +void +return_function_info(void) +{ + struct cell_spu_function_info funcs ALIGN16_ATTRIB; + int tag = TAG_MISC; + + ASSERT(sizeof(funcs) == 256); /* must be multiple of 16 bytes */ + + funcs.num = 0; + export_func(&funcs, "spu_cos", &spu_cos); + export_func(&funcs, "spu_sin", &spu_sin); + export_func(&funcs, "spu_pow", &spu_pow); + export_func(&funcs, "spu_exp2", &spu_exp2); + export_func(&funcs, "spu_log2", &spu_log2); + export_func(&funcs, "spu_tex_2d", &spu_tex_2d); + export_func(&funcs, "spu_tex_3d", &spu_tex_3d); + export_func(&funcs, "spu_tex_cube", &spu_tex_cube); + + /* Send the function info back to the PPU / main memory */ + mfc_put((void *) &funcs, /* src in local store */ + (unsigned int) spu.init.spu_functions, /* dst in main memory */ + sizeof(funcs), /* bytes */ + tag, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << tag); +} + + + diff --git a/src/gallium/drivers/cell/spu/spu_funcs.h b/src/gallium/drivers/cell/spu/spu_funcs.h new file mode 100644 index 0000000000..3adb6ae99f --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_funcs.h @@ -0,0 +1,35 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SPU_FUNCS_H +#define SPU_FUNCS_H + +extern void +return_function_info(void); + +#endif + diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 78260c4259..97c86d194d 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -32,16 +32,15 @@ #include <stdio.h> #include <libmisc.h> +#include "pipe/p_defines.h" + +#include "spu_funcs.h" +#include "spu_command.h" #include "spu_main.h" -#include "spu_render.h" #include "spu_per_fragment_op.h" #include "spu_texture.h" -#include "spu_tile.h" //#include "spu_test.h" -#include "spu_vertex_shader.h" -#include "spu_dcache.h" #include "cell/common.h" -#include "pipe/p_defines.h" /* @@ -50,600 +49,8 @@ helpful headers: /opt/cell/sdk/usr/include/libmisc.h */ -boolean Debug = FALSE; - struct spu_global spu; -struct spu_vs_context draw; - - -/** - * Buffers containing dynamically generated SPU code: - */ -static unsigned char attribute_fetch_code_buffer[136 * PIPE_MAX_ATTRIBS] - ALIGN16_ATTRIB; - - - -/** - * Tell the PPU that this SPU has finished copying a buffer to - * local store and that it may be reused by the PPU. - * This is done by writting a 16-byte batch-buffer-status block back into - * main memory (in cell_context->buffer_status[]). - */ -static void -release_buffer(uint buffer) -{ - /* Evidently, using less than a 16-byte status doesn't work reliably */ - static const uint status[4] ALIGN16_ATTRIB - = {CELL_BUFFER_STATUS_FREE, 0, 0, 0}; - - const uint index = 4 * (spu.init.id * CELL_NUM_BUFFERS + buffer); - uint *dst = spu.init.buffer_status + index; - - ASSERT(buffer < CELL_NUM_BUFFERS); - - mfc_put((void *) &status, /* src in local memory */ - (unsigned int) dst, /* dst in main memory */ - sizeof(status), /* size */ - TAG_MISC, /* tag is unimportant */ - 0, /* tid */ - 0 /* rid */); -} - - -/** - * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled - * tiles back to the main framebuffer. - */ -static void -really_clear_tiles(uint surfaceIndex) -{ - const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; - uint i; - - if (surfaceIndex == 0) { - clear_c_tile(&spu.ctile); - - for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { - uint tx = i % spu.fb.width_tiles; - uint ty = i / spu.fb.width_tiles; - if (spu.ctile_status[ty][tx] == TILE_STATUS_CLEAR) { - put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0); - } - } - } - else { - clear_z_tile(&spu.ztile); - - for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { - uint tx = i % spu.fb.width_tiles; - uint ty = i / spu.fb.width_tiles; - if (spu.ztile_status[ty][tx] == TILE_STATUS_CLEAR) - put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 1); - } - } - -#if 0 - wait_on_mask(1 << TAG_SURFACE_CLEAR); -#endif -} - - -static void -cmd_clear_surface(const struct cell_command_clear_surface *clear) -{ - if (Debug) - printf("SPU %u: CLEAR SURF %u to 0x%08x\n", spu.init.id, - clear->surface, clear->value); - - if (clear->surface == 0) { - spu.fb.color_clear_value = clear->value; - if (spu.init.debug_flags & CELL_DEBUG_CHECKER) { - uint x = (spu.init.id << 4) | (spu.init.id << 12) | - (spu.init.id << 20) | (spu.init.id << 28); - spu.fb.color_clear_value ^= x; - } - } - else { - spu.fb.depth_clear_value = clear->value; - } - -#define CLEAR_OPT 1 -#if CLEAR_OPT - - /* Simply set all tiles' status to CLEAR. - * When we actually begin rendering into a tile, we'll initialize it to - * the clear value. If any tiles go untouched during the frame, - * really_clear_tiles() will set them to the clear value. - */ - if (clear->surface == 0) { - memset(spu.ctile_status, TILE_STATUS_CLEAR, sizeof(spu.ctile_status)); - } - else { - memset(spu.ztile_status, TILE_STATUS_CLEAR, sizeof(spu.ztile_status)); - } - -#else - - /* - * This path clears the whole framebuffer to the clear color right now. - */ - - /* - printf("SPU: %s num=%d w=%d h=%d\n", - __FUNCTION__, num_tiles, spu.fb.width_tiles, spu.fb.height_tiles); - */ - - /* init a single tile to the clear value */ - if (clear->surface == 0) { - clear_c_tile(&spu.ctile); - } - else { - clear_z_tile(&spu.ztile); - } - - /* walk over my tiles, writing the 'clear' tile's data */ - { - const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; - uint i; - for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { - uint tx = i % spu.fb.width_tiles; - uint ty = i / spu.fb.width_tiles; - if (clear->surface == 0) - put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0); - else - put_tile(tx, ty, &spu.ztile, TAG_SURFACE_CLEAR, 1); - } - } - - if (spu.init.debug_flags & CELL_DEBUG_SYNC) { - wait_on_mask(1 << TAG_SURFACE_CLEAR); - } - -#endif /* CLEAR_OPT */ - - if (Debug) - printf("SPU %u: CLEAR SURF done\n", spu.init.id); -} - - -static void -cmd_release_verts(const struct cell_command_release_verts *release) -{ - if (Debug) - printf("SPU %u: RELEASE VERTS %u\n", - spu.init.id, release->vertex_buf); - ASSERT(release->vertex_buf != ~0U); - release_buffer(release->vertex_buf); -} - - -/** - * Process a CELL_CMD_STATE_FRAGMENT_OPS command. - * This involves installing new fragment ops SPU code. - * If this function is never called, we'll use a regular C fallback function - * for fragment processing. - */ -static void -cmd_state_fragment_ops(const struct cell_command_fragment_ops *fops) -{ - if (Debug) - printf("SPU %u: CMD_STATE_FRAGMENT_OPS\n", spu.init.id); - /* Copy SPU code from batch buffer to spu buffer */ - memcpy(spu.fragment_ops_code, fops->code, SPU_MAX_FRAGMENT_OPS_INSTS * 4); - /* Copy state info (for fallback case only) */ - memcpy(&spu.depth_stencil_alpha, &fops->dsa, sizeof(fops->dsa)); - memcpy(&spu.blend, &fops->blend, sizeof(fops->blend)); - - /* Point function pointer at new code */ - spu.fragment_ops = (spu_fragment_ops_func) spu.fragment_ops_code; - - spu.read_depth = spu.depth_stencil_alpha.depth.enabled; - spu.read_stencil = spu.depth_stencil_alpha.stencil[0].enabled; -} - - -static void -cmd_state_fragment_program(const struct cell_command_fragment_program *fp) -{ - if (Debug) - printf("SPU %u: CMD_STATE_FRAGMENT_PROGRAM\n", spu.init.id); - /* Copy SPU code from batch buffer to spu buffer */ - memcpy(spu.fragment_program_code, fp->code, - SPU_MAX_FRAGMENT_PROGRAM_INSTS * 4); -#if 01 - /* Point function pointer at new code */ - spu.fragment_program = (spu_fragment_program_func)spu.fragment_program_code; -#endif -} - - -static void -cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) -{ - if (Debug) - printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", - spu.init.id, - cmd->width, - cmd->height, - cmd->color_start, - cmd->color_format, - cmd->depth_format); - - ASSERT_ALIGN16(cmd->color_start); - ASSERT_ALIGN16(cmd->depth_start); - - spu.fb.color_start = cmd->color_start; - spu.fb.depth_start = cmd->depth_start; - spu.fb.color_format = cmd->color_format; - spu.fb.depth_format = cmd->depth_format; - spu.fb.width = cmd->width; - spu.fb.height = cmd->height; - spu.fb.width_tiles = (spu.fb.width + TILE_SIZE - 1) / TILE_SIZE; - spu.fb.height_tiles = (spu.fb.height + TILE_SIZE - 1) / TILE_SIZE; - - switch (spu.fb.depth_format) { - case PIPE_FORMAT_Z32_UNORM: - spu.fb.zsize = 4; - spu.fb.zscale = (float) 0xffffffffu; - break; - case PIPE_FORMAT_Z24S8_UNORM: - case PIPE_FORMAT_S8Z24_UNORM: - case PIPE_FORMAT_Z24X8_UNORM: - case PIPE_FORMAT_X8Z24_UNORM: - spu.fb.zsize = 4; - spu.fb.zscale = (float) 0x00ffffffu; - break; - case PIPE_FORMAT_Z16_UNORM: - spu.fb.zsize = 2; - spu.fb.zscale = (float) 0xffffu; - break; - default: - spu.fb.zsize = 0; - break; - } -} - - -static void -cmd_state_sampler(const struct cell_command_sampler *sampler) -{ - if (Debug) - printf("SPU %u: SAMPLER [%u]\n", - spu.init.id, sampler->unit); - - spu.sampler[sampler->unit] = sampler->state; - if (spu.sampler[sampler->unit].min_img_filter == PIPE_TEX_FILTER_LINEAR) - spu.sample_texture[sampler->unit] = sample_texture_bilinear; - else - spu.sample_texture[sampler->unit] = sample_texture_nearest; -} - - -static void -cmd_state_texture(const struct cell_command_texture *texture) -{ - const uint unit = texture->unit; - const uint width = texture->width; - const uint height = texture->height; - - if (Debug) { - printf("SPU %u: TEXTURE [%u] at %p size %u x %u\n", spu.init.id, - texture->unit, texture->start, - texture->width, texture->height); - } - - spu.texture[unit].start = texture->start; - spu.texture[unit].width = width; - spu.texture[unit].height = height; - - spu.texture[unit].tiles_per_row = width / TILE_SIZE; - - spu.texture[unit].tex_size = (vector float) { width, height, 0.0, 0.0}; - spu.texture[unit].tex_size_mask = (vector unsigned int) - { width - 1, height - 1, 0, 0 }; - spu.texture[unit].tex_size_x_mask = spu_splats(width - 1); - spu.texture[unit].tex_size_y_mask = spu_splats(height - 1); -} - - -static void -cmd_state_vertex_info(const struct vertex_info *vinfo) -{ - if (Debug) { - printf("SPU %u: VERTEX_INFO num_attribs=%u\n", spu.init.id, - vinfo->num_attribs); - } - ASSERT(vinfo->num_attribs >= 1); - ASSERT(vinfo->num_attribs <= 8); - memcpy(&spu.vertex_info, vinfo, sizeof(*vinfo)); -} - - -static void -cmd_state_vs_array_info(const struct cell_array_info *vs_info) -{ - const unsigned attr = vs_info->attr; - - ASSERT(attr < PIPE_MAX_ATTRIBS); - draw.vertex_fetch.src_ptr[attr] = vs_info->base; - draw.vertex_fetch.pitch[attr] = vs_info->pitch; - draw.vertex_fetch.size[attr] = vs_info->size; - draw.vertex_fetch.code_offset[attr] = vs_info->function_offset; - draw.vertex_fetch.dirty = 1; -} - - -static void -cmd_state_attrib_fetch(const struct cell_attribute_fetch_code *code) -{ - mfc_get(attribute_fetch_code_buffer, - (unsigned int) code->base, /* src */ - code->size, - TAG_BATCH_BUFFER, - 0, /* tid */ - 0 /* rid */); - wait_on_mask(1 << TAG_BATCH_BUFFER); - - draw.vertex_fetch.code = attribute_fetch_code_buffer; -} - - -static void -cmd_finish(void) -{ - if (Debug) - printf("SPU %u: FINISH\n", spu.init.id); - really_clear_tiles(0); - /* wait for all outstanding DMAs to finish */ - mfc_write_tag_mask(~0); - mfc_read_tag_status_all(); - /* send mbox message to PPU */ - spu_write_out_mbox(CELL_CMD_FINISH); -} - - -/** - * Execute a batch of commands which was sent to us by the PPU. - * See the cell_emit_state.c code to see where the commands come from. - * - * The opcode param encodes the location of the buffer and its size. - */ -static void -cmd_batch(uint opcode) -{ - const uint buf = (opcode >> 8) & 0xff; - uint size = (opcode >> 16); - uint64_t buffer[CELL_BUFFER_SIZE / 8] ALIGN16_ATTRIB; - const unsigned usize = size / sizeof(buffer[0]); - uint pos; - - if (Debug) - printf("SPU %u: BATCH buffer %u, len %u, from %p\n", - spu.init.id, buf, size, spu.init.buffers[buf]); - - ASSERT((opcode & CELL_CMD_OPCODE_MASK) == CELL_CMD_BATCH); - - ASSERT_ALIGN16(spu.init.buffers[buf]); - - size = ROUNDUP16(size); - - ASSERT_ALIGN16(spu.init.buffers[buf]); - - mfc_get(buffer, /* dest */ - (unsigned int) spu.init.buffers[buf], /* src */ - size, - TAG_BATCH_BUFFER, - 0, /* tid */ - 0 /* rid */); - wait_on_mask(1 << TAG_BATCH_BUFFER); - - /* Tell PPU we're done copying the buffer to local store */ - if (Debug) - printf("SPU %u: release batch buf %u\n", spu.init.id, buf); - release_buffer(buf); - - /* - * Loop over commands in the batch buffer - */ - for (pos = 0; pos < usize; /* no incr */) { - switch (buffer[pos]) { - /* - * rendering commands - */ - case CELL_CMD_CLEAR_SURFACE: - { - struct cell_command_clear_surface *clr - = (struct cell_command_clear_surface *) &buffer[pos]; - cmd_clear_surface(clr); - pos += sizeof(*clr) / 8; - } - break; - case CELL_CMD_RENDER: - { - struct cell_command_render *render - = (struct cell_command_render *) &buffer[pos]; - uint pos_incr; - cmd_render(render, &pos_incr); - pos += pos_incr; - } - break; - /* - * state-update commands - */ - case CELL_CMD_STATE_FRAMEBUFFER: - { - struct cell_command_framebuffer *fb - = (struct cell_command_framebuffer *) &buffer[pos]; - cmd_state_framebuffer(fb); - pos += sizeof(*fb) / 8; - } - break; - case CELL_CMD_STATE_FRAGMENT_OPS: - { - struct cell_command_fragment_ops *fops - = (struct cell_command_fragment_ops *) &buffer[pos]; - cmd_state_fragment_ops(fops); - pos += sizeof(*fops) / 8; - } - break; - case CELL_CMD_STATE_FRAGMENT_PROGRAM: - { - struct cell_command_fragment_program *fp - = (struct cell_command_fragment_program *) &buffer[pos]; - cmd_state_fragment_program(fp); - pos += sizeof(*fp) / 8; - } - break; - case CELL_CMD_STATE_SAMPLER: - { - struct cell_command_sampler *sampler - = (struct cell_command_sampler *) &buffer[pos]; - cmd_state_sampler(sampler); - pos += sizeof(*sampler) / 8; - } - break; - case CELL_CMD_STATE_TEXTURE: - { - struct cell_command_texture *texture - = (struct cell_command_texture *) &buffer[pos]; - cmd_state_texture(texture); - pos += sizeof(*texture) / 8; - } - break; - case CELL_CMD_STATE_VERTEX_INFO: - cmd_state_vertex_info((struct vertex_info *) &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct vertex_info)) / 8); - break; - case CELL_CMD_STATE_VIEWPORT: - (void) memcpy(& draw.viewport, &buffer[pos+1], - sizeof(struct pipe_viewport_state)); - pos += (1 + ROUNDUP8(sizeof(struct pipe_viewport_state)) / 8); - break; - case CELL_CMD_STATE_UNIFORMS: - draw.constants = (const float (*)[4]) (uintptr_t) buffer[pos + 1]; - pos += 2; - break; - case CELL_CMD_STATE_VS_ARRAY_INFO: - cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct cell_array_info)) / 8); - break; - case CELL_CMD_STATE_BIND_VS: -#if 0 - spu_bind_vertex_shader(&draw, - (struct cell_shader_info *) &buffer[pos+1]); -#endif - pos += (1 + ROUNDUP8(sizeof(struct cell_shader_info)) / 8); - break; - case CELL_CMD_STATE_ATTRIB_FETCH: - cmd_state_attrib_fetch((struct cell_attribute_fetch_code *) - &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct cell_attribute_fetch_code)) / 8); - break; - /* - * misc commands - */ - case CELL_CMD_FINISH: - cmd_finish(); - pos += 1; - break; - case CELL_CMD_RELEASE_VERTS: - { - struct cell_command_release_verts *release - = (struct cell_command_release_verts *) &buffer[pos]; - cmd_release_verts(release); - pos += sizeof(*release) / 8; - } - break; - case CELL_CMD_FLUSH_BUFFER_RANGE: { - struct cell_buffer_range *br = (struct cell_buffer_range *) - &buffer[pos+1]; - - spu_dcache_mark_dirty((unsigned) br->base, br->size); - pos += (1 + ROUNDUP8(sizeof(struct cell_buffer_range)) / 8); - break; - } - default: - printf("SPU %u: bad opcode: 0x%llx\n", spu.init.id, buffer[pos]); - ASSERT(0); - break; - } - } - - if (Debug) - printf("SPU %u: BATCH complete\n", spu.init.id); -} - - -/** - * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. - */ -static void -main_loop(void) -{ - struct cell_command cmd; - int exitFlag = 0; - - if (Debug) - printf("SPU %u: Enter main loop\n", spu.init.id); - - ASSERT((sizeof(struct cell_command) & 0xf) == 0); - ASSERT_ALIGN16(&cmd); - - while (!exitFlag) { - unsigned opcode; - int tag = 0; - - if (Debug) - printf("SPU %u: Wait for cmd...\n", spu.init.id); - - /* read/wait from mailbox */ - opcode = (unsigned int) spu_read_in_mbox(); - - if (Debug) - printf("SPU %u: got cmd 0x%x\n", spu.init.id, opcode); - - /* command payload */ - mfc_get(&cmd, /* dest */ - (unsigned int) spu.init.cmd, /* src */ - sizeof(struct cell_command), /* bytes */ - tag, - 0, /* tid */ - 0 /* rid */); - wait_on_mask( 1 << tag ); - - /* - * NOTE: most commands should be contained in a batch buffer - */ - - switch (opcode & CELL_CMD_OPCODE_MASK) { - case CELL_CMD_EXIT: - if (Debug) - printf("SPU %u: EXIT\n", spu.init.id); - exitFlag = 1; - break; - case CELL_CMD_VS_EXECUTE: -#if 0 - spu_execute_vertex_shader(&draw, &cmd.vs); -#endif - break; - case CELL_CMD_BATCH: - cmd_batch(opcode); - break; - default: - printf("Bad opcode!\n"); - } - - } - - if (Debug) - printf("SPU %u: Exit main loop\n", spu.init.id); - - spu_dcache_report(); -} - - static void one_time_init(void) @@ -651,15 +58,8 @@ one_time_init(void) memset(spu.ctile_status, TILE_STATUS_DEFINED, sizeof(spu.ctile_status)); memset(spu.ztile_status, TILE_STATUS_DEFINED, sizeof(spu.ztile_status)); invalidate_tex_cache(); - - /* Install default/fallback fragment processing function. - * This will normally be overriden by a code-gen'd function. - */ - spu.fragment_ops = spu_fallback_fragment_ops; } - - /* In some versions of the SDK the SPE main takes 'unsigned long' as a * parameter. In others it takes 'unsigned long long'. Use a define to * select between the two. @@ -682,12 +82,16 @@ main(main_param_t speid, main_param_t argp) ASSERT(sizeof(tile_t) == TILE_SIZE * TILE_SIZE * 4); ASSERT(sizeof(struct cell_command_render) % 8 == 0); + ASSERT(sizeof(struct cell_command_fragment_ops) % 8 == 0); + ASSERT(((unsigned long) &spu.fragment_program_code) % 8 == 0); one_time_init(); + spu_command_init(); - if (Debug) - printf("SPU: main() speid=%lu\n", (unsigned long) speid); + D_PRINTF(CELL_DEBUG_CMD, "main() speid=%lu\n", (unsigned long) speid); + D_PRINTF(CELL_DEBUG_FRAGMENT_OP_FALLBACK, "using fragment op fallback\n"); + /* get initialization data */ mfc_get(&spu.init, /* dest */ (unsigned int) argp, /* src */ sizeof(struct cell_init_info), /* bytes */ @@ -696,12 +100,18 @@ main(main_param_t speid, main_param_t argp) 0 /* rid */); wait_on_mask( 1 << tag ); + if (spu.init.id == 0) { + return_function_info(); + } + #if 0 if (spu.init.id==0) - spu_test_misc(); + spu_test_misc(spu.init.id); #endif - main_loop(); + command_loop(); + + spu_command_close(); return 0; } diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index 2c7b625840..33767e7c51 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -36,9 +36,18 @@ #include "pipe/p_state.h" - -#define MAX_WIDTH 1024 -#define MAX_HEIGHT 1024 +#if DEBUG +/* These debug macros use the unusual construction ", ##__VA_ARGS__" + * which expands to the expected comma + args if variadic arguments + * are supplied, but swallows the comma if there are no variadic + * arguments (which avoids syntax errors that would otherwise occur). + */ +#define D_PRINTF(flag, format,...) \ + if (spu.init.debug_flags & (flag)) \ + printf("SPU %u: " format, spu.init.id, ##__VA_ARGS__) +#else +#define D_PRINTF(...) +#endif /** @@ -61,8 +70,11 @@ typedef union { /** Function for sampling textures */ -typedef vector float (*spu_sample_texture_func)(uint unit, - vector float texcoord); +typedef void (*spu_sample_texture_2d_func)(vector float s, + vector float t, + uint unit, uint level, uint face, + vector float colors[4]); + /** Function for performing per-fragment ops */ typedef void (*spu_fragment_ops_func)(uint x, uint y, @@ -76,9 +88,9 @@ typedef void (*spu_fragment_ops_func)(uint x, uint y, vector unsigned int mask); /** Function for running fragment program */ -typedef void (*spu_fragment_program_func)(vector float *inputs, - vector float *outputs, - vector float *constants); +typedef vector unsigned int (*spu_fragment_program_func)(vector float *inputs, + vector float *outputs, + vector float *constants); struct spu_framebuffer @@ -98,15 +110,27 @@ struct spu_framebuffer } ALIGN16_ATTRIB; -struct spu_texture +/** per-texture level info */ +struct spu_texture_level { void *start; - ushort width, height; + ushort width, height, depth; ushort tiles_per_row; - vector float tex_size; - vector unsigned int tex_size_mask; /**< == int(size - 1) */ - vector unsigned int tex_size_x_mask; /**< == int(size - 1) */ - vector unsigned int tex_size_y_mask; /**< == int(size - 1) */ + uint bytes_per_image; + /** texcoord scale factors */ + vector float scale_s, scale_t, scale_r; + /** texcoord masks (if REPEAT then size-1, else ~0) */ + vector signed int mask_s, mask_t, mask_r; + /** texcoord clamp limits */ + vector signed int max_s, max_t, max_r; +} ALIGN16_ATTRIB; + + +struct spu_texture +{ + struct spu_texture_level level[CELL_MAX_TEXTURE_LEVELS]; + uint max_level; + uint target; /**< PIPE_TEXTURE_x */ } ALIGN16_ATTRIB; @@ -124,7 +148,9 @@ struct spu_global struct spu_framebuffer fb; struct pipe_depth_stencil_alpha_state depth_stencil_alpha; struct pipe_blend_state blend; + struct pipe_blend_color blend_color; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; + struct pipe_rasterizer_state rasterizer; struct spu_texture texture[PIPE_MAX_SAMPLERS]; struct vertex_info vertex_info; @@ -133,39 +159,38 @@ struct spu_global tile_t ztile ALIGN16_ATTRIB; /** Read depth/stencil tiles? */ - boolean read_depth; - boolean read_stencil; + boolean read_depth_stencil; /** Current tiles' status */ ubyte cur_ctile_status, cur_ztile_status; /** Status of all tiles in framebuffer */ - ubyte ctile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; - ubyte ztile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + ubyte ctile_status[CELL_MAX_HEIGHT/TILE_SIZE][CELL_MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + ubyte ztile_status[CELL_MAX_HEIGHT/TILE_SIZE][CELL_MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; - /** Current fragment ops machine code */ - uint fragment_ops_code[SPU_MAX_FRAGMENT_OPS_INSTS]; - /** Current fragment ops function */ - spu_fragment_ops_func fragment_ops; + /** Current fragment ops machine code, at 8-byte boundary */ + uint *fragment_ops_code; + uint fragment_ops_code_size; + /** Current fragment ops functions, 0 = frontfacing, 1 = backfacing */ + spu_fragment_ops_func fragment_ops[2]; - /** Current fragment program machine code */ - uint fragment_program_code[SPU_MAX_FRAGMENT_PROGRAM_INSTS]; + /** Current fragment program machine code, at 8-byte boundary */ + uint fragment_program_code[SPU_MAX_FRAGMENT_PROGRAM_INSTS] ALIGN8_ATTRIB; /** Current fragment ops function */ spu_fragment_program_func fragment_program; /** Current texture sampler function */ - spu_sample_texture_func sample_texture[CELL_MAX_SAMPLERS]; + spu_sample_texture_2d_func sample_texture_2d[CELL_MAX_SAMPLERS]; + spu_sample_texture_2d_func min_sample_texture_2d[CELL_MAX_SAMPLERS]; + spu_sample_texture_2d_func mag_sample_texture_2d[CELL_MAX_SAMPLERS]; - /** Fragment program constants (XXX preliminary/used) */ -#define MAX_CONSTANTS 32 - vector float constants[MAX_CONSTANTS]; + /** Fragment program constants */ + vector float constants[4 * CELL_MAX_CONSTANTS]; } ALIGN16_ATTRIB; extern struct spu_global spu; -extern boolean Debug; - @@ -184,7 +209,7 @@ extern boolean Debug; #define TAG_DCACHE1 21 #define TAG_DCACHE2 22 #define TAG_DCACHE3 23 - +#define TAG_FENCE 24 static INLINE void diff --git a/src/gallium/drivers/cell/spu/spu_per_fragment_op.c b/src/gallium/drivers/cell/spu/spu_per_fragment_op.c index 03dd547845..683664e8a4 100644 --- a/src/gallium/drivers/cell/spu/spu_per_fragment_op.c +++ b/src/gallium/drivers/cell/spu/spu_per_fragment_op.c @@ -40,6 +40,24 @@ #define LINEAR_QUAD_LAYOUT 1 +static INLINE vector float +spu_min(vector float a, vector float b) +{ + vector unsigned int m; + m = spu_cmpgt(a, b); /* m = a > b ? ~0 : 0 */ + return spu_sel(a, b, m); +} + + +static INLINE vector float +spu_max(vector float a, vector float b) +{ + vector unsigned int m; + m = spu_cmpgt(a, b); /* m = a > b ? ~0 : 0 */ + return spu_sel(b, a, m); +} + + /** * Called by rasterizer for each quad after the shader has run. Do * all the per-fragment operations including alpha test, z test, @@ -60,9 +78,12 @@ spu_fallback_fragment_ops(uint x, uint y, vector unsigned int mask) { vector float frag_aos[4]; - unsigned int c0, c1, c2, c3; + unsigned int fbc0, fbc1, fbc2, fbc3 ; /* framebuffer/tile colors */ + unsigned int fragc0, fragc1, fragc2, fragc3; /* fragment colors */ - /* do alpha test */ + /* + * Do alpha test + */ if (spu.depth_stencil_alpha.alpha.enabled) { vector float ref = spu_splats(spu.depth_stencil_alpha.alpha.ref); vector unsigned int amask; @@ -102,7 +123,10 @@ spu_fallback_fragment_ops(uint x, uint y, mask = spu_and(mask, amask); } - /* Z and/or stencil testing... */ + + /* + * Z and/or stencil testing... + */ if (spu.depth_stencil_alpha.depth.enabled || spu.depth_stencil_alpha.stencil[0].enabled) { @@ -178,6 +202,32 @@ spu_fallback_fragment_ops(uint x, uint y, } } + + /* + * If we'll need the current framebuffer/tile colors for blending + * or logicop or colormask, fetch them now. + */ + if (spu.blend.blend_enable || + spu.blend.logicop_enable || + spu.blend.colormask != 0xf) { + +#if LINEAR_QUAD_LAYOUT /* See comments/diagram below */ + fbc0 = colorTile->ui[y][x*2+0]; + fbc1 = colorTile->ui[y][x*2+1]; + fbc2 = colorTile->ui[y][x*2+2]; + fbc3 = colorTile->ui[y][x*2+3]; +#else + fbc0 = colorTile->ui[y+0][x+0]; + fbc1 = colorTile->ui[y+0][x+1]; + fbc2 = colorTile->ui[y+1][x+0]; + fbc3 = colorTile->ui[y+1][x+1]; +#endif + } + + + /* + * Do blending + */ if (spu.blend.blend_enable) { /* blending terms, misc regs */ vector float term1r, term1g, term1b, term1a; @@ -186,43 +236,30 @@ spu_fallback_fragment_ops(uint x, uint y, vector float fbRGBA[4]; /* current framebuffer colors */ - /* get colors from framebuffer/tile */ + /* convert framebuffer colors from packed int to vector float */ { - vector float fc[4]; - uint c0, c1, c2, c3; - -#if LINEAR_QUAD_LAYOUT /* See comments/diagram below */ - c0 = colorTile->ui[y][x*2+0]; - c1 = colorTile->ui[y][x*2+1]; - c2 = colorTile->ui[y][x*2+2]; - c3 = colorTile->ui[y][x*2+3]; -#else - c0 = colorTile->ui[y+0][x+0]; - c1 = colorTile->ui[y+0][x+1]; - c2 = colorTile->ui[y+1][x+0]; - c3 = colorTile->ui[y+1][x+1]; -#endif + vector float temp[4]; /* float colors in AOS form */ switch (spu.fb.color_format) { case PIPE_FORMAT_B8G8R8A8_UNORM: - fc[0] = spu_unpack_B8G8R8A8(c0); - fc[1] = spu_unpack_B8G8R8A8(c1); - fc[2] = spu_unpack_B8G8R8A8(c2); - fc[3] = spu_unpack_B8G8R8A8(c3); + temp[0] = spu_unpack_B8G8R8A8(fbc0); + temp[1] = spu_unpack_B8G8R8A8(fbc1); + temp[2] = spu_unpack_B8G8R8A8(fbc2); + temp[3] = spu_unpack_B8G8R8A8(fbc3); break; case PIPE_FORMAT_A8R8G8B8_UNORM: - fc[0] = spu_unpack_A8R8G8B8(c0); - fc[1] = spu_unpack_A8R8G8B8(c1); - fc[2] = spu_unpack_A8R8G8B8(c2); - fc[3] = spu_unpack_A8R8G8B8(c3); + temp[0] = spu_unpack_A8R8G8B8(fbc0); + temp[1] = spu_unpack_A8R8G8B8(fbc1); + temp[2] = spu_unpack_A8R8G8B8(fbc2); + temp[3] = spu_unpack_A8R8G8B8(fbc3); break; default: ASSERT(0); } - _transpose_matrix4x4(fbRGBA, fc); + _transpose_matrix4x4(fbRGBA, temp); /* fbRGBA = transpose(temp) */ } /* - * Compute Src RGB terms + * Compute Src RGB terms (fragment color * factor) */ switch (spu.blend.rgb_src_factor) { case PIPE_BLENDFACTOR_ONE: @@ -245,13 +282,33 @@ spu_fallback_fragment_ops(uint x, uint y, term1g = spu_mul(fragG, fragA); term1b = spu_mul(fragB, fragA); break; + case PIPE_BLENDFACTOR_DST_COLOR: + term1r = spu_mul(fragR, fbRGBA[0]); + term1g = spu_mul(fragG, fbRGBA[1]); + term1b = spu_mul(fragB, fbRGBA[1]); + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + term1r = spu_mul(fragR, fbRGBA[3]); + term1g = spu_mul(fragG, fbRGBA[3]); + term1b = spu_mul(fragB, fbRGBA[3]); + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + term1r = spu_mul(fragR, spu_splats(spu.blend_color.color[0])); + term1g = spu_mul(fragG, spu_splats(spu.blend_color.color[1])); + term1b = spu_mul(fragB, spu_splats(spu.blend_color.color[2])); + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + term1r = spu_mul(fragR, spu_splats(spu.blend_color.color[3])); + term1g = spu_mul(fragG, spu_splats(spu.blend_color.color[3])); + term1b = spu_mul(fragB, spu_splats(spu.blend_color.color[3])); + break; /* XXX more cases */ default: ASSERT(0); } /* - * Compute Src Alpha term + * Compute Src Alpha term (fragment alpha * factor) */ switch (spu.blend.alpha_src_factor) { case PIPE_BLENDFACTOR_ONE: @@ -263,19 +320,29 @@ spu_fallback_fragment_ops(uint x, uint y, case PIPE_BLENDFACTOR_SRC_ALPHA: term1a = spu_mul(fragA, fragA); break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_DST_ALPHA: + term1a = spu_mul(fragA, fbRGBA[3]); + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + term1a = spu_mul(fragR, spu_splats(spu.blend_color.color[3])); + break; /* XXX more cases */ default: ASSERT(0); } /* - * Compute Dest RGB terms + * Compute Dest RGB terms (framebuffer color * factor) */ switch (spu.blend.rgb_dst_factor) { case PIPE_BLENDFACTOR_ONE: - term2r = fragR; - term2g = fragG; - term2b = fragB; + term2r = fbRGBA[0]; + term2g = fbRGBA[1]; + term2b = fbRGBA[2]; break; case PIPE_BLENDFACTOR_ZERO: term2r = @@ -299,17 +366,37 @@ spu_fallback_fragment_ops(uint x, uint y, term2g = spu_mul(fbRGBA[1], tmp); term2b = spu_mul(fbRGBA[2], tmp); break; - /* XXX more cases */ + case PIPE_BLENDFACTOR_DST_COLOR: + term2r = spu_mul(fbRGBA[0], fbRGBA[0]); + term2g = spu_mul(fbRGBA[1], fbRGBA[1]); + term2b = spu_mul(fbRGBA[2], fbRGBA[2]); + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + term2r = spu_mul(fbRGBA[0], fbRGBA[3]); + term2g = spu_mul(fbRGBA[1], fbRGBA[3]); + term2b = spu_mul(fbRGBA[2], fbRGBA[3]); + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + term2r = spu_mul(fbRGBA[0], spu_splats(spu.blend_color.color[0])); + term2g = spu_mul(fbRGBA[1], spu_splats(spu.blend_color.color[1])); + term2b = spu_mul(fbRGBA[2], spu_splats(spu.blend_color.color[2])); + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + term2r = spu_mul(fbRGBA[0], spu_splats(spu.blend_color.color[3])); + term2g = spu_mul(fbRGBA[1], spu_splats(spu.blend_color.color[3])); + term2b = spu_mul(fbRGBA[2], spu_splats(spu.blend_color.color[3])); + break; + /* XXX more cases */ default: ASSERT(0); } /* - * Compute Dest Alpha term + * Compute Dest Alpha term (framebuffer alpha * factor) */ switch (spu.blend.alpha_dst_factor) { case PIPE_BLENDFACTOR_ONE: - term2a = fragA; + term2a = fbRGBA[3]; break; case PIPE_BLENDFACTOR_SRC_COLOR: term2a = spu_splats(0.0f); @@ -322,6 +409,16 @@ spu_fallback_fragment_ops(uint x, uint y, tmp = spu_sub(one, fragA); term2a = spu_mul(fbRGBA[3], tmp); break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_DST_ALPHA: + term2a = spu_mul(fbRGBA[3], fbRGBA[3]); + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + term2a = spu_mul(fbRGBA[3], spu_splats(spu.blend_color.color[3])); + break; /* XXX more cases */ default: ASSERT(0); @@ -341,7 +438,21 @@ spu_fallback_fragment_ops(uint x, uint y, fragG = spu_sub(term1g, term2g); fragB = spu_sub(term1b, term2b); break; - /* XXX more cases */ + case PIPE_BLEND_REVERSE_SUBTRACT: + fragR = spu_sub(term2r, term1r); + fragG = spu_sub(term2g, term1g); + fragB = spu_sub(term2b, term1b); + break; + case PIPE_BLEND_MIN: + fragR = spu_min(term1r, term2r); + fragG = spu_min(term1g, term2g); + fragB = spu_min(term1b, term2b); + break; + case PIPE_BLEND_MAX: + fragR = spu_max(term1r, term2r); + fragG = spu_max(term1g, term2g); + fragB = spu_max(term1b, term2b); + break; default: ASSERT(0); } @@ -356,7 +467,15 @@ spu_fallback_fragment_ops(uint x, uint y, case PIPE_BLEND_SUBTRACT: fragA = spu_sub(term1a, term2a); break; - /* XXX more cases */ + case PIPE_BLEND_REVERSE_SUBTRACT: + fragA = spu_sub(term2a, term1a); + break; + case PIPE_BLEND_MIN: + fragA = spu_min(term1a, term2a); + break; + case PIPE_BLEND_MAX: + fragA = spu_max(term1a, term2a); + break; default: ASSERT(0); } @@ -384,21 +503,20 @@ spu_fallback_fragment_ops(uint x, uint y, #endif /* - * Pack float colors into 32-bit RGBA words. + * Pack fragment float colors into 32-bit RGBA words. */ switch (spu.fb.color_format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - c0 = spu_pack_A8R8G8B8(frag_aos[0]); - c1 = spu_pack_A8R8G8B8(frag_aos[1]); - c2 = spu_pack_A8R8G8B8(frag_aos[2]); - c3 = spu_pack_A8R8G8B8(frag_aos[3]); + fragc0 = spu_pack_A8R8G8B8(frag_aos[0]); + fragc1 = spu_pack_A8R8G8B8(frag_aos[1]); + fragc2 = spu_pack_A8R8G8B8(frag_aos[2]); + fragc3 = spu_pack_A8R8G8B8(frag_aos[3]); break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - c0 = spu_pack_B8G8R8A8(frag_aos[0]); - c1 = spu_pack_B8G8R8A8(frag_aos[1]); - c2 = spu_pack_B8G8R8A8(frag_aos[2]); - c3 = spu_pack_B8G8R8A8(frag_aos[3]); + fragc0 = spu_pack_B8G8R8A8(frag_aos[0]); + fragc1 = spu_pack_B8G8R8A8(frag_aos[1]); + fragc2 = spu_pack_B8G8R8A8(frag_aos[2]); + fragc3 = spu_pack_B8G8R8A8(frag_aos[3]); break; default: fprintf(stderr, "SPU: Bad pixel format in spu_default_fragment_ops\n"); @@ -407,20 +525,57 @@ spu_fallback_fragment_ops(uint x, uint y, /* - * Color masking + * Do color masking */ if (spu.blend.colormask != 0xf) { - /* XXX to do */ - /* apply color mask to 32-bit packed colors */ + uint cmask = 0x0; /* each byte corresponds to a color channel */ + + /* Form bitmask depending on color buffer format and colormask bits */ + switch (spu.fb.color_format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + if (spu.blend.colormask & PIPE_MASK_R) + cmask |= 0x00ff0000; /* red */ + if (spu.blend.colormask & PIPE_MASK_G) + cmask |= 0x0000ff00; /* green */ + if (spu.blend.colormask & PIPE_MASK_B) + cmask |= 0x000000ff; /* blue */ + if (spu.blend.colormask & PIPE_MASK_A) + cmask |= 0xff000000; /* alpha */ + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + if (spu.blend.colormask & PIPE_MASK_R) + cmask |= 0x0000ff00; /* red */ + if (spu.blend.colormask & PIPE_MASK_G) + cmask |= 0x00ff0000; /* green */ + if (spu.blend.colormask & PIPE_MASK_B) + cmask |= 0xff000000; /* blue */ + if (spu.blend.colormask & PIPE_MASK_A) + cmask |= 0x000000ff; /* alpha */ + break; + default: + ASSERT(0); + } + + /* + * Apply color mask to the 32-bit packed colors. + * if (cmask[i]) + * frag color[i] = frag color[i]; + * else + * frag color[i] = framebuffer color[i]; + */ + fragc0 = (fragc0 & cmask) | (fbc0 & ~cmask); + fragc1 = (fragc1 & cmask) | (fbc1 & ~cmask); + fragc2 = (fragc2 & cmask) | (fbc2 & ~cmask); + fragc3 = (fragc3 & cmask) | (fbc3 & ~cmask); } /* - * Logic Ops + * Do logic ops */ if (spu.blend.logicop_enable) { /* XXX to do */ - /* apply logicop to 32-bit packed colors */ + /* apply logicop to 32-bit packed colors (fragcx and fbcx) */ } @@ -431,45 +586,46 @@ spu_fallback_fragment_ops(uint x, uint y, spu.cur_ctile_status = TILE_STATUS_DIRTY; } else { + /* write no fragments */ return; } /* - * Write new quad colors to the framebuffer/tile. + * Write new fragment/quad colors to the framebuffer/tile. * Only write pixels where the corresponding mask word is set. */ #if LINEAR_QUAD_LAYOUT /* * Quad layout: * +--+--+--+--+ - * |p0|p1|p2|p3| + * |p0|p1|p2|p3|... * +--+--+--+--+ */ if (spu_extract(mask, 0)) - colorTile->ui[y][x*2] = c0; + colorTile->ui[y][x*2] = fragc0; if (spu_extract(mask, 1)) - colorTile->ui[y][x*2+1] = c1; + colorTile->ui[y][x*2+1] = fragc1; if (spu_extract(mask, 2)) - colorTile->ui[y][x*2+2] = c2; + colorTile->ui[y][x*2+2] = fragc2; if (spu_extract(mask, 3)) - colorTile->ui[y][x*2+3] = c3; + colorTile->ui[y][x*2+3] = fragc3; #else /* * Quad layout: * +--+--+ - * |p0|p1| + * |p0|p1|... * +--+--+ - * |p2|p3| + * |p2|p3|... * +--+--+ */ if (spu_extract(mask, 0)) - colorTile->ui[y+0][x+0] = c0; + colorTile->ui[y+0][x+0] = fragc0; if (spu_extract(mask, 1)) - colorTile->ui[y+0][x+1] = c1; + colorTile->ui[y+0][x+1] = fragc1; if (spu_extract(mask, 2)) - colorTile->ui[y+1][x+0] = c2; + colorTile->ui[y+1][x+0] = fragc2; if (spu_extract(mask, 3)) - colorTile->ui[y+1][x+1] = c3; + colorTile->ui[y+1][x+1] = fragc3; #endif } diff --git a/src/gallium/drivers/cell/spu/spu_render.c b/src/gallium/drivers/cell/spu/spu_render.c index 305dc98881..7c225e2f27 100644 --- a/src/gallium/drivers/cell/spu/spu_render.c +++ b/src/gallium/drivers/cell/spu/spu_render.c @@ -98,7 +98,7 @@ my_tile(uint tx, uint ty) static INLINE void get_cz_tiles(uint tx, uint ty) { - if (spu.read_depth) { + if (spu.read_depth_stencil) { if (spu.cur_ztile_status != TILE_STATUS_CLEAR) { //printf("SPU %u: getting Z tile %u, %u\n", spu.init.id, tx, ty); get_tile(tx, ty, &spu.ztile, TAG_READ_TILE_Z, 1); @@ -153,7 +153,7 @@ static INLINE void wait_put_cz_tiles(void) { wait_on_mask(1 << TAG_WRITE_TILE_COLOR); - if (spu.read_depth) { + if (spu.read_depth_stencil) { wait_on_mask(1 << TAG_WRITE_TILE_Z); } } @@ -175,22 +175,14 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) const ubyte *vertices; const ushort *indexes; uint i, j; + uint num_tiles; - - if (Debug) { - printf("SPU %u: RENDER prim %u, num_vert=%u num_ind=%u " - "inline_vert=%u\n", - spu.init.id, - render->prim_type, - render->num_verts, - render->num_indexes, - render->inline_verts); - - /* - printf(" bound: %g, %g .. %g, %g\n", - render->xmin, render->ymin, render->xmax, render->ymax); - */ - } + D_PRINTF(CELL_DEBUG_CMD, + "RENDER prim=%u num_vert=%u num_ind=%u inline_vert=%u\n", + render->prim_type, + render->num_verts, + render->num_indexes, + render->inline_verts); ASSERT(sizeof(*render) % 4 == 0); ASSERT(total_vertex_bytes % 16 == 0); @@ -251,6 +243,8 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) wait_on_mask(1 << TAG_SURFACE_CLEAR); /* XXX temporary */ + num_tiles = 0; + /** ** loop over tiles, rendering tris **/ @@ -264,6 +258,8 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) if (!my_tile(tx, ty)) continue; + num_tiles++; + spu.cur_ctile_status = spu.ctile_status[ty][tx]; spu.cur_ztile_status = spu.ztile_status[ty][tx]; @@ -293,9 +289,7 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) spu.ztile_status[ty][tx] = spu.cur_ztile_status; } - if (Debug) - printf("SPU %u: RENDER done\n", - spu.init.id); + D_PRINTF(CELL_DEBUG_CMD, + "RENDER done (%u tiles hit)\n", + num_tiles); } - - diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c index 117b8a36f8..69784c8978 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.c +++ b/src/gallium/drivers/cell/spu/spu_texture.c @@ -26,6 +26,8 @@ **************************************************************************/ +#include <math.h> + #include "pipe/p_compiler.h" #include "spu_main.h" #include "spu_texture.h" @@ -40,37 +42,19 @@ void invalidate_tex_cache(void) { - uint unit = 0; - uint bytes = 4 * spu.texture[unit].width - * spu.texture[unit].height; - - spu_dcache_mark_dirty((unsigned) spu.texture[unit].start, bytes); -} + uint lvl; + for (lvl = 0; lvl < CELL_MAX_TEXTURE_LEVELS; lvl++) { + uint unit = 0; + uint bytes = 4 * spu.texture[unit].level[lvl].width + * spu.texture[unit].level[lvl].height; + if (spu.texture[unit].target == PIPE_TEXTURE_CUBE) + bytes *= 6; + else if (spu.texture[unit].target == PIPE_TEXTURE_3D) + bytes *= spu.texture[unit].level[lvl].depth; -/** - * XXX look into getting texels for all four pixels in a quad at once. - */ -static uint -get_texel(uint unit, vec_uint4 coordinate) -{ - /* - * XXX we could do the "/ TILE_SIZE" and "% TILE_SIZE" operations as - * SIMD since X and Y are already in a SIMD register. - */ - const unsigned texture_ea = (uintptr_t) spu.texture[unit].start; - ushort x = spu_extract(coordinate, 0); - ushort y = spu_extract(coordinate, 1); - unsigned tile_offset = sizeof(tile_t) - * ((y / TILE_SIZE * spu.texture[unit].tiles_per_row) + (x / TILE_SIZE)); - ushort texel_offset = (ushort) 4 - * (ushort) (((ushort) (y % TILE_SIZE) * (ushort) TILE_SIZE) + (x % TILE_SIZE)); - vec_uint4 tmp; - - spu_dcache_fetch_unaligned((qword *) & tmp, - texture_ea + tile_offset + texel_offset, - 4); - return spu_extract(tmp, 0); + spu_dcache_mark_dirty((unsigned) spu.texture[unit].level[lvl].start, bytes); + } } @@ -88,15 +72,17 @@ get_texel(uint unit, vec_uint4 coordinate) * a time. */ static void -get_four_texels(uint unit, vec_uint4 x, vec_uint4 y, vec_uint4 *texels) +get_four_texels(const struct spu_texture_level *tlevel, uint face, + vec_int4 x, vec_int4 y, + vec_uint4 *texels) { - const unsigned texture_ea = (uintptr_t) spu.texture[unit].start; - vec_uint4 tile_x = spu_rlmask(x, -5); - vec_uint4 tile_y = spu_rlmask(y, -5); - const qword offset_x = si_andi((qword) x, 0x1f); - const qword offset_y = si_andi((qword) y, 0x1f); + unsigned texture_ea = (uintptr_t) tlevel->start; + const vec_int4 tile_x = spu_rlmask(x, -5); /* tile_x = x / 32 */ + const vec_int4 tile_y = spu_rlmask(y, -5); /* tile_y = y / 32 */ + const qword offset_x = si_andi((qword) x, 0x1f); /* offset_x = x & 0x1f */ + const qword offset_y = si_andi((qword) y, 0x1f); /* offset_y = y & 0x1f */ - const qword tiles_per_row = (qword) spu_splats(spu.texture[unit].tiles_per_row); + const qword tiles_per_row = (qword) spu_splats(tlevel->tiles_per_row); const qword tile_size = (qword) spu_splats((unsigned) sizeof(tile_t)); qword tile_offset = si_mpya((qword) tile_y, tiles_per_row, (qword) tile_x); @@ -107,6 +93,8 @@ get_four_texels(uint unit, vec_uint4 x, vec_uint4 y, vec_uint4 *texels) vec_uint4 offset = (vec_uint4) si_a(tile_offset, texel_offset); + texture_ea = texture_ea + face * tlevel->bytes_per_image; + spu_dcache_fetch_unaligned((qword *) & texels[0], texture_ea + spu_extract(offset, 0), 4); spu_dcache_fetch_unaligned((qword *) & texels[1], @@ -118,83 +106,536 @@ get_four_texels(uint unit, vec_uint4 x, vec_uint4 y, vec_uint4 *texels) } +/** clamp vec to [0, max] */ +static INLINE vector signed int +spu_clamp(vector signed int vec, vector signed int max) +{ + static const vector signed int zero = {0,0,0,0}; + vector unsigned int c; + c = spu_cmpgt(vec, zero); /* c = vec > zero ? ~0 : 0 */ + vec = spu_sel(zero, vec, c); + c = spu_cmpgt(vec, max); /* c = vec > max ? ~0 : 0 */ + vec = spu_sel(vec, max, c); + return vec; +} + + + /** - * Get texture sample at texcoord. + * Do nearest texture sampling for four pixels. + * \param colors returned colors in SOA format (rrrr, gggg, bbbb, aaaa). */ -vector float -sample_texture_nearest(uint unit, vector float texcoord) +void +sample_texture_2d_nearest(vector float s, vector float t, + uint unit, uint level, uint face, + vector float colors[4]) { - vector float tc = spu_mul(texcoord, spu.texture[unit].tex_size); - vector unsigned int itc = spu_convtu(tc, 0); /* convert to int */ - itc = spu_and(itc, spu.texture[unit].tex_size_mask); /* mask (GL_REPEAT) */ - uint texel = get_texel(unit, itc); - return spu_unpack_A8R8G8B8(texel); + const struct spu_texture_level *tlevel = &spu.texture[unit].level[level]; + vector float ss = spu_mul(s, tlevel->scale_s); + vector float tt = spu_mul(t, tlevel->scale_t); + vector signed int is = spu_convts(ss, 0); + vector signed int it = spu_convts(tt, 0); + vec_uint4 texels[4]; + + /* PIPE_TEX_WRAP_REPEAT */ + is = spu_and(is, tlevel->mask_s); + it = spu_and(it, tlevel->mask_t); + + /* PIPE_TEX_WRAP_CLAMP */ + is = spu_clamp(is, tlevel->max_s); + it = spu_clamp(it, tlevel->max_t); + + get_four_texels(tlevel, face, is, it, texels); + + /* convert four packed ARGBA pixels to float RRRR,GGGG,BBBB,AAAA */ + spu_unpack_A8R8G8B8_transpose4(texels, colors); } -vector float -sample_texture_bilinear(uint unit, vector float texcoord) +/** + * Do bilinear texture sampling for four pixels. + * \param colors returned colors in SOA format (rrrr, gggg, bbbb, aaaa). + */ +void +sample_texture_2d_bilinear(vector float s, vector float t, + uint unit, uint level, uint face, + vector float colors[4]) { - static const vec_uint4 offset_x = {0, 0, 1, 1}; - static const vec_uint4 offset_y = {0, 1, 0, 1}; + const struct spu_texture_level *tlevel = &spu.texture[unit].level[level]; + static const vector float half = {-0.5f, -0.5f, -0.5f, -0.5f}; - vector float tc = spu_mul(texcoord, spu.texture[unit].tex_size); - tc = spu_add(tc, spu_splats(-0.5f)); /* half texel bias */ + vector float ss = spu_madd(s, tlevel->scale_s, half); + vector float tt = spu_madd(t, tlevel->scale_t, half); - /* integer texcoords S,T: */ - vec_uint4 itc = spu_convtu(tc, 0); /* convert to int */ + vector signed int is0 = spu_convts(ss, 0); + vector signed int it0 = spu_convts(tt, 0); - vec_uint4 texels[4]; - - /* setup texcoords for quad: - * +-----+-----+ - * |x0,y0|x1,y1| - * +-----+-----+ - * |x2,y2|x3,y3| - * +-----+-----+ - */ - vec_uint4 x = spu_splats(spu_extract(itc, 0)); - vec_uint4 y = spu_splats(spu_extract(itc, 1)); - x = spu_add(x, offset_x); - y = spu_add(y, offset_y); + /* is + 1, it + 1 */ + vector signed int is1 = spu_add(is0, 1); + vector signed int it1 = spu_add(it0, 1); - /* GL_REPEAT wrap mode: */ - x = spu_and(x, spu.texture[unit].tex_size_x_mask); - y = spu_and(y, spu.texture[unit].tex_size_y_mask); + /* PIPE_TEX_WRAP_REPEAT */ + is0 = spu_and(is0, tlevel->mask_s); + it0 = spu_and(it0, tlevel->mask_t); + is1 = spu_and(is1, tlevel->mask_s); + it1 = spu_and(it1, tlevel->mask_t); - get_four_texels(unit, x, y, texels); + /* PIPE_TEX_WRAP_CLAMP */ + is0 = spu_clamp(is0, tlevel->max_s); + it0 = spu_clamp(it0, tlevel->max_t); + is1 = spu_clamp(is1, tlevel->max_s); + it1 = spu_clamp(it1, tlevel->max_t); - /* integer A8R8G8B8 to float texel conversion */ - vector float texel00 = spu_unpack_A8R8G8B8(spu_extract(texels[0], 0)); - vector float texel01 = spu_unpack_A8R8G8B8(spu_extract(texels[1], 0)); - vector float texel10 = spu_unpack_A8R8G8B8(spu_extract(texels[2], 0)); - vector float texel11 = spu_unpack_A8R8G8B8(spu_extract(texels[3], 0)); + /* get packed int texels */ + vector unsigned int texels[16]; + get_four_texels(tlevel, face, is0, it0, texels + 0); /* upper-left */ + get_four_texels(tlevel, face, is1, it0, texels + 4); /* upper-right */ + get_four_texels(tlevel, face, is0, it1, texels + 8); /* lower-left */ + get_four_texels(tlevel, face, is1, it1, texels + 12); /* lower-right */ + /* convert packed int texels to float colors */ + vector float ftexels[16]; + spu_unpack_A8R8G8B8_transpose4(texels + 0, ftexels + 0); + spu_unpack_A8R8G8B8_transpose4(texels + 4, ftexels + 4); + spu_unpack_A8R8G8B8_transpose4(texels + 8, ftexels + 8); + spu_unpack_A8R8G8B8_transpose4(texels + 12, ftexels + 12); /* Compute weighting factors in [0,1] * Multiply texcoord by 1024, AND with 1023, convert back to float. */ - vector float tc1024 = spu_mul(tc, spu_splats(1024.0f)); - vector signed int itc1024 = spu_convts(tc1024, 0); - itc1024 = spu_and(itc1024, spu_splats((1 << 10) - 1)); - vector float weight = spu_convtf(itc1024, 10); - - /* smeared frac and 1-frac */ - vector float sfrac = spu_splats(spu_extract(weight, 0)); - vector float tfrac = spu_splats(spu_extract(weight, 1)); - vector float sfrac1 = spu_sub(spu_splats(1.0f), sfrac); - vector float tfrac1 = spu_sub(spu_splats(1.0f), tfrac); - - /* multiply the samples (colors) by the S/T weights */ - texel00 = spu_mul(spu_mul(texel00, sfrac1), tfrac1); - texel10 = spu_mul(spu_mul(texel10, sfrac ), tfrac1); - texel01 = spu_mul(spu_mul(texel01, sfrac1), tfrac ); - texel11 = spu_mul(spu_mul(texel11, sfrac ), tfrac ); - - /* compute sum of weighted samples */ - vector float texel_sum = spu_add(texel00, texel01); - texel_sum = spu_add(texel_sum, texel10); - texel_sum = spu_add(texel_sum, texel11); - - return texel_sum; + vector float ss1024 = spu_mul(ss, spu_splats(1024.0f)); + vector signed int iss1024 = spu_convts(ss1024, 0); + iss1024 = spu_and(iss1024, 1023); + vector float sWeights0 = spu_convtf(iss1024, 10); + + vector float tt1024 = spu_mul(tt, spu_splats(1024.0f)); + vector signed int itt1024 = spu_convts(tt1024, 0); + itt1024 = spu_and(itt1024, 1023); + vector float tWeights0 = spu_convtf(itt1024, 10); + + /* 1 - sWeight and 1 - tWeight */ + vector float sWeights1 = spu_sub(spu_splats(1.0f), sWeights0); + vector float tWeights1 = spu_sub(spu_splats(1.0f), tWeights0); + + /* reds, for four pixels */ + ftexels[ 0] = spu_mul(ftexels[ 0], spu_mul(sWeights1, tWeights1)); /*ul*/ + ftexels[ 4] = spu_mul(ftexels[ 4], spu_mul(sWeights0, tWeights1)); /*ur*/ + ftexels[ 8] = spu_mul(ftexels[ 8], spu_mul(sWeights1, tWeights0)); /*ll*/ + ftexels[12] = spu_mul(ftexels[12], spu_mul(sWeights0, tWeights0)); /*lr*/ + colors[0] = spu_add(spu_add(ftexels[0], ftexels[4]), + spu_add(ftexels[8], ftexels[12])); + + /* greens, for four pixels */ + ftexels[ 1] = spu_mul(ftexels[ 1], spu_mul(sWeights1, tWeights1)); /*ul*/ + ftexels[ 5] = spu_mul(ftexels[ 5], spu_mul(sWeights0, tWeights1)); /*ur*/ + ftexels[ 9] = spu_mul(ftexels[ 9], spu_mul(sWeights1, tWeights0)); /*ll*/ + ftexels[13] = spu_mul(ftexels[13], spu_mul(sWeights0, tWeights0)); /*lr*/ + colors[1] = spu_add(spu_add(ftexels[1], ftexels[5]), + spu_add(ftexels[9], ftexels[13])); + + /* blues, for four pixels */ + ftexels[ 2] = spu_mul(ftexels[ 2], spu_mul(sWeights1, tWeights1)); /*ul*/ + ftexels[ 6] = spu_mul(ftexels[ 6], spu_mul(sWeights0, tWeights1)); /*ur*/ + ftexels[10] = spu_mul(ftexels[10], spu_mul(sWeights1, tWeights0)); /*ll*/ + ftexels[14] = spu_mul(ftexels[14], spu_mul(sWeights0, tWeights0)); /*lr*/ + colors[2] = spu_add(spu_add(ftexels[2], ftexels[6]), + spu_add(ftexels[10], ftexels[14])); + + /* alphas, for four pixels */ + ftexels[ 3] = spu_mul(ftexels[ 3], spu_mul(sWeights1, tWeights1)); /*ul*/ + ftexels[ 7] = spu_mul(ftexels[ 7], spu_mul(sWeights0, tWeights1)); /*ur*/ + ftexels[11] = spu_mul(ftexels[11], spu_mul(sWeights1, tWeights0)); /*ll*/ + ftexels[15] = spu_mul(ftexels[15], spu_mul(sWeights0, tWeights0)); /*lr*/ + colors[3] = spu_add(spu_add(ftexels[3], ftexels[7]), + spu_add(ftexels[11], ftexels[15])); +} + + + +/** + * Adapted from /opt/cell/sdk/usr/spu/include/transpose_matrix4x4.h + */ +static INLINE void +transpose(vector unsigned int *mOut0, + vector unsigned int *mOut1, + vector unsigned int *mOut2, + vector unsigned int *mOut3, + vector unsigned int *mIn) +{ + vector unsigned int abcd, efgh, ijkl, mnop; /* input vectors */ + vector unsigned int aeim, bfjn, cgko, dhlp; /* output vectors */ + vector unsigned int aibj, ckdl, emfn, gohp; /* intermediate vectors */ + + vector unsigned char shufflehi = ((vector unsigned char) { + 0x00, 0x01, 0x02, 0x03, + 0x10, 0x11, 0x12, 0x13, + 0x04, 0x05, 0x06, 0x07, + 0x14, 0x15, 0x16, 0x17}); + vector unsigned char shufflelo = ((vector unsigned char) { + 0x08, 0x09, 0x0A, 0x0B, + 0x18, 0x19, 0x1A, 0x1B, + 0x0C, 0x0D, 0x0E, 0x0F, + 0x1C, 0x1D, 0x1E, 0x1F}); + abcd = *(mIn+0); + efgh = *(mIn+1); + ijkl = *(mIn+2); + mnop = *(mIn+3); + + aibj = spu_shuffle(abcd, ijkl, shufflehi); + ckdl = spu_shuffle(abcd, ijkl, shufflelo); + emfn = spu_shuffle(efgh, mnop, shufflehi); + gohp = spu_shuffle(efgh, mnop, shufflelo); + + aeim = spu_shuffle(aibj, emfn, shufflehi); + bfjn = spu_shuffle(aibj, emfn, shufflelo); + cgko = spu_shuffle(ckdl, gohp, shufflehi); + dhlp = spu_shuffle(ckdl, gohp, shufflelo); + + *mOut0 = aeim; + *mOut1 = bfjn; + *mOut2 = cgko; + *mOut3 = dhlp; +} + + +/** + * Bilinear filtering, using int instead of float arithmetic for computing + * sample weights. + */ +void +sample_texture_2d_bilinear_int(vector float s, vector float t, + uint unit, uint level, uint face, + vector float colors[4]) +{ + const struct spu_texture_level *tlevel = &spu.texture[unit].level[level]; + static const vector float half = {-0.5f, -0.5f, -0.5f, -0.5f}; + + /* Scale texcoords by size of texture, and add half pixel bias */ + vector float ss = spu_madd(s, tlevel->scale_s, half); + vector float tt = spu_madd(t, tlevel->scale_t, half); + + /* convert float coords to fixed-pt coords with 7 fraction bits */ + vector signed int is = spu_convts(ss, 7); /* XXX really need floor() here */ + vector signed int it = spu_convts(tt, 7); /* XXX really need floor() here */ + + /* compute integer texel weights in [0, 127] */ + vector signed int sWeights0 = spu_and(is, 127); + vector signed int tWeights0 = spu_and(it, 127); + vector signed int sWeights1 = spu_sub(127, sWeights0); + vector signed int tWeights1 = spu_sub(127, tWeights0); + + /* texel coords: is0 = is / 128, it0 = is / 128 */ + vector signed int is0 = spu_rlmask(is, -7); + vector signed int it0 = spu_rlmask(it, -7); + + /* texel coords: i1 = is0 + 1, it1 = it0 + 1 */ + vector signed int is1 = spu_add(is0, 1); + vector signed int it1 = spu_add(it0, 1); + + /* PIPE_TEX_WRAP_REPEAT */ + is0 = spu_and(is0, tlevel->mask_s); + it0 = spu_and(it0, tlevel->mask_t); + is1 = spu_and(is1, tlevel->mask_s); + it1 = spu_and(it1, tlevel->mask_t); + + /* PIPE_TEX_WRAP_CLAMP */ + is0 = spu_clamp(is0, tlevel->max_s); + it0 = spu_clamp(it0, tlevel->max_t); + is1 = spu_clamp(is1, tlevel->max_s); + it1 = spu_clamp(it1, tlevel->max_t); + + /* get packed int texels */ + vector unsigned int texels[16]; + get_four_texels(tlevel, face, is0, it0, texels + 0); /* upper-left */ + get_four_texels(tlevel, face, is1, it0, texels + 4); /* upper-right */ + get_four_texels(tlevel, face, is0, it1, texels + 8); /* lower-left */ + get_four_texels(tlevel, face, is1, it1, texels + 12); /* lower-right */ + + /* twiddle packed 32-bit BGRA pixels into RGBA as four unsigned ints */ + { + static const unsigned char ZERO = 0x80; + int i; + for (i = 0; i < 16; i++) { + texels[i] = spu_shuffle(texels[i], texels[i], + ((vector unsigned char) { + ZERO, ZERO, ZERO, 1, + ZERO, ZERO, ZERO, 2, + ZERO, ZERO, ZERO, 3, + ZERO, ZERO, ZERO, 0})); + } + } + + /* convert RGBA,RGBA,RGBA,RGBA to RRRR,GGGG,BBBB,AAAA */ + vector unsigned int texel0, texel1, texel2, texel3, texel4, texel5, texel6, texel7, + texel8, texel9, texel10, texel11, texel12, texel13, texel14, texel15; + transpose(&texel0, &texel1, &texel2, &texel3, texels + 0); + transpose(&texel4, &texel5, &texel6, &texel7, texels + 4); + transpose(&texel8, &texel9, &texel10, &texel11, texels + 8); + transpose(&texel12, &texel13, &texel14, &texel15, texels + 12); + + /* computed weighted colors */ + vector unsigned int c0, c1, c2, c3, cSum; + + /* red */ + c0 = (vector unsigned int) si_mpy((qword) texel0, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/ + c1 = (vector unsigned int) si_mpy((qword) texel4, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/ + c2 = (vector unsigned int) si_mpy((qword) texel8, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/ + c3 = (vector unsigned int) si_mpy((qword) texel12, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/ + cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3)); + colors[0] = spu_convtf(cSum, 22); + + /* green */ + c0 = (vector unsigned int) si_mpy((qword) texel1, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/ + c1 = (vector unsigned int) si_mpy((qword) texel5, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/ + c2 = (vector unsigned int) si_mpy((qword) texel9, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/ + c3 = (vector unsigned int) si_mpy((qword) texel13, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/ + cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3)); + colors[1] = spu_convtf(cSum, 22); + + /* blue */ + c0 = (vector unsigned int) si_mpy((qword) texel2, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/ + c1 = (vector unsigned int) si_mpy((qword) texel6, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/ + c2 = (vector unsigned int) si_mpy((qword) texel10, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/ + c3 = (vector unsigned int) si_mpy((qword) texel14, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/ + cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3)); + colors[2] = spu_convtf(cSum, 22); + + /* alpha */ + c0 = (vector unsigned int) si_mpy((qword) texel3, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/ + c1 = (vector unsigned int) si_mpy((qword) texel7, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/ + c2 = (vector unsigned int) si_mpy((qword) texel11, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/ + c3 = (vector unsigned int) si_mpy((qword) texel15, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/ + cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3)); + colors[3] = spu_convtf(cSum, 22); +} + + + +/** + * Compute level of detail factor from texcoords. + */ +static INLINE float +compute_lambda_2d(uint unit, vector float s, vector float t) +{ + uint baseLevel = 0; + float width = spu.texture[unit].level[baseLevel].width; + float height = spu.texture[unit].level[baseLevel].width; + float dsdx = width * (spu_extract(s, 1) - spu_extract(s, 0)); + float dsdy = width * (spu_extract(s, 2) - spu_extract(s, 0)); + float dtdx = height * (spu_extract(t, 1) - spu_extract(t, 0)); + float dtdy = height * (spu_extract(t, 2) - spu_extract(t, 0)); +#if 0 + /* ideal value */ + float x = dsdx * dsdx + dtdx * dtdx; + float y = dsdy * dsdy + dtdy * dtdy; + float rho = x > y ? x : y; + rho = sqrtf(rho); +#else + /* approximation */ + dsdx = fabsf(dsdx); + dsdy = fabsf(dsdy); + dtdx = fabsf(dtdx); + dtdy = fabsf(dtdy); + float rho = (dsdx + dsdy + dtdx + dtdy) * 0.5; +#endif + float lambda = logf(rho) * 1.442695f; /* compute logbase2(rho) */ + return lambda; +} + + +/** + * Blend two sets of colors according to weight. + */ +static void +blend_colors(vector float c0[4], const vector float c1[4], float weight) +{ + vector float t = spu_splats(weight); + vector float dc0 = spu_sub(c1[0], c0[0]); + vector float dc1 = spu_sub(c1[1], c0[1]); + vector float dc2 = spu_sub(c1[2], c0[2]); + vector float dc3 = spu_sub(c1[3], c0[3]); + c0[0] = spu_madd(dc0, t, c0[0]); + c0[1] = spu_madd(dc1, t, c0[1]); + c0[2] = spu_madd(dc2, t, c0[2]); + c0[3] = spu_madd(dc3, t, c0[3]); +} + + +/** + * Texture sampling with level of detail selection and possibly mipmap + * interpolation. + */ +void +sample_texture_2d_lod(vector float s, vector float t, + uint unit, uint level_ignored, uint face, + vector float colors[4]) +{ + /* + * Note that we're computing a lambda/lod here that's used for all + * four pixels in the quad. + */ + float lambda = compute_lambda_2d(unit, s, t); + + (void) face; + (void) level_ignored; + + /* apply lod bias */ + lambda += spu.sampler[unit].lod_bias; + + /* clamp */ + if (lambda < spu.sampler[unit].min_lod) + lambda = spu.sampler[unit].min_lod; + else if (lambda > spu.sampler[unit].max_lod) + lambda = spu.sampler[unit].max_lod; + + if (lambda <= 0.0f) { + /* magnify */ + spu.mag_sample_texture_2d[unit](s, t, unit, 0, face, colors); + } + else { + /* minify */ + if (spu.sampler[unit].min_img_filter == PIPE_TEX_FILTER_LINEAR) { + /* sample two mipmap levels and interpolate */ + int level = (int) lambda; + if (level > (int) spu.texture[unit].max_level) + level = spu.texture[unit].max_level; + spu.min_sample_texture_2d[unit](s, t, unit, level, face, colors); + if (spu.sampler[unit].min_img_filter == PIPE_TEX_FILTER_LINEAR) { + /* sample second mipmap level */ + float weight = lambda - (float) level; + level++; + if (level <= (int) spu.texture[unit].max_level) { + vector float colors2[4]; + spu.min_sample_texture_2d[unit](s, t, unit, level, face, colors2); + blend_colors(colors, colors2, weight); + } + } + } + else { + /* sample one mipmap level */ + int level = (int) (lambda + 0.5f); + if (level > (int) spu.texture[unit].max_level) + level = spu.texture[unit].max_level; + spu.min_sample_texture_2d[unit](s, t, unit, level, face, colors); + } + } +} + + +/** XXX need a SIMD version of this */ +static unsigned +choose_cube_face(float rx, float ry, float rz, float *newS, float *newT) +{ + /* + major axis + direction target sc tc ma + ---------- ------------------------------- --- --- --- + +rx TEXTURE_CUBE_MAP_POSITIVE_X_EXT -rz -ry rx + -rx TEXTURE_CUBE_MAP_NEGATIVE_X_EXT +rz -ry rx + +ry TEXTURE_CUBE_MAP_POSITIVE_Y_EXT +rx +rz ry + -ry TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT +rx -rz ry + +rz TEXTURE_CUBE_MAP_POSITIVE_Z_EXT +rx -ry rz + -rz TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT -rx -ry rz + */ + const float arx = fabsf(rx); + const float ary = fabsf(ry); + const float arz = fabsf(rz); + unsigned face; + float sc, tc, ma; + + if (arx > ary && arx > arz) { + if (rx >= 0.0F) { + face = PIPE_TEX_FACE_POS_X; + sc = -rz; + tc = -ry; + ma = arx; + } + else { + face = PIPE_TEX_FACE_NEG_X; + sc = rz; + tc = -ry; + ma = arx; + } + } + else if (ary > arx && ary > arz) { + if (ry >= 0.0F) { + face = PIPE_TEX_FACE_POS_Y; + sc = rx; + tc = rz; + ma = ary; + } + else { + face = PIPE_TEX_FACE_NEG_Y; + sc = rx; + tc = -rz; + ma = ary; + } + } + else { + if (rz > 0.0F) { + face = PIPE_TEX_FACE_POS_Z; + sc = rx; + tc = -ry; + ma = arz; + } + else { + face = PIPE_TEX_FACE_NEG_Z; + sc = -rx; + tc = -ry; + ma = arz; + } + } + + *newS = (sc / ma + 1.0F) * 0.5F; + *newT = (tc / ma + 1.0F) * 0.5F; + + return face; +} + + + +void +sample_texture_cube(vector float s, vector float t, vector float r, + uint unit, vector float colors[4]) +{ + uint p, faces[4], level = 0; + float newS[4], newT[4]; + + /* Compute cube faces referenced by the four sets of texcoords. + * XXX we should SIMD-ize this. + */ + for (p = 0; p < 4; p++) { + float rx = spu_extract(s, p); + float ry = spu_extract(t, p); + float rz = spu_extract(r, p); + faces[p] = choose_cube_face(rx, ry, rz, &newS[p], &newT[p]); + } + + if (faces[0] == faces[1] && + faces[0] == faces[2] && + faces[0] == faces[3]) { + /* GOOD! All four texcoords refer to the same cube face */ + s = (vector float) {newS[0], newS[1], newS[2], newS[3]}; + t = (vector float) {newT[0], newT[1], newT[2], newT[3]}; + spu.sample_texture_2d[unit](s, t, unit, level, faces[0], colors); + } + else { + /* BAD! The four texcoords refer to different faces */ + for (p = 0; p < 4; p++) { + vector float c[4]; + + spu.sample_texture_2d[unit](spu_splats(newS[p]), spu_splats(newT[p]), + unit, level, faces[p], c); + + float red = spu_extract(c[0], p); + float green = spu_extract(c[1], p); + float blue = spu_extract(c[2], p); + float alpha = spu_extract(c[3], p); + + colors[0] = spu_insert(red, colors[0], p); + colors[1] = spu_insert(green, colors[1], p); + colors[2] = spu_insert(blue, colors[2], p); + colors[3] = spu_insert(alpha, colors[3], p); + } + } } diff --git a/src/gallium/drivers/cell/spu/spu_texture.h b/src/gallium/drivers/cell/spu/spu_texture.h index f7c9738be8..7b75b007b5 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.h +++ b/src/gallium/drivers/cell/spu/spu_texture.h @@ -36,12 +36,32 @@ extern void invalidate_tex_cache(void); -extern vector float -sample_texture_nearest(uint unit, vector float texcoord); +extern void +sample_texture_2d_nearest(vector float s, vector float t, + uint unit, uint level, uint face, + vector float colors[4]); + + +extern void +sample_texture_2d_bilinear(vector float s, vector float t, + uint unit, uint level, uint face, + vector float colors[4]); + +extern void +sample_texture_2d_bilinear_int(vector float s, vector float t, + uint unit, uint level, uint face, + vector float colors[4]); + +extern void +sample_texture_2d_lod(vector float s, vector float t, + uint unit, uint level, uint face, + vector float colors[4]); -extern vector float -sample_texture_bilinear(uint unit, vector float texcoord); + +extern void +sample_texture_cube(vector float s, vector float t, vector float r, + uint unit, vector float colors[4]); #endif /* SPU_TEXTURE_H */ diff --git a/src/gallium/drivers/cell/spu/spu_tile.c b/src/gallium/drivers/cell/spu/spu_tile.c index 216a33126b..6905015a48 100644 --- a/src/gallium/drivers/cell/spu/spu_tile.c +++ b/src/gallium/drivers/cell/spu/spu_tile.c @@ -87,3 +87,40 @@ put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf) 0 /* rid */); } + +/** + * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled + * tiles back to the main framebuffer. + */ +void +really_clear_tiles(uint surfaceIndex) +{ + const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; + uint i; + + if (surfaceIndex == 0) { + clear_c_tile(&spu.ctile); + + for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { + uint tx = i % spu.fb.width_tiles; + uint ty = i / spu.fb.width_tiles; + if (spu.ctile_status[ty][tx] == TILE_STATUS_CLEAR) { + put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0); + } + } + } + else { + clear_z_tile(&spu.ztile); + + for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { + uint tx = i % spu.fb.width_tiles; + uint ty = i / spu.fb.width_tiles; + if (spu.ztile_status[ty][tx] == TILE_STATUS_CLEAR) + put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 1); + } + } + +#if 0 + wait_on_mask(1 << TAG_SURFACE_CLEAR); +#endif +} diff --git a/src/gallium/drivers/cell/spu/spu_tile.h b/src/gallium/drivers/cell/spu/spu_tile.h index 1b5491112d..7bfb52be8f 100644 --- a/src/gallium/drivers/cell/spu/spu_tile.h +++ b/src/gallium/drivers/cell/spu/spu_tile.h @@ -36,12 +36,14 @@ -void +extern void get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf); -void +extern void put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf); +extern void +really_clear_tiles(uint surfaceIndex); static INLINE void diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index 8b93878192..22e51a86ae 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -43,11 +43,6 @@ /** Masks are uint[4] vectors with each element being 0 or 0xffffffff */ typedef vector unsigned int mask_t; -typedef union -{ - vector float v; - float f[4]; -} float4; /** @@ -91,9 +86,9 @@ struct edge { struct interp_coef { - float4 a0; - float4 dadx; - float4 dady; + vector float a0; + vector float dadx; + vector float dady; }; @@ -116,21 +111,15 @@ struct setup_stage { struct edge etop; struct edge emaj; - float oneoverarea; + float oneOverArea; /* XXX maybe make into vector? */ + + uint facing; - uint tx, ty; + uint tx, ty; /**< position of current tile (x, y) */ int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; -#if 0 - struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; -#else struct interp_coef coef[PIPE_MAX_SHADER_INPUTS]; -#endif - -#if 0 - struct quad_header quad; -#endif struct { int left[2]; /**< [0] = row0, [1] = row1 */ @@ -142,118 +131,105 @@ struct setup_stage { }; - static struct setup_stage setup; - - -#if 0 -/** - * Basically a cast wrapper. - */ -static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) -{ - return (struct setup_stage *)stage; -} -#endif - -#if 0 -/** - * Clip setup.quad against the scissor/surface bounds. - */ -static INLINE void -quad_clip(struct setup_stage *setup) -{ - const struct pipe_scissor_state *cliprect = &setup.softpipe->cliprect; - const int minx = (int) cliprect->minx; - const int maxx = (int) cliprect->maxx; - const int miny = (int) cliprect->miny; - const int maxy = (int) cliprect->maxy; - - if (setup.quad.x0 >= maxx || - setup.quad.y0 >= maxy || - setup.quad.x0 + 1 < minx || - setup.quad.y0 + 1 < miny) { - /* totally clipped */ - setup.quad.mask = 0x0; - return; - } - if (setup.quad.x0 < minx) - setup.quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - if (setup.quad.y0 < miny) - setup.quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - if (setup.quad.x0 == maxx - 1) - setup.quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - if (setup.quad.y0 == maxy - 1) - setup.quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); -} -#endif - -#if 0 -/** - * Emit a quad (pass to next stage) with clipping. - */ -static INLINE void -clip_emit_quad(struct setup_stage *setup) -{ - quad_clip(setup); - if (setup.quad.mask) { - struct softpipe_context *sp = setup.softpipe; - sp->quad.first->run(sp->quad.first, &setup.quad); - } -} -#endif - /** * Evaluate attribute coefficients (plane equations) to compute * attribute values for the four fragments in a quad. * Eg: four colors will be computed (in AoS format). */ static INLINE void -eval_coeff(uint slot, float x, float y, vector float result[4]) +eval_coeff(uint slot, float x, float y, vector float w, vector float result[4]) { - switch (spu.vertex_info.interp_mode[slot]) { + switch (spu.vertex_info.attrib[slot].interp_mode) { case INTERP_CONSTANT: result[QUAD_TOP_LEFT] = result[QUAD_TOP_RIGHT] = result[QUAD_BOTTOM_LEFT] = - result[QUAD_BOTTOM_RIGHT] = setup.coef[slot].a0.v; + result[QUAD_BOTTOM_RIGHT] = setup.coef[slot].a0; break; - case INTERP_LINEAR: - /* fall-through, for now */ - default: { - register vector float dadx = setup.coef[slot].dadx.v; - register vector float dady = setup.coef[slot].dady.v; - register vector float topLeft - = spu_add(setup.coef[slot].a0.v, - spu_add(spu_mul(spu_splats(x), dadx), - spu_mul(spu_splats(y), dady))); + vector float dadx = setup.coef[slot].dadx; + vector float dady = setup.coef[slot].dady; + vector float topLeft = + spu_add(setup.coef[slot].a0, + spu_add(spu_mul(spu_splats(x), dadx), + spu_mul(spu_splats(y), dady))); result[QUAD_TOP_LEFT] = topLeft; result[QUAD_TOP_RIGHT] = spu_add(topLeft, dadx); result[QUAD_BOTTOM_LEFT] = spu_add(topLeft, dady); result[QUAD_BOTTOM_RIGHT] = spu_add(spu_add(topLeft, dadx), dady); } + break; + case INTERP_PERSPECTIVE: + { + vector float dadx = setup.coef[slot].dadx; + vector float dady = setup.coef[slot].dady; + vector float topLeft = + spu_add(setup.coef[slot].a0, + spu_add(spu_mul(spu_splats(x), dadx), + spu_mul(spu_splats(y), dady))); + + vector float wInv = spu_re(w); /* 1.0 / w */ + + result[QUAD_TOP_LEFT] = spu_mul(topLeft, wInv); + result[QUAD_TOP_RIGHT] = spu_mul(spu_add(topLeft, dadx), wInv); + result[QUAD_BOTTOM_LEFT] = spu_mul(spu_add(topLeft, dady), wInv); + result[QUAD_BOTTOM_RIGHT] = spu_mul(spu_add(spu_add(topLeft, dadx), dady), wInv); + } + break; + case INTERP_POS: + case INTERP_NONE: + break; + default: + ASSERT(0); } } +/** + * As above, but return 4 vectors in SOA format. + * XXX this will all be re-written someday. + */ +static INLINE void +eval_coeff_soa(uint slot, float x, float y, vector float w, vector float result[4]) +{ + eval_coeff(slot, x, y, w, result); + _transpose_matrix4x4(result, result); +} + + +/** Evalute coefficients to get Z for four pixels in a quad */ static INLINE vector float eval_z(float x, float y) { const uint slot = 0; - const float dzdx = setup.coef[slot].dadx.f[2]; - const float dzdy = setup.coef[slot].dady.f[2]; - const float topLeft = setup.coef[slot].a0.f[2] + x * dzdx + y * dzdy; + const float dzdx = spu_extract(setup.coef[slot].dadx, 2); + const float dzdy = spu_extract(setup.coef[slot].dady, 2); + const float topLeft = spu_extract(setup.coef[slot].a0, 2) + x * dzdx + y * dzdy; const vector float topLeftv = spu_splats(topLeft); const vector float derivs = (vector float) { 0.0, dzdx, dzdy, dzdx + dzdy }; return spu_add(topLeftv, derivs); } +/** Evalute coefficients to get W for four pixels in a quad */ +static INLINE vector float +eval_w(float x, float y) +{ + const uint slot = 0; + const float dwdx = spu_extract(setup.coef[slot].dadx, 3); + const float dwdy = spu_extract(setup.coef[slot].dady, 3); + const float topLeft = spu_extract(setup.coef[slot].a0, 3) + x * dwdx + y * dwdy; + const vector float topLeftv = spu_splats(topLeft); + const vector float derivs = (vector float) { 0.0, dwdx, dwdy, dwdx + dwdy }; + return spu_add(topLeftv, derivs); +} + + /** * Emit a quad (pass to next stage). No clipping is done. * Note: about 1/5 to 1/7 of the time, mask is zero and this function @@ -261,120 +237,59 @@ eval_z(float x, float y) * overall. */ static INLINE void -emit_quad( int x, int y, mask_t mask ) +emit_quad( int x, int y, mask_t mask) { /* If any bits in mask are set... */ if (spu_extract(spu_orx(mask), 0)) { const int ix = x - setup.cliprect_minx; const int iy = y - setup.cliprect_miny; - vector float colors[4]; spu.cur_ctile_status = TILE_STATUS_DIRTY; spu.cur_ztile_status = TILE_STATUS_DIRTY; - if (spu.texture[0].start) { - /* texture mapping */ - const uint unit = 0; - vector float texcoords[4]; - eval_coeff(2, (float) x, (float) y, texcoords); - - if (spu_extract(mask, 0)) - colors[0] = spu.sample_texture[unit](unit, texcoords[0]); - if (spu_extract(mask, 1)) - colors[1] = spu.sample_texture[unit](unit, texcoords[1]); - if (spu_extract(mask, 2)) - colors[2] = spu.sample_texture[unit](unit, texcoords[2]); - if (spu_extract(mask, 3)) - colors[3] = spu.sample_texture[unit](unit, texcoords[3]); - - - if (spu.texture[1].start) { - /* multi-texture mapping */ - const uint unit = 1; - vector float colors1[4]; - - eval_coeff(2, (float) x, (float) y, texcoords); - - if (spu_extract(mask, 0)) - colors1[0] = spu.sample_texture[unit](unit, texcoords[0]); - if (spu_extract(mask, 1)) - colors1[1] = spu.sample_texture[unit](unit, texcoords[1]); - if (spu_extract(mask, 2)) - colors1[2] = spu.sample_texture[unit](unit, texcoords[2]); - if (spu_extract(mask, 3)) - colors1[3] = spu.sample_texture[unit](unit, texcoords[3]); - - /* hack: modulate first texture by second */ - colors[0] = spu_mul(colors[0], colors1[0]); - colors[1] = spu_mul(colors[1], colors1[1]); - colors[2] = spu_mul(colors[2], colors1[2]); - colors[3] = spu_mul(colors[3], colors1[3]); - } + { + /* + * Run fragment shader, execute per-fragment ops, update fb/tile. + */ + vector float inputs[4*4], outputs[2*4]; + vector float fragZ = eval_z((float) x, (float) y); + vector float fragW = eval_w((float) x, (float) y); + vector unsigned int kill_mask; - } - else { - /* simple shading */ + /* setup inputs */ #if 0 - eval_coeff(1, (float) x, (float) y, colors); - + eval_coeff_soa(1, (float) x, (float) y, fragW, inputs); #else - /* XXX new fragment program code */ - - if (spu.fragment_program) { - vector float inputs[4*4], outputs[2*4]; - - /* setup inputs */ - eval_coeff(1, (float) x, (float) y, inputs); - - /* Execute the current fragment program */ - spu.fragment_program(inputs, outputs, spu.constants); - - /* Copy outputs */ - colors[0] = outputs[0*4+0]; - colors[1] = outputs[0*4+1]; - colors[2] = outputs[0*4+2]; - colors[3] = outputs[0*4+3]; - - if (0 && spu.init.id==0 && y == 48) { - printf("colors[0] = %f %f %f %f\n", - spu_extract(colors[0], 0), - spu_extract(colors[0], 1), - spu_extract(colors[0], 2), - spu_extract(colors[0], 3)); - printf("colors[1] = %f %f %f %f\n", - spu_extract(colors[1], 0), - spu_extract(colors[1], 1), - spu_extract(colors[1], 2), - spu_extract(colors[1], 3)); - } - + uint i; + for (i = 0; i < spu.vertex_info.num_attribs; i++) { + eval_coeff_soa(i+1, (float) x, (float) y, fragW, inputs + i * 4); } #endif - } - - - { - /* Convert fragment data from AoS to SoA format. - * I.e. (RGBA,RGBA,RGBA,RGBA) -> (RRRR,GGGG,BBBB,AAAA) - * This is temporary! + ASSERT(spu.fragment_program); + ASSERT(spu.fragment_ops); + + /* Execute the current fragment program */ + kill_mask = spu.fragment_program(inputs, outputs, spu.constants); + + mask = spu_andc(mask, kill_mask); + + /* Execute per-fragment/quad operations, including: + * alpha test, z test, stencil test, blend and framebuffer writing. + * Note that there are two different fragment operations functions + * that can be called, one for front-facing fragments, and one + * for back-facing fragments. (Often the two are the same; + * but in some cases, like two-sided stenciling, they can be + * very different.) So choose the correct function depending + * on the calculated facing. */ - vector float soa_frag[4]; - _transpose_matrix4x4(soa_frag, colors); - - float4 fragZ; - - fragZ.v = eval_z((float) x, (float) y); - - /* Do all per-fragment/quad operations here, including: - * alpha test, z test, stencil test, blend and framebuffer writing. - */ - spu.fragment_ops(ix, iy, &spu.ctile, &spu.ztile, - fragZ.v, - soa_frag[0], soa_frag[1], - soa_frag[2], soa_frag[3], + spu.fragment_ops[setup.facing](ix, iy, &spu.ctile, &spu.ztile, + fragZ, + outputs[0*4+0], + outputs[0*4+1], + outputs[0*4+2], + outputs[0*4+3], mask); } - } } @@ -383,7 +298,8 @@ emit_quad( int x, int y, mask_t mask ) * Given an X or Y coordinate, return the block/quad coordinate that it * belongs to. */ -static INLINE int block( int x ) +static INLINE int +block(int x) { return x & ~1; } @@ -394,7 +310,8 @@ static INLINE int block( int x ) * the triangle's bounds. * The mask is a uint4 vector and each element will be 0 or 0xffffffff. */ -static INLINE mask_t calculate_mask( int x ) +static INLINE mask_t +calculate_mask(int x) { /* This is a little tricky. * Use & instead of && to avoid branches. @@ -412,7 +329,8 @@ static INLINE mask_t calculate_mask( int x ) /** * Render a horizontal span of quads */ -static void flush_spans( void ) +static void +flush_spans(void) { int minleft, maxright; int x; @@ -440,7 +358,6 @@ static void flush_spans( void ) return; } - /* OK, we're very likely to need the tile data now. * clear or finish waiting if needed. */ @@ -457,7 +374,7 @@ static void flush_spans( void ) } ASSERT(spu.cur_ctile_status != TILE_STATUS_DEFINED); - if (spu.read_depth) { + if (spu.read_depth_stencil) { if (spu.cur_ztile_status == TILE_STATUS_GETTING) { /* wait for mfc_get() to complete */ //printf("SPU: %u: waiting for ztile\n", spu.init.id); @@ -476,9 +393,7 @@ static void flush_spans( void ) * calculate_mask() could be simplified a bit... */ for (x = block(minleft); x <= block(maxright); x += 2) { -#if 1 - emit_quad( x, setup.span.y, calculate_mask( x ) ); -#endif + emit_quad( x, setup.span.y, calculate_mask( x )); } setup.span.y = 0; @@ -487,33 +402,46 @@ static void flush_spans( void ) setup.span.right[1] = 0; } + #if DEBUG_VERTS -static void print_vertex(const struct vertex_header *v) +static void +print_vertex(const struct vertex_header *v) { - int i; - fprintf(stderr, "Vertex: (%p)\n", v); - for (i = 0; i < setup.quad.nr_attrs; i++) { - fprintf(stderr, " %d: %f %f %f %f\n", i, - v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); + uint i; + fprintf(stderr, " Vertex: (%p)\n", v); + for (i = 0; i < spu.vertex_info.num_attribs; i++) { + fprintf(stderr, " %d: %f %f %f %f\n", i, + spu_extract(v->data[i], 0), + spu_extract(v->data[i], 1), + spu_extract(v->data[i], 2), + spu_extract(v->data[i], 3)); } } #endif -static boolean setup_sort_vertices(const struct vertex_header *v0, - const struct vertex_header *v1, - const struct vertex_header *v2) +/** + * Sort vertices from top to bottom. + * Compute area and determine front vs. back facing. + * Do coarse clip test against tile bounds + * \return FALSE if tri is totally outside tile, TRUE otherwise + */ +static boolean +setup_sort_vertices(const struct vertex_header *v0, + const struct vertex_header *v1, + const struct vertex_header *v2) { + float area, sign; #if DEBUG_VERTS - fprintf(stderr, "Triangle:\n"); - print_vertex(v0); - print_vertex(v1); - print_vertex(v2); + if (spu.init.id==0) { + fprintf(stderr, "SPU %u: Triangle:\n", spu.init.id); + print_vertex(v0); + print_vertex(v1); + print_vertex(v2); + } #endif - setup.vprovoke = v2; - /* determine bottom to top order of vertices */ { float y0 = spu_extract(v0->data[0], 1); @@ -525,18 +453,21 @@ static boolean setup_sort_vertices(const struct vertex_header *v0, setup.vmin = v0; setup.vmid = v1; setup.vmax = v2; + sign = -1.0f; } else if (y2 <= y0) { /* y2<=y0<=y1 */ setup.vmin = v2; setup.vmid = v0; setup.vmax = v1; + sign = -1.0f; } else { /* y0<=y2<=y1 */ setup.vmin = v0; setup.vmid = v2; setup.vmax = v1; + sign = 1.0f; } } else { @@ -545,18 +476,21 @@ static boolean setup_sort_vertices(const struct vertex_header *v0, setup.vmin = v1; setup.vmid = v0; setup.vmax = v2; + sign = 1.0f; } else if (y2 <= y1) { /* y2<=y1<=y0 */ setup.vmin = v2; setup.vmid = v1; setup.vmax = v0; + sign = 1.0f; } else { /* y1<=y2<=y0 */ setup.vmin = v1; setup.vmid = v2; setup.vmax = v0; + sign = -1.0f; } } } @@ -585,31 +519,23 @@ static boolean setup_sort_vertices(const struct vertex_header *v0, /* * Compute triangle's area. Use 1/area to compute partial * derivatives of attributes later. - * - * The area will be the same as prim->det, but the sign may be - * different depending on how the vertices get sorted above. - * - * To determine whether the primitive is front or back facing we - * use the prim->det value because its sign is correct. */ - { - const float area = (setup.emaj.dx * setup.ebot.dy - - setup.ebot.dx * setup.emaj.dy); - - setup.oneoverarea = 1.0f / area; - /* - _mesa_printf("%s one-over-area %f area %f det %f\n", - __FUNCTION__, setup.oneoverarea, area, prim->det ); - */ - } + area = setup.emaj.dx * setup.ebot.dy - setup.ebot.dx * setup.emaj.dy; -#if 0 - /* We need to know if this is a front or back-facing triangle for: - * - the GLSL gl_FrontFacing fragment attribute (bool) - * - two-sided stencil test + setup.oneOverArea = 1.0f / area; + + /* The product of area * sign indicates front/back orientation (0/1). + * Just in case someone gets the bright idea of switching the front + * and back constants without noticing that we're assuming their + * values in this operation, also assert that the values are + * what we think they are. */ - setup.quad.facing = (prim->det > 0.0) ^ (setup.softpipe->rasterizer->front_winding == PIPE_WINDING_CW); -#endif + ASSERT(CELL_FACING_FRONT == 0); + ASSERT(CELL_FACING_BACK == 1); + setup.facing = (area * sign > 0.0f) + ^ (spu.rasterizer.front_winding == PIPE_WINDING_CW); + + setup.vprovoke = v2; return TRUE; } @@ -622,63 +548,11 @@ static boolean setup_sort_vertices(const struct vertex_header *v0, * \param slot which attribute slot */ static INLINE void -const_coeff(uint slot) +const_coeff4(uint slot) { - setup.coef[slot].dadx.v = (vector float) {0.0, 0.0, 0.0, 0.0}; - setup.coef[slot].dady.v = (vector float) {0.0, 0.0, 0.0, 0.0}; - setup.coef[slot].a0.v = setup.vprovoke->data[slot]; -} - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a triangle. - */ -static INLINE void -tri_linear_coeff(uint slot, uint firstComp, uint lastComp) -{ - uint i; - const float *vmin_d = (float *) &setup.vmin->data[slot]; - const float *vmid_d = (float *) &setup.vmid->data[slot]; - const float *vmax_d = (float *) &setup.vmax->data[slot]; - const float x = spu_extract(setup.vmin->data[0], 0) - 0.5f; - const float y = spu_extract(setup.vmin->data[0], 1) - 0.5f; - - for (i = firstComp; i < lastComp; i++) { - float botda = vmid_d[i] - vmin_d[i]; - float majda = vmax_d[i] - vmin_d[i]; - float a = setup.ebot.dy * majda - botda * setup.emaj.dy; - float b = setup.emaj.dx * botda - majda * setup.ebot.dx; - - ASSERT(slot < PIPE_MAX_SHADER_INPUTS); - - setup.coef[slot].dadx.f[i] = a * setup.oneoverarea; - setup.coef[slot].dady.f[i] = b * setup.oneoverarea; - - /* calculate a0 as the value which would be sampled for the - * fragment at (0,0), taking into account that we want to sample at - * pixel centers, in other words (0.5, 0.5). - * - * this is neat but unfortunately not a good way to do things for - * triangles with very large values of dadx or dady as it will - * result in the subtraction and re-addition from a0 of a very - * large number, which means we'll end up loosing a lot of the - * fractional bits and precision from a0. the way to fix this is - * to define a0 as the sample at a pixel center somewhere near vmin - * instead - i'll switch to this later. - */ - setup.coef[slot].a0.f[i] = (vmin_d[i] - - (setup.coef[slot].dadx.f[i] * x + - setup.coef[slot].dady.f[i] * y)); - } - - /* - _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", - slot, "xyzw"[i], - setup.coef[slot].a0[i], - setup.coef[slot].dadx.f[i], - setup.coef[slot].dady.f[i]); - */ + setup.coef[slot].dadx = (vector float) {0.0, 0.0, 0.0, 0.0}; + setup.coef[slot].dady = (vector float) {0.0, 0.0, 0.0, 0.0}; + setup.coef[slot].a0 = setup.vprovoke->data[slot]; } @@ -702,18 +576,16 @@ tri_linear_coeff4(uint slot) vector float b = spu_sub(spu_mul(spu_splats(setup.emaj.dx), botda), spu_mul(majda, spu_splats(setup.ebot.dx))); - setup.coef[slot].dadx.v = spu_mul(a, spu_splats(setup.oneoverarea)); - setup.coef[slot].dady.v = spu_mul(b, spu_splats(setup.oneoverarea)); + setup.coef[slot].dadx = spu_mul(a, spu_splats(setup.oneOverArea)); + setup.coef[slot].dady = spu_mul(b, spu_splats(setup.oneOverArea)); - vector float tempx = spu_mul(setup.coef[slot].dadx.v, xxxx); - vector float tempy = spu_mul(setup.coef[slot].dady.v, yyyy); + vector float tempx = spu_mul(setup.coef[slot].dadx, xxxx); + vector float tempy = spu_mul(setup.coef[slot].dady, yyyy); - setup.coef[slot].a0.v = spu_sub(vmin_d, spu_add(tempx, tempy)); + setup.coef[slot].a0 = spu_sub(vmin_d, spu_add(tempx, tempy)); } - -#if 0 /** * Compute a0, dadx and dady for a perspective-corrected interpolant, * for a triangle. @@ -722,82 +594,76 @@ tri_linear_coeff4(uint slot) * Later, when we compute the value at a particular fragment position we'll * divide the interpolated value by the interpolated W at that fragment. */ -static void tri_persp_coeff( unsigned slot, - unsigned i ) +static void +tri_persp_coeff4(uint slot) { - /* premultiply by 1/w: - */ - float mina = setup.vmin->data[slot][i] * setup.vmin->data[0][3]; - float mida = setup.vmid->data[slot][i] * setup.vmid->data[0][3]; - float maxa = setup.vmax->data[slot][i] * setup.vmax->data[0][3]; - - float botda = mida - mina; - float majda = maxa - mina; - float a = setup.ebot.dy * majda - botda * setup.emaj.dy; - float b = setup.emaj.dx * botda - majda * setup.ebot.dx; - - /* - printf("tri persp %d,%d: %f %f %f\n", slot, i, - setup.vmin->data[slot][i], - setup.vmid->data[slot][i], - setup.vmax->data[slot][i] - ); - */ + const vector float xxxx = spu_splats(spu_extract(setup.vmin->data[0], 0) - 0.5f); + const vector float yyyy = spu_splats(spu_extract(setup.vmin->data[0], 1) - 0.5f); + + const vector float vmin_w = spu_splats(spu_extract(setup.vmin->data[0], 3)); + const vector float vmid_w = spu_splats(spu_extract(setup.vmid->data[0], 3)); + const vector float vmax_w = spu_splats(spu_extract(setup.vmax->data[0], 3)); + + vector float vmin_d = setup.vmin->data[slot]; + vector float vmid_d = setup.vmid->data[slot]; + vector float vmax_d = setup.vmax->data[slot]; + + vmin_d = spu_mul(vmin_d, vmin_w); + vmid_d = spu_mul(vmid_d, vmid_w); + vmax_d = spu_mul(vmax_d, vmax_w); + + vector float botda = vmid_d - vmin_d; + vector float majda = vmax_d - vmin_d; - assert(slot < PIPE_MAX_SHADER_INPUTS); - assert(i <= 3); + vector float a = spu_sub(spu_mul(spu_splats(setup.ebot.dy), majda), + spu_mul(botda, spu_splats(setup.emaj.dy))); + vector float b = spu_sub(spu_mul(spu_splats(setup.emaj.dx), botda), + spu_mul(majda, spu_splats(setup.ebot.dx))); - setup.coef[slot].dadx.f[i] = a * setup.oneoverarea; - setup.coef[slot].dady.f[i] = b * setup.oneoverarea; - setup.coef[slot].a0.f[i] = (mina - - (setup.coef[slot].dadx.f[i] * (setup.vmin->data[0][0] - 0.5f) + - setup.coef[slot].dady.f[i] * (setup.vmin->data[0][1] - 0.5f))); + setup.coef[slot].dadx = spu_mul(a, spu_splats(setup.oneOverArea)); + setup.coef[slot].dady = spu_mul(b, spu_splats(setup.oneOverArea)); + + vector float tempx = spu_mul(setup.coef[slot].dadx, xxxx); + vector float tempy = spu_mul(setup.coef[slot].dady, yyyy); + + setup.coef[slot].a0 = spu_sub(vmin_d, spu_add(tempx, tempy)); } -#endif + /** * Compute the setup.coef[] array dadx, dady, a0 values. * Must be called after setup.vmin,vmid,vmax,vprovoke are initialized. */ -static void setup_tri_coefficients(void) +static void +setup_tri_coefficients(void) { -#if 1 uint i; for (i = 0; i < spu.vertex_info.num_attribs; i++) { - switch (spu.vertex_info.interp_mode[i]) { + switch (spu.vertex_info.attrib[i].interp_mode) { case INTERP_NONE: break; - case INTERP_POS: - /*tri_linear_coeff(i, 2, 3);*/ - /* XXX interp W if PERSPECTIVE... */ - tri_linear_coeff4(i); - break; case INTERP_CONSTANT: - const_coeff(i); + const_coeff4(i); break; + case INTERP_POS: + /* fall-through */ case INTERP_LINEAR: tri_linear_coeff4(i); break; case INTERP_PERSPECTIVE: - tri_linear_coeff4(i); /* temporary */ + tri_persp_coeff4(i); break; default: ASSERT(0); } } -#else - ASSERT(spu.vertex_info.interp_mode[0] == INTERP_POS); - ASSERT(spu.vertex_info.interp_mode[1] == INTERP_LINEAR || - spu.vertex_info.interp_mode[1] == INTERP_CONSTANT); - tri_linear_coeff(0, 2, 3); /* slot 0, z */ - tri_linear_coeff(1, 0, 4); /* slot 1, color */ -#endif } -static void setup_tri_edges(void) +static void +setup_tri_edges(void) { float vmin_x = spu_extract(setup.vmin->data[0], 0) + 0.5f; float vmid_x = spu_extract(setup.vmid->data[0], 0) + 0.5f; @@ -827,9 +693,8 @@ static void setup_tri_edges(void) * Render the upper or lower half of a triangle. * Scissoring/cliprect is applied here too. */ -static void subtriangle( struct edge *eleft, - struct edge *eright, - unsigned lines ) +static void +subtriangle(struct edge *eleft, struct edge *eright, unsigned lines) { const int minx = setup.cliprect_minx; const int maxx = setup.cliprect_maxx; @@ -902,7 +767,8 @@ static void subtriangle( struct edge *eleft, * The tile data should have already been fetched. */ boolean -tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) +tri_draw(const float *v0, const float *v1, const float *v2, + uint tx, uint ty) { setup.tx = tx; setup.ty = ty; @@ -926,19 +792,14 @@ tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) setup.span.y_flags = 0; setup.span.right[0] = 0; setup.span.right[1] = 0; - /* setup.span.z_mode = tri_z_mode( setup.ctx ); */ - /* init_constant_attribs( setup ); */ - - if (setup.oneoverarea < 0.0) { - /* emaj on left: - */ + if (setup.oneOverArea < 0.0) { + /* emaj on left */ subtriangle( &setup.emaj, &setup.ebot, setup.ebot.lines ); subtriangle( &setup.emaj, &setup.etop, setup.etop.lines ); } else { - /* emaj on right: - */ + /* emaj on right */ subtriangle( &setup.ebot, &setup.emaj, setup.ebot.lines ); subtriangle( &setup.etop, &setup.emaj, setup.etop.lines ); } diff --git a/src/gallium/drivers/i915simple/i915_debug.c b/src/gallium/drivers/i915simple/i915_debug.c index 5e26d1b905..7a4e7051d2 100644 --- a/src/gallium/drivers/i915simple/i915_debug.c +++ b/src/gallium/drivers/i915simple/i915_debug.c @@ -210,6 +210,7 @@ BITS( PRINTF(stream, ": 0x%x\n", ((dw) & himask) >> (lo)); } +#ifdef DEBUG #define MBZ( dw, hi, lo) do { \ unsigned x = (dw) >> (lo); \ unsigned lomask = (1 << (lo)) - 1; \ @@ -217,6 +218,10 @@ BITS( himask = (1UL << (hi)) - 1; \ assert ((x & himask & ~lomask) == 0); \ } while (0) +#else +#define MBZ( dw, hi, lo) do { \ +} while (0) +#endif static void FLAG( diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 34b4a846c1..43d62c5176 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -144,7 +144,7 @@ src_vector(struct i915_fp_compile *p, const struct tgsi_full_src_register *source) { uint index = source->SrcRegister.Index; - uint src, sem_name, sem_ind; + uint src = 0, sem_name, sem_ind; switch (source->SrcRegister.File) { case TGSI_FILE_TEMPORARY: diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index bd87217063..2f5459af67 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -178,7 +178,9 @@ i915_displaytarget_layout(struct i915_texture *tex) if (tex->base.width[0] >= 128) { tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size); tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); +#if 0 /* used for tiled display targets */ tex->tiled = 1; +#endif } else { tex->stride = round_up(tex->base.nblocksx[0] * pt->block.size, 64); tex->total_nblocksy = tex->base.nblocksy[0]; @@ -206,11 +208,10 @@ i945_miptree_layout_2d( struct i915_texture *tex ) unsigned nblocksx = pt->nblocksx[0]; unsigned nblocksy = pt->nblocksy[0]; -#if 0 /* used for tiled display targets */ - if (pt->last_level == 0 && pt->block.size == 4) + /* used for tiled display targets */ + if (0) if (i915_displaytarget_layout(tex)) return; -#endif tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); diff --git a/src/gallium/drivers/nouveau/nouveau_bo.h b/src/gallium/drivers/nouveau/nouveau_bo.h new file mode 100644 index 0000000000..65b138283c --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_bo.h @@ -0,0 +1,53 @@ +/* + * Copyright 2007 Nouveau Project + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef __NOUVEAU_BO_H__ +#define __NOUVEAU_BO_H__ + +/* Relocation/Buffer type flags */ +#define NOUVEAU_BO_VRAM (1 << 0) +#define NOUVEAU_BO_GART (1 << 1) +#define NOUVEAU_BO_RD (1 << 2) +#define NOUVEAU_BO_WR (1 << 3) +#define NOUVEAU_BO_RDWR (NOUVEAU_BO_RD | NOUVEAU_BO_WR) +#define NOUVEAU_BO_MAP (1 << 4) +#define NOUVEAU_BO_PIN (1 << 5) +#define NOUVEAU_BO_LOW (1 << 6) +#define NOUVEAU_BO_HIGH (1 << 7) +#define NOUVEAU_BO_OR (1 << 8) +#define NOUVEAU_BO_LOCAL (1 << 9) +#define NOUVEAU_BO_TILED (1 << 10) +#define NOUVEAU_BO_ZTILE (1 << 11) +#define NOUVEAU_BO_DUMMY (1 << 31) + +struct nouveau_bo { + struct nouveau_device *device; + uint64_t handle; + + uint64_t size; + void *map; + + uint32_t flags; + uint64_t offset; +}; + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_channel.h b/src/gallium/drivers/nouveau/nouveau_channel.h new file mode 100644 index 0000000000..cd99a676bd --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_channel.h @@ -0,0 +1,40 @@ +/* + * Copyright 2007 Nouveau Project + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef __NOUVEAU_CHANNEL_H__ +#define __NOUVEAU_CHANNEL_H__ + +struct nouveau_channel { + struct nouveau_device *device; + int id; + + struct nouveau_pushbuf *pushbuf; + + struct nouveau_grobj *nullobj; + struct nouveau_grobj *vram; + struct nouveau_grobj *gart; + + void *user_private; + void (*hang_notify)(struct nouveau_channel *); +}; + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_class.h b/src/gallium/drivers/nouveau/nouveau_class.h new file mode 100644 index 0000000000..3df3d7b2b8 --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_class.h @@ -0,0 +1,8006 @@ +/************************************************************************* + + Autogenerated file, do not edit ! + +************************************************************************** + + Copyright (C) 2006-2008 : + Dmitry Baryshkov, + Laurent Carlier, + Matthieu Castet, + Dawid Gajownik, + Jeremy Kolb, + Stephane Loeuillet, + Patrice Mandin, + Stephane Marchesin, + Serge Martin, + Sylvain Munaut, + Simon Raffeiner, + Ben Skeggs, + Erik Waling, + koala_br, + +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +*************************************************************************/ + + +#ifndef NOUVEAU_REG_H +#define NOUVEAU_REG_H 1 + + +#define NV01_ROOT 0x00000001 + + + +#define NV01_CONTEXT_DMA 0x00000002 + + + +#define NV01_DEVICE 0x00000003 + + + +#define NV01_TIMER 0x00000004 + +#define NV01_TIMER_SYNCHRONIZE 0x00000100 +#define NV01_TIMER_STOP_ALARM 0x00000104 +#define NV01_TIMER_DMA_NOTIFY 0x00000180 +#define NV01_TIMER_TIME(x) (0x00000300+((x)*4)) +#define NV01_TIMER_TIME__SIZE 0x00000002 +#define NV01_TIMER_ALARM_NOTIFY 0x00000308 + + +#define NV_IMAGE_STENCIL 0x00000010 + +#define NV_IMAGE_STENCIL_NOTIFY 0x00000104 +#define NV_IMAGE_STENCIL_DMA_NOTIFY 0x00000180 +#define NV_IMAGE_STENCIL_IMAGE_OUTPUT 0x00000200 +#define NV_IMAGE_STENCIL_IMAGE_INPUT(x) (0x00000204+((x)*4)) +#define NV_IMAGE_STENCIL_IMAGE_INPUT__SIZE 0x00000002 + + +#define NV_IMAGE_BLEND_AND 0x00000011 + +#define NV_IMAGE_BLEND_AND_NOP 0x00000100 +#define NV_IMAGE_BLEND_AND_NOTIFY 0x00000104 +#define NV_IMAGE_BLEND_AND_DMA_NOTIFY 0x00000180 +#define NV_IMAGE_BLEND_AND_IMAGE_OUTPUT 0x00000200 +#define NV_IMAGE_BLEND_AND_BETA_INPUT 0x00000204 +#define NV_IMAGE_BLEND_AND_IMAGE_INPUT 0x00000208 + + +#define NV01_CONTEXT_BETA1 0x00000012 + +#define NV01_CONTEXT_BETA1_NOP 0x00000100 +#define NV01_CONTEXT_BETA1_NOTIFY 0x00000104 +#define NV01_CONTEXT_BETA1_DMA_NOTIFY 0x00000180 +#define NV01_CONTEXT_BETA1_BETA_1D31 0x00000300 + + +#define NV_IMAGE_ROP_AND 0x00000013 + +#define NV_IMAGE_ROP_AND_NOTIFY 0x00000104 +#define NV_IMAGE_ROP_AND_DMA_NOTIFY 0x00000180 +#define NV_IMAGE_ROP_AND_IMAGE_OUTPUT 0x00000200 +#define NV_IMAGE_ROP_AND_ROP_INPUT 0x00000204 +#define NV_IMAGE_ROP_AND_IMAGE_INPUT(x) (0x00000208+((x)*4)) +#define NV_IMAGE_ROP_AND_IMAGE_INPUT__SIZE 0x00000002 + + +#define NV_IMAGE_COLOR_KEY 0x00000015 + + + +#define NV01_CONTEXT_COLOR_KEY 0x00000017 + +#define NV01_CONTEXT_COLOR_KEY_NOP 0x00000100 +#define NV01_CONTEXT_COLOR_KEY_NOTIFY 0x00000104 +#define NV01_CONTEXT_COLOR_KEY_DMA_NOTIFY 0x00000180 +#define NV01_CONTEXT_COLOR_KEY_COLOR_FORMAT 0x00000300 +#define NV01_CONTEXT_COLOR_KEY_COLOR_FORMAT_X16A8Y8 0x00000001 +#define NV01_CONTEXT_COLOR_KEY_COLOR_FORMAT_X24Y8 0x00000002 +#define NV01_CONTEXT_COLOR_KEY_COLOR_FORMAT_X16A1R5G5B5 0x00000003 +#define NV01_CONTEXT_COLOR_KEY_COLOR_FORMAT_X17R5G5B5 0x00000004 +#define NV01_CONTEXT_COLOR_KEY_COLOR_FORMAT_A8R8G8B8 0x00000005 +#define NV01_CONTEXT_COLOR_KEY_COLOR_FORMAT_X8R8G8B8 0x00000006 +#define NV01_CONTEXT_COLOR_KEY_COLOR_FORMAT_A16Y16 0x00000007 +#define NV01_CONTEXT_COLOR_KEY_COLOR_FORMAT_X16Y16 0x00000008 +#define NV01_CONTEXT_COLOR_KEY_COLOR 0x00000304 + + +#define NV01_CONTEXT_PATTERN 0x00000018 + +#define NV01_CONTEXT_PATTERN_NOP 0x00000100 +#define NV01_CONTEXT_PATTERN_NOTIFY 0x00000104 +#define NV01_CONTEXT_PATTERN_DMA_NOTIFY 0x00000180 +#define NV01_CONTEXT_PATTERN_COLOR_FORMAT 0x00000300 +#define NV01_CONTEXT_PATTERN_MONOCHROME_FORMAT 0x00000304 +#define NV01_CONTEXT_PATTERN_SHAPE 0x00000308 +#define NV01_CONTEXT_PATTERN_COLOR(x) (0x00000310+((x)*4)) +#define NV01_CONTEXT_PATTERN_COLOR__SIZE 0x00000002 +#define NV01_CONTEXT_PATTERN_PATTERN(x) (0x00000318+((x)*4)) +#define NV01_CONTEXT_PATTERN_PATTERN__SIZE 0x00000002 + + +#define NV01_CONTEXT_CLIP_RECTANGLE 0x00000019 + +#define NV01_CONTEXT_CLIP_RECTANGLE_NOP 0x00000100 +#define NV01_CONTEXT_CLIP_RECTANGLE_NOTIFY 0x00000104 +#define NV01_CONTEXT_CLIP_RECTANGLE_DMA_NOTIFY 0x00000180 +#define NV01_CONTEXT_CLIP_RECTANGLE_POINT 0x00000300 +#define NV01_CONTEXT_CLIP_RECTANGLE_POINT_X_SHIFT 0 +#define NV01_CONTEXT_CLIP_RECTANGLE_POINT_X_MASK 0x0000ffff +#define NV01_CONTEXT_CLIP_RECTANGLE_POINT_Y_SHIFT 16 +#define NV01_CONTEXT_CLIP_RECTANGLE_POINT_Y_MASK 0xffff0000 +#define NV01_CONTEXT_CLIP_RECTANGLE_SIZE 0x00000304 +#define NV01_CONTEXT_CLIP_RECTANGLE_SIZE_W_SHIFT 0 +#define NV01_CONTEXT_CLIP_RECTANGLE_SIZE_W_MASK 0x0000ffff +#define NV01_CONTEXT_CLIP_RECTANGLE_SIZE_H_SHIFT 16 +#define NV01_CONTEXT_CLIP_RECTANGLE_SIZE_H_MASK 0xffff0000 + + +#define NV01_RENDER_SOLID_LINE 0x0000001c + +#define NV01_RENDER_SOLID_LINE_NOP 0x00000100 +#define NV01_RENDER_SOLID_LINE_NOTIFY 0x00000104 +#define NV01_RENDER_SOLID_LINE_PATCH 0x0000010c +#define NV01_RENDER_SOLID_LINE_DMA_NOTIFY 0x00000180 +#define NV01_RENDER_SOLID_LINE_CLIP_RECTANGLE 0x00000184 +#define NV01_RENDER_SOLID_LINE_PATTERN 0x00000188 +#define NV01_RENDER_SOLID_LINE_ROP 0x0000018c +#define NV01_RENDER_SOLID_LINE_BETA1 0x00000190 +#define NV01_RENDER_SOLID_LINE_SURFACE 0x00000194 +#define NV01_RENDER_SOLID_LINE_OPERATION 0x000002fc +#define NV01_RENDER_SOLID_LINE_OPERATION_SRCCOPY_AND 0x00000000 +#define NV01_RENDER_SOLID_LINE_OPERATION_ROP_AND 0x00000001 +#define NV01_RENDER_SOLID_LINE_OPERATION_BLEND_AND 0x00000002 +#define NV01_RENDER_SOLID_LINE_OPERATION_SRCCOPY 0x00000003 +#define NV01_RENDER_SOLID_LINE_OPERATION_SRCCOPY_PREMULT 0x00000004 +#define NV01_RENDER_SOLID_LINE_OPERATION_BLEND_PREMULT 0x00000005 +#define NV01_RENDER_SOLID_LINE_COLOR_FORMAT 0x00000300 +#define NV01_RENDER_SOLID_LINE_COLOR_FORMAT_X16A8Y8 0x00000001 +#define NV01_RENDER_SOLID_LINE_COLOR_FORMAT_X24Y8 0x00000002 +#define NV01_RENDER_SOLID_LINE_COLOR_FORMAT_X16A1R5G5B5 0x00000003 +#define NV01_RENDER_SOLID_LINE_COLOR_FORMAT_X17R5G5B5 0x00000004 +#define NV01_RENDER_SOLID_LINE_COLOR_FORMAT_A8R8G8B8 0x00000005 +#define NV01_RENDER_SOLID_LINE_COLOR_FORMAT_X8R8G8B8 0x00000006 +#define NV01_RENDER_SOLID_LINE_COLOR_FORMAT_A16Y16 0x00000007 +#define NV01_RENDER_SOLID_LINE_COLOR_FORMAT_X16Y16 0x00000008 +#define NV01_RENDER_SOLID_LINE_COLOR 0x00000304 +#define NV01_RENDER_SOLID_LINE_LINE_POINT0(x) (0x00000400+((x)*8)) +#define NV01_RENDER_SOLID_LINE_LINE_POINT0__SIZE 0x00000010 +#define NV01_RENDER_SOLID_LINE_LINE_POINT0_X_SHIFT 0 +#define NV01_RENDER_SOLID_LINE_LINE_POINT0_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_LINE_LINE_POINT0_Y_SHIFT 16 +#define NV01_RENDER_SOLID_LINE_LINE_POINT0_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_LINE_LINE_POINT1(x) (0x00000404+((x)*8)) +#define NV01_RENDER_SOLID_LINE_LINE_POINT1__SIZE 0x00000010 +#define NV01_RENDER_SOLID_LINE_LINE_POINT1_X_SHIFT 0 +#define NV01_RENDER_SOLID_LINE_LINE_POINT1_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_LINE_LINE_POINT1_Y_SHIFT 16 +#define NV01_RENDER_SOLID_LINE_LINE_POINT1_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_LINE_LINE32_POINT0_X(x) (0x00000480+((x)*16)) +#define NV01_RENDER_SOLID_LINE_LINE32_POINT0_X__SIZE 0x00000010 +#define NV01_RENDER_SOLID_LINE_LINE32_POINT0_Y(x) (0x00000484+((x)*16)) +#define NV01_RENDER_SOLID_LINE_LINE32_POINT0_Y__SIZE 0x00000010 +#define NV01_RENDER_SOLID_LINE_LINE32_POINT1_X(x) (0x00000488+((x)*16)) +#define NV01_RENDER_SOLID_LINE_LINE32_POINT1_X__SIZE 0x00000010 +#define NV01_RENDER_SOLID_LINE_LINE32_POINT1_Y(x) (0x0000048c+((x)*16)) +#define NV01_RENDER_SOLID_LINE_LINE32_POINT1_Y__SIZE 0x00000010 +#define NV01_RENDER_SOLID_LINE_POLYLINE(x) (0x00000500+((x)*4)) +#define NV01_RENDER_SOLID_LINE_POLYLINE__SIZE 0x00000020 +#define NV01_RENDER_SOLID_LINE_POLYLINE_X_SHIFT 0 +#define NV01_RENDER_SOLID_LINE_POLYLINE_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_LINE_POLYLINE_Y_SHIFT 16 +#define NV01_RENDER_SOLID_LINE_POLYLINE_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_LINE_POLYLINE32_POINT_X(x) (0x00000580+((x)*8)) +#define NV01_RENDER_SOLID_LINE_POLYLINE32_POINT_X__SIZE 0x00000010 +#define NV01_RENDER_SOLID_LINE_POLYLINE32_POINT_Y(x) (0x00000584+((x)*8)) +#define NV01_RENDER_SOLID_LINE_POLYLINE32_POINT_Y__SIZE 0x00000010 +#define NV01_RENDER_SOLID_LINE_CPOLYLINE_COLOR(x) (0x00000600+((x)*8)) +#define NV01_RENDER_SOLID_LINE_CPOLYLINE_COLOR__SIZE 0x00000010 +#define NV01_RENDER_SOLID_LINE_CPOLYLINE_POINT(x) (0x00000604+((x)*8)) +#define NV01_RENDER_SOLID_LINE_CPOLYLINE_POINT__SIZE 0x00000010 +#define NV01_RENDER_SOLID_LINE_CPOLYLINE_POINT_X_SHIFT 0 +#define NV01_RENDER_SOLID_LINE_CPOLYLINE_POINT_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_LINE_CPOLYLINE_POINT_Y_SHIFT 16 +#define NV01_RENDER_SOLID_LINE_CPOLYLINE_POINT_Y_MASK 0xffff0000 + + +#define NV01_RENDER_SOLID_TRIANGLE 0x0000001d + +#define NV01_RENDER_SOLID_TRIANGLE_NOP 0x00000100 +#define NV01_RENDER_SOLID_TRIANGLE_NOTIFY 0x00000104 +#define NV01_RENDER_SOLID_TRIANGLE_PATCH 0x0000010c +#define NV01_RENDER_SOLID_TRIANGLE_DMA_NOTIFY 0x00000180 +#define NV01_RENDER_SOLID_TRIANGLE_CLIP_RECTANGLE 0x00000184 +#define NV01_RENDER_SOLID_TRIANGLE_PATTERN 0x00000188 +#define NV01_RENDER_SOLID_TRIANGLE_ROP 0x0000018c +#define NV01_RENDER_SOLID_TRIANGLE_BETA1 0x00000190 +#define NV01_RENDER_SOLID_TRIANGLE_SURFACE 0x00000194 +#define NV01_RENDER_SOLID_TRIANGLE_OPERATION 0x000002fc +#define NV01_RENDER_SOLID_TRIANGLE_OPERATION_SRCCOPY_AND 0x00000000 +#define NV01_RENDER_SOLID_TRIANGLE_OPERATION_ROP_AND 0x00000001 +#define NV01_RENDER_SOLID_TRIANGLE_OPERATION_BLEND_AND 0x00000002 +#define NV01_RENDER_SOLID_TRIANGLE_OPERATION_SRCCOPY 0x00000003 +#define NV01_RENDER_SOLID_TRIANGLE_OPERATION_SRCCOPY_PREMULT 0x00000004 +#define NV01_RENDER_SOLID_TRIANGLE_OPERATION_BLEND_PREMULT 0x00000005 +#define NV01_RENDER_SOLID_TRIANGLE_COLOR_FORMAT 0x00000300 +#define NV01_RENDER_SOLID_TRIANGLE_COLOR 0x00000304 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT0 0x00000310 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT0_X_SHIFT 0 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT0_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT0_Y_SHIFT 16 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT0_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT1 0x00000314 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT1_X_SHIFT 0 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT1_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT1_Y_SHIFT 16 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT1_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT2 0x00000318 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT2_X_SHIFT 0 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT2_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT2_Y_SHIFT 16 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE_POINT2_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE32_POINT0_X 0x00000320 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE32_POINT0_Y 0x00000324 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE32_POINT1_X 0x00000328 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE32_POINT1_Y 0x0000032c +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE32_POINT2_X 0x00000330 +#define NV01_RENDER_SOLID_TRIANGLE_TRIANGLE32_POINT2_Y 0x00000334 +#define NV01_RENDER_SOLID_TRIANGLE_TRIMESH(x) (0x00000400+((x)*4)) +#define NV01_RENDER_SOLID_TRIANGLE_TRIMESH__SIZE 0x00000020 +#define NV01_RENDER_SOLID_TRIANGLE_TRIMESH_X_SHIFT 0 +#define NV01_RENDER_SOLID_TRIANGLE_TRIMESH_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_TRIANGLE_TRIMESH_Y_SHIFT 16 +#define NV01_RENDER_SOLID_TRIANGLE_TRIMESH_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_TRIANGLE_TRIMESH32_POINT_X(x) (0x00000480+((x)*8)) +#define NV01_RENDER_SOLID_TRIANGLE_TRIMESH32_POINT_X__SIZE 0x00000010 +#define NV01_RENDER_SOLID_TRIANGLE_TRIMESH32_POINT_Y(x) (0x00000484+((x)*8)) +#define NV01_RENDER_SOLID_TRIANGLE_TRIMESH32_POINT_Y__SIZE 0x00000010 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_COLOR(x) (0x00000500+((x)*16)) +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_COLOR__SIZE 0x00000008 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT0(x) (0x00000504+((x)*16)) +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT0__SIZE 0x00000008 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT0_X_SHIFT 0 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT0_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT0_Y_SHIFT 16 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT0_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT1(x) (0x00000508+((x)*16)) +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT1__SIZE 0x00000008 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT1_X_SHIFT 0 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT1_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT1_Y_SHIFT 16 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT1_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT2(x) (0x0000050c+((x)*16)) +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT2__SIZE 0x00000008 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT2_X_SHIFT 0 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT2_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT2_Y_SHIFT 16 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIANGLE_POINT2_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIMESH_COLOR(x) (0x00000580+((x)*8)) +#define NV01_RENDER_SOLID_TRIANGLE_CTRIMESH_COLOR__SIZE 0x00000010 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIMESH_POINT(x) (0x00000584+((x)*8)) +#define NV01_RENDER_SOLID_TRIANGLE_CTRIMESH_POINT__SIZE 0x00000010 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIMESH_POINT_X_SHIFT 0 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIMESH_POINT_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_TRIANGLE_CTRIMESH_POINT_Y_SHIFT 16 +#define NV01_RENDER_SOLID_TRIANGLE_CTRIMESH_POINT_Y_MASK 0xffff0000 + + +#define NV01_RENDER_SOLID_RECTANGLE 0x0000001e + +#define NV01_RENDER_SOLID_RECTANGLE_NOP 0x00000100 +#define NV01_RENDER_SOLID_RECTANGLE_NOTIFY 0x00000104 +#define NV01_RENDER_SOLID_RECTANGLE_PATCH 0x0000010c +#define NV01_RENDER_SOLID_RECTANGLE_DMA_NOTIFY 0x00000180 +#define NV01_RENDER_SOLID_RECTANGLE_CLIP_RECTANGLE 0x00000184 +#define NV01_RENDER_SOLID_RECTANGLE_PATTERN 0x00000188 +#define NV01_RENDER_SOLID_RECTANGLE_ROP 0x0000018c +#define NV01_RENDER_SOLID_RECTANGLE_BETA1 0x00000190 +#define NV01_RENDER_SOLID_RECTANGLE_SURFACE 0x00000194 +#define NV01_RENDER_SOLID_RECTANGLE_OPERATION 0x000002fc +#define NV01_RENDER_SOLID_RECTANGLE_OPERATION_SRCCOPY_AND 0x00000000 +#define NV01_RENDER_SOLID_RECTANGLE_OPERATION_ROP_AND 0x00000001 +#define NV01_RENDER_SOLID_RECTANGLE_OPERATION_BLEND_AND 0x00000002 +#define NV01_RENDER_SOLID_RECTANGLE_OPERATION_SRCCOPY 0x00000003 +#define NV01_RENDER_SOLID_RECTANGLE_OPERATION_SRCCOPY_PREMULT 0x00000004 +#define NV01_RENDER_SOLID_RECTANGLE_OPERATION_BLEND_PREMULT 0x00000005 +#define NV01_RENDER_SOLID_RECTANGLE_COLOR_FORMAT 0x00000300 +#define NV01_RENDER_SOLID_RECTANGLE_COLOR 0x00000304 +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_POINT(x) (0x00000400+((x)*8)) +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_POINT__SIZE 0x00000010 +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_POINT_X_SHIFT 0 +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_POINT_X_MASK 0x0000ffff +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_POINT_Y_SHIFT 16 +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_POINT_Y_MASK 0xffff0000 +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_SIZE(x) (0x00000404+((x)*8)) +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_SIZE__SIZE 0x00000010 +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_SIZE_W_SHIFT 0 +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_SIZE_W_MASK 0x0000ffff +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_SIZE_H_SHIFT 16 +#define NV01_RENDER_SOLID_RECTANGLE_RECTANGLE_SIZE_H_MASK 0xffff0000 + + +#define NV01_IMAGE_BLIT 0x0000001f + +#define NV01_IMAGE_BLIT_NOP 0x00000100 +#define NV01_IMAGE_BLIT_NOTIFY 0x00000104 +#define NV01_IMAGE_BLIT_PATCH 0x0000010c +#define NV01_IMAGE_BLIT_DMA_NOTIFY 0x00000180 +#define NV01_IMAGE_BLIT_COLOR_KEY 0x00000184 +#define NV01_IMAGE_BLIT_CLIP_RECTANGLE 0x00000188 +#define NV01_IMAGE_BLIT_PATTERN 0x0000018c +#define NV01_IMAGE_BLIT_ROP 0x00000190 +#define NV01_IMAGE_BLIT_BETA1 0x00000194 +#define NV01_IMAGE_BLIT_SURFACE 0x0000019c +#define NV01_IMAGE_BLIT_OPERATION 0x000002fc +#define NV01_IMAGE_BLIT_IMAGE_INPUT 0x00000204 +#define NV01_IMAGE_BLIT_POINT_IN 0x00000300 +#define NV01_IMAGE_BLIT_POINT_IN_X_SHIFT 0 +#define NV01_IMAGE_BLIT_POINT_IN_X_MASK 0x0000ffff +#define NV01_IMAGE_BLIT_POINT_IN_Y_SHIFT 16 +#define NV01_IMAGE_BLIT_POINT_IN_Y_MASK 0xffff0000 +#define NV01_IMAGE_BLIT_POINT_OUT 0x00000304 +#define NV01_IMAGE_BLIT_POINT_OUT_X_SHIFT 0 +#define NV01_IMAGE_BLIT_POINT_OUT_X_MASK 0x0000ffff +#define NV01_IMAGE_BLIT_POINT_OUT_Y_SHIFT 16 +#define NV01_IMAGE_BLIT_POINT_OUT_Y_MASK 0xffff0000 +#define NV01_IMAGE_BLIT_SIZE 0x00000308 +#define NV01_IMAGE_BLIT_SIZE_W_SHIFT 0 +#define NV01_IMAGE_BLIT_SIZE_W_MASK 0x0000ffff +#define NV01_IMAGE_BLIT_SIZE_H_SHIFT 16 +#define NV01_IMAGE_BLIT_SIZE_H_MASK 0xffff0000 + + +#define NV01_IMAGE_FROM_CPU 0x00000021 + +#define NV01_IMAGE_FROM_CPU_NOP 0x00000100 +#define NV01_IMAGE_FROM_CPU_NOTIFY 0x00000104 +#define NV01_IMAGE_FROM_CPU_PATCH 0x0000010c +#define NV01_IMAGE_FROM_CPU_DMA_NOTIFY 0x00000180 +#define NV01_IMAGE_FROM_CPU_COLOR_KEY 0x00000184 +#define NV01_IMAGE_FROM_CPU_CLIP_RECTANGLE 0x00000188 +#define NV01_IMAGE_FROM_CPU_PATTERN 0x0000018c +#define NV01_IMAGE_FROM_CPU_ROP 0x00000190 +#define NV01_IMAGE_FROM_CPU_BETA1 0x00000194 +#define NV01_IMAGE_FROM_CPU_SURFACE 0x00000198 +#define NV01_IMAGE_FROM_CPU_OPERATION 0x000002fc +#define NV01_IMAGE_FROM_CPU_OPERATION_SRCCOPY_AND 0x00000000 +#define NV01_IMAGE_FROM_CPU_OPERATION_ROP_AND 0x00000001 +#define NV01_IMAGE_FROM_CPU_OPERATION_BLEND_AND 0x00000002 +#define NV01_IMAGE_FROM_CPU_OPERATION_SRCCOPY 0x00000003 +#define NV01_IMAGE_FROM_CPU_OPERATION_SRCCOPY_PREMULT 0x00000004 +#define NV01_IMAGE_FROM_CPU_OPERATION_BLEND_PREMULT 0x00000005 +#define NV01_IMAGE_FROM_CPU_COLOR_FORMAT 0x00000300 +#define NV01_IMAGE_FROM_CPU_COLOR_FORMAT_Y8 0x00000001 +#define NV01_IMAGE_FROM_CPU_COLOR_FORMAT_A1R5G5B5 0x00000002 +#define NV01_IMAGE_FROM_CPU_COLOR_FORMAT_X1R5G5B5 0x00000003 +#define NV01_IMAGE_FROM_CPU_COLOR_FORMAT_A8R8G8B8 0x00000004 +#define NV01_IMAGE_FROM_CPU_COLOR_FORMAT_X8R8G8B8 0x00000005 +#define NV01_IMAGE_FROM_CPU_POINT 0x00000304 +#define NV01_IMAGE_FROM_CPU_POINT_X_SHIFT 0 +#define NV01_IMAGE_FROM_CPU_POINT_X_MASK 0x0000ffff +#define NV01_IMAGE_FROM_CPU_POINT_Y_SHIFT 16 +#define NV01_IMAGE_FROM_CPU_POINT_Y_MASK 0xffff0000 +#define NV01_IMAGE_FROM_CPU_SIZE_OUT 0x00000308 +#define NV01_IMAGE_FROM_CPU_SIZE_OUT_W_SHIFT 0 +#define NV01_IMAGE_FROM_CPU_SIZE_OUT_W_MASK 0x0000ffff +#define NV01_IMAGE_FROM_CPU_SIZE_OUT_H_SHIFT 16 +#define NV01_IMAGE_FROM_CPU_SIZE_OUT_H_MASK 0xffff0000 +#define NV01_IMAGE_FROM_CPU_SIZE_IN 0x0000030c +#define NV01_IMAGE_FROM_CPU_SIZE_IN_W_SHIFT 0 +#define NV01_IMAGE_FROM_CPU_SIZE_IN_W_MASK 0x0000ffff +#define NV01_IMAGE_FROM_CPU_SIZE_IN_H_SHIFT 16 +#define NV01_IMAGE_FROM_CPU_SIZE_IN_H_MASK 0xffff0000 +#define NV01_IMAGE_FROM_CPU_COLOR(x) (0x00000400+((x)*4)) +#define NV01_IMAGE_FROM_CPU_COLOR__SIZE 0x00000020 + + +#define NV01_NULL 0x00000030 + + + +#define NV03_STRETCHED_IMAGE_FROM_CPU 0x00000036 + +#define NV03_STRETCHED_IMAGE_FROM_CPU_NOP 0x00000100 +#define NV03_STRETCHED_IMAGE_FROM_CPU_NOTIFY 0x00000104 +#define NV03_STRETCHED_IMAGE_FROM_CPU_PATCH 0x0000010c +#define NV03_STRETCHED_IMAGE_FROM_CPU_DMA_NOTIFY 0x00000180 +#define NV03_STRETCHED_IMAGE_FROM_CPU_COLOR_KEY 0x00000184 +#define NV03_STRETCHED_IMAGE_FROM_CPU_PATTERN 0x00000188 +#define NV03_STRETCHED_IMAGE_FROM_CPU_ROP 0x0000018c +#define NV03_STRETCHED_IMAGE_FROM_CPU_BETA1 0x00000190 +#define NV03_STRETCHED_IMAGE_FROM_CPU_SURFACE 0x00000194 +#define NV03_STRETCHED_IMAGE_FROM_CPU_OPERATION 0x000002fc +#define NV03_STRETCHED_IMAGE_FROM_CPU_COLOR_FORMAT 0x00000300 +#define NV03_STRETCHED_IMAGE_FROM_CPU_SIZE_IN 0x00000304 +#define NV03_STRETCHED_IMAGE_FROM_CPU_SIZE_IN_W_SHIFT 0 +#define NV03_STRETCHED_IMAGE_FROM_CPU_SIZE_IN_W_MASK 0x0000ffff +#define NV03_STRETCHED_IMAGE_FROM_CPU_SIZE_IN_H_SHIFT 16 +#define NV03_STRETCHED_IMAGE_FROM_CPU_SIZE_IN_H_MASK 0xffff0000 +#define NV03_STRETCHED_IMAGE_FROM_CPU_DX_DU 0x00000308 +#define NV03_STRETCHED_IMAGE_FROM_CPU_DY_DV 0x0000030c +#define NV03_STRETCHED_IMAGE_FROM_CPU_CLIP_POINT 0x00000310 +#define NV03_STRETCHED_IMAGE_FROM_CPU_CLIP_POINT_X_SHIFT 0 +#define NV03_STRETCHED_IMAGE_FROM_CPU_CLIP_POINT_X_MASK 0x0000ffff +#define NV03_STRETCHED_IMAGE_FROM_CPU_CLIP_POINT_Y_SHIFT 16 +#define NV03_STRETCHED_IMAGE_FROM_CPU_CLIP_POINT_Y_MASK 0xffff0000 +#define NV03_STRETCHED_IMAGE_FROM_CPU_CLIP_SIZE 0x00000314 +#define NV03_STRETCHED_IMAGE_FROM_CPU_CLIP_SIZE_W_SHIFT 0 +#define NV03_STRETCHED_IMAGE_FROM_CPU_CLIP_SIZE_W_MASK 0x0000ffff +#define NV03_STRETCHED_IMAGE_FROM_CPU_CLIP_SIZE_H_SHIFT 16 +#define NV03_STRETCHED_IMAGE_FROM_CPU_CLIP_SIZE_H_MASK 0xffff0000 +#define NV03_STRETCHED_IMAGE_FROM_CPU_POINT12D4 0x00000318 +#define NV03_STRETCHED_IMAGE_FROM_CPU_POINT12D4_X_SHIFT 0 +#define NV03_STRETCHED_IMAGE_FROM_CPU_POINT12D4_X_MASK 0x0000ffff +#define NV03_STRETCHED_IMAGE_FROM_CPU_POINT12D4_Y_SHIFT 16 +#define NV03_STRETCHED_IMAGE_FROM_CPU_POINT12D4_Y_MASK 0xffff0000 +#define NV03_STRETCHED_IMAGE_FROM_CPU_COLOR(x) (0x00000400+((x)*4)) +#define NV03_STRETCHED_IMAGE_FROM_CPU_COLOR__SIZE 0x00000020 + + +#define NV03_SCALED_IMAGE_FROM_MEMORY 0x00000037 + +#define NV03_SCALED_IMAGE_FROM_MEMORY_NOP 0x00000100 +#define NV03_SCALED_IMAGE_FROM_MEMORY_NOTIFY 0x00000104 +#define NV03_SCALED_IMAGE_FROM_MEMORY_DMA_NOTIFY 0x00000180 +#define NV03_SCALED_IMAGE_FROM_MEMORY_DMA_IMAGE 0x00000184 +#define NV03_SCALED_IMAGE_FROM_MEMORY_PATTERN 0x00000188 +#define NV03_SCALED_IMAGE_FROM_MEMORY_ROP 0x0000018c +#define NV03_SCALED_IMAGE_FROM_MEMORY_BETA1 0x00000190 +#define NV03_SCALED_IMAGE_FROM_MEMORY_SURFACE 0x00000194 +#define NV03_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT 0x00000300 +#define NV03_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_A1R5G5B5 0x00000001 +#define NV03_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_X1R5G5B5 0x00000002 +#define NV03_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_A8R8G8B8 0x00000003 +#define NV03_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_X8R8G8B8 0x00000004 +#define NV03_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_V8YB8U8YA8 0x00000005 +#define NV03_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_YB8V8YA8U8 0x00000006 +#define NV03_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_R5G6B5 0x00000007 +#define NV03_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_Y8 0x00000008 +#define NV03_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_AY8 0x00000009 +#define NV03_SCALED_IMAGE_FROM_MEMORY_OPERATION 0x00000304 +#define NV03_SCALED_IMAGE_FROM_MEMORY_OPERATION_SRCCOPY_AND 0x00000000 +#define NV03_SCALED_IMAGE_FROM_MEMORY_OPERATION_ROP_AND 0x00000001 +#define NV03_SCALED_IMAGE_FROM_MEMORY_OPERATION_BLEND_AND 0x00000002 +#define NV03_SCALED_IMAGE_FROM_MEMORY_OPERATION_SRCCOPY 0x00000003 +#define NV03_SCALED_IMAGE_FROM_MEMORY_OPERATION_SRCCOPY_PREMULT 0x00000004 +#define NV03_SCALED_IMAGE_FROM_MEMORY_OPERATION_BLEND_PREMULT 0x00000005 +#define NV03_SCALED_IMAGE_FROM_MEMORY_CLIP_POINT 0x00000308 +#define NV03_SCALED_IMAGE_FROM_MEMORY_CLIP_POINT_X_SHIFT 0 +#define NV03_SCALED_IMAGE_FROM_MEMORY_CLIP_POINT_X_MASK 0x0000ffff +#define NV03_SCALED_IMAGE_FROM_MEMORY_CLIP_POINT_Y_SHIFT 16 +#define NV03_SCALED_IMAGE_FROM_MEMORY_CLIP_POINT_Y_MASK 0xffff0000 +#define NV03_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE 0x0000030c +#define NV03_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE_W_SHIFT 0 +#define NV03_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE_W_MASK 0x0000ffff +#define NV03_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE_H_SHIFT 16 +#define NV03_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE_H_MASK 0xffff0000 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_OUT_POINT 0x00000310 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_OUT_POINT_X_SHIFT 0 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_OUT_POINT_X_MASK 0x0000ffff +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_OUT_POINT_Y_SHIFT 16 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_OUT_POINT_Y_MASK 0xffff0000 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_OUT_SIZE 0x00000314 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_OUT_SIZE_W_SHIFT 0 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_OUT_SIZE_W_MASK 0x0000ffff +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_OUT_SIZE_H_SHIFT 16 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_OUT_SIZE_H_MASK 0xffff0000 +#define NV03_SCALED_IMAGE_FROM_MEMORY_DELTA_DU_DX 0x00000318 +#define NV03_SCALED_IMAGE_FROM_MEMORY_DELTA_DV_DY 0x0000031c +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_SIZE 0x00000400 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_SIZE_W_SHIFT 0 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_SIZE_W_MASK 0x0000ffff +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_SIZE_H_SHIFT 16 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_SIZE_H_MASK 0xffff0000 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_FORMAT 0x00000404 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_FORMAT_PITCH_SHIFT 0 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_FORMAT_PITCH_MASK 0x0000ffff +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_FORMAT_ORIGIN_SHIFT 16 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_FORMAT_ORIGIN_MASK 0x00ff0000 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_FORMAT_ORIGIN_CENTER 0x00010000 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_FORMAT_ORIGIN_CORNER 0x00020000 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_FORMAT_INTERPOLATOR_SHIFT 24 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_FORMAT_INTERPOLATOR_MASK 0xff000000 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_OFFSET 0x00000408 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_POINT 0x0000040c +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_POINT_U_SHIFT 0 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_POINT_U_MASK 0x0000ffff +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_POINT_V_SHIFT 16 +#define NV03_SCALED_IMAGE_FROM_MEMORY_IMAGE_IN_POINT_V_MASK 0xffff0000 + + +#define NV04_DVD_SUBPICTURE 0x00000038 + +#define NV04_DVD_SUBPICTURE_NOP 0x00000100 +#define NV04_DVD_SUBPICTURE_NOTIFY 0x00000104 +#define NV04_DVD_SUBPICTURE_WAIT_FOR_IDLE 0x00000108 +#define NV04_DVD_SUBPICTURE_DMA_NOTIFY 0x00000180 +#define NV04_DVD_SUBPICTURE_DMA_OVERLAY 0x00000184 +#define NV04_DVD_SUBPICTURE_DMA_IMAGEIN 0x00000188 +#define NV04_DVD_SUBPICTURE_DMA_IMAGEOUT 0x0000018c +#define NV04_DVD_SUBPICTURE_IMAGEOUT_POINT 0x00000300 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_POINT_X_SHIFT 0 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_POINT_X_MASK 0x0000ffff +#define NV04_DVD_SUBPICTURE_IMAGEOUT_POINT_Y_SHIFT 16 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_POINT_Y_MASK 0xffff0000 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_SIZE 0x00000304 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_SIZE_W_SHIFT 0 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_SIZE_W_MASK 0x0000ffff +#define NV04_DVD_SUBPICTURE_IMAGEOUT_SIZE_H_SHIFT 16 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_SIZE_H_MASK 0xffff0000 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_FORMAT 0x00000308 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_FORMAT_PITCH_SHIFT 0 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_FORMAT_PITCH_MASK 0x0000ffff +#define NV04_DVD_SUBPICTURE_IMAGEOUT_FORMAT_COLOR_SHIFT 16 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_FORMAT_COLOR_MASK 0xffff0000 +#define NV04_DVD_SUBPICTURE_IMAGEOUT_OFFSET 0x0000030c +#define NV04_DVD_SUBPICTURE_IMAGEIN_DELTA_DU_DX 0x00000310 +#define NV04_DVD_SUBPICTURE_IMAGEIN_DELTA_DV_DY 0x00000314 +#define NV04_DVD_SUBPICTURE_IMAGEIN_SIZE 0x00000318 +#define NV04_DVD_SUBPICTURE_IMAGEIN_SIZE_W_SHIFT 0 +#define NV04_DVD_SUBPICTURE_IMAGEIN_SIZE_W_MASK 0x0000ffff +#define NV04_DVD_SUBPICTURE_IMAGEIN_SIZE_H_SHIFT 16 +#define NV04_DVD_SUBPICTURE_IMAGEIN_SIZE_H_MASK 0xffff0000 +#define NV04_DVD_SUBPICTURE_IMAGEIN_FORMAT 0x0000031c +#define NV04_DVD_SUBPICTURE_IMAGEIN_FORMAT_PITCH_SHIFT 0 +#define NV04_DVD_SUBPICTURE_IMAGEIN_FORMAT_PITCH_MASK 0x0000ffff +#define NV04_DVD_SUBPICTURE_IMAGEIN_FORMAT_COLOR_SHIFT 16 +#define NV04_DVD_SUBPICTURE_IMAGEIN_FORMAT_COLOR_MASK 0xffff0000 +#define NV04_DVD_SUBPICTURE_IMAGEIN_OFFSET 0x00000320 +#define NV04_DVD_SUBPICTURE_IMAGEIN_POINT 0x00000324 +#define NV04_DVD_SUBPICTURE_IMAGEIN_POINT_U_SHIFT 0 +#define NV04_DVD_SUBPICTURE_IMAGEIN_POINT_U_MASK 0x0000ffff +#define NV04_DVD_SUBPICTURE_IMAGEIN_POINT_V_SHIFT 16 +#define NV04_DVD_SUBPICTURE_IMAGEIN_POINT_V_MASK 0xffff0000 +#define NV04_DVD_SUBPICTURE_OVERLAY_DELTA_DU_DX 0x00000328 +#define NV04_DVD_SUBPICTURE_OVERLAY_DELTA_DV_DY 0x0000032c +#define NV04_DVD_SUBPICTURE_OVERLAY_SIZE 0x00000330 +#define NV04_DVD_SUBPICTURE_OVERLAY_SIZE_W_SHIFT 0 +#define NV04_DVD_SUBPICTURE_OVERLAY_SIZE_W_MASK 0x0000ffff +#define NV04_DVD_SUBPICTURE_OVERLAY_SIZE_H_SHIFT 16 +#define NV04_DVD_SUBPICTURE_OVERLAY_SIZE_H_MASK 0xffff0000 +#define NV04_DVD_SUBPICTURE_OVERLAY_FORMAT 0x00000334 +#define NV04_DVD_SUBPICTURE_OVERLAY_FORMAT_PITCH_SHIFT 0 +#define NV04_DVD_SUBPICTURE_OVERLAY_FORMAT_PITCH_MASK 0x0000ffff +#define NV04_DVD_SUBPICTURE_OVERLAY_FORMAT_COLOR_SHIFT 16 +#define NV04_DVD_SUBPICTURE_OVERLAY_FORMAT_COLOR_MASK 0xffff0000 +#define NV04_DVD_SUBPICTURE_OVERLAY_OFFSET 0x00000338 +#define NV04_DVD_SUBPICTURE_OVERLAY_POINT 0x0000033c +#define NV04_DVD_SUBPICTURE_OVERLAY_POINT_U_SHIFT 0 +#define NV04_DVD_SUBPICTURE_OVERLAY_POINT_U_MASK 0x0000ffff +#define NV04_DVD_SUBPICTURE_OVERLAY_POINT_V_SHIFT 16 +#define NV04_DVD_SUBPICTURE_OVERLAY_POINT_V_MASK 0xffff0000 + + +#define NV04_MEMORY_TO_MEMORY_FORMAT 0x00000039 + +#define NV04_MEMORY_TO_MEMORY_FORMAT_NOP 0x00000100 +#define NV04_MEMORY_TO_MEMORY_FORMAT_NOTIFY 0x00000104 +#define NV04_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY 0x00000180 +#define NV04_MEMORY_TO_MEMORY_FORMAT_DMA_BUFFER_IN 0x00000184 +#define NV04_MEMORY_TO_MEMORY_FORMAT_DMA_BUFFER_OUT 0x00000188 +#define NV04_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN 0x0000030c +#define NV04_MEMORY_TO_MEMORY_FORMAT_OFFSET_OUT 0x00000310 +#define NV04_MEMORY_TO_MEMORY_FORMAT_PITCH_IN 0x00000314 +#define NV04_MEMORY_TO_MEMORY_FORMAT_PITCH_OUT 0x00000318 +#define NV04_MEMORY_TO_MEMORY_FORMAT_LINE_LENGTH_IN 0x0000031c +#define NV04_MEMORY_TO_MEMORY_FORMAT_LINE_COUNT 0x00000320 +#define NV04_MEMORY_TO_MEMORY_FORMAT_FORMAT 0x00000324 +#define NV04_MEMORY_TO_MEMORY_FORMAT_FORMAT_INPUT_INC_SHIFT 0 +#define NV04_MEMORY_TO_MEMORY_FORMAT_FORMAT_INPUT_INC_MASK 0x0000000f +#define NV04_MEMORY_TO_MEMORY_FORMAT_FORMAT_OUTPUT_INC_SHIFT 8 +#define NV04_MEMORY_TO_MEMORY_FORMAT_FORMAT_OUTPUT_INC_MASK 0x00000f00 +#define NV04_MEMORY_TO_MEMORY_FORMAT_BUF_NOTIFY 0x00000328 + + +#define NV01_MEMORY_LOCAL_BANKED 0x0000003d + + + +#define NV01_MAPPING_SYSTEM 0x0000003e + + + +#define NV03_MEMORY_LOCAL_CURSOR 0x0000003f + + + +#define NV01_MEMORY_LOCAL_LINEAR 0x00000040 + + + +#define NV01_MAPPING_LOCAL 0x00000041 + + + +#define NV04_CONTEXT_SURFACES_2D 0x00000042 + +#define NV04_CONTEXT_SURFACES_2D_NOP 0x00000100 +#define NV04_CONTEXT_SURFACES_2D_NOTIFY 0x00000104 +#define NV04_CONTEXT_SURFACES_2D_PM_TRIGGER 0x00000140 +#define NV04_CONTEXT_SURFACES_2D_DMA_NOTIFY 0x00000180 +#define NV04_CONTEXT_SURFACES_2D_DMA_IMAGE_SOURCE 0x00000184 +#define NV04_CONTEXT_SURFACES_2D_DMA_IMAGE_DESTIN 0x00000188 +#define NV04_CONTEXT_SURFACES_2D_FORMAT 0x00000300 +#define NV04_CONTEXT_SURFACES_2D_FORMAT_Y8 0x00000001 +#define NV04_CONTEXT_SURFACES_2D_FORMAT_X1R5G5B5_Z1R5G5B5 0x00000002 +#define NV04_CONTEXT_SURFACES_2D_FORMAT_X1R5G5B5_X1R5G5B5 0x00000003 +#define NV04_CONTEXT_SURFACES_2D_FORMAT_R5G6B5 0x00000004 +#define NV04_CONTEXT_SURFACES_2D_FORMAT_Y16 0x00000005 +#define NV04_CONTEXT_SURFACES_2D_FORMAT_X8R8G8B8_Z8R8G8B8 0x00000006 +#define NV04_CONTEXT_SURFACES_2D_FORMAT_X8R8G8B8_X8R8G8B8 0x00000007 +#define NV04_CONTEXT_SURFACES_2D_FORMAT_X1A7R8G8B8_Z1A7R8G8B8 0x00000008 +#define NV04_CONTEXT_SURFACES_2D_FORMAT_X1A7R8G8B8_X1A7R8G8B8 0x00000009 +#define NV04_CONTEXT_SURFACES_2D_FORMAT_A8R8G8B8 0x0000000a +#define NV04_CONTEXT_SURFACES_2D_FORMAT_Y32 0x0000000b +#define NV04_CONTEXT_SURFACES_2D_PITCH 0x00000304 +#define NV04_CONTEXT_SURFACES_2D_PITCH_SOURCE_SHIFT 0 +#define NV04_CONTEXT_SURFACES_2D_PITCH_SOURCE_MASK 0x0000ffff +#define NV04_CONTEXT_SURFACES_2D_PITCH_DESTIN_SHIFT 16 +#define NV04_CONTEXT_SURFACES_2D_PITCH_DESTIN_MASK 0xffff0000 +#define NV04_CONTEXT_SURFACES_2D_OFFSET_SOURCE 0x00000308 +#define NV04_CONTEXT_SURFACES_2D_OFFSET_DESTIN 0x0000030c + + +#define NV03_CONTEXT_ROP 0x00000043 + +#define NV03_CONTEXT_ROP_NOP 0x00000100 +#define NV03_CONTEXT_ROP_NOTIFY 0x00000104 +#define NV03_CONTEXT_ROP_DMA_NOTIFY 0x00000180 +#define NV03_CONTEXT_ROP_ROP 0x00000300 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_SHIFT 0 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_MASK 0x0000000f +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_CLEAR 0x00000000 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_NOR 0x00000001 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_AND_INVERTED 0x00000002 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_COPY_INVERTED 0x00000003 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_AND_REVERSE 0x00000004 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_INVERT 0x00000005 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_XOR 0x00000006 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_NAND 0x00000007 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_AND 0x00000008 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_EQUI 0x00000009 +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_NOOP 0x0000000a +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_OR_INVERTED 0x0000000b +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_COPY 0x0000000c +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_OR_REVERSE 0x0000000d +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_OR 0x0000000e +#define NV03_CONTEXT_ROP_ROP_DST_LOGIC_OP_SET 0x0000000f +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_SHIFT 4 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_MASK 0x000000f0 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_CLEAR 0x00000000 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_NOR 0x00000010 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_AND_INVERTED 0x00000020 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_COPY_INVERTED 0x00000030 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_AND_REVERSE 0x00000040 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_INVERT 0x00000050 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_XOR 0x00000060 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_NAND 0x00000070 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_AND 0x00000080 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_EQUI 0x00000090 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_NOOP 0x000000a0 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_OR_INVERTED 0x000000b0 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_COPY 0x000000c0 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_OR_REVERSE 0x000000d0 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_OR 0x000000e0 +#define NV03_CONTEXT_ROP_ROP_SRC_LOGIC_OP_SET 0x000000f0 + + +#define NV04_IMAGE_PATTERN 0x00000044 + +#define NV04_IMAGE_PATTERN_NOP 0x00000100 +#define NV04_IMAGE_PATTERN_NOTIFY 0x00000104 +#define NV04_IMAGE_PATTERN_DMA_NOTIFY 0x00000180 +#define NV04_IMAGE_PATTERN_COLOR_FORMAT 0x00000300 +#define NV04_IMAGE_PATTERN_COLOR_FORMAT_A16R5G6B5 0x00000001 +#define NV04_IMAGE_PATTERN_COLOR_FORMAT_X16A1R5G5B5 0x00000002 +#define NV04_IMAGE_PATTERN_COLOR_FORMAT_A8R8G8B8 0x00000003 +#define NV04_IMAGE_PATTERN_MONOCHROME_FORMAT 0x00000304 +#define NV04_IMAGE_PATTERN_MONOCHROME_FORMAT_CGA6 0x00000001 +#define NV04_IMAGE_PATTERN_MONOCHROME_FORMAT_LE 0x00000002 +#define NV04_IMAGE_PATTERN_MONOCHROME_SHAPE 0x00000308 +#define NV04_IMAGE_PATTERN_MONOCHROME_SHAPE_8X8 0x00000000 +#define NV04_IMAGE_PATTERN_MONOCHROME_SHAPE_64X1 0x00000001 +#define NV04_IMAGE_PATTERN_MONOCHROME_SHAPE_1X64 0x00000002 +#define NV04_IMAGE_PATTERN_PATTERN_SELECT 0x0000030c +#define NV04_IMAGE_PATTERN_PATTERN_SELECT_MONO 0x00000001 +#define NV04_IMAGE_PATTERN_PATTERN_SELECT_COLOR 0x00000002 +#define NV04_IMAGE_PATTERN_MONOCHROME_COLOR0 0x00000310 +#define NV04_IMAGE_PATTERN_MONOCHROME_COLOR1 0x00000314 +#define NV04_IMAGE_PATTERN_MONOCHROME_PATTERN0 0x00000318 +#define NV04_IMAGE_PATTERN_MONOCHROME_PATTERN1 0x0000031c +#define NV04_IMAGE_PATTERN_PATTERN_Y8(x) (0x00000400+((x)*4)) +#define NV04_IMAGE_PATTERN_PATTERN_Y8__SIZE 0x00000010 +#define NV04_IMAGE_PATTERN_PATTERN_Y8_Y0_SHIFT 0 +#define NV04_IMAGE_PATTERN_PATTERN_Y8_Y0_MASK 0x000000ff +#define NV04_IMAGE_PATTERN_PATTERN_Y8_Y1_SHIFT 8 +#define NV04_IMAGE_PATTERN_PATTERN_Y8_Y1_MASK 0x0000ff00 +#define NV04_IMAGE_PATTERN_PATTERN_Y8_Y2_SHIFT 16 +#define NV04_IMAGE_PATTERN_PATTERN_Y8_Y2_MASK 0x00ff0000 +#define NV04_IMAGE_PATTERN_PATTERN_Y8_Y3_SHIFT 24 +#define NV04_IMAGE_PATTERN_PATTERN_Y8_Y3_MASK 0xff000000 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5(x) (0x00000500+((x)*4)) +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5__SIZE 0x00000020 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_B0_SHIFT 0 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_B0_MASK 0x0000001f +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_G0_SHIFT 5 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_G0_MASK 0x000007e0 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_R0_SHIFT 11 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_R0_MASK 0x0000f800 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_B1_SHIFT 16 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_B1_MASK 0x001f0000 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_G1_SHIFT 21 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_G1_MASK 0x07e00000 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_R1_SHIFT 27 +#define NV04_IMAGE_PATTERN_PATTERN_R5G6B5_R1_MASK 0xf8000000 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5(x) (0x00000600+((x)*4)) +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5__SIZE 0x00000020 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_B0_SHIFT 0 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_B0_MASK 0x0000001f +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_G0_SHIFT 5 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_G0_MASK 0x000003e0 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_R0_SHIFT 10 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_R0_MASK 0x00007c00 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_B1_SHIFT 16 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_B1_MASK 0x001f0000 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_G1_SHIFT 21 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_G1_MASK 0x03e00000 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_R1_SHIFT 26 +#define NV04_IMAGE_PATTERN_PATTERN_X1R5G5B5_R1_MASK 0x7c000000 +#define NV04_IMAGE_PATTERN_PATTERN_X8R8G8B8(x) (0x00000700+((x)*4)) +#define NV04_IMAGE_PATTERN_PATTERN_X8R8G8B8__SIZE 0x00000040 +#define NV04_IMAGE_PATTERN_PATTERN_X8R8G8B8_B_SHIFT 0 +#define NV04_IMAGE_PATTERN_PATTERN_X8R8G8B8_B_MASK 0x000000ff +#define NV04_IMAGE_PATTERN_PATTERN_X8R8G8B8_G_SHIFT 8 +#define NV04_IMAGE_PATTERN_PATTERN_X8R8G8B8_G_MASK 0x0000ff00 +#define NV04_IMAGE_PATTERN_PATTERN_X8R8G8B8_R_SHIFT 16 +#define NV04_IMAGE_PATTERN_PATTERN_X8R8G8B8_R_MASK 0x00ff0000 + + +#define NV03_VIDEO_LUT_CURSOR_DAC 0x00000046 + +#define NV03_VIDEO_LUT_CURSOR_DAC_SYNCHRONIZE 0x00000100 +#define NV03_VIDEO_LUT_CURSOR_DAC_STOP_IMAGE 0x00000104 +#define NV03_VIDEO_LUT_CURSOR_DAC_STOP_CURSOR 0x00000108 +#define NV03_VIDEO_LUT_CURSOR_DAC_STOP_DAC 0x0000010c +#define NV03_VIDEO_LUT_CURSOR_DAC_DMA_NOTIFY 0x00000180 +#define NV03_VIDEO_LUT_CURSOR_DAC_DMA_IMAGE(x) (0x00000184+((x)*4)) +#define NV03_VIDEO_LUT_CURSOR_DAC_DMA_IMAGE__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_DMA_LUT(x) (0x0000018c+((x)*4)) +#define NV03_VIDEO_LUT_CURSOR_DAC_DMA_LUT__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_DMA_CURSOR(x) (0x00000194+((x)*4)) +#define NV03_VIDEO_LUT_CURSOR_DAC_DMA_CURSOR__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_GET 0x000002fc +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_IMAGE_OFFSET(x) (0x00000300+((x)*8)) +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_IMAGE_OFFSET__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_IMAGE_FORMAT(x) (0x00000304+((x)*8)) +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_IMAGE_FORMAT__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_IMAGE_FORMAT_PITCH_SHIFT 0 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_IMAGE_FORMAT_PITCH_MASK 0x0000ffff +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_IMAGE_FORMAT_COLOR_SHIFT 16 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_IMAGE_FORMAT_COLOR_MASK 0x0fff0000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_IMAGE_FORMAT_NOTIFY_SHIFT 28 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_IMAGE_FORMAT_NOTIFY_MASK 0xf0000000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_OFFSET(x) (0x00000340+((x)*12)) +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_OFFSET__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT(x) (0x00000344+((x)*12)) +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT_X_SHIFT 0 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT_X_MASK 0x0000ffff +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT_Y_SHIFT 16 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT_Y_MASK 0xffff0000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_FORMAT(x) (0x00000348+((x)*12)) +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_FORMAT__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT_A 0x00000358 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT_A_X_SHIFT 0 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT_A_X_MASK 0x0000ffff +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT_A_Y_SHIFT 16 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_CURSOR_POINT_OUT_A_Y_MASK 0xffff0000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_IMAGE_SIZE(x) (0x00000380+((x)*16)) +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_IMAGE_SIZE__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_IMAGE_SIZE_W_SHIFT 0 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_IMAGE_SIZE_W_MASK 0x0000ffff +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_IMAGE_SIZE_H_SHIFT 16 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_IMAGE_SIZE_H_MASK 0xffff0000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_HSYNC(x) (0x00000384+((x)*16)) +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_HSYNC__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_HSYNC_START_SHIFT 0 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_HSYNC_START_MASK 0x0000ffff +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_HSYNC_WIDTH_SHIFT 16 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_HSYNC_WIDTH_MASK 0x0fff0000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_HSYNC_POLARITY_SHIFT 28 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_HSYNC_POLARITY_MASK 0xf0000000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_VSYNC(x) (0x00000388+((x)*16)) +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_VSYNC__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_VSYNC_START_SHIFT 0 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_VSYNC_START_MASK 0x0000ffff +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_VSYNC_WIDTH_SHIFT 16 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_VSYNC_WIDTH_MASK 0x0fff0000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_VSYNC_POLARITY_SHIFT 28 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_VSYNC_POLARITY_MASK 0xf0000000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_TOTAL_SIZE(x) (0x0000038c+((x)*16)) +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_TOTAL_SIZE__SIZE 0x00000002 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_TOTAL_SIZE_WIDTH_SHIFT 0 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_TOTAL_SIZE_WIDTH_MASK 0x0000ffff +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_TOTAL_SIZE_HEIGHT_SHIFT 16 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_TOTAL_SIZE_HEIGHT_MASK 0x0fff0000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_TOTAL_SIZE_NOTIFY_SHIFT 28 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_DAC_TOTAL_SIZE_NOTIFY_MASK 0xf0000000 +#define NV03_VIDEO_LUT_CURSOR_DAC_SET_PIXEL_CLOCK 0x000003a0 + + +#define NV03_DX3_TEXTURED_TRIANGLE 0x00000048 + +#define NV03_DX3_TEXTURED_TRIANGLE_NOP 0x00000100 +#define NV03_DX3_TEXTURED_TRIANGLE_NOTIFY 0x00000104 +#define NV03_DX3_TEXTURED_TRIANGLE_PATCH 0x0000010c +#define NV03_DX3_TEXTURED_TRIANGLE_DMA_NOTIFY 0x00000180 +#define NV03_DX3_TEXTURED_TRIANGLE_DMA_TEXTURE 0x00000184 +#define NV03_DX3_TEXTURED_TRIANGLE_CLIP_RECTANGLE 0x00000188 +#define NV03_DX3_TEXTURED_TRIANGLE_SURFACE 0x0000018c +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_OFFSET 0x00000304 +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT 0x00000308 +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT_COLOR_KEY_MASK_SHIFT 0 +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT_COLOR_KEY_MASK_MASK 0x0000ffff +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT_COLOR_KEY_ENABLE_SHIFT 16 +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT_COLOR_KEY_ENABLE_MASK 0x000f0000 +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT_COLOR_SHIFT 20 +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT_COLOR_MASK 0x00f00000 +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT_SIZE_MIN_SHIFT 24 +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT_SIZE_MIN_MASK 0x0f000000 +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT_SIZE_MAX_SHIFT 28 +#define NV03_DX3_TEXTURED_TRIANGLE_TEXTURE_FORMAT_SIZE_MAX_MASK 0xf0000000 +#define NV03_DX3_TEXTURED_TRIANGLE_FILTER 0x0000030c +#define NV03_DX3_TEXTURED_TRIANGLE_FILTER_SPREAD_X_SHIFT 0 +#define NV03_DX3_TEXTURED_TRIANGLE_FILTER_SPREAD_X_MASK 0x0000001f +#define NV03_DX3_TEXTURED_TRIANGLE_FILTER_SPREAD_Y_SHIFT 8 +#define NV03_DX3_TEXTURED_TRIANGLE_FILTER_SPREAD_Y_MASK 0x00001f00 +#define NV03_DX3_TEXTURED_TRIANGLE_FILTER_SIZE_ADJUST_SHIFT 16 +#define NV03_DX3_TEXTURED_TRIANGLE_FILTER_SIZE_ADJUST_MASK 0x00ff0000 +#define NV03_DX3_TEXTURED_TRIANGLE_FOG_COLOR 0x00000310 +#define NV03_DX3_TEXTURED_TRIANGLE_FOG_COLOR_B_SHIFT 0 +#define NV03_DX3_TEXTURED_TRIANGLE_FOG_COLOR_B_MASK 0x000000ff +#define NV03_DX3_TEXTURED_TRIANGLE_FOG_COLOR_G_SHIFT 8 +#define NV03_DX3_TEXTURED_TRIANGLE_FOG_COLOR_G_MASK 0x0000ff00 +#define NV03_DX3_TEXTURED_TRIANGLE_FOG_COLOR_R_SHIFT 16 +#define NV03_DX3_TEXTURED_TRIANGLE_FOG_COLOR_R_MASK 0x00ff0000 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT 0x00000314 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_INTERPOLATOR_SHIFT 0 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_INTERPOLATOR_MASK 0x0000000f +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_WRAP_U_SHIFT 4 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_WRAP_U_MASK 0x00000030 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_WRAP_V_SHIFT 6 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_WRAP_V_MASK 0x000000c0 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_SOURCE_COLOR_SHIFT 8 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_SOURCE_COLOR_MASK 0x00000f00 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_CULLING_SHIFT 12 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_CULLING_MASK 0x00007000 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_Z_PERSPECTIVE_ENABLE (1 << 15) +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_Z_FUNC_SHIFT 16 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_Z_FUNC_MASK 0x000f0000 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_Z_WRITE_ENABLE_SHIFT 20 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_Z_WRITE_ENABLE_MASK 0x00f00000 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_COLOR_WRITE_ENABLE_SHIFT 24 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_COLOR_WRITE_ENABLE_MASK 0x07000000 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_ROP_SHIFT 27 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_ROP_MASK 0x18000000 +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_BETA (1 << 29) +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_DST_BLEND (1 << 30) +#define NV03_DX3_TEXTURED_TRIANGLE_CONTROL_OUT_SRC_BLEND (1 << 31) +#define NV03_DX3_TEXTURED_TRIANGLE_ALPHA_CONTROL 0x00000318 +#define NV03_DX3_TEXTURED_TRIANGLE_ALPHA_CONTROL_ALPHA_REF_SHIFT 0 +#define NV03_DX3_TEXTURED_TRIANGLE_ALPHA_CONTROL_ALPHA_REF_MASK 0x000000ff +#define NV03_DX3_TEXTURED_TRIANGLE_ALPHA_CONTROL_ALPHA_FUNC_SHIFT 8 +#define NV03_DX3_TEXTURED_TRIANGLE_ALPHA_CONTROL_ALPHA_FUNC_MASK 0xffffff00 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR(x) (0x00001000+((x)*32)) +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR__SIZE 0x00000040 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I0_SHIFT 0 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I0_MASK 0x0000000f +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I1_SHIFT 4 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I1_MASK 0x000000f0 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I2_SHIFT 8 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I2_MASK 0x00000f00 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I3_SHIFT 12 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I3_MASK 0x0000f000 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I4_SHIFT 16 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I4_MASK 0x000f0000 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I5_SHIFT 20 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_I5_MASK 0x00f00000 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_FOG_SHIFT 24 +#define NV03_DX3_TEXTURED_TRIANGLE_SPECULAR_FOG_MASK 0xff000000 +#define NV03_DX3_TEXTURED_TRIANGLE_COLOR(x) (0x00001004+((x)*32)) +#define NV03_DX3_TEXTURED_TRIANGLE_COLOR__SIZE 0x00000040 +#define NV03_DX3_TEXTURED_TRIANGLE_X(x) (0x00001008+((x)*32)) +#define NV03_DX3_TEXTURED_TRIANGLE_X__SIZE 0x00000040 +#define NV03_DX3_TEXTURED_TRIANGLE_Y(x) (0x0000100c+((x)*32)) +#define NV03_DX3_TEXTURED_TRIANGLE_Y__SIZE 0x00000040 +#define NV03_DX3_TEXTURED_TRIANGLE_Z(x) (0x00001010+((x)*32)) +#define NV03_DX3_TEXTURED_TRIANGLE_Z__SIZE 0x00000040 +#define NV03_DX3_TEXTURED_TRIANGLE_M(x) (0x00001014+((x)*32)) +#define NV03_DX3_TEXTURED_TRIANGLE_M__SIZE 0x00000040 +#define NV03_DX3_TEXTURED_TRIANGLE_U(x) (0x00001018+((x)*32)) +#define NV03_DX3_TEXTURED_TRIANGLE_U__SIZE 0x00000040 +#define NV03_DX3_TEXTURED_TRIANGLE_V(x) (0x0000101c+((x)*32)) +#define NV03_DX3_TEXTURED_TRIANGLE_V__SIZE 0x00000040 + + +#define NV04_GDI_RECTANGLE_TEXT 0x0000004a + +#define NV04_GDI_RECTANGLE_TEXT_NOP 0x00000100 +#define NV04_GDI_RECTANGLE_TEXT_NOTIFY 0x00000104 +#define NV04_GDI_RECTANGLE_TEXT_PATCH 0x0000010c +#define NV04_GDI_RECTANGLE_TEXT_PM_TRIGGER 0x00000140 +#define NV04_GDI_RECTANGLE_TEXT_DMA_NOTIFY 0x00000180 +#define NV04_GDI_RECTANGLE_TEXT_DMA_FONTS 0x00000184 +#define NV04_GDI_RECTANGLE_TEXT_PATTERN 0x00000188 +#define NV04_GDI_RECTANGLE_TEXT_ROP 0x0000018c +#define NV04_GDI_RECTANGLE_TEXT_BETA1 0x00000190 +#define NV04_GDI_RECTANGLE_TEXT_BETA4 0x00000194 +#define NV04_GDI_RECTANGLE_TEXT_SURFACE 0x00000198 +#define NV04_GDI_RECTANGLE_TEXT_OPERATION 0x000002fc +#define NV04_GDI_RECTANGLE_TEXT_OPERATION_SRCCOPY_AND 0x00000000 +#define NV04_GDI_RECTANGLE_TEXT_OPERATION_ROP_AND 0x00000001 +#define NV04_GDI_RECTANGLE_TEXT_OPERATION_BLEND_AND 0x00000002 +#define NV04_GDI_RECTANGLE_TEXT_OPERATION_SRCCOPY 0x00000003 +#define NV04_GDI_RECTANGLE_TEXT_OPERATION_SRCCOPY_PREMULT 0x00000004 +#define NV04_GDI_RECTANGLE_TEXT_OPERATION_BLEND_PREMULT 0x00000005 +#define NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT 0x00000300 +#define NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A16R5G6B5 0x00000001 +#define NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_X16A1R5G5B5 0x00000002 +#define NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A8R8G8B8 0x00000003 +#define NV04_GDI_RECTANGLE_TEXT_MONOCHROME_FORMAT 0x00000304 +#define NV04_GDI_RECTANGLE_TEXT_MONOCHROME_FORMAT_CGA6 0x00000001 +#define NV04_GDI_RECTANGLE_TEXT_MONOCHROME_FORMAT_LE 0x00000002 +#define NV04_GDI_RECTANGLE_TEXT_COLOR1_A 0x000003fc +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT(x) (0x00000400+((x)*8)) +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT__SIZE 0x00000020 +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT_Y_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT_Y_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT_X_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT_X_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE(x) (0x00000404+((x)*8)) +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE__SIZE 0x00000020 +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE_H_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE_H_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE_W_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE_W_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_B_POINT0 0x000005f4 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_B_POINT0_L_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_B_POINT0_L_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIP_B_POINT0_T_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_B_POINT0_T_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_B_POINT1 0x000005f8 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_B_POINT1_R_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_B_POINT1_R_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIP_B_POINT1_B_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_B_POINT1_B_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_COLOR1_B 0x000005fc +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0(x) (0x00000600+((x)*8)) +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0__SIZE 0x00000020 +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0_L_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0_L_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0_T_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0_T_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1(x) (0x00000604+((x)*8)) +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1__SIZE 0x00000020 +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1_R_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1_R_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1_B_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1_B_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_C_POINT0 0x000007ec +#define NV04_GDI_RECTANGLE_TEXT_CLIP_C_POINT0_L_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_C_POINT0_L_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIP_C_POINT0_T_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_C_POINT0_T_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_C_POINT1 0x000007f0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_C_POINT1_R_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_C_POINT1_R_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIP_C_POINT1_B_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_C_POINT1_B_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_COLOR1_C 0x000007f4 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_C 0x000007f8 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_C_W_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_C_W_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_SIZE_C_H_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_C_H_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_POINT_C 0x000007fc +#define NV04_GDI_RECTANGLE_TEXT_POINT_C_X_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_POINT_C_X_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_POINT_C_Y_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_POINT_C_Y_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_MONOCHROME_COLOR1_C(x) (0x00000800+((x)*4)) +#define NV04_GDI_RECTANGLE_TEXT_MONOCHROME_COLOR1_C__SIZE 0x00000080 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_E_POINT0 0x00000be4 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_E_POINT0_L_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_E_POINT0_L_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIP_E_POINT0_T_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_E_POINT0_T_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_E_POINT1 0x00000be8 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_E_POINT1_R_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_E_POINT1_R_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIP_E_POINT1_B_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_E_POINT1_B_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_COLOR0_E 0x00000bec +#define NV04_GDI_RECTANGLE_TEXT_COLOR1_E 0x00000bf0 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_IN_E 0x00000bf4 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_IN_E_W_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_IN_E_W_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_SIZE_IN_E_H_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_IN_E_H_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_OUT_E 0x00000bf8 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_OUT_E_W_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_OUT_E_W_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_SIZE_OUT_E_H_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_SIZE_OUT_E_H_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_POINT_E 0x00000bfc +#define NV04_GDI_RECTANGLE_TEXT_POINT_E_X_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_POINT_E_X_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_POINT_E_Y_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_POINT_E_Y_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_MONOCHROME_COLOR01_E(x) (0x00000c00+((x)*4)) +#define NV04_GDI_RECTANGLE_TEXT_MONOCHROME_COLOR01_E__SIZE 0x00000080 +#define NV04_GDI_RECTANGLE_TEXT_FONT_F 0x00000ff0 +#define NV04_GDI_RECTANGLE_TEXT_FONT_F_OFFSET_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_FONT_F_OFFSET_MASK 0x0fffffff +#define NV04_GDI_RECTANGLE_TEXT_FONT_F_PITCH_SHIFT 28 +#define NV04_GDI_RECTANGLE_TEXT_FONT_F_PITCH_MASK 0xf0000000 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_F_POINT0 0x00000ff4 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_F_POINT0_L_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_F_POINT0_L_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIP_F_POINT0_T_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_F_POINT0_T_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_F_POINT1 0x00000ff8 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_F_POINT1_R_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_F_POINT1_R_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIP_F_POINT1_B_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_F_POINT1_B_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_COLOR1_F 0x00000ffc +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_F(x) (0x00001000+((x)*4)) +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_F__SIZE 0x00000100 +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_F_INDEX_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_F_INDEX_MASK 0x000000ff +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_F_X_SHIFT 8 +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_F_X_MASK 0x000fff00 +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_F_Y_SHIFT 20 +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_F_Y_MASK 0xfff00000 +#define NV04_GDI_RECTANGLE_TEXT_FONT_G 0x000017f0 +#define NV04_GDI_RECTANGLE_TEXT_FONT_G_OFFSET_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_FONT_G_OFFSET_MASK 0x0fffffff +#define NV04_GDI_RECTANGLE_TEXT_FONT_G_PITCH_SHIFT 28 +#define NV04_GDI_RECTANGLE_TEXT_FONT_G_PITCH_MASK 0xf0000000 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_G_POINT0 0x000017f4 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_G_POINT0_L_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_G_POINT0_L_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIP_G_POINT0_T_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_G_POINT0_T_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_G_POINT1 0x000017f8 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_G_POINT1_R_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_G_POINT1_R_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CLIP_G_POINT1_B_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CLIP_G_POINT1_B_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_COLOR1_G 0x000017fc +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_G_POINT(x) (0x00001800+((x)*8)) +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_G_POINT__SIZE 0x00000100 +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_G_POINT_X_SHIFT 0 +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_G_POINT_X_MASK 0x0000ffff +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_G_POINT_Y_SHIFT 16 +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_G_POINT_Y_MASK 0xffff0000 +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_G_INDEX(x) (0x00001804+((x)*8)) +#define NV04_GDI_RECTANGLE_TEXT_CHARACTER_COLOR1_G_INDEX__SIZE 0x00000100 + + +#define NV03_GDI_RECTANGLE_TEXT 0x0000004b + +#define NV03_GDI_RECTANGLE_TEXT_NOP 0x00000100 +#define NV03_GDI_RECTANGLE_TEXT_NOTIFY 0x00000104 +#define NV03_GDI_RECTANGLE_TEXT_DMA_NOTIFY 0x00000180 +#define NV03_GDI_RECTANGLE_TEXT_PATTERN 0x00000184 +#define NV03_GDI_RECTANGLE_TEXT_ROP 0x00000188 +#define NV03_GDI_RECTANGLE_TEXT_BETA1 0x0000018c +#define NV03_GDI_RECTANGLE_TEXT_SURFACE 0x00000190 +#define NV03_GDI_RECTANGLE_TEXT_OPERATION 0x000002fc +#define NV03_GDI_RECTANGLE_TEXT_COLOR_FORMAT 0x00000300 +#define NV03_GDI_RECTANGLE_TEXT_MONOCHROME_FORMAT 0x00000304 +#define NV03_GDI_RECTANGLE_TEXT_COLOR1_A 0x000003fc +#define NV03_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT 0x00000400 +#define NV03_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT_Y_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT_Y_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT_X_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT_X_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE 0x00000404 +#define NV03_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE_H_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE_H_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE_W_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_SIZE_W_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_POINT0_B 0x000007f4 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_POINT0_B_L_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_POINT0_B_L_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_CLIP_POINT0_B_T_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_POINT0_B_T_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_POINT1_B 0x000007f8 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_POINT1_B_R_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_POINT1_B_R_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_CLIP_POINT1_B_B_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_POINT1_B_B_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_COLOR1_B 0x000007fc +#define NV03_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0 0x00000800 +#define NV03_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0_L_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0_L_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0_T_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_0_T_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1 0x00000804 +#define NV03_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1_R_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1_R_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1_B_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_CLIPPED_RECTANGLE_POINT_1_B_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_C_POINT0 0x00000bec +#define NV03_GDI_RECTANGLE_TEXT_CLIP_C_POINT0_L_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_C_POINT0_L_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_CLIP_C_POINT0_T_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_C_POINT0_T_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_C_POINT1 0x00000bf0 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_C_POINT1_R_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_C_POINT1_R_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_CLIP_C_POINT1_B_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_C_POINT1_B_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_COLOR1_C 0x00000bf4 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_C 0x00000bf8 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_C_W_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_C_W_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_SIZE_C_H_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_C_H_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_POINT_C 0x00000bfc +#define NV03_GDI_RECTANGLE_TEXT_POINT_C_X_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_POINT_C_X_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_POINT_C_Y_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_POINT_C_Y_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_MONOCHROME_COLOR1_C(x) (0x00000c00+((x)*4)) +#define NV03_GDI_RECTANGLE_TEXT_MONOCHROME_COLOR1_C__SIZE 0x00000020 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_D_POINT0 0x00000fe8 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_D_POINT0_L_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_D_POINT0_L_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_CLIP_D_POINT0_T_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_D_POINT0_T_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_D_POINT1 0x00000fec +#define NV03_GDI_RECTANGLE_TEXT_CLIP_D_POINT1_R_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_D_POINT1_R_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_CLIP_D_POINT1_B_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_D_POINT1_B_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_COLOR1_D 0x00000ff0 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_IN_D 0x00000ff4 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_IN_D_W_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_IN_D_W_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_SIZE_IN_D_H_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_IN_D_H_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_OUT_D 0x00000ff8 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_OUT_D_W_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_OUT_D_W_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_SIZE_OUT_D_H_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_OUT_D_H_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_POINT_D 0x00000ffc +#define NV03_GDI_RECTANGLE_TEXT_POINT_D_X_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_POINT_D_X_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_POINT_D_Y_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_POINT_D_Y_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_MONOCHROME_COLOR1_D(x) (0x00001000+((x)*4)) +#define NV03_GDI_RECTANGLE_TEXT_MONOCHROME_COLOR1_D__SIZE 0x00000020 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_E_POINT0 0x000013e4 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_E_POINT0_L_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_E_POINT0_L_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_CLIP_E_POINT0_T_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_E_POINT0_T_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_E_POINT1 0x000013e8 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_E_POINT1_R_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_E_POINT1_R_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_CLIP_E_POINT1_B_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_CLIP_E_POINT1_B_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_COLOR0_E 0x000013ec +#define NV03_GDI_RECTANGLE_TEXT_COLOR1_E 0x000013f0 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_IN_E 0x000013f4 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_IN_E_W_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_IN_E_W_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_SIZE_IN_E_H_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_IN_E_H_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_OUT_E 0x000013f8 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_OUT_E_W_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_OUT_E_W_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_SIZE_OUT_E_H_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_SIZE_OUT_E_H_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_POINT_E 0x000013fc +#define NV03_GDI_RECTANGLE_TEXT_POINT_E_X_SHIFT 0 +#define NV03_GDI_RECTANGLE_TEXT_POINT_E_X_MASK 0x0000ffff +#define NV03_GDI_RECTANGLE_TEXT_POINT_E_Y_SHIFT 16 +#define NV03_GDI_RECTANGLE_TEXT_POINT_E_Y_MASK 0xffff0000 +#define NV03_GDI_RECTANGLE_TEXT_MONOCHROME_COLOR01_E(x) (0x00001400+((x)*4)) +#define NV03_GDI_RECTANGLE_TEXT_MONOCHROME_COLOR01_E__SIZE 0x00000020 + + +#define NV04_SWIZZLED_SURFACE 0x00000052 + +#define NV04_SWIZZLED_SURFACE_NOP 0x00000100 +#define NV04_SWIZZLED_SURFACE_NOTIFY 0x00000104 +#define NV04_SWIZZLED_SURFACE_DMA_NOTIFY 0x00000180 +#define NV04_SWIZZLED_SURFACE_DMA_IMAGE 0x00000184 +#define NV04_SWIZZLED_SURFACE_FORMAT 0x00000300 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_SHIFT 0 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_MASK 0x000000ff +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_Y8 0x00000001 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_X1R5G5B5_Z1R5G5B5 0x00000002 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_X1R5G5B5_X1R5G5B5 0x00000003 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_R5G6B5 0x00000004 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_Y16 0x00000005 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_X8R8G8B8_Z8R8G8B8 0x00000006 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_X8R8G8B8_X8R8G8B8 0x00000007 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_X1A7R8G8B8_Z1A7R8G8B8 0x00000008 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_X1A7R8G8B8_X1A7R8G8B8 0x00000009 +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_A8R8G8B8 0x0000000a +#define NV04_SWIZZLED_SURFACE_FORMAT_COLOR_Y32 0x0000000b +#define NV04_SWIZZLED_SURFACE_FORMAT_BASE_SIZE_U_SHIFT 16 +#define NV04_SWIZZLED_SURFACE_FORMAT_BASE_SIZE_U_MASK 0x00ff0000 +#define NV04_SWIZZLED_SURFACE_FORMAT_BASE_SIZE_V_SHIFT 24 +#define NV04_SWIZZLED_SURFACE_FORMAT_BASE_SIZE_V_MASK 0xff000000 +#define NV04_SWIZZLED_SURFACE_OFFSET 0x00000304 + + +#define NV04_CONTEXT_SURFACES_3D 0x00000053 + +#define NV04_CONTEXT_SURFACES_3D_NOP 0x00000100 +#define NV04_CONTEXT_SURFACES_3D_NOTIFY 0x00000104 +#define NV04_CONTEXT_SURFACES_3D_DMA_NOTIFY 0x00000180 +#define NV04_CONTEXT_SURFACES_3D_DMA_COLOR 0x00000184 +#define NV04_CONTEXT_SURFACES_3D_DMA_ZETA 0x00000188 +#define NV04_CONTEXT_SURFACES_3D_CLIP_HORIZONTAL 0x000002f8 +#define NV04_CONTEXT_SURFACES_3D_CLIP_HORIZONTAL_X_SHIFT 0 +#define NV04_CONTEXT_SURFACES_3D_CLIP_HORIZONTAL_X_MASK 0x0000ffff +#define NV04_CONTEXT_SURFACES_3D_CLIP_HORIZONTAL_W_SHIFT 16 +#define NV04_CONTEXT_SURFACES_3D_CLIP_HORIZONTAL_W_MASK 0xffff0000 +#define NV04_CONTEXT_SURFACES_3D_CLIP_VERTICAL 0x000002fc +#define NV04_CONTEXT_SURFACES_3D_CLIP_VERTICAL_Y_SHIFT 0 +#define NV04_CONTEXT_SURFACES_3D_CLIP_VERTICAL_Y_MASK 0x0000ffff +#define NV04_CONTEXT_SURFACES_3D_CLIP_VERTICAL_H_SHIFT 16 +#define NV04_CONTEXT_SURFACES_3D_CLIP_VERTICAL_H_MASK 0xffff0000 +#define NV04_CONTEXT_SURFACES_3D_FORMAT 0x00000300 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_COLOR_SHIFT 0 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_COLOR_MASK 0x000000ff +#define NV04_CONTEXT_SURFACES_3D_FORMAT_COLOR_X1R5G5B5_Z1R5G5B5 0x00000001 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_COLOR_X1R5G5B5_X1R5G5B5 0x00000002 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_COLOR_R5G6B5 0x00000003 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_COLOR_X8R8G8B8_Z8R8G8B8 0x00000004 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_COLOR_X8R8G8B8_X8R8G8B8 0x00000005 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_COLOR_X1A7R8G8B8_Z1A7R8G8B8 0x00000006 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_COLOR_X1A7R8G8B8_X1A7R8G8B8 0x00000007 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_COLOR_A8R8G8B8 0x00000008 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_TYPE_SHIFT 8 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_TYPE_MASK 0x0000ff00 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_TYPE_PITCH 0x00000100 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_TYPE_SWIZZLE 0x00000200 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_BASE_SIZE_U_SHIFT 16 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_BASE_SIZE_U_MASK 0x00ff0000 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_BASE_SIZE_V_SHIFT 24 +#define NV04_CONTEXT_SURFACES_3D_FORMAT_BASE_SIZE_V_MASK 0xff000000 +#define NV04_CONTEXT_SURFACES_3D_CLIP_SIZE 0x00000304 +#define NV04_CONTEXT_SURFACES_3D_CLIP_SIZE_W_SHIFT 0 +#define NV04_CONTEXT_SURFACES_3D_CLIP_SIZE_W_MASK 0x0000ffff +#define NV04_CONTEXT_SURFACES_3D_CLIP_SIZE_H_SHIFT 16 +#define NV04_CONTEXT_SURFACES_3D_CLIP_SIZE_H_MASK 0xffff0000 +#define NV04_CONTEXT_SURFACES_3D_PITCH 0x00000308 +#define NV04_CONTEXT_SURFACES_3D_PITCH_COLOR_SHIFT 0 +#define NV04_CONTEXT_SURFACES_3D_PITCH_COLOR_MASK 0x0000ffff +#define NV04_CONTEXT_SURFACES_3D_PITCH_ZETA_SHIFT 16 +#define NV04_CONTEXT_SURFACES_3D_PITCH_ZETA_MASK 0xffff0000 +#define NV04_CONTEXT_SURFACES_3D_OFFSET_COLOR 0x0000030c +#define NV04_CONTEXT_SURFACES_3D_OFFSET_ZETA 0x00000310 + + +#define NV04_DX5_TEXTURED_TRIANGLE 0x00000054 + +#define NV04_DX5_TEXTURED_TRIANGLE_NOP 0x00000100 +#define NV04_DX5_TEXTURED_TRIANGLE_NOTIFY 0x00000104 +#define NV04_DX5_TEXTURED_TRIANGLE_DMA_NOTIFY 0x00000180 +#define NV04_DX5_TEXTURED_TRIANGLE_DMA_A 0x00000184 +#define NV04_DX5_TEXTURED_TRIANGLE_DMA_B 0x00000188 +#define NV04_DX5_TEXTURED_TRIANGLE_SURFACE 0x0000018c +#define NV04_DX5_TEXTURED_TRIANGLE_COLORKEY 0x00000300 +#define NV04_DX5_TEXTURED_TRIANGLE_OFFSET 0x00000304 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT 0x00000308 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_DMA_SHIFT 0 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_DMA_MASK 0x00000003 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_KEY_MATCH_SHIFT 2 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_KEY_MATCH_MASK 0x0000000c +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_ZOH_SHIFT 4 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_ZOH_MASK 0x00000030 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_ZOH_CENTER 0x00000010 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_ZOH_CORNER 0x00000020 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_FOH_SHIFT 6 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_FOH_MASK 0x000000c0 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_FOH_CENTER 0x00000040 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_FOH_CORNER 0x00000080 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_SHIFT 8 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_MASK 0x00000f00 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_Y8 0x00000100 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_A1R5G5B5 0x00000200 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_X1R5G5B5 0x00000300 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_A4R4G4B4 0x00000400 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_R5G6B5 0x00000500 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_A8R8G8B8 0x00000600 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_X8R8G8B8 0x00000700 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_MIPMAP_LEVELS_SHIFT 12 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_MIPMAP_LEVELS_MASK 0x0000f000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_U_SHIFT 16 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_U_MASK 0x000f0000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_V_SHIFT 20 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_V_MASK 0x00f00000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_SHIFT 24 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_MASK 0x07000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_REPEAT 0x01000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_MIRRORED_REPEAT 0x02000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP_TO_EDGE 0x03000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP_TO_BORDER 0x04000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP 0x05000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_WRAPU (1 << 27) +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_SHIFT 28 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_MASK 0x70000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_REPEAT 0x10000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_MIRRORED_REPEAT 0x20000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_CLAMP_TO_EDGE 0x30000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_CLAMP_TO_BORDER 0x40000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_CLAMP 0x50000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FORMAT_WRAPV (1 << 31) +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER 0x0000030c +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_KERNEL_SIZE_X_SHIFT 0 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_KERNEL_SIZE_X_MASK 0x000000ff +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_KERNEL_SIZE_Y_SHIFT 8 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_KERNEL_SIZE_Y_MASK 0x00007f00 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MIPMAP_DITHER_ENABLE (1 << 15) +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MIPMAP_LODBIAS_SHIFT 16 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MIPMAP_LODBIAS_MASK 0x00ff0000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_SHIFT 24 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_MASK 0x07000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_NEAREST 0x01000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_LINEAR 0x02000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST 0x03000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST 0x04000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR 0x05000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR 0x06000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_ANISOTROPIC_MINIFY_ENABLE (1 << 27) +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MAGNIFY_SHIFT 28 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MAGNIFY_MASK 0x70000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MAGNIFY_NEAREST 0x10000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_MAGNIFY_LINEAR 0x20000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FILTER_ANISOTROPIC_MAGNIFY_ENABLE (1 << 31) +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND 0x00000310 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_TEXTURE_MAP_SHIFT 0 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_TEXTURE_MAP_MASK 0x0000000f +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_MASK_BIT_SHIFT 4 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_MASK_BIT_MASK 0x00000030 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_SHADE_MODE_SHIFT 6 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_SHADE_MODE_MASK 0x000000c0 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_SHADE_MODE_FLAT 0x00000040 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_SHADE_MODE_GOURAUD 0x00000080 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_SHADE_MODE_PHONG 0x000000c0 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_TEXTURE_PERSPECTIVE_ENABLE_SHIFT 8 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_TEXTURE_PERSPECTIVE_ENABLE_MASK 0x00000f00 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_SPECULAR_ENABLE_SHIFT 12 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_SPECULAR_ENABLE_MASK 0x0000f000 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_FOG_ENABLE_SHIFT 16 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_FOG_ENABLE_MASK 0x000f0000 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_ALPHA_ENABLE_SHIFT 20 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_ALPHA_ENABLE_MASK 0x00f00000 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_SRC_SHIFT 24 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_SRC_MASK 0x0f000000 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_DST_SHIFT 28 +#define NV04_DX5_TEXTURED_TRIANGLE_BLEND_DST_MASK 0xf0000000 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL 0x00000314 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ALPHA_REF_SHIFT 0 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ALPHA_REF_MASK 0x000000ff +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ALPHA_FUNC_SHIFT 8 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ALPHA_FUNC_MASK 0x00000f00 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ALPHA_TEST_ENABLE (1 << 12) +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ORIGIN (1 << 13) +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_ENABLE_SHIFT 14 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_ENABLE_MASK 0x0000c000 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_FUNC_SHIFT 16 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_FUNC_MASK 0x000f0000 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_CULL_MODE_SHIFT 20 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_CULL_MODE_MASK 0x00300000 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_DITHER_ENABLE (1 << 22) +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_PERSPECTIVE_ENABLE (1 << 23) +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_WRITE_ENABLE_SHIFT 24 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_WRITE_ENABLE_MASK 0x3f000000 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_FORMAT_SHIFT 30 +#define NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_FORMAT_MASK 0xc0000000 +#define NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR 0x00000318 +#define NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR_B_SHIFT 0 +#define NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR_B_MASK 0x000000ff +#define NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR_G_SHIFT 8 +#define NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR_G_MASK 0x0000ff00 +#define NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR_R_SHIFT 16 +#define NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR_R_MASK 0x00ff0000 +#define NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR_A_SHIFT 24 +#define NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR_A_MASK 0xff000000 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(x) (0x00000400+((x)*32)) +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX__SIZE 0x00000010 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SY(x) (0x00000404+((x)*32)) +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SY__SIZE 0x00000010 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SZ(x) (0x00000408+((x)*32)) +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SZ__SIZE 0x00000010 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_RHW(x) (0x0000040c+((x)*32)) +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_RHW__SIZE 0x00000010 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_COLOR(x) (0x00000410+((x)*32)) +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_COLOR__SIZE 0x00000010 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_COLOR_B_SHIFT 0 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_COLOR_B_MASK 0x000000ff +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_COLOR_G_SHIFT 8 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_COLOR_G_MASK 0x0000ff00 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_COLOR_R_SHIFT 16 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_COLOR_R_MASK 0x00ff0000 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_COLOR_A_SHIFT 24 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_COLOR_A_MASK 0xff000000 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SPECULAR(x) (0x00000414+((x)*32)) +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SPECULAR__SIZE 0x00000010 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SPECULAR_B_SHIFT 0 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SPECULAR_B_MASK 0x000000ff +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SPECULAR_G_SHIFT 8 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SPECULAR_G_MASK 0x0000ff00 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SPECULAR_R_SHIFT 16 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SPECULAR_R_MASK 0x00ff0000 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SPECULAR_FOG_SHIFT 24 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SPECULAR_FOG_MASK 0xff000000 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_TU(x) (0x00000418+((x)*32)) +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_TU__SIZE 0x00000010 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_TV(x) (0x0000041c+((x)*32)) +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_TV__SIZE 0x00000010 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE(x) (0x00000600+((x)*4)) +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE__SIZE 0x00000040 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I0_SHIFT 0 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I0_MASK 0x0000000f +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I1_SHIFT 4 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I1_MASK 0x000000f0 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I2_SHIFT 8 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I2_MASK 0x00000f00 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I3_SHIFT 12 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I3_MASK 0x0000f000 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I4_SHIFT 16 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I4_MASK 0x000f0000 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I5_SHIFT 20 +#define NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE_I5_MASK 0x00f00000 + + +#define NV04_DX6_MULTITEX_TRIANGLE 0x00000055 + +#define NV04_DX6_MULTITEX_TRIANGLE_NOP 0x00000100 +#define NV04_DX6_MULTITEX_TRIANGLE_NOTIFY 0x00000104 +#define NV04_DX6_MULTITEX_TRIANGLE_DMA_NOTIFY 0x00000180 +#define NV04_DX6_MULTITEX_TRIANGLE_DMA_A 0x00000184 +#define NV04_DX6_MULTITEX_TRIANGLE_DMA_B 0x00000188 +#define NV04_DX6_MULTITEX_TRIANGLE_SURFACE 0x0000018c +#define NV04_DX6_MULTITEX_TRIANGLE_OFFSET(x) (0x00000308+((x)*4)) +#define NV04_DX6_MULTITEX_TRIANGLE_OFFSET__SIZE 0x00000002 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT(x) (0x00000310+((x)*4)) +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT__SIZE 0x00000002 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_DMA_SHIFT 0 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_DMA_MASK 0x0000000f +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_ORIGIN_ZOH_SHIFT 4 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_ORIGIN_ZOH_MASK 0x00000030 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_ORIGIN_FOH_SHIFT 6 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_ORIGIN_FOH_MASK 0x000000c0 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_COLOR_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_COLOR_MASK 0x00000f00 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_MIPMAP_LEVELS_SHIFT 12 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_MIPMAP_LEVELS_MASK 0x0000f000 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_BASE_SIZE_U_SHIFT 16 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_BASE_SIZE_U_MASK 0x000f0000 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_BASE_SIZE_V_SHIFT 20 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_BASE_SIZE_V_MASK 0x00f00000 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_ADDRESSU_SHIFT 24 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_ADDRESSU_MASK 0x07000000 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_WRAPU (1 << 27) +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_ADDRESSV_SHIFT 28 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_ADDRESSV_MASK 0x70000000 +#define NV04_DX6_MULTITEX_TRIANGLE_FORMAT_WRAPV (1 << 31) +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER(x) (0x00000318+((x)*4)) +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER__SIZE 0x00000002 +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_KERNEL_SIZE_X_SHIFT 0 +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_KERNEL_SIZE_X_MASK 0x000000ff +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_KERNEL_SIZE_Y_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_KERNEL_SIZE_Y_MASK 0x00007f00 +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_MIPMAP_DITHER_ENABLE (1 << 15) +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_MIPMAP_LODBIAS_SHIFT 16 +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_MIPMAP_LODBIAS_MASK 0x00ff0000 +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_MINIFY_SHIFT 24 +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_MINIFY_MASK 0x07000000 +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_ANISOTROPIC_MINIFY_ENABLE (1 << 27) +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_MAGNIFY_SHIFT 28 +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_MAGNIFY_MASK 0x70000000 +#define NV04_DX6_MULTITEX_TRIANGLE_FILTER_ANISOTROPIC_MAGNIFY_ENABLE (1 << 31) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA 0x00000320 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_INVERSE0 (1 << 0) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ALPHA0 (1 << 1) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ARGUMENT0_SHIFT 2 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ARGUMENT0_MASK 0x000000fc +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_INVERSE1 (1 << 8) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ALPHA1 (1 << 9) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ARGUMENT1_SHIFT 10 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ARGUMENT1_MASK 0x0000fc00 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_INVERSE2 (1 << 16) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ALPHA2 (1 << 17) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ARGUMENT2_SHIFT 18 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ARGUMENT2_MASK 0x00fc0000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_INVERSE3 (1 << 24) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ALPHA3 (1 << 25) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ARGUMENT3_SHIFT 26 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_ARGUMENT3_MASK 0x1c000000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_OPERATION_SHIFT 29 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA_OPERATION_MASK 0xe0000000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR 0x00000324 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_INVERSE0 (1 << 0) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ALPHA0 (1 << 1) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ARGUMENT0_SHIFT 2 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ARGUMENT0_MASK 0x000000fc +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_INVERSE1 (1 << 8) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ALPHA1 (1 << 9) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ARGUMENT1_SHIFT 10 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ARGUMENT1_MASK 0x0000fc00 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_INVERSE2 (1 << 16) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ALPHA2 (1 << 17) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ARGUMENT2_SHIFT 18 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ARGUMENT2_MASK 0x00fc0000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_INVERSE3 (1 << 24) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ALPHA3 (1 << 25) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ARGUMENT3_SHIFT 26 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_ARGUMENT3_MASK 0x1c000000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_OPERATION_SHIFT 29 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR_OPERATION_MASK 0xe0000000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA 0x0000032c +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_INVERSE0 (1 << 0) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ALPHA0 (1 << 1) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ARGUMENT0_SHIFT 2 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ARGUMENT0_MASK 0x000000fc +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_INVERSE1 (1 << 8) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ALPHA1 (1 << 9) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ARGUMENT1_SHIFT 10 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ARGUMENT1_MASK 0x0000fc00 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_INVERSE2 (1 << 16) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ALPHA2 (1 << 17) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ARGUMENT2_SHIFT 18 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ARGUMENT2_MASK 0x00fc0000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_INVERSE3 (1 << 24) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ALPHA3 (1 << 25) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ARGUMENT3_SHIFT 26 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_ARGUMENT3_MASK 0x1c000000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_OPERATION_SHIFT 29 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA_OPERATION_MASK 0xe0000000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR 0x00000330 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_INVERSE0 (1 << 0) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ALPHA0 (1 << 1) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ARGUMENT0_SHIFT 2 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ARGUMENT0_MASK 0x000000fc +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_INVERSE1 (1 << 8) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ALPHA1 (1 << 9) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ARGUMENT1_SHIFT 10 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ARGUMENT1_MASK 0x0000fc00 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_INVERSE2 (1 << 16) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ALPHA2 (1 << 17) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ARGUMENT2_SHIFT 18 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ARGUMENT2_MASK 0x00fc0000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_INVERSE3 (1 << 24) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ALPHA3 (1 << 25) +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ARGUMENT3_SHIFT 26 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_ARGUMENT3_MASK 0x1c000000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_OPERATION_SHIFT 29 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR_OPERATION_MASK 0xe0000000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_FACTOR 0x00000334 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_FACTOR_B_SHIFT 0 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_FACTOR_B_MASK 0x000000ff +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_FACTOR_G_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_FACTOR_G_MASK 0x0000ff00 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_FACTOR_R_SHIFT 16 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_FACTOR_R_MASK 0x00ff0000 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_FACTOR_A_SHIFT 24 +#define NV04_DX6_MULTITEX_TRIANGLE_COMBINE_FACTOR_A_MASK 0xff000000 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND 0x00000338 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_MASK_BIT_SHIFT 4 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_MASK_BIT_MASK 0x00000030 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_SHADE_MODE_SHIFT 6 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_SHADE_MODE_MASK 0x000000c0 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_TEXTURE_PERSPECTIVE_ENABLE_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_TEXTURE_PERSPECTIVE_ENABLE_MASK 0x00000f00 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_SPECULAR_ENABLE_SHIFT 12 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_SPECULAR_ENABLE_MASK 0x0000f000 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_FOG_ENABLE_SHIFT 16 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_FOG_ENABLE_MASK 0x000f0000 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_ALPHA_ENABLE_SHIFT 20 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_ALPHA_ENABLE_MASK 0x00f00000 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_SRC_SHIFT 24 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_SRC_MASK 0x0f000000 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_DST_SHIFT 28 +#define NV04_DX6_MULTITEX_TRIANGLE_BLEND_DST_MASK 0xf0000000 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0 0x0000033c +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_ALPHA_REF_SHIFT 0 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_ALPHA_REF_MASK 0x000000ff +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_ALPHA_FUNC_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_ALPHA_FUNC_MASK 0x00000f00 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_ALPHA_TEST_ENABLE (1 << 12) +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_ORIGIN (1 << 13) +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_Z_ENABLE_SHIFT 14 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_Z_ENABLE_MASK 0x0000c000 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_Z_FUNC_SHIFT 16 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_Z_FUNC_MASK 0x000f0000 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_CULL_MODE_SHIFT 20 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_CULL_MODE_MASK 0x00300000 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_DITHER_ENABLE (1 << 22) +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_Z_PERSPECTIVE_ENABLE (1 << 23) +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_Z_WRITE_ENABLE (1 << 24) +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_STENCIL_WRITE_ENABLE (1 << 25) +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_ALPHA_WRITE_ENABLE (1 << 26) +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_RED_WRITE_ENABLE (1 << 27) +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_GREEN_WRITE_ENABLE (1 << 28) +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_BLUE_WRITE_ENABLE (1 << 29) +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_Z_FORMAT_SHIFT 30 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL0_Z_FORMAT_MASK 0xc0000000 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1 0x00000340 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1_STENCIL_TEST_ENABLE_SHIFT 0 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1_STENCIL_TEST_ENABLE_MASK 0x0000000f +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1_STENCIL_FUNC_SHIFT 4 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1_STENCIL_FUNC_MASK 0x000000f0 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1_STENCIL_REF_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1_STENCIL_REF_MASK 0x0000ff00 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1_STENCIL_MASK_READ_SHIFT 16 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1_STENCIL_MASK_READ_MASK 0x00ff0000 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1_STENCIL_MASK_WRITE_SHIFT 24 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL1_STENCIL_MASK_WRITE_MASK 0xff000000 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL2 0x00000344 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL2_STENCIL_OP_FAIL_SHIFT 0 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL2_STENCIL_OP_FAIL_MASK 0x0000000f +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL2_STENCIL_OP_ZFAIL_SHIFT 4 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL2_STENCIL_OP_ZFAIL_MASK 0x000000f0 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL2_STENCIL_OP_ZPASS_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_CONTROL2_STENCIL_OP_ZPASS_MASK 0x00000f00 +#define NV04_DX6_MULTITEX_TRIANGLE_FOGCOLOR 0x00000348 +#define NV04_DX6_MULTITEX_TRIANGLE_FOGCOLOR_B_SHIFT 0 +#define NV04_DX6_MULTITEX_TRIANGLE_FOGCOLOR_B_MASK 0x000000ff +#define NV04_DX6_MULTITEX_TRIANGLE_FOGCOLOR_G_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_FOGCOLOR_G_MASK 0x0000ff00 +#define NV04_DX6_MULTITEX_TRIANGLE_FOGCOLOR_R_SHIFT 16 +#define NV04_DX6_MULTITEX_TRIANGLE_FOGCOLOR_R_MASK 0x00ff0000 +#define NV04_DX6_MULTITEX_TRIANGLE_FOGCOLOR_A_SHIFT 24 +#define NV04_DX6_MULTITEX_TRIANGLE_FOGCOLOR_A_MASK 0xff000000 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SX(x) (0x00000400+((x)*40)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SX__SIZE 0x00000008 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SY(x) (0x00000404+((x)*40)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SY__SIZE 0x00000008 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SZ(x) (0x00000408+((x)*40)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SZ__SIZE 0x00000008 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_RHW(x) (0x0000040c+((x)*40)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_RHW__SIZE 0x00000008 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_COLOR(x) (0x00000410+((x)*40)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_COLOR__SIZE 0x00000008 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_COLOR_B_SHIFT 0 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_COLOR_B_MASK 0x000000ff +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_COLOR_G_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_COLOR_G_MASK 0x0000ff00 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_COLOR_R_SHIFT 16 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_COLOR_R_MASK 0x00ff0000 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_COLOR_A_SHIFT 24 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_COLOR_A_MASK 0xff000000 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SPECULAR(x) (0x00000414+((x)*40)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SPECULAR__SIZE 0x00000008 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SPECULAR_B_SHIFT 0 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SPECULAR_B_MASK 0x000000ff +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SPECULAR_G_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SPECULAR_G_MASK 0x0000ff00 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SPECULAR_R_SHIFT 16 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SPECULAR_R_MASK 0x00ff0000 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SPECULAR_FOG_SHIFT 24 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_SPECULAR_FOG_MASK 0xff000000 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_TU0(x) (0x00000418+((x)*40)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_TU0__SIZE 0x00000008 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_TV0(x) (0x0000041c+((x)*40)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_TV0__SIZE 0x00000008 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_TU1(x) (0x00000420+((x)*40)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_TU1__SIZE 0x00000008 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_TV1(x) (0x00000424+((x)*40)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_TV1__SIZE 0x00000008 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE(x) (0x00000540+((x)*4)) +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE__SIZE 0x00000030 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I0_SHIFT 0 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I0_MASK 0x0000000f +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I1_SHIFT 4 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I1_MASK 0x000000f0 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I2_SHIFT 8 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I2_MASK 0x00000f00 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I3_SHIFT 12 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I3_MASK 0x0000f000 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I4_SHIFT 16 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I4_MASK 0x000f0000 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I5_SHIFT 20 +#define NV04_DX6_MULTITEX_TRIANGLE_TLMTVERTEX_DRAWPRIMITIVE_I5_MASK 0x00f00000 + + +#define NV10_DX5_TEXTURED_TRIANGLE 0x00000094 + + + +#define NV10TCL 0x00000056 + +#define NV10TCL_NOP 0x00000100 +#define NV10TCL_NOTIFY 0x00000104 +#define NV10TCL_DMA_NOTIFY 0x00000180 +#define NV10TCL_DMA_IN_MEMORY0 0x00000184 +#define NV10TCL_DMA_IN_MEMORY1 0x00000188 +#define NV10TCL_DMA_VTXBUF0 0x0000018c +#define NV10TCL_DMA_IN_MEMORY2 0x00000194 +#define NV10TCL_DMA_IN_MEMORY3 0x00000198 +#define NV10TCL_RT_HORIZ 0x00000200 +#define NV10TCL_RT_HORIZ_X_SHIFT 0 +#define NV10TCL_RT_HORIZ_X_MASK 0x0000ffff +#define NV10TCL_RT_HORIZ_W_SHIFT 16 +#define NV10TCL_RT_HORIZ_W_MASK 0xffff0000 +#define NV10TCL_RT_VERT 0x00000204 +#define NV10TCL_RT_VERT_Y_SHIFT 0 +#define NV10TCL_RT_VERT_Y_MASK 0x0000ffff +#define NV10TCL_RT_VERT_H_SHIFT 16 +#define NV10TCL_RT_VERT_H_MASK 0xffff0000 +#define NV10TCL_RT_FORMAT 0x00000208 +#define NV10TCL_RT_FORMAT_TYPE_SHIFT 8 +#define NV10TCL_RT_FORMAT_TYPE_MASK 0x00000f00 +#define NV10TCL_RT_FORMAT_TYPE_LINEAR 0x00000100 +#define NV10TCL_RT_FORMAT_TYPE_SWIZZLED 0x00000200 +#define NV10TCL_RT_FORMAT_COLOR_SHIFT 0 +#define NV10TCL_RT_FORMAT_COLOR_MASK 0x0000001f +#define NV10TCL_RT_FORMAT_COLOR_R5G6B5 0x00000003 +#define NV10TCL_RT_FORMAT_COLOR_X8R8G8B8 0x00000005 +#define NV10TCL_RT_FORMAT_COLOR_A8R8G8B8 0x00000008 +#define NV10TCL_RT_FORMAT_COLOR_B8 0x00000009 +#define NV10TCL_RT_FORMAT_COLOR_UNKNOWN 0x0000000d +#define NV10TCL_RT_FORMAT_COLOR_X8B8G8R8 0x0000000f +#define NV10TCL_RT_FORMAT_COLOR_A8B8G8R8 0x00000010 +#define NV10TCL_RT_PITCH 0x0000020c +#define NV10TCL_RT_PITCH_COLOR_PITCH_SHIFT 0 +#define NV10TCL_RT_PITCH_COLOR_PITCH_MASK 0x0000ffff +#define NV10TCL_RT_PITCH_ZETA_PITCH_SHIFT 16 +#define NV10TCL_RT_PITCH_ZETA_PITCH_MASK 0xffff0000 +#define NV10TCL_COLOR_OFFSET 0x00000210 +#define NV10TCL_ZETA_OFFSET 0x00000214 +#define NV10TCL_TX_OFFSET(x) (0x00000218+((x)*4)) +#define NV10TCL_TX_OFFSET__SIZE 0x00000002 +#define NV10TCL_TX_FORMAT(x) (0x00000220+((x)*4)) +#define NV10TCL_TX_FORMAT__SIZE 0x00000002 +#define NV10TCL_TX_FORMAT_DMA0 (1 << 0) +#define NV10TCL_TX_FORMAT_DMA1 (1 << 1) +#define NV10TCL_TX_FORMAT_CUBE_MAP (1 << 2) +#define NV10TCL_TX_FORMAT_FORMAT_SHIFT 7 +#define NV10TCL_TX_FORMAT_FORMAT_MASK 0x00000780 +#define NV10TCL_TX_FORMAT_FORMAT_L8 0x00000000 +#define NV10TCL_TX_FORMAT_FORMAT_A8 0x00000080 +#define NV10TCL_TX_FORMAT_FORMAT_A1R5G5B5 0x00000100 +#define NV10TCL_TX_FORMAT_FORMAT_A8_RECT 0x00000180 +#define NV10TCL_TX_FORMAT_FORMAT_A4R4G4B4 0x00000200 +#define NV10TCL_TX_FORMAT_FORMAT_R5G6B5 0x00000280 +#define NV10TCL_TX_FORMAT_FORMAT_A8R8G8B8 0x00000300 +#define NV10TCL_TX_FORMAT_FORMAT_X8R8G8B8 0x00000380 +#define NV10TCL_TX_FORMAT_FORMAT_INDEX8 0x00000580 +#define NV10TCL_TX_FORMAT_FORMAT_DXT1 0x00000600 +#define NV10TCL_TX_FORMAT_FORMAT_DXT3 0x00000700 +#define NV10TCL_TX_FORMAT_FORMAT_DXT5 0x00000780 +#define NV10TCL_TX_FORMAT_FORMAT_A1R5G5B5_RECT 0x00000800 +#define NV10TCL_TX_FORMAT_FORMAT_R5G6B5_RECT 0x00000880 +#define NV10TCL_TX_FORMAT_FORMAT_A8R8G8B8_RECT 0x00000900 +#define NV10TCL_TX_FORMAT_FORMAT_L8_RECT 0x00000980 +#define NV10TCL_TX_FORMAT_FORMAT_A8L8 0x00000d00 +#define NV10TCL_TX_FORMAT_FORMAT_A8_RECT2 0x00000d80 +#define NV10TCL_TX_FORMAT_FORMAT_A4R4G4B4_RECT 0x00000e80 +#define NV10TCL_TX_FORMAT_FORMAT_R8G8B8_RECT 0x00000f00 +#define NV10TCL_TX_FORMAT_FORMAT_L8A8_RECT 0x00001000 +#define NV10TCL_TX_FORMAT_FORMAT_DSDT 0x00001400 +#define NV10TCL_TX_FORMAT_FORMAT_A16 0x00001900 +#define NV10TCL_TX_FORMAT_FORMAT_HILO16 0x00001980 +#define NV10TCL_TX_FORMAT_FORMAT_A16_RECT 0x00001a80 +#define NV10TCL_TX_FORMAT_FORMAT_HILO16_RECT 0x00001b00 +#define NV10TCL_TX_FORMAT_FORMAT_HILO8 0x00002200 +#define NV10TCL_TX_FORMAT_FORMAT_SIGNED_HILO8 0x00002280 +#define NV10TCL_TX_FORMAT_FORMAT_HILO8_RECT 0x00002300 +#define NV10TCL_TX_FORMAT_FORMAT_SIGNED_HILO8_RECT 0x00002380 +#define NV10TCL_TX_FORMAT_FORMAT_FLOAT_RGBA16_NV 0x00002500 +#define NV10TCL_TX_FORMAT_FORMAT_FLOAT_RGBA32_NV 0x00002580 +#define NV10TCL_TX_FORMAT_FORMAT_FLOAT_R32_NV 0x00002600 +#define NV10TCL_TX_FORMAT_NPOT (1 << 11) +#define NV10TCL_TX_FORMAT_MIPMAP (1 << 15) +#define NV10TCL_TX_FORMAT_BASE_SIZE_U_SHIFT 16 +#define NV10TCL_TX_FORMAT_BASE_SIZE_U_MASK 0x000f0000 +#define NV10TCL_TX_FORMAT_BASE_SIZE_V_SHIFT 20 +#define NV10TCL_TX_FORMAT_BASE_SIZE_V_MASK 0x00f00000 +#define NV10TCL_TX_FORMAT_WRAP_S_SHIFT 24 +#define NV10TCL_TX_FORMAT_WRAP_S_MASK 0x0f000000 +#define NV10TCL_TX_FORMAT_WRAP_S_REPEAT 0x01000000 +#define NV10TCL_TX_FORMAT_WRAP_S_MIRRORED_REPEAT 0x02000000 +#define NV10TCL_TX_FORMAT_WRAP_S_CLAMP_TO_EDGE 0x03000000 +#define NV10TCL_TX_FORMAT_WRAP_S_CLAMP_TO_BORDER 0x04000000 +#define NV10TCL_TX_FORMAT_WRAP_S_CLAMP 0x05000000 +#define NV10TCL_TX_FORMAT_WRAP_T_SHIFT 28 +#define NV10TCL_TX_FORMAT_WRAP_T_MASK 0xf0000000 +#define NV10TCL_TX_FORMAT_WRAP_T_REPEAT 0x10000000 +#define NV10TCL_TX_FORMAT_WRAP_T_MIRRORED_REPEAT 0x20000000 +#define NV10TCL_TX_FORMAT_WRAP_T_CLAMP_TO_EDGE 0x30000000 +#define NV10TCL_TX_FORMAT_WRAP_T_CLAMP_TO_BORDER 0x40000000 +#define NV10TCL_TX_FORMAT_WRAP_T_CLAMP 0x50000000 +#define NV10TCL_TX_ENABLE(x) (0x00000228+((x)*4)) +#define NV10TCL_TX_ENABLE__SIZE 0x00000002 +#define NV10TCL_TX_ENABLE_ANISOTROPY_SHIFT 4 +#define NV10TCL_TX_ENABLE_ANISOTROPY_MASK 0x00000030 +#define NV10TCL_TX_ENABLE_MIPMAP_MAX_LOD_SHIFT 14 +#define NV10TCL_TX_ENABLE_MIPMAP_MAX_LOD_MASK 0x0003c000 +#define NV10TCL_TX_ENABLE_MIPMAP_MIN_LOD_SHIFT 26 +#define NV10TCL_TX_ENABLE_MIPMAP_MIN_LOD_MASK 0x3c000000 +#define NV10TCL_TX_ENABLE_ENABLE (1 << 30) +#define NV10TCL_TX_NPOT_PITCH(x) (0x00000230+((x)*4)) +#define NV10TCL_TX_NPOT_PITCH__SIZE 0x00000002 +#define NV10TCL_TX_NPOT_PITCH_PITCH_SHIFT 16 +#define NV10TCL_TX_NPOT_PITCH_PITCH_MASK 0xffff0000 +#define NV10TCL_TX_NPOT_SIZE(x) (0x00000240+((x)*4)) +#define NV10TCL_TX_NPOT_SIZE__SIZE 0x00000002 +#define NV10TCL_TX_NPOT_SIZE_H_SHIFT 0 +#define NV10TCL_TX_NPOT_SIZE_H_MASK 0x0000ffff +#define NV10TCL_TX_NPOT_SIZE_W_SHIFT 16 +#define NV10TCL_TX_NPOT_SIZE_W_MASK 0xffff0000 +#define NV10TCL_TX_FILTER(x) (0x00000248+((x)*4)) +#define NV10TCL_TX_FILTER__SIZE 0x00000002 +#define NV10TCL_TX_FILTER_LOD_BIAS_SHIFT 8 +#define NV10TCL_TX_FILTER_LOD_BIAS_MASK 0x00000f00 +#define NV10TCL_TX_FILTER_MINIFY_SHIFT 24 +#define NV10TCL_TX_FILTER_MINIFY_MASK 0x0f000000 +#define NV10TCL_TX_FILTER_MINIFY_NEAREST 0x01000000 +#define NV10TCL_TX_FILTER_MINIFY_LINEAR 0x02000000 +#define NV10TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST 0x03000000 +#define NV10TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST 0x04000000 +#define NV10TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR 0x05000000 +#define NV10TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR 0x06000000 +#define NV10TCL_TX_FILTER_MAGNIFY_SHIFT 28 +#define NV10TCL_TX_FILTER_MAGNIFY_MASK 0xf0000000 +#define NV10TCL_TX_FILTER_MAGNIFY_NEAREST 0x10000000 +#define NV10TCL_TX_FILTER_MAGNIFY_LINEAR 0x20000000 +#define NV10TCL_TX_PALETTE_OFFSET(x) (0x00000250+((x)*4)) +#define NV10TCL_TX_PALETTE_OFFSET__SIZE 0x00000002 +#define NV10TCL_RC_IN_ALPHA(x) (0x00000260+((x)*4)) +#define NV10TCL_RC_IN_ALPHA__SIZE 0x00000002 +#define NV10TCL_RC_IN_ALPHA_D_INPUT_SHIFT 0 +#define NV10TCL_RC_IN_ALPHA_D_INPUT_MASK 0x0000000f +#define NV10TCL_RC_IN_ALPHA_D_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_IN_ALPHA_D_INPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV10TCL_RC_IN_ALPHA_D_INPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV10TCL_RC_IN_ALPHA_D_INPUT_FOG 0x00000003 +#define NV10TCL_RC_IN_ALPHA_D_INPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV10TCL_RC_IN_ALPHA_D_INPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV10TCL_RC_IN_ALPHA_D_INPUT_TEXTURE0_ARB 0x00000008 +#define NV10TCL_RC_IN_ALPHA_D_INPUT_TEXTURE1_ARB 0x00000009 +#define NV10TCL_RC_IN_ALPHA_D_INPUT_SPARE0_NV 0x0000000c +#define NV10TCL_RC_IN_ALPHA_D_INPUT_SPARE1_NV 0x0000000d +#define NV10TCL_RC_IN_ALPHA_D_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV10TCL_RC_IN_ALPHA_D_INPUT_E_TIMES_F_NV 0x0000000f +#define NV10TCL_RC_IN_ALPHA_D_COMPONENT_USAGE (1 << 4) +#define NV10TCL_RC_IN_ALPHA_D_COMPONENT_USAGE_BLUE 0x00000000 +#define NV10TCL_RC_IN_ALPHA_D_COMPONENT_USAGE_ALPHA 0x00000010 +#define NV10TCL_RC_IN_ALPHA_D_MAPPING_SHIFT 5 +#define NV10TCL_RC_IN_ALPHA_D_MAPPING_MASK 0x000000e0 +#define NV10TCL_RC_IN_ALPHA_D_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_IN_ALPHA_D_MAPPING_UNSIGNED_INVERT_NV 0x00000020 +#define NV10TCL_RC_IN_ALPHA_D_MAPPING_EXPAND_NORMAL_NV 0x00000040 +#define NV10TCL_RC_IN_ALPHA_D_MAPPING_EXPAND_NEGATE_NV 0x00000060 +#define NV10TCL_RC_IN_ALPHA_D_MAPPING_HALF_BIAS_NORMAL_NV 0x00000080 +#define NV10TCL_RC_IN_ALPHA_D_MAPPING_HALF_BIAS_NEGATE_NV 0x000000a0 +#define NV10TCL_RC_IN_ALPHA_D_MAPPING_SIGNED_IDENTITY_NV 0x000000c0 +#define NV10TCL_RC_IN_ALPHA_D_MAPPING_SIGNED_NEGATE_NV 0x000000e0 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_SHIFT 8 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_MASK 0x00000f00 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_FOG 0x00000300 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_TEXTURE0_ARB 0x00000800 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_TEXTURE1_ARB 0x00000900 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_SPARE0_NV 0x00000c00 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_SPARE1_NV 0x00000d00 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV10TCL_RC_IN_ALPHA_C_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV10TCL_RC_IN_ALPHA_C_COMPONENT_USAGE (1 << 12) +#define NV10TCL_RC_IN_ALPHA_C_COMPONENT_USAGE_BLUE 0x00000000 +#define NV10TCL_RC_IN_ALPHA_C_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV10TCL_RC_IN_ALPHA_C_MAPPING_SHIFT 13 +#define NV10TCL_RC_IN_ALPHA_C_MAPPING_MASK 0x0000e000 +#define NV10TCL_RC_IN_ALPHA_C_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_IN_ALPHA_C_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV10TCL_RC_IN_ALPHA_C_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV10TCL_RC_IN_ALPHA_C_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV10TCL_RC_IN_ALPHA_C_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV10TCL_RC_IN_ALPHA_C_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV10TCL_RC_IN_ALPHA_C_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV10TCL_RC_IN_ALPHA_C_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_SHIFT 16 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_MASK 0x000f0000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_FOG 0x00030000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_TEXTURE0_ARB 0x00080000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_TEXTURE1_ARB 0x00090000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_SPARE0_NV 0x000c0000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_SPARE1_NV 0x000d0000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV10TCL_RC_IN_ALPHA_B_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV10TCL_RC_IN_ALPHA_B_COMPONENT_USAGE (1 << 20) +#define NV10TCL_RC_IN_ALPHA_B_COMPONENT_USAGE_BLUE 0x00000000 +#define NV10TCL_RC_IN_ALPHA_B_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV10TCL_RC_IN_ALPHA_B_MAPPING_SHIFT 21 +#define NV10TCL_RC_IN_ALPHA_B_MAPPING_MASK 0x00e00000 +#define NV10TCL_RC_IN_ALPHA_B_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_IN_ALPHA_B_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV10TCL_RC_IN_ALPHA_B_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV10TCL_RC_IN_ALPHA_B_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV10TCL_RC_IN_ALPHA_B_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV10TCL_RC_IN_ALPHA_B_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV10TCL_RC_IN_ALPHA_B_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV10TCL_RC_IN_ALPHA_B_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_SHIFT 24 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_MASK 0x0f000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_FOG 0x03000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_TEXTURE0_ARB 0x08000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_TEXTURE1_ARB 0x09000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_SPARE0_NV 0x0c000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_SPARE1_NV 0x0d000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV10TCL_RC_IN_ALPHA_A_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV10TCL_RC_IN_ALPHA_A_COMPONENT_USAGE (1 << 28) +#define NV10TCL_RC_IN_ALPHA_A_COMPONENT_USAGE_BLUE 0x00000000 +#define NV10TCL_RC_IN_ALPHA_A_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV10TCL_RC_IN_ALPHA_A_MAPPING_SHIFT 29 +#define NV10TCL_RC_IN_ALPHA_A_MAPPING_MASK 0xe0000000 +#define NV10TCL_RC_IN_ALPHA_A_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_IN_ALPHA_A_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV10TCL_RC_IN_ALPHA_A_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV10TCL_RC_IN_ALPHA_A_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV10TCL_RC_IN_ALPHA_A_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV10TCL_RC_IN_ALPHA_A_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV10TCL_RC_IN_ALPHA_A_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV10TCL_RC_IN_ALPHA_A_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV10TCL_RC_IN_RGB(x) (0x00000268+((x)*4)) +#define NV10TCL_RC_IN_RGB__SIZE 0x00000002 +#define NV10TCL_RC_IN_RGB_D_INPUT_SHIFT 0 +#define NV10TCL_RC_IN_RGB_D_INPUT_MASK 0x0000000f +#define NV10TCL_RC_IN_RGB_D_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_IN_RGB_D_INPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV10TCL_RC_IN_RGB_D_INPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV10TCL_RC_IN_RGB_D_INPUT_FOG 0x00000003 +#define NV10TCL_RC_IN_RGB_D_INPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV10TCL_RC_IN_RGB_D_INPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV10TCL_RC_IN_RGB_D_INPUT_TEXTURE0_ARB 0x00000008 +#define NV10TCL_RC_IN_RGB_D_INPUT_TEXTURE1_ARB 0x00000009 +#define NV10TCL_RC_IN_RGB_D_INPUT_SPARE0_NV 0x0000000c +#define NV10TCL_RC_IN_RGB_D_INPUT_SPARE1_NV 0x0000000d +#define NV10TCL_RC_IN_RGB_D_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV10TCL_RC_IN_RGB_D_INPUT_E_TIMES_F_NV 0x0000000f +#define NV10TCL_RC_IN_RGB_D_COMPONENT_USAGE (1 << 4) +#define NV10TCL_RC_IN_RGB_D_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_IN_RGB_D_COMPONENT_USAGE_ALPHA 0x00000010 +#define NV10TCL_RC_IN_RGB_D_MAPPING_SHIFT 5 +#define NV10TCL_RC_IN_RGB_D_MAPPING_MASK 0x000000e0 +#define NV10TCL_RC_IN_RGB_D_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_IN_RGB_D_MAPPING_UNSIGNED_INVERT_NV 0x00000020 +#define NV10TCL_RC_IN_RGB_D_MAPPING_EXPAND_NORMAL_NV 0x00000040 +#define NV10TCL_RC_IN_RGB_D_MAPPING_EXPAND_NEGATE_NV 0x00000060 +#define NV10TCL_RC_IN_RGB_D_MAPPING_HALF_BIAS_NORMAL_NV 0x00000080 +#define NV10TCL_RC_IN_RGB_D_MAPPING_HALF_BIAS_NEGATE_NV 0x000000a0 +#define NV10TCL_RC_IN_RGB_D_MAPPING_SIGNED_IDENTITY_NV 0x000000c0 +#define NV10TCL_RC_IN_RGB_D_MAPPING_SIGNED_NEGATE_NV 0x000000e0 +#define NV10TCL_RC_IN_RGB_C_INPUT_SHIFT 8 +#define NV10TCL_RC_IN_RGB_C_INPUT_MASK 0x00000f00 +#define NV10TCL_RC_IN_RGB_C_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_IN_RGB_C_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV10TCL_RC_IN_RGB_C_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV10TCL_RC_IN_RGB_C_INPUT_FOG 0x00000300 +#define NV10TCL_RC_IN_RGB_C_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV10TCL_RC_IN_RGB_C_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV10TCL_RC_IN_RGB_C_INPUT_TEXTURE0_ARB 0x00000800 +#define NV10TCL_RC_IN_RGB_C_INPUT_TEXTURE1_ARB 0x00000900 +#define NV10TCL_RC_IN_RGB_C_INPUT_SPARE0_NV 0x00000c00 +#define NV10TCL_RC_IN_RGB_C_INPUT_SPARE1_NV 0x00000d00 +#define NV10TCL_RC_IN_RGB_C_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV10TCL_RC_IN_RGB_C_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV10TCL_RC_IN_RGB_C_COMPONENT_USAGE (1 << 12) +#define NV10TCL_RC_IN_RGB_C_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_IN_RGB_C_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV10TCL_RC_IN_RGB_C_MAPPING_SHIFT 13 +#define NV10TCL_RC_IN_RGB_C_MAPPING_MASK 0x0000e000 +#define NV10TCL_RC_IN_RGB_C_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_IN_RGB_C_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV10TCL_RC_IN_RGB_C_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV10TCL_RC_IN_RGB_C_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV10TCL_RC_IN_RGB_C_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV10TCL_RC_IN_RGB_C_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV10TCL_RC_IN_RGB_C_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV10TCL_RC_IN_RGB_C_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV10TCL_RC_IN_RGB_B_INPUT_SHIFT 16 +#define NV10TCL_RC_IN_RGB_B_INPUT_MASK 0x000f0000 +#define NV10TCL_RC_IN_RGB_B_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_IN_RGB_B_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV10TCL_RC_IN_RGB_B_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV10TCL_RC_IN_RGB_B_INPUT_FOG 0x00030000 +#define NV10TCL_RC_IN_RGB_B_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV10TCL_RC_IN_RGB_B_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV10TCL_RC_IN_RGB_B_INPUT_TEXTURE0_ARB 0x00080000 +#define NV10TCL_RC_IN_RGB_B_INPUT_TEXTURE1_ARB 0x00090000 +#define NV10TCL_RC_IN_RGB_B_INPUT_SPARE0_NV 0x000c0000 +#define NV10TCL_RC_IN_RGB_B_INPUT_SPARE1_NV 0x000d0000 +#define NV10TCL_RC_IN_RGB_B_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV10TCL_RC_IN_RGB_B_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV10TCL_RC_IN_RGB_B_COMPONENT_USAGE (1 << 20) +#define NV10TCL_RC_IN_RGB_B_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_IN_RGB_B_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV10TCL_RC_IN_RGB_B_MAPPING_SHIFT 21 +#define NV10TCL_RC_IN_RGB_B_MAPPING_MASK 0x00e00000 +#define NV10TCL_RC_IN_RGB_B_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_IN_RGB_B_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV10TCL_RC_IN_RGB_B_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV10TCL_RC_IN_RGB_B_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV10TCL_RC_IN_RGB_B_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV10TCL_RC_IN_RGB_B_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV10TCL_RC_IN_RGB_B_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV10TCL_RC_IN_RGB_B_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV10TCL_RC_IN_RGB_A_INPUT_SHIFT 24 +#define NV10TCL_RC_IN_RGB_A_INPUT_MASK 0x0f000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_FOG 0x03000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_TEXTURE0_ARB 0x08000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_TEXTURE1_ARB 0x09000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_SPARE0_NV 0x0c000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_SPARE1_NV 0x0d000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV10TCL_RC_IN_RGB_A_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV10TCL_RC_IN_RGB_A_COMPONENT_USAGE (1 << 28) +#define NV10TCL_RC_IN_RGB_A_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_IN_RGB_A_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV10TCL_RC_IN_RGB_A_MAPPING_SHIFT 29 +#define NV10TCL_RC_IN_RGB_A_MAPPING_MASK 0xe0000000 +#define NV10TCL_RC_IN_RGB_A_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_IN_RGB_A_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV10TCL_RC_IN_RGB_A_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV10TCL_RC_IN_RGB_A_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV10TCL_RC_IN_RGB_A_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV10TCL_RC_IN_RGB_A_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV10TCL_RC_IN_RGB_A_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV10TCL_RC_IN_RGB_A_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV10TCL_RC_COLOR(x) (0x00000270+((x)*4)) +#define NV10TCL_RC_COLOR__SIZE 0x00000002 +#define NV10TCL_RC_COLOR_B_SHIFT 0 +#define NV10TCL_RC_COLOR_B_MASK 0x000000ff +#define NV10TCL_RC_COLOR_G_SHIFT 8 +#define NV10TCL_RC_COLOR_G_MASK 0x0000ff00 +#define NV10TCL_RC_COLOR_R_SHIFT 16 +#define NV10TCL_RC_COLOR_R_MASK 0x00ff0000 +#define NV10TCL_RC_COLOR_A_SHIFT 24 +#define NV10TCL_RC_COLOR_A_MASK 0xff000000 +#define NV10TCL_RC_OUT_ALPHA(x) (0x00000278+((x)*4)) +#define NV10TCL_RC_OUT_ALPHA__SIZE 0x00000002 +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_SHIFT 0 +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_MASK 0x0000000f +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_ZERO 0x00000000 +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_FOG 0x00000003 +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_TEXTURE0_ARB 0x00000008 +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_TEXTURE1_ARB 0x00000009 +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_SPARE0_NV 0x0000000c +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_SPARE1_NV 0x0000000d +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV10TCL_RC_OUT_ALPHA_CD_OUTPUT_E_TIMES_F_NV 0x0000000f +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_SHIFT 4 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_MASK 0x000000f0 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_ZERO 0x00000000 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_CONSTANT_COLOR0_NV 0x00000010 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_CONSTANT_COLOR1_NV 0x00000020 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_FOG 0x00000030 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_PRIMARY_COLOR_NV 0x00000040 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_SECONDARY_COLOR_NV 0x00000050 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_TEXTURE0_ARB 0x00000080 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_TEXTURE1_ARB 0x00000090 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_SPARE0_NV 0x000000c0 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_SPARE1_NV 0x000000d0 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000000e0 +#define NV10TCL_RC_OUT_ALPHA_AB_OUTPUT_E_TIMES_F_NV 0x000000f0 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_SHIFT 8 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_MASK 0x00000f00 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_ZERO 0x00000000 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_FOG 0x00000300 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_TEXTURE0_ARB 0x00000800 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_TEXTURE1_ARB 0x00000900 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_SPARE0_NV 0x00000c00 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_SPARE1_NV 0x00000d00 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV10TCL_RC_OUT_ALPHA_SUM_OUTPUT_E_TIMES_F_NV 0x00000f00 +#define NV10TCL_RC_OUT_ALPHA_CD_DOT_PRODUCT (1 << 12) +#define NV10TCL_RC_OUT_ALPHA_AB_DOT_PRODUCT (1 << 13) +#define NV10TCL_RC_OUT_ALPHA_MUX_SUM (1 << 14) +#define NV10TCL_RC_OUT_ALPHA_BIAS (1 << 15) +#define NV10TCL_RC_OUT_ALPHA_BIAS_NONE 0x00000000 +#define NV10TCL_RC_OUT_ALPHA_BIAS_BIAS_BY_NEGATIVE_ONE_HALF_NV 0x00008000 +#define NV10TCL_RC_OUT_ALPHA_SCALE_SHIFT 17 +#define NV10TCL_RC_OUT_ALPHA_SCALE_MASK 0x00000000 +#define NV10TCL_RC_OUT_ALPHA_SCALE_NONE 0x00000000 +#define NV10TCL_RC_OUT_ALPHA_SCALE_SCALE_BY_TWO_NV 0x00020000 +#define NV10TCL_RC_OUT_ALPHA_SCALE_SCALE_BY_FOUR_NV 0x00040000 +#define NV10TCL_RC_OUT_ALPHA_SCALE_SCALE_BY_ONE_HALF_NV 0x00060000 +#define NV10TCL_RC_OUT_RGB(x) (0x00000280+((x)*4)) +#define NV10TCL_RC_OUT_RGB__SIZE 0x00000002 +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_SHIFT 0 +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_MASK 0x0000000f +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_ZERO 0x00000000 +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_FOG 0x00000003 +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_TEXTURE0_ARB 0x00000008 +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_TEXTURE1_ARB 0x00000009 +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_SPARE0_NV 0x0000000c +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_SPARE1_NV 0x0000000d +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV10TCL_RC_OUT_RGB_CD_OUTPUT_E_TIMES_F_NV 0x0000000f +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_SHIFT 4 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_MASK 0x000000f0 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_ZERO 0x00000000 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_CONSTANT_COLOR0_NV 0x00000010 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_CONSTANT_COLOR1_NV 0x00000020 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_FOG 0x00000030 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_PRIMARY_COLOR_NV 0x00000040 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_SECONDARY_COLOR_NV 0x00000050 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_TEXTURE0_ARB 0x00000080 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_TEXTURE1_ARB 0x00000090 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_SPARE0_NV 0x000000c0 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_SPARE1_NV 0x000000d0 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000000e0 +#define NV10TCL_RC_OUT_RGB_AB_OUTPUT_E_TIMES_F_NV 0x000000f0 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_SHIFT 8 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_MASK 0x00000f00 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_ZERO 0x00000000 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_FOG 0x00000300 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_TEXTURE0_ARB 0x00000800 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_TEXTURE1_ARB 0x00000900 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_SPARE0_NV 0x00000c00 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_SPARE1_NV 0x00000d00 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV10TCL_RC_OUT_RGB_SUM_OUTPUT_E_TIMES_F_NV 0x00000f00 +#define NV10TCL_RC_OUT_RGB_CD_DOT_PRODUCT (1 << 12) +#define NV10TCL_RC_OUT_RGB_AB_DOT_PRODUCT (1 << 13) +#define NV10TCL_RC_OUT_RGB_MUX_SUM (1 << 14) +#define NV10TCL_RC_OUT_RGB_BIAS (1 << 15) +#define NV10TCL_RC_OUT_RGB_BIAS_NONE 0x00000000 +#define NV10TCL_RC_OUT_RGB_BIAS_BIAS_BY_NEGATIVE_ONE_HALF_NV 0x00008000 +#define NV10TCL_RC_OUT_RGB_SCALE_SHIFT 17 +#define NV10TCL_RC_OUT_RGB_SCALE_MASK 0x00000000 +#define NV10TCL_RC_OUT_RGB_SCALE_NONE 0x00000000 +#define NV10TCL_RC_OUT_RGB_SCALE_SCALE_BY_TWO_NV 0x00020000 +#define NV10TCL_RC_OUT_RGB_SCALE_SCALE_BY_FOUR_NV 0x00040000 +#define NV10TCL_RC_OUT_RGB_SCALE_SCALE_BY_ONE_HALF_NV 0x00060000 +#define NV10TCL_RC_OUT_RGB_OPERATION_SHIFT 27 +#define NV10TCL_RC_OUT_RGB_OPERATION_MASK 0x38000000 +#define NV10TCL_RC_FINAL0 0x00000288 +#define NV10TCL_RC_FINAL0_D_INPUT_SHIFT 0 +#define NV10TCL_RC_FINAL0_D_INPUT_MASK 0x0000000f +#define NV10TCL_RC_FINAL0_D_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_FINAL0_D_INPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV10TCL_RC_FINAL0_D_INPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV10TCL_RC_FINAL0_D_INPUT_FOG 0x00000003 +#define NV10TCL_RC_FINAL0_D_INPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV10TCL_RC_FINAL0_D_INPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV10TCL_RC_FINAL0_D_INPUT_TEXTURE0_ARB 0x00000008 +#define NV10TCL_RC_FINAL0_D_INPUT_TEXTURE1_ARB 0x00000009 +#define NV10TCL_RC_FINAL0_D_INPUT_SPARE0_NV 0x0000000c +#define NV10TCL_RC_FINAL0_D_INPUT_SPARE1_NV 0x0000000d +#define NV10TCL_RC_FINAL0_D_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV10TCL_RC_FINAL0_D_INPUT_E_TIMES_F_NV 0x0000000f +#define NV10TCL_RC_FINAL0_D_COMPONENT_USAGE (1 << 4) +#define NV10TCL_RC_FINAL0_D_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_FINAL0_D_COMPONENT_USAGE_ALPHA 0x00000010 +#define NV10TCL_RC_FINAL0_D_MAPPING_SHIFT 5 +#define NV10TCL_RC_FINAL0_D_MAPPING_MASK 0x000000e0 +#define NV10TCL_RC_FINAL0_D_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_FINAL0_D_MAPPING_UNSIGNED_INVERT_NV 0x00000020 +#define NV10TCL_RC_FINAL0_D_MAPPING_EXPAND_NORMAL_NV 0x00000040 +#define NV10TCL_RC_FINAL0_D_MAPPING_EXPAND_NEGATE_NV 0x00000060 +#define NV10TCL_RC_FINAL0_D_MAPPING_HALF_BIAS_NORMAL_NV 0x00000080 +#define NV10TCL_RC_FINAL0_D_MAPPING_HALF_BIAS_NEGATE_NV 0x000000a0 +#define NV10TCL_RC_FINAL0_D_MAPPING_SIGNED_IDENTITY_NV 0x000000c0 +#define NV10TCL_RC_FINAL0_D_MAPPING_SIGNED_NEGATE_NV 0x000000e0 +#define NV10TCL_RC_FINAL0_C_INPUT_SHIFT 8 +#define NV10TCL_RC_FINAL0_C_INPUT_MASK 0x00000f00 +#define NV10TCL_RC_FINAL0_C_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_FINAL0_C_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV10TCL_RC_FINAL0_C_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV10TCL_RC_FINAL0_C_INPUT_FOG 0x00000300 +#define NV10TCL_RC_FINAL0_C_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV10TCL_RC_FINAL0_C_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV10TCL_RC_FINAL0_C_INPUT_TEXTURE0_ARB 0x00000800 +#define NV10TCL_RC_FINAL0_C_INPUT_TEXTURE1_ARB 0x00000900 +#define NV10TCL_RC_FINAL0_C_INPUT_SPARE0_NV 0x00000c00 +#define NV10TCL_RC_FINAL0_C_INPUT_SPARE1_NV 0x00000d00 +#define NV10TCL_RC_FINAL0_C_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV10TCL_RC_FINAL0_C_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV10TCL_RC_FINAL0_C_COMPONENT_USAGE (1 << 12) +#define NV10TCL_RC_FINAL0_C_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_FINAL0_C_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV10TCL_RC_FINAL0_C_MAPPING_SHIFT 13 +#define NV10TCL_RC_FINAL0_C_MAPPING_MASK 0x0000e000 +#define NV10TCL_RC_FINAL0_C_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_FINAL0_C_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV10TCL_RC_FINAL0_C_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV10TCL_RC_FINAL0_C_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV10TCL_RC_FINAL0_C_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV10TCL_RC_FINAL0_C_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV10TCL_RC_FINAL0_C_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV10TCL_RC_FINAL0_C_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV10TCL_RC_FINAL0_B_INPUT_SHIFT 16 +#define NV10TCL_RC_FINAL0_B_INPUT_MASK 0x000f0000 +#define NV10TCL_RC_FINAL0_B_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_FINAL0_B_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV10TCL_RC_FINAL0_B_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV10TCL_RC_FINAL0_B_INPUT_FOG 0x00030000 +#define NV10TCL_RC_FINAL0_B_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV10TCL_RC_FINAL0_B_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV10TCL_RC_FINAL0_B_INPUT_TEXTURE0_ARB 0x00080000 +#define NV10TCL_RC_FINAL0_B_INPUT_TEXTURE1_ARB 0x00090000 +#define NV10TCL_RC_FINAL0_B_INPUT_SPARE0_NV 0x000c0000 +#define NV10TCL_RC_FINAL0_B_INPUT_SPARE1_NV 0x000d0000 +#define NV10TCL_RC_FINAL0_B_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV10TCL_RC_FINAL0_B_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV10TCL_RC_FINAL0_B_COMPONENT_USAGE (1 << 20) +#define NV10TCL_RC_FINAL0_B_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_FINAL0_B_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV10TCL_RC_FINAL0_B_MAPPING_SHIFT 21 +#define NV10TCL_RC_FINAL0_B_MAPPING_MASK 0x00e00000 +#define NV10TCL_RC_FINAL0_B_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_FINAL0_B_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV10TCL_RC_FINAL0_B_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV10TCL_RC_FINAL0_B_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV10TCL_RC_FINAL0_B_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV10TCL_RC_FINAL0_B_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV10TCL_RC_FINAL0_B_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV10TCL_RC_FINAL0_B_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV10TCL_RC_FINAL0_A_INPUT_SHIFT 24 +#define NV10TCL_RC_FINAL0_A_INPUT_MASK 0x0f000000 +#define NV10TCL_RC_FINAL0_A_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_FINAL0_A_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV10TCL_RC_FINAL0_A_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV10TCL_RC_FINAL0_A_INPUT_FOG 0x03000000 +#define NV10TCL_RC_FINAL0_A_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV10TCL_RC_FINAL0_A_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV10TCL_RC_FINAL0_A_INPUT_TEXTURE0_ARB 0x08000000 +#define NV10TCL_RC_FINAL0_A_INPUT_TEXTURE1_ARB 0x09000000 +#define NV10TCL_RC_FINAL0_A_INPUT_SPARE0_NV 0x0c000000 +#define NV10TCL_RC_FINAL0_A_INPUT_SPARE1_NV 0x0d000000 +#define NV10TCL_RC_FINAL0_A_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV10TCL_RC_FINAL0_A_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV10TCL_RC_FINAL0_A_COMPONENT_USAGE (1 << 28) +#define NV10TCL_RC_FINAL0_A_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_FINAL0_A_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV10TCL_RC_FINAL0_A_MAPPING_SHIFT 29 +#define NV10TCL_RC_FINAL0_A_MAPPING_MASK 0xe0000000 +#define NV10TCL_RC_FINAL0_A_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_FINAL0_A_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV10TCL_RC_FINAL0_A_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV10TCL_RC_FINAL0_A_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV10TCL_RC_FINAL0_A_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV10TCL_RC_FINAL0_A_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV10TCL_RC_FINAL0_A_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV10TCL_RC_FINAL0_A_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV10TCL_RC_FINAL1 0x0000028c +#define NV10TCL_RC_FINAL1_COLOR_SUM_CLAMP (1 << 7) +#define NV10TCL_RC_FINAL1_G_INPUT_SHIFT 8 +#define NV10TCL_RC_FINAL1_G_INPUT_MASK 0x00000f00 +#define NV10TCL_RC_FINAL1_G_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_FINAL1_G_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV10TCL_RC_FINAL1_G_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV10TCL_RC_FINAL1_G_INPUT_FOG 0x00000300 +#define NV10TCL_RC_FINAL1_G_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV10TCL_RC_FINAL1_G_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV10TCL_RC_FINAL1_G_INPUT_TEXTURE0_ARB 0x00000800 +#define NV10TCL_RC_FINAL1_G_INPUT_TEXTURE1_ARB 0x00000900 +#define NV10TCL_RC_FINAL1_G_INPUT_SPARE0_NV 0x00000c00 +#define NV10TCL_RC_FINAL1_G_INPUT_SPARE1_NV 0x00000d00 +#define NV10TCL_RC_FINAL1_G_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV10TCL_RC_FINAL1_G_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV10TCL_RC_FINAL1_G_COMPONENT_USAGE (1 << 12) +#define NV10TCL_RC_FINAL1_G_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_FINAL1_G_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV10TCL_RC_FINAL1_G_MAPPING_SHIFT 13 +#define NV10TCL_RC_FINAL1_G_MAPPING_MASK 0x0000e000 +#define NV10TCL_RC_FINAL1_G_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_FINAL1_G_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV10TCL_RC_FINAL1_G_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV10TCL_RC_FINAL1_G_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV10TCL_RC_FINAL1_G_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV10TCL_RC_FINAL1_G_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV10TCL_RC_FINAL1_G_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV10TCL_RC_FINAL1_G_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV10TCL_RC_FINAL1_F_INPUT_SHIFT 16 +#define NV10TCL_RC_FINAL1_F_INPUT_MASK 0x000f0000 +#define NV10TCL_RC_FINAL1_F_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_FINAL1_F_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV10TCL_RC_FINAL1_F_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV10TCL_RC_FINAL1_F_INPUT_FOG 0x00030000 +#define NV10TCL_RC_FINAL1_F_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV10TCL_RC_FINAL1_F_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV10TCL_RC_FINAL1_F_INPUT_TEXTURE0_ARB 0x00080000 +#define NV10TCL_RC_FINAL1_F_INPUT_TEXTURE1_ARB 0x00090000 +#define NV10TCL_RC_FINAL1_F_INPUT_SPARE0_NV 0x000c0000 +#define NV10TCL_RC_FINAL1_F_INPUT_SPARE1_NV 0x000d0000 +#define NV10TCL_RC_FINAL1_F_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV10TCL_RC_FINAL1_F_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV10TCL_RC_FINAL1_F_COMPONENT_USAGE (1 << 20) +#define NV10TCL_RC_FINAL1_F_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_FINAL1_F_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV10TCL_RC_FINAL1_F_MAPPING_SHIFT 21 +#define NV10TCL_RC_FINAL1_F_MAPPING_MASK 0x00e00000 +#define NV10TCL_RC_FINAL1_F_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_FINAL1_F_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV10TCL_RC_FINAL1_F_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV10TCL_RC_FINAL1_F_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV10TCL_RC_FINAL1_F_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV10TCL_RC_FINAL1_F_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV10TCL_RC_FINAL1_F_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV10TCL_RC_FINAL1_F_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV10TCL_RC_FINAL1_E_INPUT_SHIFT 24 +#define NV10TCL_RC_FINAL1_E_INPUT_MASK 0x0f000000 +#define NV10TCL_RC_FINAL1_E_INPUT_ZERO 0x00000000 +#define NV10TCL_RC_FINAL1_E_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV10TCL_RC_FINAL1_E_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV10TCL_RC_FINAL1_E_INPUT_FOG 0x03000000 +#define NV10TCL_RC_FINAL1_E_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV10TCL_RC_FINAL1_E_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV10TCL_RC_FINAL1_E_INPUT_TEXTURE0_ARB 0x08000000 +#define NV10TCL_RC_FINAL1_E_INPUT_TEXTURE1_ARB 0x09000000 +#define NV10TCL_RC_FINAL1_E_INPUT_SPARE0_NV 0x0c000000 +#define NV10TCL_RC_FINAL1_E_INPUT_SPARE1_NV 0x0d000000 +#define NV10TCL_RC_FINAL1_E_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV10TCL_RC_FINAL1_E_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV10TCL_RC_FINAL1_E_COMPONENT_USAGE (1 << 28) +#define NV10TCL_RC_FINAL1_E_COMPONENT_USAGE_RGB 0x00000000 +#define NV10TCL_RC_FINAL1_E_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV10TCL_RC_FINAL1_E_MAPPING_SHIFT 29 +#define NV10TCL_RC_FINAL1_E_MAPPING_MASK 0xe0000000 +#define NV10TCL_RC_FINAL1_E_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV10TCL_RC_FINAL1_E_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV10TCL_RC_FINAL1_E_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV10TCL_RC_FINAL1_E_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV10TCL_RC_FINAL1_E_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV10TCL_RC_FINAL1_E_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV10TCL_RC_FINAL1_E_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV10TCL_RC_FINAL1_E_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV10TCL_LIGHT_MODEL 0x00000294 +#define NV10TCL_LIGHT_MODEL_COLOR_CONTROL (1 << 1) +#define NV10TCL_LIGHT_MODEL_LOCAL_VIEWER (1 << 16) +#define NV10TCL_COLOR_MATERIAL_ENABLE 0x00000298 +#define NV10TCL_COLOR_MATERIAL_ENABLE_SPECULAR (1 << 0) +#define NV10TCL_COLOR_MATERIAL_ENABLE_DIFFUSE (1 << 1) +#define NV10TCL_COLOR_MATERIAL_ENABLE_AMBIENT (1 << 2) +#define NV10TCL_COLOR_MATERIAL_ENABLE_EMISSION (1 << 3) +#define NV10TCL_FOG_MODE 0x0000029c +#define NV10TCL_FOG_MODE_EXP 0x00000800 +#define NV10TCL_FOG_MODE_EXP_2 0x00000802 +#define NV10TCL_FOG_MODE_EXP2 0x00000803 +#define NV10TCL_FOG_MODE_LINEAR 0x00000804 +#define NV10TCL_FOG_MODE_LINEAR_2 0x00002601 +#define NV10TCL_FOG_COORD_DIST 0x000002a0 +#define NV10TCL_FOG_COORD_DIST_COORD_FALSE 0x00000000 +#define NV10TCL_FOG_COORD_DIST_COORD_FRAGMENT_DEPTH_DISTANCE_EYE_RADIAL_NV 0x00000001 +#define NV10TCL_FOG_COORD_DIST_COORD_FRAGMENT_DEPTH_DISTANCE_EYE_PLANE_ABSOLUTE_NV 0x00000002 +#define NV10TCL_FOG_COORD_DIST_COORD_FOG 0x00000003 +#define NV10TCL_FOG_ENABLE 0x000002a4 +#define NV10TCL_FOG_COLOR 0x000002a8 +#define NV10TCL_FOG_COLOR_R_SHIFT 0 +#define NV10TCL_FOG_COLOR_R_MASK 0x000000ff +#define NV10TCL_FOG_COLOR_G_SHIFT 8 +#define NV10TCL_FOG_COLOR_G_MASK 0x0000ff00 +#define NV10TCL_FOG_COLOR_B_SHIFT 16 +#define NV10TCL_FOG_COLOR_B_MASK 0x00ff0000 +#define NV10TCL_FOG_COLOR_A_SHIFT 24 +#define NV10TCL_FOG_COLOR_A_MASK 0xff000000 +#define NV10TCL_VIEWPORT_CLIP_MODE 0x000002b4 +#define NV10TCL_VIEWPORT_CLIP_HORIZ(x) (0x000002c0+((x)*4)) +#define NV10TCL_VIEWPORT_CLIP_HORIZ__SIZE 0x00000008 +#define NV10TCL_VIEWPORT_CLIP_HORIZ_CLIP_L_SHIFT 0 +#define NV10TCL_VIEWPORT_CLIP_HORIZ_CLIP_L_MASK 0x000007ff +#define NV10TCL_VIEWPORT_CLIP_HORIZ_CLIP_LEFT_ENABLE (1 << 11) +#define NV10TCL_VIEWPORT_CLIP_HORIZ_CLIP_R_SHIFT 16 +#define NV10TCL_VIEWPORT_CLIP_HORIZ_CLIP_R_MASK 0x07ff0000 +#define NV10TCL_VIEWPORT_CLIP_HORIZ_CLIP_RIGHT_ENABLE (1 << 27) +#define NV10TCL_VIEWPORT_CLIP_VERT(x) (0x000002e0+((x)*4)) +#define NV10TCL_VIEWPORT_CLIP_VERT__SIZE 0x00000008 +#define NV10TCL_VIEWPORT_CLIP_VERT_CLIP_T_SHIFT 0 +#define NV10TCL_VIEWPORT_CLIP_VERT_CLIP_T_MASK 0x000007ff +#define NV10TCL_VIEWPORT_CLIP_VERT_CLIP_TOP_ENABLE (1 << 11) +#define NV10TCL_VIEWPORT_CLIP_VERT_CLIP_B_SHIFT 16 +#define NV10TCL_VIEWPORT_CLIP_VERT_CLIP_B_MASK 0x07ff0000 +#define NV10TCL_VIEWPORT_CLIP_VERT_CLIP_BOTTOM_ENABLE (1 << 27) +#define NV10TCL_ALPHA_FUNC_ENABLE 0x00000300 +#define NV10TCL_BLEND_FUNC_ENABLE 0x00000304 +#define NV10TCL_CULL_FACE_ENABLE 0x00000308 +#define NV10TCL_DEPTH_TEST_ENABLE 0x0000030c +#define NV10TCL_DITHER_ENABLE 0x00000310 +#define NV10TCL_LIGHTING_ENABLE 0x00000314 +#define NV10TCL_POINT_PARAMETERS_ENABLE 0x00000318 +#define NV10TCL_POINT_SMOOTH_ENABLE 0x0000031c +#define NV10TCL_LINE_SMOOTH_ENABLE 0x00000320 +#define NV10TCL_POLYGON_SMOOTH_ENABLE 0x00000324 +#define NV10TCL_VERTEX_WEIGHT_ENABLE 0x00000328 +#define NV10TCL_STENCIL_ENABLE 0x0000032c +#define NV10TCL_POLYGON_OFFSET_POINT_ENABLE 0x00000330 +#define NV10TCL_POLYGON_OFFSET_LINE_ENABLE 0x00000334 +#define NV10TCL_POLYGON_OFFSET_FILL_ENABLE 0x00000338 +#define NV10TCL_ALPHA_FUNC_FUNC 0x0000033c +#define NV10TCL_ALPHA_FUNC_FUNC_NEVER 0x00000200 +#define NV10TCL_ALPHA_FUNC_FUNC_LESS 0x00000201 +#define NV10TCL_ALPHA_FUNC_FUNC_EQUAL 0x00000202 +#define NV10TCL_ALPHA_FUNC_FUNC_LEQUAL 0x00000203 +#define NV10TCL_ALPHA_FUNC_FUNC_GREATER 0x00000204 +#define NV10TCL_ALPHA_FUNC_FUNC_GREATER 0x00000204 +#define NV10TCL_ALPHA_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV10TCL_ALPHA_FUNC_FUNC_GEQUAL 0x00000206 +#define NV10TCL_ALPHA_FUNC_FUNC_ALWAYS 0x00000207 +#define NV10TCL_ALPHA_FUNC_REF 0x00000340 +#define NV10TCL_BLEND_FUNC_SRC 0x00000344 +#define NV10TCL_BLEND_FUNC_SRC_ZERO 0x00000000 +#define NV10TCL_BLEND_FUNC_SRC_ONE 0x00000001 +#define NV10TCL_BLEND_FUNC_SRC_SRC_COLOR 0x00000300 +#define NV10TCL_BLEND_FUNC_SRC_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV10TCL_BLEND_FUNC_SRC_SRC_ALPHA 0x00000302 +#define NV10TCL_BLEND_FUNC_SRC_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV10TCL_BLEND_FUNC_SRC_DST_ALPHA 0x00000304 +#define NV10TCL_BLEND_FUNC_SRC_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV10TCL_BLEND_FUNC_SRC_DST_COLOR 0x00000306 +#define NV10TCL_BLEND_FUNC_SRC_ONE_MINUS_DST_COLOR 0x00000307 +#define NV10TCL_BLEND_FUNC_SRC_SRC_ALPHA_SATURATE 0x00000308 +#define NV10TCL_BLEND_FUNC_SRC_CONSTANT_COLOR 0x00008001 +#define NV10TCL_BLEND_FUNC_SRC_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV10TCL_BLEND_FUNC_SRC_CONSTANT_ALPHA 0x00008003 +#define NV10TCL_BLEND_FUNC_SRC_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV10TCL_BLEND_FUNC_DST 0x00000348 +#define NV10TCL_BLEND_FUNC_DST_ZERO 0x00000000 +#define NV10TCL_BLEND_FUNC_DST_ONE 0x00000001 +#define NV10TCL_BLEND_FUNC_DST_SRC_COLOR 0x00000300 +#define NV10TCL_BLEND_FUNC_DST_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV10TCL_BLEND_FUNC_DST_SRC_ALPHA 0x00000302 +#define NV10TCL_BLEND_FUNC_DST_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV10TCL_BLEND_FUNC_DST_DST_ALPHA 0x00000304 +#define NV10TCL_BLEND_FUNC_DST_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV10TCL_BLEND_FUNC_DST_DST_COLOR 0x00000306 +#define NV10TCL_BLEND_FUNC_DST_ONE_MINUS_DST_COLOR 0x00000307 +#define NV10TCL_BLEND_FUNC_DST_SRC_ALPHA_SATURATE 0x00000308 +#define NV10TCL_BLEND_FUNC_DST_CONSTANT_COLOR 0x00008001 +#define NV10TCL_BLEND_FUNC_DST_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV10TCL_BLEND_FUNC_DST_CONSTANT_ALPHA 0x00008003 +#define NV10TCL_BLEND_FUNC_DST_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV10TCL_BLEND_COLOR 0x0000034c +#define NV10TCL_BLEND_COLOR_B_SHIFT 0 +#define NV10TCL_BLEND_COLOR_B_MASK 0x000000ff +#define NV10TCL_BLEND_COLOR_G_SHIFT 8 +#define NV10TCL_BLEND_COLOR_G_MASK 0x0000ff00 +#define NV10TCL_BLEND_COLOR_R_SHIFT 16 +#define NV10TCL_BLEND_COLOR_R_MASK 0x00ff0000 +#define NV10TCL_BLEND_COLOR_A_SHIFT 24 +#define NV10TCL_BLEND_COLOR_A_MASK 0xff000000 +#define NV10TCL_BLEND_EQUATION 0x00000350 +#define NV10TCL_BLEND_EQUATION_FUNC_ADD 0x00008006 +#define NV10TCL_BLEND_EQUATION_MIN 0x00008007 +#define NV10TCL_BLEND_EQUATION_MAX 0x00008008 +#define NV10TCL_BLEND_EQUATION_FUNC_SUBTRACT 0x0000800a +#define NV10TCL_BLEND_EQUATION_FUNC_REVERSE_SUBTRACT 0x0000800b +#define NV10TCL_DEPTH_FUNC 0x00000354 +#define NV10TCL_DEPTH_FUNC_NEVER 0x00000200 +#define NV10TCL_DEPTH_FUNC_LESS 0x00000201 +#define NV10TCL_DEPTH_FUNC_EQUAL 0x00000202 +#define NV10TCL_DEPTH_FUNC_LEQUAL 0x00000203 +#define NV10TCL_DEPTH_FUNC_GREATER 0x00000204 +#define NV10TCL_DEPTH_FUNC_GREATER 0x00000204 +#define NV10TCL_DEPTH_FUNC_NOTEQUAL 0x00000205 +#define NV10TCL_DEPTH_FUNC_GEQUAL 0x00000206 +#define NV10TCL_DEPTH_FUNC_ALWAYS 0x00000207 +#define NV10TCL_COLOR_MASK 0x00000358 +#define NV10TCL_COLOR_MASK_B (1 << 0) +#define NV10TCL_COLOR_MASK_G (1 << 8) +#define NV10TCL_COLOR_MASK_R (1 << 16) +#define NV10TCL_COLOR_MASK_A (1 << 24) +#define NV10TCL_DEPTH_WRITE_ENABLE 0x0000035c +#define NV10TCL_STENCIL_MASK 0x00000360 +#define NV10TCL_STENCIL_FUNC_FUNC 0x00000364 +#define NV10TCL_STENCIL_FUNC_FUNC_NEVER 0x00000200 +#define NV10TCL_STENCIL_FUNC_FUNC_LESS 0x00000201 +#define NV10TCL_STENCIL_FUNC_FUNC_EQUAL 0x00000202 +#define NV10TCL_STENCIL_FUNC_FUNC_LEQUAL 0x00000203 +#define NV10TCL_STENCIL_FUNC_FUNC_GREATER 0x00000204 +#define NV10TCL_STENCIL_FUNC_FUNC_GREATER 0x00000204 +#define NV10TCL_STENCIL_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV10TCL_STENCIL_FUNC_FUNC_GEQUAL 0x00000206 +#define NV10TCL_STENCIL_FUNC_FUNC_ALWAYS 0x00000207 +#define NV10TCL_STENCIL_FUNC_REF 0x00000368 +#define NV10TCL_STENCIL_FUNC_MASK 0x0000036c +#define NV10TCL_STENCIL_OP_FAIL 0x00000370 +#define NV10TCL_STENCIL_OP_FAIL_ZERO 0x00000000 +#define NV10TCL_STENCIL_OP_FAIL_INVERT 0x0000150a +#define NV10TCL_STENCIL_OP_FAIL_KEEP 0x00001e00 +#define NV10TCL_STENCIL_OP_FAIL_REPLACE 0x00001e01 +#define NV10TCL_STENCIL_OP_FAIL_INCR 0x00001e02 +#define NV10TCL_STENCIL_OP_FAIL_DECR 0x00001e03 +#define NV10TCL_STENCIL_OP_FAIL_INCR_WRAP 0x00008507 +#define NV10TCL_STENCIL_OP_FAIL_DECR_WRAP 0x00008508 +#define NV10TCL_STENCIL_OP_ZFAIL 0x00000374 +#define NV10TCL_STENCIL_OP_ZFAIL_ZERO 0x00000000 +#define NV10TCL_STENCIL_OP_ZFAIL_INVERT 0x0000150a +#define NV10TCL_STENCIL_OP_ZFAIL_KEEP 0x00001e00 +#define NV10TCL_STENCIL_OP_ZFAIL_REPLACE 0x00001e01 +#define NV10TCL_STENCIL_OP_ZFAIL_INCR 0x00001e02 +#define NV10TCL_STENCIL_OP_ZFAIL_DECR 0x00001e03 +#define NV10TCL_STENCIL_OP_ZFAIL_INCR_WRAP 0x00008507 +#define NV10TCL_STENCIL_OP_ZFAIL_DECR_WRAP 0x00008508 +#define NV10TCL_STENCIL_OP_ZPASS 0x00000378 +#define NV10TCL_STENCIL_OP_ZPASS_ZERO 0x00000000 +#define NV10TCL_STENCIL_OP_ZPASS_INVERT 0x0000150a +#define NV10TCL_STENCIL_OP_ZPASS_KEEP 0x00001e00 +#define NV10TCL_STENCIL_OP_ZPASS_REPLACE 0x00001e01 +#define NV10TCL_STENCIL_OP_ZPASS_INCR 0x00001e02 +#define NV10TCL_STENCIL_OP_ZPASS_DECR 0x00001e03 +#define NV10TCL_STENCIL_OP_ZPASS_INCR_WRAP 0x00008507 +#define NV10TCL_STENCIL_OP_ZPASS_DECR_WRAP 0x00008508 +#define NV10TCL_SHADE_MODEL 0x0000037c +#define NV10TCL_SHADE_MODEL_FLAT 0x00001d00 +#define NV10TCL_SHADE_MODEL_SMOOTH 0x00001d01 +#define NV10TCL_LINE_WIDTH 0x00000380 +#define NV10TCL_POLYGON_OFFSET_FACTOR 0x00000384 +#define NV10TCL_POLYGON_OFFSET_UNITS 0x00000388 +#define NV10TCL_POLYGON_MODE_FRONT 0x0000038c +#define NV10TCL_POLYGON_MODE_FRONT_POINT 0x00001b00 +#define NV10TCL_POLYGON_MODE_FRONT_LINE 0x00001b01 +#define NV10TCL_POLYGON_MODE_FRONT_FILL 0x00001b02 +#define NV10TCL_POLYGON_MODE_BACK 0x00000390 +#define NV10TCL_POLYGON_MODE_BACK_POINT 0x00001b00 +#define NV10TCL_POLYGON_MODE_BACK_LINE 0x00001b01 +#define NV10TCL_POLYGON_MODE_BACK_FILL 0x00001b02 +#define NV10TCL_DEPTH_RANGE_NEAR 0x00000394 +#define NV10TCL_DEPTH_RANGE_FAR 0x00000398 +#define NV10TCL_CULL_FACE 0x0000039c +#define NV10TCL_CULL_FACE_FRONT 0x00000404 +#define NV10TCL_CULL_FACE_BACK 0x00000405 +#define NV10TCL_CULL_FACE_FRONT_AND_BACK 0x00000408 +#define NV10TCL_FRONT_FACE 0x000003a0 +#define NV10TCL_FRONT_FACE_CW 0x00000900 +#define NV10TCL_FRONT_FACE_CCW 0x00000901 +#define NV10TCL_NORMALIZE_ENABLE 0x000003a4 +#define NV10TCL_COLOR_MATERIAL_R 0x000003a8 +#define NV10TCL_COLOR_MATERIAL_G 0x000003ac +#define NV10TCL_COLOR_MATERIAL_B 0x000003b0 +#define NV10TCL_COLOR_MATERIAL_A 0x000003b4 +#define NV10TCL_COLOR_CONTROL 0x000003b8 +#define NV10TCL_ENABLED_LIGHTS 0x000003bc +#define NV10TCL_ENABLED_LIGHTS_LIGHT0 (1 << 0) +#define NV10TCL_ENABLED_LIGHTS_LIGHT1 (1 << 2) +#define NV10TCL_ENABLED_LIGHTS_LIGHT2 (1 << 4) +#define NV10TCL_ENABLED_LIGHTS_LIGHT3 (1 << 6) +#define NV10TCL_ENABLED_LIGHTS_LIGHT4 (1 << 8) +#define NV10TCL_ENABLED_LIGHTS_LIGHT5 (1 << 10) +#define NV10TCL_ENABLED_LIGHTS_LIGHT6 (1 << 12) +#define NV10TCL_ENABLED_LIGHTS_LIGHT7 (1 << 14) +#define NV10TCL_TX_GEN_S(x) (0x000003c0+((x)*16)) +#define NV10TCL_TX_GEN_S__SIZE 0x00000002 +#define NV10TCL_TX_GEN_S_FALSE 0x00000000 +#define NV10TCL_TX_GEN_S_EYE_LINEAR 0x00002400 +#define NV10TCL_TX_GEN_S_OBJECT_LINEAR 0x00002401 +#define NV10TCL_TX_GEN_S_SPHERE_MAP 0x00002402 +#define NV10TCL_TX_GEN_S_NORMAL_MAP 0x00008511 +#define NV10TCL_TX_GEN_S_REFLECTION_MAP 0x00008512 +#define NV10TCL_TX_GEN_T(x) (0x000003c4+((x)*16)) +#define NV10TCL_TX_GEN_T__SIZE 0x00000002 +#define NV10TCL_TX_GEN_T_FALSE 0x00000000 +#define NV10TCL_TX_GEN_T_EYE_LINEAR 0x00002400 +#define NV10TCL_TX_GEN_T_OBJECT_LINEAR 0x00002401 +#define NV10TCL_TX_GEN_T_SPHERE_MAP 0x00002402 +#define NV10TCL_TX_GEN_T_NORMAL_MAP 0x00008511 +#define NV10TCL_TX_GEN_T_REFLECTION_MAP 0x00008512 +#define NV10TCL_TX_GEN_R(x) (0x000003c8+((x)*16)) +#define NV10TCL_TX_GEN_R__SIZE 0x00000002 +#define NV10TCL_TX_GEN_R_FALSE 0x00000000 +#define NV10TCL_TX_GEN_R_EYE_LINEAR 0x00002400 +#define NV10TCL_TX_GEN_R_OBJECT_LINEAR 0x00002401 +#define NV10TCL_TX_GEN_R_SPHERE_MAP 0x00002402 +#define NV10TCL_TX_GEN_R_NORMAL_MAP 0x00008511 +#define NV10TCL_TX_GEN_R_REFLECTION_MAP 0x00008512 +#define NV10TCL_TX_GEN_Q(x) (0x000003cc+((x)*16)) +#define NV10TCL_TX_GEN_Q__SIZE 0x00000002 +#define NV10TCL_TX_GEN_Q_FALSE 0x00000000 +#define NV10TCL_TX_GEN_Q_EYE_LINEAR 0x00002400 +#define NV10TCL_TX_GEN_Q_OBJECT_LINEAR 0x00002401 +#define NV10TCL_TX_GEN_Q_SPHERE_MAP 0x00002402 +#define NV10TCL_TX_GEN_Q_NORMAL_MAP 0x00008511 +#define NV10TCL_TX_GEN_Q_REFLECTION_MAP 0x00008512 +#define NV10TCL_TX_MATRIX_ENABLE(x) (0x000003e0+((x)*4)) +#define NV10TCL_TX_MATRIX_ENABLE__SIZE 0x00000002 +#define NV10TCL_VIEW_MATRIX_ENABLE 0x000003e8 +#define NV10TCL_VIEW_MATRIX_ENABLE_MODELVIEW1 (1 << 0) +#define NV10TCL_VIEW_MATRIX_ENABLE_MODELVIEW0 (1 << 1) +#define NV10TCL_VIEW_MATRIX_ENABLE_PROJECTION (1 << 2) +#define NV10TCL_POINT_SIZE 0x000003ec +#define NV10TCL_MODELVIEW0_MATRIX(x) (0x00000400+((x)*4)) +#define NV10TCL_MODELVIEW0_MATRIX__SIZE 0x00000010 +#define NV10TCL_MODELVIEW1_MATRIX(x) (0x00000440+((x)*4)) +#define NV10TCL_MODELVIEW1_MATRIX__SIZE 0x00000010 +#define NV10TCL_INVERSE_MODELVIEW0_MATRIX(x) (0x00000480+((x)*4)) +#define NV10TCL_INVERSE_MODELVIEW0_MATRIX__SIZE 0x00000010 +#define NV10TCL_INVERSE_MODELVIEW1_MATRIX(x) (0x000004c0+((x)*4)) +#define NV10TCL_INVERSE_MODELVIEW1_MATRIX__SIZE 0x00000010 +#define NV10TCL_PROJECTION_MATRIX(x) (0x00000500+((x)*4)) +#define NV10TCL_PROJECTION_MATRIX__SIZE 0x00000010 +#define NV10TCL_TX0_MATRIX(x) (0x00000540+((x)*4)) +#define NV10TCL_TX0_MATRIX__SIZE 0x00000010 +#define NV10TCL_TX1_MATRIX(x) (0x00000580+((x)*4)) +#define NV10TCL_TX1_MATRIX__SIZE 0x00000010 +#define NV10TCL_CLIP_PLANE_A(x) (0x00000600+((x)*16)) +#define NV10TCL_CLIP_PLANE_A__SIZE 0x00000008 +#define NV10TCL_CLIP_PLANE_B(x) (0x00000604+((x)*16)) +#define NV10TCL_CLIP_PLANE_B__SIZE 0x00000008 +#define NV10TCL_CLIP_PLANE_C(x) (0x00000608+((x)*16)) +#define NV10TCL_CLIP_PLANE_C__SIZE 0x00000008 +#define NV10TCL_CLIP_PLANE_D(x) (0x0000060c+((x)*16)) +#define NV10TCL_CLIP_PLANE_D__SIZE 0x00000008 +#define NV10TCL_FOG_EQUATION_CONSTANT 0x00000680 +#define NV10TCL_FOG_EQUATION_LINEAR 0x00000684 +#define NV10TCL_FOG_EQUATION_QUADRATIC 0x00000688 +#define NV10TCL_FRONT_MATERIAL_SHININESS(x) (0x000006a0+((x)*4)) +#define NV10TCL_FRONT_MATERIAL_SHININESS__SIZE 0x00000006 +#define NV10TCL_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_R 0x000006c4 +#define NV10TCL_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_G 0x000006c8 +#define NV10TCL_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_B 0x000006cc +#define NV10TCL_VIEWPORT_SCALE_X 0x000006e8 +#define NV10TCL_VIEWPORT_SCALE_Y 0x000006ec +#define NV10TCL_VIEWPORT_SCALE_Z 0x000006f0 +#define NV10TCL_VIEWPORT_SCALE_W 0x000006f4 +#define NV10TCL_POINT_PARAMETER(x) (0x000006f8+((x)*4)) +#define NV10TCL_POINT_PARAMETER__SIZE 0x00000008 +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R(x) (0x00000800+((x)*128)) +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R__SIZE 0x00000008 +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_G(x) (0x00000804+((x)*128)) +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_G__SIZE 0x00000008 +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_B(x) (0x00000808+((x)*128)) +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_B__SIZE 0x00000008 +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R(x) (0x0000080c+((x)*128)) +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R__SIZE 0x00000008 +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_G(x) (0x00000810+((x)*128)) +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_G__SIZE 0x00000008 +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_B(x) (0x00000814+((x)*128)) +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_B__SIZE 0x00000008 +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R(x) (0x00000818+((x)*128)) +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R__SIZE 0x00000008 +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_G(x) (0x0000081c+((x)*128)) +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_G__SIZE 0x00000008 +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_B(x) (0x00000820+((x)*128)) +#define NV10TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_B__SIZE 0x00000008 +#define NV10TCL_LIGHT_HALF_VECTOR_X(x) (0x00000828+((x)*128)) +#define NV10TCL_LIGHT_HALF_VECTOR_X__SIZE 0x00000008 +#define NV10TCL_LIGHT_HALF_VECTOR_Y(x) (0x0000082c+((x)*128)) +#define NV10TCL_LIGHT_HALF_VECTOR_Y__SIZE 0x00000008 +#define NV10TCL_LIGHT_HALF_VECTOR_Z(x) (0x00000830+((x)*128)) +#define NV10TCL_LIGHT_HALF_VECTOR_Z__SIZE 0x00000008 +#define NV10TCL_LIGHT_DIRECTION_X(x) (0x00000834+((x)*128)) +#define NV10TCL_LIGHT_DIRECTION_X__SIZE 0x00000008 +#define NV10TCL_LIGHT_DIRECTION_Y(x) (0x00000838+((x)*128)) +#define NV10TCL_LIGHT_DIRECTION_Y__SIZE 0x00000008 +#define NV10TCL_LIGHT_DIRECTION_Z(x) (0x0000083c+((x)*128)) +#define NV10TCL_LIGHT_DIRECTION_Z__SIZE 0x00000008 +#define NV10TCL_LIGHT_SPOT_CUTOFF_A(x) (0x00000840+((x)*128)) +#define NV10TCL_LIGHT_SPOT_CUTOFF_A__SIZE 0x00000008 +#define NV10TCL_LIGHT_SPOT_CUTOFF_B(x) (0x00000844+((x)*128)) +#define NV10TCL_LIGHT_SPOT_CUTOFF_B__SIZE 0x00000008 +#define NV10TCL_LIGHT_SPOT_CUTOFF_C(x) (0x00000848+((x)*128)) +#define NV10TCL_LIGHT_SPOT_CUTOFF_C__SIZE 0x00000008 +#define NV10TCL_LIGHT_SPOT_DIR_X(x) (0x0000084c+((x)*128)) +#define NV10TCL_LIGHT_SPOT_DIR_X__SIZE 0x00000008 +#define NV10TCL_LIGHT_SPOT_DIR_Y(x) (0x00000850+((x)*128)) +#define NV10TCL_LIGHT_SPOT_DIR_Y__SIZE 0x00000008 +#define NV10TCL_LIGHT_SPOT_DIR_Z(x) (0x00000854+((x)*128)) +#define NV10TCL_LIGHT_SPOT_DIR_Z__SIZE 0x00000008 +#define NV10TCL_LIGHT_SPOT_CUTOFF_D(x) (0x00000858+((x)*128)) +#define NV10TCL_LIGHT_SPOT_CUTOFF_D__SIZE 0x00000008 +#define NV10TCL_LIGHT_POSITION_X(x) (0x0000085c+((x)*128)) +#define NV10TCL_LIGHT_POSITION_X__SIZE 0x00000008 +#define NV10TCL_LIGHT_POSITION_Y(x) (0x00000860+((x)*128)) +#define NV10TCL_LIGHT_POSITION_Y__SIZE 0x00000008 +#define NV10TCL_LIGHT_POSITION_Z(x) (0x00000864+((x)*128)) +#define NV10TCL_LIGHT_POSITION_Z__SIZE 0x00000008 +#define NV10TCL_LIGHT_ATTENUATION_CONSTANT(x) (0x00000868+((x)*128)) +#define NV10TCL_LIGHT_ATTENUATION_CONSTANT__SIZE 0x00000008 +#define NV10TCL_LIGHT_ATTENUATION_LINEAR(x) (0x0000086c+((x)*128)) +#define NV10TCL_LIGHT_ATTENUATION_LINEAR__SIZE 0x00000008 +#define NV10TCL_LIGHT_ATTENUATION_QUADRATIC(x) (0x00000870+((x)*128)) +#define NV10TCL_LIGHT_ATTENUATION_QUADRATIC__SIZE 0x00000008 +#define NV10TCL_VERTEX_POS_3F_X 0x00000c00 +#define NV10TCL_VERTEX_POS_3F_Y 0x00000c04 +#define NV10TCL_VERTEX_POS_3F_Z 0x00000c08 +#define NV10TCL_VERTEX_POS_4F_X 0x00000c18 +#define NV10TCL_VERTEX_POS_4F_Y 0x00000c1c +#define NV10TCL_VERTEX_POS_4F_Z 0x00000c20 +#define NV10TCL_VERTEX_POS_4F_W 0x00000c24 +#define NV10TCL_VERTEX_NOR_3F_X 0x00000c30 +#define NV10TCL_VERTEX_NOR_3F_Y 0x00000c34 +#define NV10TCL_VERTEX_NOR_3F_Z 0x00000c38 +#define NV10TCL_VERTEX_NOR_3I_XY 0x00000c40 +#define NV10TCL_VERTEX_NOR_3I_XY_X_SHIFT 0 +#define NV10TCL_VERTEX_NOR_3I_XY_X_MASK 0x0000ffff +#define NV10TCL_VERTEX_NOR_3I_XY_Y_SHIFT 16 +#define NV10TCL_VERTEX_NOR_3I_XY_Y_MASK 0xffff0000 +#define NV10TCL_VERTEX_NOR_3I_Z 0x00000c44 +#define NV10TCL_VERTEX_NOR_3I_Z_Z_SHIFT 0 +#define NV10TCL_VERTEX_NOR_3I_Z_Z_MASK 0x0000ffff +#define NV10TCL_VERTEX_COL_4F_R 0x00000c50 +#define NV10TCL_VERTEX_COL_4F_G 0x00000c54 +#define NV10TCL_VERTEX_COL_4F_B 0x00000c58 +#define NV10TCL_VERTEX_COL_4F_A 0x00000c5c +#define NV10TCL_VERTEX_COL_3F_R 0x00000c60 +#define NV10TCL_VERTEX_COL_3F_G 0x00000c64 +#define NV10TCL_VERTEX_COL_3F_B 0x00000c68 +#define NV10TCL_VERTEX_COL_4I 0x00000c6c +#define NV10TCL_VERTEX_COL_4I_R_SHIFT 0 +#define NV10TCL_VERTEX_COL_4I_R_MASK 0x000000ff +#define NV10TCL_VERTEX_COL_4I_G_SHIFT 8 +#define NV10TCL_VERTEX_COL_4I_G_MASK 0x0000ff00 +#define NV10TCL_VERTEX_COL_4I_B_SHIFT 16 +#define NV10TCL_VERTEX_COL_4I_B_MASK 0x00ff0000 +#define NV10TCL_VERTEX_COL_4I_A_SHIFT 24 +#define NV10TCL_VERTEX_COL_4I_A_MASK 0xff000000 +#define NV10TCL_VERTEX_COL2_3F_R 0x00000c80 +#define NV10TCL_VERTEX_COL2_3F_G 0x00000c84 +#define NV10TCL_VERTEX_COL2_3F_B 0x00000c88 +#define NV10TCL_VERTEX_COL2_3I 0x00000c8c +#define NV10TCL_VERTEX_COL2_3I_R_SHIFT 0 +#define NV10TCL_VERTEX_COL2_3I_R_MASK 0x000000ff +#define NV10TCL_VERTEX_COL2_3I_G_SHIFT 8 +#define NV10TCL_VERTEX_COL2_3I_G_MASK 0x0000ff00 +#define NV10TCL_VERTEX_COL2_3I_B_SHIFT 16 +#define NV10TCL_VERTEX_COL2_3I_B_MASK 0x00ff0000 +#define NV10TCL_VERTEX_TX0_2F_S 0x00000c90 +#define NV10TCL_VERTEX_TX0_2F_T 0x00000c94 +#define NV10TCL_VERTEX_TX0_2I 0x00000c98 +#define NV10TCL_VERTEX_TX0_2I_S_SHIFT 0 +#define NV10TCL_VERTEX_TX0_2I_S_MASK 0x0000ffff +#define NV10TCL_VERTEX_TX0_2I_T_SHIFT 16 +#define NV10TCL_VERTEX_TX0_2I_T_MASK 0xffff0000 +#define NV10TCL_VERTEX_TX0_4F_S 0x00000ca0 +#define NV10TCL_VERTEX_TX0_4F_T 0x00000ca4 +#define NV10TCL_VERTEX_TX0_4F_R 0x00000ca8 +#define NV10TCL_VERTEX_TX0_4F_Q 0x00000cac +#define NV10TCL_VERTEX_TX0_4I_ST 0x00000cb0 +#define NV10TCL_VERTEX_TX0_4I_ST_S_SHIFT 0 +#define NV10TCL_VERTEX_TX0_4I_ST_S_MASK 0x0000ffff +#define NV10TCL_VERTEX_TX0_4I_ST_T_SHIFT 16 +#define NV10TCL_VERTEX_TX0_4I_ST_T_MASK 0xffff0000 +#define NV10TCL_VERTEX_TX0_4I_RQ 0x00000cb4 +#define NV10TCL_VERTEX_TX0_4I_RQ_R_SHIFT 0 +#define NV10TCL_VERTEX_TX0_4I_RQ_R_MASK 0x0000ffff +#define NV10TCL_VERTEX_TX0_4I_RQ_Q_SHIFT 16 +#define NV10TCL_VERTEX_TX0_4I_RQ_Q_MASK 0xffff0000 +#define NV10TCL_VERTEX_TX1_2F_S 0x00000cb8 +#define NV10TCL_VERTEX_TX1_2F_T 0x00000cbc +#define NV10TCL_VERTEX_TX1_2I 0x00000cc0 +#define NV10TCL_VERTEX_TX1_2I_S_SHIFT 0 +#define NV10TCL_VERTEX_TX1_2I_S_MASK 0x0000ffff +#define NV10TCL_VERTEX_TX1_2I_T_SHIFT 16 +#define NV10TCL_VERTEX_TX1_2I_T_MASK 0xffff0000 +#define NV10TCL_VERTEX_TX1_4F_S 0x00000cc8 +#define NV10TCL_VERTEX_TX1_4F_T 0x00000ccc +#define NV10TCL_VERTEX_TX1_4F_R 0x00000cd0 +#define NV10TCL_VERTEX_TX1_4F_Q 0x00000cd4 +#define NV10TCL_VERTEX_TX1_4I_ST 0x00000cd8 +#define NV10TCL_VERTEX_TX1_4I_ST_S_SHIFT 0 +#define NV10TCL_VERTEX_TX1_4I_ST_S_MASK 0x0000ffff +#define NV10TCL_VERTEX_TX1_4I_ST_T_SHIFT 16 +#define NV10TCL_VERTEX_TX1_4I_ST_T_MASK 0xffff0000 +#define NV10TCL_VERTEX_TX1_4I_RQ 0x00000cdc +#define NV10TCL_VERTEX_TX1_4I_RQ_R_SHIFT 0 +#define NV10TCL_VERTEX_TX1_4I_RQ_R_MASK 0x0000ffff +#define NV10TCL_VERTEX_TX1_4I_RQ_Q_SHIFT 16 +#define NV10TCL_VERTEX_TX1_4I_RQ_Q_MASK 0xffff0000 +#define NV10TCL_VERTEX_FOG_1F 0x00000ce0 +#define NV10TCL_VERTEX_WGH_1F 0x00000ce4 +#define NV10TCL_EDGEFLAG_ENABLE 0x00000cec +#define NV10TCL_VERTEX_ARRAY_VALIDATE 0x00000cf0 +#define NV10TCL_VERTEX_ARRAY_ATTRIB_OFFSET(x) (0x00000d00+((x)*8)) +#define NV10TCL_VERTEX_ARRAY_ATTRIB_OFFSET__SIZE 0x00000008 +#define NV10TCL_VERTEX_ARRAY_ATTRIB_FORMAT(x) (0x00000d04+((x)*8)) +#define NV10TCL_VERTEX_ARRAY_ATTRIB_FORMAT__SIZE 0x00000008 +#define NV10TCL_VERTEX_ARRAY_ATTRIB_FORMAT_TYPE_SHIFT 0 +#define NV10TCL_VERTEX_ARRAY_ATTRIB_FORMAT_TYPE_MASK 0x0000000f +#define NV10TCL_VERTEX_ARRAY_ATTRIB_FORMAT_FIELDS_SHIFT 4 +#define NV10TCL_VERTEX_ARRAY_ATTRIB_FORMAT_FIELDS_MASK 0x000000f0 +#define NV10TCL_VERTEX_ARRAY_ATTRIB_FORMAT_STRIDE_SHIFT 8 +#define NV10TCL_VERTEX_ARRAY_ATTRIB_FORMAT_STRIDE_MASK 0x0000ff00 +#define NV10TCL_VERTEX_ARRAY_OFFSET_POS 0x00000d00 +#define NV10TCL_VERTEX_ARRAY_FORMAT_POS 0x00000d04 +#define NV10TCL_VERTEX_ARRAY_FORMAT_POS_TYPE_SHIFT 0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_POS_TYPE_MASK 0x0000000f +#define NV10TCL_VERTEX_ARRAY_FORMAT_POS_FIELDS_SHIFT 4 +#define NV10TCL_VERTEX_ARRAY_FORMAT_POS_FIELDS_MASK 0x000000f0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_POS_STRIDE_SHIFT 8 +#define NV10TCL_VERTEX_ARRAY_FORMAT_POS_STRIDE_MASK 0x0000ff00 +#define NV10TCL_VERTEX_ARRAY_OFFSET_COL 0x00000d08 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL 0x00000d0c +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL_TYPE_SHIFT 0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL_TYPE_MASK 0x0000000f +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL_FIELDS_SHIFT 4 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL_FIELDS_MASK 0x000000f0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL_STRIDE_SHIFT 8 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL_STRIDE_MASK 0x0000ff00 +#define NV10TCL_VERTEX_ARRAY_OFFSET_COL2 0x00000d10 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL2 0x00000d14 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL2_TYPE_SHIFT 0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL2_TYPE_MASK 0x0000000f +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL2_FIELDS_SHIFT 4 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL2_FIELDS_MASK 0x000000f0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL2_STRIDE_SHIFT 8 +#define NV10TCL_VERTEX_ARRAY_FORMAT_COL2_STRIDE_MASK 0x0000ff00 +#define NV10TCL_VERTEX_ARRAY_OFFSET_TX0 0x00000d18 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX0 0x00000d1c +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX0_TYPE_SHIFT 0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX0_TYPE_MASK 0x0000000f +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX0_FIELDS_SHIFT 4 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX0_FIELDS_MASK 0x000000f0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX0_STRIDE_SHIFT 8 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX0_STRIDE_MASK 0x0000ff00 +#define NV10TCL_VERTEX_ARRAY_OFFSET_TX1 0x00000d20 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX1 0x00000d24 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX1_TYPE_SHIFT 0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX1_TYPE_MASK 0x0000000f +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX1_FIELDS_SHIFT 4 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX1_FIELDS_MASK 0x000000f0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX1_STRIDE_SHIFT 8 +#define NV10TCL_VERTEX_ARRAY_FORMAT_TX1_STRIDE_MASK 0x0000ff00 +#define NV10TCL_VERTEX_ARRAY_OFFSET_NOR 0x00000d28 +#define NV10TCL_VERTEX_ARRAY_FORMAT_NOR 0x00000d2c +#define NV10TCL_VERTEX_ARRAY_FORMAT_NOR_TYPE_SHIFT 0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_NOR_TYPE_MASK 0x0000000f +#define NV10TCL_VERTEX_ARRAY_FORMAT_NOR_FIELDS_SHIFT 4 +#define NV10TCL_VERTEX_ARRAY_FORMAT_NOR_FIELDS_MASK 0x000000f0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_NOR_STRIDE_SHIFT 8 +#define NV10TCL_VERTEX_ARRAY_FORMAT_NOR_STRIDE_MASK 0x0000ff00 +#define NV10TCL_VERTEX_ARRAY_OFFSET_WGH 0x00000d30 +#define NV10TCL_VERTEX_ARRAY_FORMAT_WGH 0x00000d34 +#define NV10TCL_VERTEX_ARRAY_FORMAT_WGH_TYPE_SHIFT 0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_WGH_TYPE_MASK 0x0000000f +#define NV10TCL_VERTEX_ARRAY_FORMAT_WGH_FIELDS_SHIFT 4 +#define NV10TCL_VERTEX_ARRAY_FORMAT_WGH_FIELDS_MASK 0x000000f0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_WGH_STRIDE_SHIFT 8 +#define NV10TCL_VERTEX_ARRAY_FORMAT_WGH_STRIDE_MASK 0x0000ff00 +#define NV10TCL_VERTEX_ARRAY_OFFSET_FOG 0x00000d38 +#define NV10TCL_VERTEX_ARRAY_FORMAT_FOG 0x00000d3c +#define NV10TCL_VERTEX_ARRAY_FORMAT_FOG_TYPE_SHIFT 0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_FOG_TYPE_MASK 0x0000000f +#define NV10TCL_VERTEX_ARRAY_FORMAT_FOG_FIELDS_SHIFT 4 +#define NV10TCL_VERTEX_ARRAY_FORMAT_FOG_FIELDS_MASK 0x000000f0 +#define NV10TCL_VERTEX_ARRAY_FORMAT_FOG_STRIDE_SHIFT 8 +#define NV10TCL_VERTEX_ARRAY_FORMAT_FOG_STRIDE_MASK 0x0000ff00 +#define NV10TCL_VERTEX_BEGIN_END 0x00000dfc +#define NV10TCL_VERTEX_BEGIN_END_STOP 0x00000000 +#define NV10TCL_VERTEX_BEGIN_END_POINTS 0x00000001 +#define NV10TCL_VERTEX_BEGIN_END_LINES 0x00000002 +#define NV10TCL_VERTEX_BEGIN_END_LINE_LOOP 0x00000003 +#define NV10TCL_VERTEX_BEGIN_END_LINE_STRIP 0x00000004 +#define NV10TCL_VERTEX_BEGIN_END_TRIANGLES 0x00000005 +#define NV10TCL_VERTEX_BEGIN_END_TRIANGLE_STRIP 0x00000006 +#define NV10TCL_VERTEX_BEGIN_END_TRIANGLE_FAN 0x00000007 +#define NV10TCL_VERTEX_BEGIN_END_QUADS 0x00000008 +#define NV10TCL_VERTEX_BEGIN_END_QUAD_STRIP 0x00000009 +#define NV10TCL_VERTEX_BEGIN_END_POLYGON 0x0000000a +#define NV10TCL_VB_ELEMENT_U16 0x00000e00 +#define NV10TCL_VB_ELEMENT_U16_I0_SHIFT 0 +#define NV10TCL_VB_ELEMENT_U16_I0_MASK 0x0000ffff +#define NV10TCL_VB_ELEMENT_U16_I1_SHIFT 16 +#define NV10TCL_VB_ELEMENT_U16_I1_MASK 0xffff0000 +#define NV10TCL_VB_ELEMENT_U32 0x00001100 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END 0x000013fc +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_STOP 0x00000000 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_POINTS 0x00000001 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_LINES 0x00000002 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_LINE_LOOP 0x00000003 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_LINE_STRIP 0x00000004 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_TRIANGLES 0x00000005 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_TRIANGLE_STRIP 0x00000006 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_TRIANGLE_FAN 0x00000007 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_QUADS 0x00000008 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_QUAD_STRIP 0x00000009 +#define NV10TCL_VERTEX_BUFFER_BEGIN_END_POLYGON 0x0000000a +#define NV10TCL_VERTEX_BUFFER_DRAW_ARRAYS 0x00001400 +#define NV10TCL_VERTEX_BUFFER_DRAW_ARRAYS_FIRST_SHIFT 0 +#define NV10TCL_VERTEX_BUFFER_DRAW_ARRAYS_FIRST_MASK 0x0000ffff +#define NV10TCL_VERTEX_BUFFER_DRAW_ARRAYS_LAST_SHIFT 24 +#define NV10TCL_VERTEX_BUFFER_DRAW_ARRAYS_LAST_MASK 0xff000000 +#define NV10TCL_VERTEX_ARRAY_DATA 0x00001800 + + +#define NV04_CONTEXT_COLOR_KEY 0x00000057 + + + +#define NV03_CONTEXT_SURFACES_2D 0x00000058 + +#define NV03_CONTEXT_SURFACES_2D_SYNCHRONIZE 0x00000100 +#define NV03_CONTEXT_SURFACES_2D_DMA_NOTIFY 0x00000180 +#define NV03_CONTEXT_SURFACES_2D_DMA_SOURCE 0x00000184 +#define NV03_CONTEXT_SURFACES_2D_DMA_DESTIN 0x00000188 +#define NV03_CONTEXT_SURFACES_2D_COLOR_FORMAT 0x00000300 +#define NV03_CONTEXT_SURFACES_2D_PITCH 0x00000304 +#define NV03_CONTEXT_SURFACES_2D_PITCH_SOURCE_SHIFT 0 +#define NV03_CONTEXT_SURFACES_2D_PITCH_SOURCE_MASK 0x0000ffff +#define NV03_CONTEXT_SURFACES_2D_PITCH_DESTIN_SHIFT 16 +#define NV03_CONTEXT_SURFACES_2D_PITCH_DESTIN_MASK 0xffff0000 +#define NV03_CONTEXT_SURFACES_2D_OFFSET_SOURCE 0x00000308 +#define NV03_CONTEXT_SURFACES_2D_OFFSET_DESTIN 0x0000030c + + +#define NV03_CONTEXT_SURFACES_3D 0x0000005a + +#define NV03_CONTEXT_SURFACES_3D_SYNCHRONIZE 0x00000100 +#define NV03_CONTEXT_SURFACES_3D_DMA_NOTIFY 0x00000180 +#define NV03_CONTEXT_SURFACES_3D_DMA_SURFACE 0x00000184 +#define NV03_CONTEXT_SURFACES_3D_PITCH 0x00000300 +#define NV03_CONTEXT_SURFACES_3D_OFFSET_COLOR 0x00000304 +#define NV03_CONTEXT_SURFACES_3D_OFFSET_ZETA 0x00000308 + + +#define NV04_RENDER_SOLID_LINE 0x0000005c + +#define NV04_RENDER_SOLID_LINE_SURFACE 0x00000198 + + +#define NV04_RENDER_SOLID_TRIANGLE 0x0000005d + + + +#define NV04_RENDER_SOLID_RECTANGLE 0x0000005e + +#define NV04_RENDER_SOLID_RECTANGLE_SURFACE 0x00000198 + + +#define NV04_IMAGE_BLIT 0x0000005f + +#define NV04_IMAGE_BLIT_NOP 0x00000100 +#define NV04_IMAGE_BLIT_NOTIFY 0x00000104 +#define NV04_IMAGE_BLIT_DMA_NOTIFY 0x00000180 +#define NV04_IMAGE_BLIT_COLOR_KEY 0x00000184 +#define NV04_IMAGE_BLIT_CLIP_RECTANGLE 0x00000188 +#define NV04_IMAGE_BLIT_PATTERN 0x0000018c +#define NV04_IMAGE_BLIT_ROP 0x00000190 +#define NV04_IMAGE_BLIT_BETA4 0x00000198 +#define NV04_IMAGE_BLIT_SURFACE 0x0000019c +#define NV04_IMAGE_BLIT_OPERATION 0x000002fc +#define NV04_IMAGE_BLIT_OPERATION_SRCCOPY_AND 0x00000000 +#define NV04_IMAGE_BLIT_OPERATION_ROP_AND 0x00000001 +#define NV04_IMAGE_BLIT_OPERATION_BLEND_AND 0x00000002 +#define NV04_IMAGE_BLIT_OPERATION_SRCCOPY 0x00000003 +#define NV04_IMAGE_BLIT_OPERATION_SRCCOPY_PREMULT 0x00000004 +#define NV04_IMAGE_BLIT_OPERATION_BLEND_PREMULT 0x00000005 + + +#define NV04_INDEXED_IMAGE_FROM_CPU 0x00000060 + +#define NV04_INDEXED_IMAGE_FROM_CPU_NOP 0x00000100 +#define NV04_INDEXED_IMAGE_FROM_CPU_NOTIFY 0x00000104 +#define NV04_INDEXED_IMAGE_FROM_CPU_PATCH 0x0000010c +#define NV04_INDEXED_IMAGE_FROM_CPU_DMA_NOTIFY 0x00000180 +#define NV04_INDEXED_IMAGE_FROM_CPU_DMA_LUT 0x00000184 +#define NV04_INDEXED_IMAGE_FROM_CPU_COLOR_FORMAT 0x000003e8 +#define NV04_INDEXED_IMAGE_FROM_CPU_INDEX_FORMAT 0x000003ec +#define NV04_INDEXED_IMAGE_FROM_CPU_LUT_OFFSET 0x000003f0 +#define NV04_INDEXED_IMAGE_FROM_CPU_POINT 0x000003f4 +#define NV04_INDEXED_IMAGE_FROM_CPU_SIZE_OUT 0x000003f8 +#define NV04_INDEXED_IMAGE_FROM_CPU_SIZE_IN 0x000003fc +#define NV04_INDEXED_IMAGE_FROM_CPU_COLOR 0x00000400 + + +#define NV04_IMAGE_FROM_CPU 0x00000061 + +#define NV04_IMAGE_FROM_CPU_BETA4 0x00000198 +#define NV04_IMAGE_FROM_CPU_SURFACE 0x0000019c + + +#define NV10_CONTEXT_SURFACES_2D 0x00000062 + + + +#define NV05_SCALED_IMAGE_FROM_MEMORY 0x00000063 + +#define NV05_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION 0x000002fc +#define NV05_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION_DITHER 0x00000000 +#define NV05_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION_TRUNCATE 0x00000001 +#define NV05_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION_SUBTR_TRUNCATE 0x00000002 + + +#define NV01_IMAGE_SRCCOPY_AND 0x00000064 + +#define NV01_IMAGE_SRCCOPY_AND_NOTIFY 0x00000104 +#define NV01_IMAGE_SRCCOPY_AND_DMA_NOTIFY 0x00000180 +#define NV01_IMAGE_SRCCOPY_AND_IMAGE_OUTPUT 0x00000200 +#define NV01_IMAGE_SRCCOPY_AND_IMAGE_INPUT 0x00000204 + + +#define NV05_INDEXED_IMAGE_FROM_CPU 0x00000064 + +#define NV05_INDEXED_IMAGE_FROM_CPU_COLOR_KEY 0x00000188 +#define NV05_INDEXED_IMAGE_FROM_CPU_CLIP_RECTANGLE 0x0000018c +#define NV05_INDEXED_IMAGE_FROM_CPU_PATTERN 0x00000190 +#define NV05_INDEXED_IMAGE_FROM_CPU_ROP 0x00000194 +#define NV05_INDEXED_IMAGE_FROM_CPU_BETA1 0x00000198 +#define NV05_INDEXED_IMAGE_FROM_CPU_BETA4 0x0000019c +#define NV05_INDEXED_IMAGE_FROM_CPU_SURFACE 0x000001a0 +#define NV05_INDEXED_IMAGE_FROM_CPU_COLOR_CONVERSION 0x000003e0 +#define NV05_INDEXED_IMAGE_FROM_CPU_OPERATION 0x000003e4 +#define NV05_INDEXED_IMAGE_FROM_CPU_INDICES 0x00000400 + + +#define NV05_IMAGE_FROM_CPU 0x00000065 + +#define NV05_IMAGE_FROM_CPU_BETA4 0x00000198 +#define NV05_IMAGE_FROM_CPU_SURFACE 0x0000019c + + +#define NV05_STRETCHED_IMAGE_FROM_CPU 0x00000066 + +#define NV05_STRETCHED_IMAGE_FROM_CPU_BETA4 0x00000194 +#define NV05_STRETCHED_IMAGE_FROM_CPU_SURFACE 0x00000198 +#define NV05_STRETCHED_IMAGE_FROM_CPU_COLOR_CONVERSION 0x000002f8 + + +#define NV04_IMAGE_BLEND_PREMULT 0x00000067 + +#define NV04_IMAGE_BLEND_PREMULT_NOP 0x00000100 +#define NV04_IMAGE_BLEND_PREMULT_NOTIFY 0x00000104 +#define NV04_IMAGE_BLEND_PREMULT_DMA_NOTIFY 0x00000180 +#define NV04_IMAGE_BLEND_PREMULT_IMAGE_OUTPUT 0x00000200 +#define NV04_IMAGE_BLEND_PREMULT_BETA_INPUT 0x00000204 +#define NV04_IMAGE_BLEND_PREMULT_IMAGE_INPUT 0x00000208 + + +#define NV03_CHANNEL_PIO 0x0000006a + + + +#define NV03_CHANNEL_DMA 0x0000006b + + + +#define NV04_BETA_SOLID 0x00000072 + +#define NV04_BETA_SOLID_NOP 0x00000100 +#define NV04_BETA_SOLID_NOTIFY 0x00000104 +#define NV04_BETA_SOLID_DMA_NOTIFY 0x00000180 +#define NV04_BETA_SOLID_BETA_OUTPUT 0x00000200 +#define NV04_BETA_SOLID_BETA_FACTOR 0x00000300 + + +#define NV04_STRETCHED_IMAGE_FROM_CPU 0x00000076 + + + +#define NV04_SCALED_IMAGE_FROM_MEMORY 0x00000077 + +#define NV04_SCALED_IMAGE_FROM_MEMORY_NOP 0x00000100 +#define NV04_SCALED_IMAGE_FROM_MEMORY_NOTIFY 0x00000104 +#define NV04_SCALED_IMAGE_FROM_MEMORY_DMA_NOTIFY 0x00000180 +#define NV04_SCALED_IMAGE_FROM_MEMORY_DMA_IMAGE 0x00000184 +#define NV04_SCALED_IMAGE_FROM_MEMORY_PATTERN 0x00000188 +#define NV04_SCALED_IMAGE_FROM_MEMORY_ROP 0x0000018c +#define NV04_SCALED_IMAGE_FROM_MEMORY_BETA1 0x00000190 +#define NV04_SCALED_IMAGE_FROM_MEMORY_BETA4 0x00000194 +#define NV04_SCALED_IMAGE_FROM_MEMORY_SURFACE 0x00000198 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION 0x000002fc +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION_DITHER 0x00000000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION_TRUNCATE 0x00000001 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION_SUBTR_TRUNCATE 0x00000002 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT 0x00000300 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_A1R5G5B5 0x00000001 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_X1R5G5B5 0x00000002 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_A8R8G8B8 0x00000003 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_X8R8G8B8 0x00000004 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_V8YB8U8YA8 0x00000005 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_YB8V8YA8U8 0x00000006 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_R5G6B5 0x00000007 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_Y8 0x00000008 +#define NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_AY8 0x00000009 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OPERATION 0x00000304 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OPERATION_SRCCOPY_AND 0x00000000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OPERATION_ROP_AND 0x00000001 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OPERATION_BLEND_AND 0x00000002 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OPERATION_SRCCOPY 0x00000003 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OPERATION_SRCCOPY_PREMULT 0x00000004 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OPERATION_BLEND_PREMULT 0x00000005 +#define NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_POINT 0x00000308 +#define NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_POINT_X_SHIFT 0 +#define NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_POINT_X_MASK 0x0000ffff +#define NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_POINT_Y_SHIFT 16 +#define NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_POINT_Y_MASK 0xffff0000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE 0x0000030c +#define NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE_W_SHIFT 0 +#define NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE_W_MASK 0x0000ffff +#define NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE_H_SHIFT 16 +#define NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE_H_MASK 0xffff0000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OUT_POINT 0x00000310 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OUT_POINT_X_SHIFT 0 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OUT_POINT_X_MASK 0x0000ffff +#define NV04_SCALED_IMAGE_FROM_MEMORY_OUT_POINT_Y_SHIFT 16 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OUT_POINT_Y_MASK 0xffff0000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OUT_SIZE 0x00000314 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OUT_SIZE_W_SHIFT 0 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OUT_SIZE_W_MASK 0x0000ffff +#define NV04_SCALED_IMAGE_FROM_MEMORY_OUT_SIZE_H_SHIFT 16 +#define NV04_SCALED_IMAGE_FROM_MEMORY_OUT_SIZE_H_MASK 0xffff0000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_DU_DX 0x00000318 +#define NV04_SCALED_IMAGE_FROM_MEMORY_DV_DY 0x0000031c +#define NV04_SCALED_IMAGE_FROM_MEMORY_SIZE 0x00000400 +#define NV04_SCALED_IMAGE_FROM_MEMORY_SIZE_W_SHIFT 0 +#define NV04_SCALED_IMAGE_FROM_MEMORY_SIZE_W_MASK 0x0000ffff +#define NV04_SCALED_IMAGE_FROM_MEMORY_SIZE_H_SHIFT 16 +#define NV04_SCALED_IMAGE_FROM_MEMORY_SIZE_H_MASK 0xffff0000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT 0x00000404 +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_PITCH_SHIFT 0 +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_PITCH_MASK 0x0000ffff +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_SHIFT 16 +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_MASK 0x00ff0000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_CENTER 0x00010000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_CORNER 0x00020000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_SHIFT 24 +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_MASK 0xff000000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_POINT_SAMPLE 0x00000000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_BILINEAR 0x01000000 +#define NV04_SCALED_IMAGE_FROM_MEMORY_ADDRESS 0x00000408 +#define NV04_SCALED_IMAGE_FROM_MEMORY_POINT 0x0000040c +#define NV04_SCALED_IMAGE_FROM_MEMORY_POINT_X_SHIFT 0 +#define NV04_SCALED_IMAGE_FROM_MEMORY_POINT_X_MASK 0x0000ffff +#define NV04_SCALED_IMAGE_FROM_MEMORY_POINT_Y_SHIFT 16 +#define NV04_SCALED_IMAGE_FROM_MEMORY_POINT_Y_MASK 0xffff0000 + + +#define NV10_TEXTURE_FROM_CPU 0x0000007b + +#define NV10_TEXTURE_FROM_CPU_NOP 0x00000100 +#define NV10_TEXTURE_FROM_CPU_NOTIFY 0x00000104 +#define NV10_TEXTURE_FROM_CPU_WAIT_FOR_IDLE 0x00000108 +#define NV10_TEXTURE_FROM_CPU_PM_TRIGGER 0x00000140 +#define NV10_TEXTURE_FROM_CPU_DMA_NOTIFY 0x00000180 +#define NV10_TEXTURE_FROM_CPU_SURFACE 0x00000184 +#define NV10_TEXTURE_FROM_CPU_COLOR_FORMAT 0x00000300 +#define NV10_TEXTURE_FROM_CPU_POINT 0x00000304 +#define NV10_TEXTURE_FROM_CPU_POINT_X_SHIFT 0 +#define NV10_TEXTURE_FROM_CPU_POINT_X_MASK 0x0000ffff +#define NV10_TEXTURE_FROM_CPU_POINT_Y_SHIFT 16 +#define NV10_TEXTURE_FROM_CPU_POINT_Y_MASK 0xffff0000 +#define NV10_TEXTURE_FROM_CPU_SIZE 0x00000308 +#define NV10_TEXTURE_FROM_CPU_SIZE_W_SHIFT 0 +#define NV10_TEXTURE_FROM_CPU_SIZE_W_MASK 0x0000ffff +#define NV10_TEXTURE_FROM_CPU_SIZE_H_SHIFT 16 +#define NV10_TEXTURE_FROM_CPU_SIZE_H_MASK 0xffff0000 +#define NV10_TEXTURE_FROM_CPU_CLIP_HORIZONTAL 0x0000030c +#define NV10_TEXTURE_FROM_CPU_CLIP_HORIZONTAL_X_SHIFT 0 +#define NV10_TEXTURE_FROM_CPU_CLIP_HORIZONTAL_X_MASK 0x0000ffff +#define NV10_TEXTURE_FROM_CPU_CLIP_HORIZONTAL_W_SHIFT 16 +#define NV10_TEXTURE_FROM_CPU_CLIP_HORIZONTAL_W_MASK 0xffff0000 +#define NV10_TEXTURE_FROM_CPU_CLIP_VERTICAL 0x00000310 +#define NV10_TEXTURE_FROM_CPU_CLIP_VERTICAL_Y_SHIFT 0 +#define NV10_TEXTURE_FROM_CPU_CLIP_VERTICAL_Y_MASK 0x0000ffff +#define NV10_TEXTURE_FROM_CPU_CLIP_VERTICAL_H_SHIFT 16 +#define NV10_TEXTURE_FROM_CPU_CLIP_VERTICAL_H_MASK 0xffff0000 +#define NV10_TEXTURE_FROM_CPU_COLOR(x) (0x00000400+((x)*4)) +#define NV10_TEXTURE_FROM_CPU_COLOR__SIZE 0x00000700 + + +#define NV10_VIDEO_DISPLAY 0x0000007c + + + +#define NV10_DVD_SUBPICTURE 0x00000088 + + + +#define NV10_SCALED_IMAGE_FROM_MEMORY 0x00000089 + +#define NV10_SCALED_IMAGE_FROM_MEMORY_WAIT_FOR_IDLE 0x00000108 + + +#define NV10_IMAGE_FROM_CPU 0x0000008a + +#define NV10_IMAGE_FROM_CPU_COLOR_CONVERSION 0x000002f8 + + +#define NV10_CONTEXT_SURFACES_3D 0x00000093 + + + +#define NV10_DX5_TEXTURE_TRIANGLE 0x00000094 + + + +#define NV10_DX6_MULTI_TEXTURE_TRIANGLE 0x00000095 + + + +#define NV11TCL 0x00000096 + +#define NV11TCL_COLOR_LOGIC_OP_ENABLE 0x00000d40 +#define NV11TCL_COLOR_LOGIC_OP_OP 0x00000d44 +#define NV11TCL_COLOR_LOGIC_OP_OP_CLEAR 0x00001500 +#define NV11TCL_COLOR_LOGIC_OP_OP_AND 0x00001501 +#define NV11TCL_COLOR_LOGIC_OP_OP_AND_REVERSE 0x00001502 +#define NV11TCL_COLOR_LOGIC_OP_OP_COPY 0x00001503 +#define NV11TCL_COLOR_LOGIC_OP_OP_AND_INVERTED 0x00001504 +#define NV11TCL_COLOR_LOGIC_OP_OP_NOOP 0x00001505 +#define NV11TCL_COLOR_LOGIC_OP_OP_XOR 0x00001506 +#define NV11TCL_COLOR_LOGIC_OP_OP_OR 0x00001507 +#define NV11TCL_COLOR_LOGIC_OP_OP_NOR 0x00001508 +#define NV11TCL_COLOR_LOGIC_OP_OP_EQUIV 0x00001509 +#define NV11TCL_COLOR_LOGIC_OP_OP_INVERT 0x0000150a +#define NV11TCL_COLOR_LOGIC_OP_OP_OR_REVERSE 0x0000150b +#define NV11TCL_COLOR_LOGIC_OP_OP_COPY_INVERTED 0x0000150c +#define NV11TCL_COLOR_LOGIC_OP_OP_OR_INVERTED 0x0000150d +#define NV11TCL_COLOR_LOGIC_OP_OP_NAND 0x0000150e +#define NV11TCL_COLOR_LOGIC_OP_OP_SET 0x0000150f + + +#define NV20TCL 0x00000097 + +#define NV20TCL_NOP 0x00000100 +#define NV20TCL_NOTIFY 0x00000104 +#define NV20TCL_DMA_NOTIFY 0x00000180 +#define NV20TCL_DMA_TEXTURE0 0x00000184 +#define NV20TCL_DMA_TEXTURE1 0x00000188 +#define NV20TCL_DMA_COLOR 0x00000194 +#define NV20TCL_DMA_ZETA 0x00000198 +#define NV20TCL_DMA_VTXBUF0 0x0000019c +#define NV20TCL_DMA_VTXBUF1 0x000001a0 +#define NV20TCL_DMA_FENCE 0x000001a4 +#define NV20TCL_DMA_QUERY 0x000001a8 +#define NV20TCL_RT_HORIZ 0x00000200 +#define NV20TCL_RT_HORIZ_X_SHIFT 0 +#define NV20TCL_RT_HORIZ_X_MASK 0x0000ffff +#define NV20TCL_RT_HORIZ_W_SHIFT 16 +#define NV20TCL_RT_HORIZ_W_MASK 0xffff0000 +#define NV20TCL_RT_VERT 0x00000204 +#define NV20TCL_RT_VERT_Y_SHIFT 0 +#define NV20TCL_RT_VERT_Y_MASK 0x0000ffff +#define NV20TCL_RT_VERT_H_SHIFT 16 +#define NV20TCL_RT_VERT_H_MASK 0xffff0000 +#define NV20TCL_RT_FORMAT 0x00000208 +#define NV20TCL_RT_FORMAT_TYPE_SHIFT 8 +#define NV20TCL_RT_FORMAT_TYPE_MASK 0x00000f00 +#define NV20TCL_RT_FORMAT_TYPE_LINEAR 0x00000100 +#define NV20TCL_RT_FORMAT_TYPE_SWIZZLED 0x00000200 +#define NV20TCL_RT_FORMAT_COLOR_SHIFT 0 +#define NV20TCL_RT_FORMAT_COLOR_MASK 0x0000001f +#define NV20TCL_RT_FORMAT_COLOR_R5G6B5 0x00000003 +#define NV20TCL_RT_FORMAT_COLOR_X8R8G8B8 0x00000005 +#define NV20TCL_RT_FORMAT_COLOR_A8R8G8B8 0x00000008 +#define NV20TCL_RT_FORMAT_COLOR_B8 0x00000009 +#define NV20TCL_RT_FORMAT_COLOR_UNKNOWN 0x0000000d +#define NV20TCL_RT_FORMAT_COLOR_X8B8G8R8 0x0000000f +#define NV20TCL_RT_FORMAT_COLOR_A8B8G8R8 0x00000010 +#define NV20TCL_RT_PITCH 0x0000020c +#define NV20TCL_RT_PITCH_COLOR_PITCH_SHIFT 0 +#define NV20TCL_RT_PITCH_COLOR_PITCH_MASK 0x0000ffff +#define NV20TCL_RT_PITCH_ZETA_PITCH_SHIFT 16 +#define NV20TCL_RT_PITCH_ZETA_PITCH_MASK 0xffff0000 +#define NV20TCL_COLOR_OFFSET 0x00000210 +#define NV20TCL_ZETA_OFFSET 0x00000214 +#define NV20TCL_RC_IN_ALPHA(x) (0x00000260+((x)*4)) +#define NV20TCL_RC_IN_ALPHA__SIZE 0x00000008 +#define NV20TCL_RC_IN_ALPHA_D_INPUT_SHIFT 0 +#define NV20TCL_RC_IN_ALPHA_D_INPUT_MASK 0x0000000f +#define NV20TCL_RC_IN_ALPHA_D_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_IN_ALPHA_D_INPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV20TCL_RC_IN_ALPHA_D_INPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV20TCL_RC_IN_ALPHA_D_INPUT_FOG 0x00000003 +#define NV20TCL_RC_IN_ALPHA_D_INPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV20TCL_RC_IN_ALPHA_D_INPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV20TCL_RC_IN_ALPHA_D_INPUT_TEXTURE0_ARB 0x00000008 +#define NV20TCL_RC_IN_ALPHA_D_INPUT_TEXTURE1_ARB 0x00000009 +#define NV20TCL_RC_IN_ALPHA_D_INPUT_SPARE0_NV 0x0000000c +#define NV20TCL_RC_IN_ALPHA_D_INPUT_SPARE1_NV 0x0000000d +#define NV20TCL_RC_IN_ALPHA_D_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV20TCL_RC_IN_ALPHA_D_INPUT_E_TIMES_F_NV 0x0000000f +#define NV20TCL_RC_IN_ALPHA_D_COMPONENT_USAGE (1 << 4) +#define NV20TCL_RC_IN_ALPHA_D_COMPONENT_USAGE_BLUE 0x00000000 +#define NV20TCL_RC_IN_ALPHA_D_COMPONENT_USAGE_ALPHA 0x00000010 +#define NV20TCL_RC_IN_ALPHA_D_MAPPING_SHIFT 5 +#define NV20TCL_RC_IN_ALPHA_D_MAPPING_MASK 0x000000e0 +#define NV20TCL_RC_IN_ALPHA_D_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_IN_ALPHA_D_MAPPING_UNSIGNED_INVERT_NV 0x00000020 +#define NV20TCL_RC_IN_ALPHA_D_MAPPING_EXPAND_NORMAL_NV 0x00000040 +#define NV20TCL_RC_IN_ALPHA_D_MAPPING_EXPAND_NEGATE_NV 0x00000060 +#define NV20TCL_RC_IN_ALPHA_D_MAPPING_HALF_BIAS_NORMAL_NV 0x00000080 +#define NV20TCL_RC_IN_ALPHA_D_MAPPING_HALF_BIAS_NEGATE_NV 0x000000a0 +#define NV20TCL_RC_IN_ALPHA_D_MAPPING_SIGNED_IDENTITY_NV 0x000000c0 +#define NV20TCL_RC_IN_ALPHA_D_MAPPING_SIGNED_NEGATE_NV 0x000000e0 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_SHIFT 8 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_MASK 0x00000f00 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_FOG 0x00000300 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_TEXTURE0_ARB 0x00000800 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_TEXTURE1_ARB 0x00000900 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_SPARE0_NV 0x00000c00 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_SPARE1_NV 0x00000d00 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV20TCL_RC_IN_ALPHA_C_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV20TCL_RC_IN_ALPHA_C_COMPONENT_USAGE (1 << 12) +#define NV20TCL_RC_IN_ALPHA_C_COMPONENT_USAGE_BLUE 0x00000000 +#define NV20TCL_RC_IN_ALPHA_C_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV20TCL_RC_IN_ALPHA_C_MAPPING_SHIFT 13 +#define NV20TCL_RC_IN_ALPHA_C_MAPPING_MASK 0x0000e000 +#define NV20TCL_RC_IN_ALPHA_C_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_IN_ALPHA_C_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV20TCL_RC_IN_ALPHA_C_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV20TCL_RC_IN_ALPHA_C_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV20TCL_RC_IN_ALPHA_C_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV20TCL_RC_IN_ALPHA_C_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV20TCL_RC_IN_ALPHA_C_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV20TCL_RC_IN_ALPHA_C_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_SHIFT 16 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_MASK 0x000f0000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_FOG 0x00030000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_TEXTURE0_ARB 0x00080000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_TEXTURE1_ARB 0x00090000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_SPARE0_NV 0x000c0000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_SPARE1_NV 0x000d0000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV20TCL_RC_IN_ALPHA_B_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV20TCL_RC_IN_ALPHA_B_COMPONENT_USAGE (1 << 20) +#define NV20TCL_RC_IN_ALPHA_B_COMPONENT_USAGE_BLUE 0x00000000 +#define NV20TCL_RC_IN_ALPHA_B_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV20TCL_RC_IN_ALPHA_B_MAPPING_SHIFT 21 +#define NV20TCL_RC_IN_ALPHA_B_MAPPING_MASK 0x00e00000 +#define NV20TCL_RC_IN_ALPHA_B_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_IN_ALPHA_B_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV20TCL_RC_IN_ALPHA_B_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV20TCL_RC_IN_ALPHA_B_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV20TCL_RC_IN_ALPHA_B_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV20TCL_RC_IN_ALPHA_B_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV20TCL_RC_IN_ALPHA_B_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV20TCL_RC_IN_ALPHA_B_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_SHIFT 24 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_MASK 0x0f000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_FOG 0x03000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_TEXTURE0_ARB 0x08000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_TEXTURE1_ARB 0x09000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_SPARE0_NV 0x0c000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_SPARE1_NV 0x0d000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV20TCL_RC_IN_ALPHA_A_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV20TCL_RC_IN_ALPHA_A_COMPONENT_USAGE (1 << 28) +#define NV20TCL_RC_IN_ALPHA_A_COMPONENT_USAGE_BLUE 0x00000000 +#define NV20TCL_RC_IN_ALPHA_A_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV20TCL_RC_IN_ALPHA_A_MAPPING_SHIFT 29 +#define NV20TCL_RC_IN_ALPHA_A_MAPPING_MASK 0xe0000000 +#define NV20TCL_RC_IN_ALPHA_A_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_IN_ALPHA_A_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV20TCL_RC_IN_ALPHA_A_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV20TCL_RC_IN_ALPHA_A_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV20TCL_RC_IN_ALPHA_A_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV20TCL_RC_IN_ALPHA_A_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV20TCL_RC_IN_ALPHA_A_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV20TCL_RC_IN_ALPHA_A_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV20TCL_RC_FINAL0 0x00000288 +#define NV20TCL_RC_FINAL0_D_INPUT_SHIFT 0 +#define NV20TCL_RC_FINAL0_D_INPUT_MASK 0x0000000f +#define NV20TCL_RC_FINAL0_D_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_FINAL0_D_INPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV20TCL_RC_FINAL0_D_INPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV20TCL_RC_FINAL0_D_INPUT_FOG 0x00000003 +#define NV20TCL_RC_FINAL0_D_INPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV20TCL_RC_FINAL0_D_INPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV20TCL_RC_FINAL0_D_INPUT_TEXTURE0_ARB 0x00000008 +#define NV20TCL_RC_FINAL0_D_INPUT_TEXTURE1_ARB 0x00000009 +#define NV20TCL_RC_FINAL0_D_INPUT_SPARE0_NV 0x0000000c +#define NV20TCL_RC_FINAL0_D_INPUT_SPARE1_NV 0x0000000d +#define NV20TCL_RC_FINAL0_D_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV20TCL_RC_FINAL0_D_INPUT_E_TIMES_F_NV 0x0000000f +#define NV20TCL_RC_FINAL0_D_COMPONENT_USAGE (1 << 4) +#define NV20TCL_RC_FINAL0_D_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_FINAL0_D_COMPONENT_USAGE_ALPHA 0x00000010 +#define NV20TCL_RC_FINAL0_D_MAPPING_SHIFT 5 +#define NV20TCL_RC_FINAL0_D_MAPPING_MASK 0x000000e0 +#define NV20TCL_RC_FINAL0_D_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_FINAL0_D_MAPPING_UNSIGNED_INVERT_NV 0x00000020 +#define NV20TCL_RC_FINAL0_D_MAPPING_EXPAND_NORMAL_NV 0x00000040 +#define NV20TCL_RC_FINAL0_D_MAPPING_EXPAND_NEGATE_NV 0x00000060 +#define NV20TCL_RC_FINAL0_D_MAPPING_HALF_BIAS_NORMAL_NV 0x00000080 +#define NV20TCL_RC_FINAL0_D_MAPPING_HALF_BIAS_NEGATE_NV 0x000000a0 +#define NV20TCL_RC_FINAL0_D_MAPPING_SIGNED_IDENTITY_NV 0x000000c0 +#define NV20TCL_RC_FINAL0_D_MAPPING_SIGNED_NEGATE_NV 0x000000e0 +#define NV20TCL_RC_FINAL0_C_INPUT_SHIFT 8 +#define NV20TCL_RC_FINAL0_C_INPUT_MASK 0x00000f00 +#define NV20TCL_RC_FINAL0_C_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_FINAL0_C_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV20TCL_RC_FINAL0_C_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV20TCL_RC_FINAL0_C_INPUT_FOG 0x00000300 +#define NV20TCL_RC_FINAL0_C_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV20TCL_RC_FINAL0_C_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV20TCL_RC_FINAL0_C_INPUT_TEXTURE0_ARB 0x00000800 +#define NV20TCL_RC_FINAL0_C_INPUT_TEXTURE1_ARB 0x00000900 +#define NV20TCL_RC_FINAL0_C_INPUT_SPARE0_NV 0x00000c00 +#define NV20TCL_RC_FINAL0_C_INPUT_SPARE1_NV 0x00000d00 +#define NV20TCL_RC_FINAL0_C_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV20TCL_RC_FINAL0_C_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV20TCL_RC_FINAL0_C_COMPONENT_USAGE (1 << 12) +#define NV20TCL_RC_FINAL0_C_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_FINAL0_C_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV20TCL_RC_FINAL0_C_MAPPING_SHIFT 13 +#define NV20TCL_RC_FINAL0_C_MAPPING_MASK 0x0000e000 +#define NV20TCL_RC_FINAL0_C_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_FINAL0_C_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV20TCL_RC_FINAL0_C_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV20TCL_RC_FINAL0_C_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV20TCL_RC_FINAL0_C_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV20TCL_RC_FINAL0_C_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV20TCL_RC_FINAL0_C_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV20TCL_RC_FINAL0_C_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV20TCL_RC_FINAL0_B_INPUT_SHIFT 16 +#define NV20TCL_RC_FINAL0_B_INPUT_MASK 0x000f0000 +#define NV20TCL_RC_FINAL0_B_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_FINAL0_B_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV20TCL_RC_FINAL0_B_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV20TCL_RC_FINAL0_B_INPUT_FOG 0x00030000 +#define NV20TCL_RC_FINAL0_B_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV20TCL_RC_FINAL0_B_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV20TCL_RC_FINAL0_B_INPUT_TEXTURE0_ARB 0x00080000 +#define NV20TCL_RC_FINAL0_B_INPUT_TEXTURE1_ARB 0x00090000 +#define NV20TCL_RC_FINAL0_B_INPUT_SPARE0_NV 0x000c0000 +#define NV20TCL_RC_FINAL0_B_INPUT_SPARE1_NV 0x000d0000 +#define NV20TCL_RC_FINAL0_B_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV20TCL_RC_FINAL0_B_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV20TCL_RC_FINAL0_B_COMPONENT_USAGE (1 << 20) +#define NV20TCL_RC_FINAL0_B_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_FINAL0_B_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV20TCL_RC_FINAL0_B_MAPPING_SHIFT 21 +#define NV20TCL_RC_FINAL0_B_MAPPING_MASK 0x00e00000 +#define NV20TCL_RC_FINAL0_B_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_FINAL0_B_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV20TCL_RC_FINAL0_B_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV20TCL_RC_FINAL0_B_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV20TCL_RC_FINAL0_B_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV20TCL_RC_FINAL0_B_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV20TCL_RC_FINAL0_B_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV20TCL_RC_FINAL0_B_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV20TCL_RC_FINAL0_A_INPUT_SHIFT 24 +#define NV20TCL_RC_FINAL0_A_INPUT_MASK 0x0f000000 +#define NV20TCL_RC_FINAL0_A_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_FINAL0_A_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV20TCL_RC_FINAL0_A_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV20TCL_RC_FINAL0_A_INPUT_FOG 0x03000000 +#define NV20TCL_RC_FINAL0_A_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV20TCL_RC_FINAL0_A_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV20TCL_RC_FINAL0_A_INPUT_TEXTURE0_ARB 0x08000000 +#define NV20TCL_RC_FINAL0_A_INPUT_TEXTURE1_ARB 0x09000000 +#define NV20TCL_RC_FINAL0_A_INPUT_SPARE0_NV 0x0c000000 +#define NV20TCL_RC_FINAL0_A_INPUT_SPARE1_NV 0x0d000000 +#define NV20TCL_RC_FINAL0_A_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV20TCL_RC_FINAL0_A_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV20TCL_RC_FINAL0_A_COMPONENT_USAGE (1 << 28) +#define NV20TCL_RC_FINAL0_A_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_FINAL0_A_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV20TCL_RC_FINAL0_A_MAPPING_SHIFT 29 +#define NV20TCL_RC_FINAL0_A_MAPPING_MASK 0xe0000000 +#define NV20TCL_RC_FINAL0_A_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_FINAL0_A_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV20TCL_RC_FINAL0_A_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV20TCL_RC_FINAL0_A_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV20TCL_RC_FINAL0_A_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV20TCL_RC_FINAL0_A_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV20TCL_RC_FINAL0_A_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV20TCL_RC_FINAL0_A_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV20TCL_RC_FINAL1 0x0000028c +#define NV20TCL_RC_FINAL1_COLOR_SUM_CLAMP (1 << 7) +#define NV20TCL_RC_FINAL1_G_INPUT_SHIFT 8 +#define NV20TCL_RC_FINAL1_G_INPUT_MASK 0x00000f00 +#define NV20TCL_RC_FINAL1_G_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_FINAL1_G_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV20TCL_RC_FINAL1_G_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV20TCL_RC_FINAL1_G_INPUT_FOG 0x00000300 +#define NV20TCL_RC_FINAL1_G_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV20TCL_RC_FINAL1_G_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV20TCL_RC_FINAL1_G_INPUT_TEXTURE0_ARB 0x00000800 +#define NV20TCL_RC_FINAL1_G_INPUT_TEXTURE1_ARB 0x00000900 +#define NV20TCL_RC_FINAL1_G_INPUT_SPARE0_NV 0x00000c00 +#define NV20TCL_RC_FINAL1_G_INPUT_SPARE1_NV 0x00000d00 +#define NV20TCL_RC_FINAL1_G_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV20TCL_RC_FINAL1_G_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV20TCL_RC_FINAL1_G_COMPONENT_USAGE (1 << 12) +#define NV20TCL_RC_FINAL1_G_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_FINAL1_G_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV20TCL_RC_FINAL1_G_MAPPING_SHIFT 13 +#define NV20TCL_RC_FINAL1_G_MAPPING_MASK 0x0000e000 +#define NV20TCL_RC_FINAL1_G_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_FINAL1_G_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV20TCL_RC_FINAL1_G_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV20TCL_RC_FINAL1_G_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV20TCL_RC_FINAL1_G_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV20TCL_RC_FINAL1_G_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV20TCL_RC_FINAL1_G_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV20TCL_RC_FINAL1_G_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV20TCL_RC_FINAL1_F_INPUT_SHIFT 16 +#define NV20TCL_RC_FINAL1_F_INPUT_MASK 0x000f0000 +#define NV20TCL_RC_FINAL1_F_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_FINAL1_F_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV20TCL_RC_FINAL1_F_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV20TCL_RC_FINAL1_F_INPUT_FOG 0x00030000 +#define NV20TCL_RC_FINAL1_F_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV20TCL_RC_FINAL1_F_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV20TCL_RC_FINAL1_F_INPUT_TEXTURE0_ARB 0x00080000 +#define NV20TCL_RC_FINAL1_F_INPUT_TEXTURE1_ARB 0x00090000 +#define NV20TCL_RC_FINAL1_F_INPUT_SPARE0_NV 0x000c0000 +#define NV20TCL_RC_FINAL1_F_INPUT_SPARE1_NV 0x000d0000 +#define NV20TCL_RC_FINAL1_F_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV20TCL_RC_FINAL1_F_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV20TCL_RC_FINAL1_F_COMPONENT_USAGE (1 << 20) +#define NV20TCL_RC_FINAL1_F_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_FINAL1_F_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV20TCL_RC_FINAL1_F_MAPPING_SHIFT 21 +#define NV20TCL_RC_FINAL1_F_MAPPING_MASK 0x00e00000 +#define NV20TCL_RC_FINAL1_F_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_FINAL1_F_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV20TCL_RC_FINAL1_F_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV20TCL_RC_FINAL1_F_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV20TCL_RC_FINAL1_F_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV20TCL_RC_FINAL1_F_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV20TCL_RC_FINAL1_F_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV20TCL_RC_FINAL1_F_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV20TCL_RC_FINAL1_E_INPUT_SHIFT 24 +#define NV20TCL_RC_FINAL1_E_INPUT_MASK 0x0f000000 +#define NV20TCL_RC_FINAL1_E_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_FINAL1_E_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV20TCL_RC_FINAL1_E_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV20TCL_RC_FINAL1_E_INPUT_FOG 0x03000000 +#define NV20TCL_RC_FINAL1_E_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV20TCL_RC_FINAL1_E_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV20TCL_RC_FINAL1_E_INPUT_TEXTURE0_ARB 0x08000000 +#define NV20TCL_RC_FINAL1_E_INPUT_TEXTURE1_ARB 0x09000000 +#define NV20TCL_RC_FINAL1_E_INPUT_SPARE0_NV 0x0c000000 +#define NV20TCL_RC_FINAL1_E_INPUT_SPARE1_NV 0x0d000000 +#define NV20TCL_RC_FINAL1_E_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV20TCL_RC_FINAL1_E_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV20TCL_RC_FINAL1_E_COMPONENT_USAGE (1 << 28) +#define NV20TCL_RC_FINAL1_E_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_FINAL1_E_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV20TCL_RC_FINAL1_E_MAPPING_SHIFT 29 +#define NV20TCL_RC_FINAL1_E_MAPPING_MASK 0xe0000000 +#define NV20TCL_RC_FINAL1_E_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_FINAL1_E_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV20TCL_RC_FINAL1_E_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV20TCL_RC_FINAL1_E_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV20TCL_RC_FINAL1_E_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV20TCL_RC_FINAL1_E_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV20TCL_RC_FINAL1_E_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV20TCL_RC_FINAL1_E_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV20TCL_LIGHT_CONTROL 0x00000294 +#define NV20TCL_FOG_MODE 0x0000029c +#define NV20TCL_FOG_MODE_EXP 0x00000800 +#define NV20TCL_FOG_MODE_EXP_2 0x00000802 +#define NV20TCL_FOG_MODE_EXP2 0x00000803 +#define NV20TCL_FOG_MODE_LINEAR 0x00000804 +#define NV20TCL_FOG_MODE_LINEAR_2 0x00002601 +#define NV20TCL_FOG_COORD_DIST 0x000002a0 +#define NV20TCL_FOG_COORD_DIST_COORD_FALSE 0x00000000 +#define NV20TCL_FOG_COORD_DIST_COORD_FRAGMENT_DEPTH_DISTANCE_EYE_RADIAL_NV 0x00000001 +#define NV20TCL_FOG_COORD_DIST_COORD_FRAGMENT_DEPTH_DISTANCE_EYE_PLANE_ABSOLUTE_NV 0x00000002 +#define NV20TCL_FOG_COORD_DIST_COORD_FOG 0x00000003 +#define NV20TCL_FOG_ENABLE 0x000002a4 +#define NV20TCL_FOG_COLOR 0x000002a8 +#define NV20TCL_FOG_COLOR_R_SHIFT 0 +#define NV20TCL_FOG_COLOR_R_MASK 0x000000ff +#define NV20TCL_FOG_COLOR_G_SHIFT 8 +#define NV20TCL_FOG_COLOR_G_MASK 0x0000ff00 +#define NV20TCL_FOG_COLOR_B_SHIFT 16 +#define NV20TCL_FOG_COLOR_B_MASK 0x00ff0000 +#define NV20TCL_FOG_COLOR_A_SHIFT 24 +#define NV20TCL_FOG_COLOR_A_MASK 0xff000000 +#define NV20TCL_VIEWPORT_CLIP_MODE 0x000002b4 +#define NV20TCL_VIEWPORT_CLIP_HORIZ(x) (0x000002c0+((x)*4)) +#define NV20TCL_VIEWPORT_CLIP_HORIZ__SIZE 0x00000008 +#define NV20TCL_VIEWPORT_CLIP_VERT(x) (0x000002e0+((x)*4)) +#define NV20TCL_VIEWPORT_CLIP_VERT__SIZE 0x00000008 +#define NV20TCL_ALPHA_FUNC_ENABLE 0x00000300 +#define NV20TCL_BLEND_FUNC_ENABLE 0x00000304 +#define NV20TCL_CULL_FACE_ENABLE 0x00000308 +#define NV20TCL_DEPTH_TEST_ENABLE 0x0000030c +#define NV20TCL_DITHER_ENABLE 0x00000310 +#define NV20TCL_LIGHTING_ENABLE 0x00000314 +#define NV20TCL_POINT_PARAMETERS_ENABLE 0x00000318 +#define NV20TCL_POINT_SMOOTH_ENABLE 0x0000031c +#define NV20TCL_LINE_SMOOTH_ENABLE 0x00000320 +#define NV20TCL_POLYGON_SMOOTH_ENABLE 0x00000324 +#define NV20TCL_STENCIL_ENABLE 0x0000032c +#define NV20TCL_POLYGON_OFFSET_POINT_ENABLE 0x00000330 +#define NV20TCL_POLYGON_OFFSET_LINE_ENABLE 0x00000334 +#define NV20TCL_POLYGON_OFFSET_FILL_ENABLE 0x00000338 +#define NV20TCL_ALPHA_FUNC_FUNC 0x0000033c +#define NV20TCL_ALPHA_FUNC_FUNC_NEVER 0x00000200 +#define NV20TCL_ALPHA_FUNC_FUNC_LESS 0x00000201 +#define NV20TCL_ALPHA_FUNC_FUNC_EQUAL 0x00000202 +#define NV20TCL_ALPHA_FUNC_FUNC_LEQUAL 0x00000203 +#define NV20TCL_ALPHA_FUNC_FUNC_GREATER 0x00000204 +#define NV20TCL_ALPHA_FUNC_FUNC_GREATER 0x00000204 +#define NV20TCL_ALPHA_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV20TCL_ALPHA_FUNC_FUNC_GEQUAL 0x00000206 +#define NV20TCL_ALPHA_FUNC_FUNC_ALWAYS 0x00000207 +#define NV20TCL_ALPHA_FUNC_REF 0x00000340 +#define NV20TCL_BLEND_FUNC_SRC 0x00000344 +#define NV20TCL_BLEND_FUNC_SRC_ZERO 0x00000000 +#define NV20TCL_BLEND_FUNC_SRC_ONE 0x00000001 +#define NV20TCL_BLEND_FUNC_SRC_SRC_COLOR 0x00000300 +#define NV20TCL_BLEND_FUNC_SRC_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV20TCL_BLEND_FUNC_SRC_SRC_ALPHA 0x00000302 +#define NV20TCL_BLEND_FUNC_SRC_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV20TCL_BLEND_FUNC_SRC_DST_ALPHA 0x00000304 +#define NV20TCL_BLEND_FUNC_SRC_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV20TCL_BLEND_FUNC_SRC_DST_COLOR 0x00000306 +#define NV20TCL_BLEND_FUNC_SRC_ONE_MINUS_DST_COLOR 0x00000307 +#define NV20TCL_BLEND_FUNC_SRC_SRC_ALPHA_SATURATE 0x00000308 +#define NV20TCL_BLEND_FUNC_SRC_CONSTANT_COLOR 0x00008001 +#define NV20TCL_BLEND_FUNC_SRC_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV20TCL_BLEND_FUNC_SRC_CONSTANT_ALPHA 0x00008003 +#define NV20TCL_BLEND_FUNC_SRC_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV20TCL_BLEND_FUNC_DST 0x00000348 +#define NV20TCL_BLEND_FUNC_DST_ZERO 0x00000000 +#define NV20TCL_BLEND_FUNC_DST_ONE 0x00000001 +#define NV20TCL_BLEND_FUNC_DST_SRC_COLOR 0x00000300 +#define NV20TCL_BLEND_FUNC_DST_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV20TCL_BLEND_FUNC_DST_SRC_ALPHA 0x00000302 +#define NV20TCL_BLEND_FUNC_DST_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV20TCL_BLEND_FUNC_DST_DST_ALPHA 0x00000304 +#define NV20TCL_BLEND_FUNC_DST_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV20TCL_BLEND_FUNC_DST_DST_COLOR 0x00000306 +#define NV20TCL_BLEND_FUNC_DST_ONE_MINUS_DST_COLOR 0x00000307 +#define NV20TCL_BLEND_FUNC_DST_SRC_ALPHA_SATURATE 0x00000308 +#define NV20TCL_BLEND_FUNC_DST_CONSTANT_COLOR 0x00008001 +#define NV20TCL_BLEND_FUNC_DST_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV20TCL_BLEND_FUNC_DST_CONSTANT_ALPHA 0x00008003 +#define NV20TCL_BLEND_FUNC_DST_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV20TCL_BLEND_COLOR 0x0000034c +#define NV20TCL_BLEND_COLOR_B_SHIFT 0 +#define NV20TCL_BLEND_COLOR_B_MASK 0x000000ff +#define NV20TCL_BLEND_COLOR_G_SHIFT 8 +#define NV20TCL_BLEND_COLOR_G_MASK 0x0000ff00 +#define NV20TCL_BLEND_COLOR_R_SHIFT 16 +#define NV20TCL_BLEND_COLOR_R_MASK 0x00ff0000 +#define NV20TCL_BLEND_COLOR_A_SHIFT 24 +#define NV20TCL_BLEND_COLOR_A_MASK 0xff000000 +#define NV20TCL_BLEND_EQUATION 0x00000350 +#define NV20TCL_BLEND_EQUATION_FUNC_ADD 0x00008006 +#define NV20TCL_BLEND_EQUATION_MIN 0x00008007 +#define NV20TCL_BLEND_EQUATION_MAX 0x00008008 +#define NV20TCL_BLEND_EQUATION_FUNC_SUBTRACT 0x0000800a +#define NV20TCL_BLEND_EQUATION_FUNC_REVERSE_SUBTRACT 0x0000800b +#define NV20TCL_DEPTH_FUNC 0x00000354 +#define NV20TCL_DEPTH_FUNC_NEVER 0x00000200 +#define NV20TCL_DEPTH_FUNC_LESS 0x00000201 +#define NV20TCL_DEPTH_FUNC_EQUAL 0x00000202 +#define NV20TCL_DEPTH_FUNC_LEQUAL 0x00000203 +#define NV20TCL_DEPTH_FUNC_GREATER 0x00000204 +#define NV20TCL_DEPTH_FUNC_GREATER 0x00000204 +#define NV20TCL_DEPTH_FUNC_NOTEQUAL 0x00000205 +#define NV20TCL_DEPTH_FUNC_GEQUAL 0x00000206 +#define NV20TCL_DEPTH_FUNC_ALWAYS 0x00000207 +#define NV20TCL_COLOR_MASK 0x00000358 +#define NV20TCL_COLOR_MASK_B (1 << 0) +#define NV20TCL_COLOR_MASK_G (1 << 8) +#define NV20TCL_COLOR_MASK_R (1 << 16) +#define NV20TCL_COLOR_MASK_A (1 << 24) +#define NV20TCL_DEPTH_WRITE_ENABLE 0x0000035c +#define NV20TCL_STENCIL_MASK 0x00000360 +#define NV20TCL_STENCIL_FUNC_FUNC 0x00000364 +#define NV20TCL_STENCIL_FUNC_FUNC_NEVER 0x00000200 +#define NV20TCL_STENCIL_FUNC_FUNC_LESS 0x00000201 +#define NV20TCL_STENCIL_FUNC_FUNC_EQUAL 0x00000202 +#define NV20TCL_STENCIL_FUNC_FUNC_LEQUAL 0x00000203 +#define NV20TCL_STENCIL_FUNC_FUNC_GREATER 0x00000204 +#define NV20TCL_STENCIL_FUNC_FUNC_GREATER 0x00000204 +#define NV20TCL_STENCIL_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV20TCL_STENCIL_FUNC_FUNC_GEQUAL 0x00000206 +#define NV20TCL_STENCIL_FUNC_FUNC_ALWAYS 0x00000207 +#define NV20TCL_STENCIL_FUNC_REF 0x00000368 +#define NV20TCL_STENCIL_FUNC_MASK 0x0000036c +#define NV20TCL_STENCIL_OP_FAIL 0x00000370 +#define NV20TCL_STENCIL_OP_FAIL_ZERO 0x00000000 +#define NV20TCL_STENCIL_OP_FAIL_INVERT 0x0000150a +#define NV20TCL_STENCIL_OP_FAIL_KEEP 0x00001e00 +#define NV20TCL_STENCIL_OP_FAIL_REPLACE 0x00001e01 +#define NV20TCL_STENCIL_OP_FAIL_INCR 0x00001e02 +#define NV20TCL_STENCIL_OP_FAIL_DECR 0x00001e03 +#define NV20TCL_STENCIL_OP_FAIL_INCR_WRAP 0x00008507 +#define NV20TCL_STENCIL_OP_FAIL_DECR_WRAP 0x00008508 +#define NV20TCL_STENCIL_OP_ZFAIL 0x00000374 +#define NV20TCL_STENCIL_OP_ZFAIL_ZERO 0x00000000 +#define NV20TCL_STENCIL_OP_ZFAIL_INVERT 0x0000150a +#define NV20TCL_STENCIL_OP_ZFAIL_KEEP 0x00001e00 +#define NV20TCL_STENCIL_OP_ZFAIL_REPLACE 0x00001e01 +#define NV20TCL_STENCIL_OP_ZFAIL_INCR 0x00001e02 +#define NV20TCL_STENCIL_OP_ZFAIL_DECR 0x00001e03 +#define NV20TCL_STENCIL_OP_ZFAIL_INCR_WRAP 0x00008507 +#define NV20TCL_STENCIL_OP_ZFAIL_DECR_WRAP 0x00008508 +#define NV20TCL_STENCIL_OP_ZPASS 0x00000378 +#define NV20TCL_STENCIL_OP_ZPASS_ZERO 0x00000000 +#define NV20TCL_STENCIL_OP_ZPASS_INVERT 0x0000150a +#define NV20TCL_STENCIL_OP_ZPASS_KEEP 0x00001e00 +#define NV20TCL_STENCIL_OP_ZPASS_REPLACE 0x00001e01 +#define NV20TCL_STENCIL_OP_ZPASS_INCR 0x00001e02 +#define NV20TCL_STENCIL_OP_ZPASS_DECR 0x00001e03 +#define NV20TCL_STENCIL_OP_ZPASS_INCR_WRAP 0x00008507 +#define NV20TCL_STENCIL_OP_ZPASS_DECR_WRAP 0x00008508 +#define NV20TCL_SHADE_MODEL 0x0000037c +#define NV20TCL_SHADE_MODEL_FLAT 0x00001d00 +#define NV20TCL_SHADE_MODEL_SMOOTH 0x00001d01 +#define NV20TCL_LINE_WIDTH 0x00000380 +#define NV20TCL_POLYGON_OFFSET_FACTOR 0x00000384 +#define NV20TCL_POLYGON_OFFSET_UNITS 0x00000388 +#define NV20TCL_POLYGON_MODE_FRONT 0x0000038c +#define NV20TCL_POLYGON_MODE_FRONT_POINT 0x00001b00 +#define NV20TCL_POLYGON_MODE_FRONT_LINE 0x00001b01 +#define NV20TCL_POLYGON_MODE_FRONT_FILL 0x00001b02 +#define NV20TCL_POLYGON_MODE_BACK 0x00000390 +#define NV20TCL_POLYGON_MODE_BACK_POINT 0x00001b00 +#define NV20TCL_POLYGON_MODE_BACK_LINE 0x00001b01 +#define NV20TCL_POLYGON_MODE_BACK_FILL 0x00001b02 +#define NV20TCL_DEPTH_RANGE_NEAR 0x00000394 +#define NV20TCL_DEPTH_RANGE_FAR 0x00000398 +#define NV20TCL_CULL_FACE 0x0000039c +#define NV20TCL_CULL_FACE_FRONT 0x00000404 +#define NV20TCL_CULL_FACE_BACK 0x00000405 +#define NV20TCL_CULL_FACE_FRONT_AND_BACK 0x00000408 +#define NV20TCL_FRONT_FACE 0x000003a0 +#define NV20TCL_FRONT_FACE_CW 0x00000900 +#define NV20TCL_FRONT_FACE_CCW 0x00000901 +#define NV20TCL_NORMALIZE_ENABLE 0x000003a4 +#define NV20TCL_COLOR_MATERIAL_FRONT_R 0x000003a8 +#define NV20TCL_COLOR_MATERIAL_FRONT_G 0x000003ac +#define NV20TCL_COLOR_MATERIAL_FRONT_B 0x000003b0 +#define NV20TCL_COLOR_MATERIAL_FRONT_A 0x000003b4 +#define NV20TCL_SEPARATE_SPECULAR_ENABLE 0x000003b8 +#define NV20TCL_ENABLED_LIGHTS 0x000003bc +#define NV20TCL_TX_GEN_S(x) (0x000003c0+((x)*16)) +#define NV20TCL_TX_GEN_S__SIZE 0x00000004 +#define NV20TCL_TX_GEN_S_FALSE 0x00000000 +#define NV20TCL_TX_GEN_S_EYE_LINEAR 0x00002400 +#define NV20TCL_TX_GEN_S_OBJECT_LINEAR 0x00002401 +#define NV20TCL_TX_GEN_S_SPHERE_MAP 0x00002402 +#define NV20TCL_TX_GEN_S_NORMAL_MAP 0x00008511 +#define NV20TCL_TX_GEN_S_REFLECTION_MAP 0x00008512 +#define NV20TCL_TX_GEN_T(x) (0x000003c4+((x)*16)) +#define NV20TCL_TX_GEN_T__SIZE 0x00000004 +#define NV20TCL_TX_GEN_T_FALSE 0x00000000 +#define NV20TCL_TX_GEN_T_EYE_LINEAR 0x00002400 +#define NV20TCL_TX_GEN_T_OBJECT_LINEAR 0x00002401 +#define NV20TCL_TX_GEN_T_SPHERE_MAP 0x00002402 +#define NV20TCL_TX_GEN_T_NORMAL_MAP 0x00008511 +#define NV20TCL_TX_GEN_T_REFLECTION_MAP 0x00008512 +#define NV20TCL_TX_GEN_R(x) (0x000003c8+((x)*16)) +#define NV20TCL_TX_GEN_R__SIZE 0x00000004 +#define NV20TCL_TX_GEN_R_FALSE 0x00000000 +#define NV20TCL_TX_GEN_R_EYE_LINEAR 0x00002400 +#define NV20TCL_TX_GEN_R_OBJECT_LINEAR 0x00002401 +#define NV20TCL_TX_GEN_R_SPHERE_MAP 0x00002402 +#define NV20TCL_TX_GEN_R_NORMAL_MAP 0x00008511 +#define NV20TCL_TX_GEN_R_REFLECTION_MAP 0x00008512 +#define NV20TCL_TX_GEN_Q(x) (0x000003cc+((x)*16)) +#define NV20TCL_TX_GEN_Q__SIZE 0x00000004 +#define NV20TCL_TX_GEN_Q_FALSE 0x00000000 +#define NV20TCL_TX_GEN_Q_EYE_LINEAR 0x00002400 +#define NV20TCL_TX_GEN_Q_OBJECT_LINEAR 0x00002401 +#define NV20TCL_TX_GEN_Q_SPHERE_MAP 0x00002402 +#define NV20TCL_TX_GEN_Q_NORMAL_MAP 0x00008511 +#define NV20TCL_TX_GEN_Q_REFLECTION_MAP 0x00008512 +#define NV20TCL_TX_MATRIX_ENABLE(x) (0x00000420+((x)*4)) +#define NV20TCL_TX_MATRIX_ENABLE__SIZE 0x00000004 +#define NV20TCL_POINT_SIZE 0x0000043c +#define NV20TCL_MODELVIEW0_MATRIX(x) (0x00000480+((x)*4)) +#define NV20TCL_MODELVIEW0_MATRIX__SIZE 0x00000010 +#define NV20TCL_MODELVIEW1_MATRIX(x) (0x000004c0+((x)*4)) +#define NV20TCL_MODELVIEW1_MATRIX__SIZE 0x00000010 +#define NV20TCL_MODELVIEW2_MATRIX(x) (0x00000500+((x)*4)) +#define NV20TCL_MODELVIEW2_MATRIX__SIZE 0x00000010 +#define NV20TCL_MODELVIEW3_MATRIX(x) (0x00000540+((x)*4)) +#define NV20TCL_MODELVIEW3_MATRIX__SIZE 0x00000010 +#define NV20TCL_INVERSE_MODELVIEW0_MATRIX(x) (0x00000580+((x)*4)) +#define NV20TCL_INVERSE_MODELVIEW0_MATRIX__SIZE 0x00000010 +#define NV20TCL_INVERSE_MODELVIEW1_MATRIX(x) (0x000005c0+((x)*4)) +#define NV20TCL_INVERSE_MODELVIEW1_MATRIX__SIZE 0x00000010 +#define NV20TCL_INVERSE_MODELVIEW2_MATRIX(x) (0x00000600+((x)*4)) +#define NV20TCL_INVERSE_MODELVIEW2_MATRIX__SIZE 0x00000010 +#define NV20TCL_INVERSE_MODELVIEW3_MATRIX(x) (0x00000640+((x)*4)) +#define NV20TCL_INVERSE_MODELVIEW3_MATRIX__SIZE 0x00000010 +#define NV20TCL_PROJECTION_MATRIX(x) (0x00000680+((x)*4)) +#define NV20TCL_PROJECTION_MATRIX__SIZE 0x00000010 +#define NV20TCL_TX0_MATRIX(x) (0x000006c0+((x)*4)) +#define NV20TCL_TX0_MATRIX__SIZE 0x00000010 +#define NV20TCL_TX1_MATRIX(x) (0x00000700+((x)*4)) +#define NV20TCL_TX1_MATRIX__SIZE 0x00000010 +#define NV20TCL_TX2_MATRIX(x) (0x00000740+((x)*4)) +#define NV20TCL_TX2_MATRIX__SIZE 0x00000010 +#define NV20TCL_TX3_MATRIX(x) (0x00000780+((x)*4)) +#define NV20TCL_TX3_MATRIX__SIZE 0x00000010 +#define NV20TCL_TX0_CLIP_PLANE_A(x) (0x00000840+((x)*16)) +#define NV20TCL_TX0_CLIP_PLANE_A__SIZE 0x00000004 +#define NV20TCL_TX0_CLIP_PLANE_B(x) (0x00000844+((x)*16)) +#define NV20TCL_TX0_CLIP_PLANE_B__SIZE 0x00000004 +#define NV20TCL_TX0_CLIP_PLANE_C(x) (0x00000848+((x)*16)) +#define NV20TCL_TX0_CLIP_PLANE_C__SIZE 0x00000004 +#define NV20TCL_TX0_CLIP_PLANE_D(x) (0x0000084c+((x)*16)) +#define NV20TCL_TX0_CLIP_PLANE_D__SIZE 0x00000004 +#define NV20TCL_TX1_CLIP_PLANE_A(x) (0x00000880+((x)*16)) +#define NV20TCL_TX1_CLIP_PLANE_A__SIZE 0x00000004 +#define NV20TCL_TX1_CLIP_PLANE_B(x) (0x00000884+((x)*16)) +#define NV20TCL_TX1_CLIP_PLANE_B__SIZE 0x00000004 +#define NV20TCL_TX1_CLIP_PLANE_C(x) (0x00000888+((x)*16)) +#define NV20TCL_TX1_CLIP_PLANE_C__SIZE 0x00000004 +#define NV20TCL_TX1_CLIP_PLANE_D(x) (0x0000088c+((x)*16)) +#define NV20TCL_TX1_CLIP_PLANE_D__SIZE 0x00000004 +#define NV20TCL_TX2_CLIP_PLANE_A(x) (0x000008c0+((x)*16)) +#define NV20TCL_TX2_CLIP_PLANE_A__SIZE 0x00000004 +#define NV20TCL_TX2_CLIP_PLANE_B(x) (0x000008c4+((x)*16)) +#define NV20TCL_TX2_CLIP_PLANE_B__SIZE 0x00000004 +#define NV20TCL_TX2_CLIP_PLANE_C(x) (0x000008c8+((x)*16)) +#define NV20TCL_TX2_CLIP_PLANE_C__SIZE 0x00000004 +#define NV20TCL_TX2_CLIP_PLANE_D(x) (0x000008cc+((x)*16)) +#define NV20TCL_TX2_CLIP_PLANE_D__SIZE 0x00000004 +#define NV20TCL_TX3_CLIP_PLANE_A(x) (0x00000900+((x)*16)) +#define NV20TCL_TX3_CLIP_PLANE_A__SIZE 0x00000004 +#define NV20TCL_TX3_CLIP_PLANE_B(x) (0x00000904+((x)*16)) +#define NV20TCL_TX3_CLIP_PLANE_B__SIZE 0x00000004 +#define NV20TCL_TX3_CLIP_PLANE_C(x) (0x00000908+((x)*16)) +#define NV20TCL_TX3_CLIP_PLANE_C__SIZE 0x00000004 +#define NV20TCL_TX3_CLIP_PLANE_D(x) (0x0000090c+((x)*16)) +#define NV20TCL_TX3_CLIP_PLANE_D__SIZE 0x00000004 +#define NV20TCL_FOG_EQUATION_CONSTANT 0x000009c0 +#define NV20TCL_FOG_EQUATION_LINEAR 0x000009c4 +#define NV20TCL_FOG_EQUATION_QUADRATIC 0x000009c8 +#define NV20TCL_FRONT_MATERIAL_SHININESS(x) (0x000009e0+((x)*4)) +#define NV20TCL_FRONT_MATERIAL_SHININESS__SIZE 0x00000006 +#define NV20TCL_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_R 0x00000a10 +#define NV20TCL_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_G 0x00000a14 +#define NV20TCL_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_B 0x00000a18 +#define NV20TCL_VIEWPORT_SCALE0_X 0x00000a20 +#define NV20TCL_VIEWPORT_SCALE0_Y 0x00000a24 +#define NV20TCL_VIEWPORT_SCALE0_Z 0x00000a28 +#define NV20TCL_VIEWPORT_SCALE0_W 0x00000a2c +#define NV20TCL_POINT_PARAMETER(x) (0x00000a30+((x)*4)) +#define NV20TCL_POINT_PARAMETER__SIZE 0x00000008 +#define NV20TCL_RC_CONSTANT_COLOR0(x) (0x00000a60+((x)*4)) +#define NV20TCL_RC_CONSTANT_COLOR0__SIZE 0x00000008 +#define NV20TCL_RC_CONSTANT_COLOR0_B_SHIFT 0 +#define NV20TCL_RC_CONSTANT_COLOR0_B_MASK 0x000000ff +#define NV20TCL_RC_CONSTANT_COLOR0_G_SHIFT 8 +#define NV20TCL_RC_CONSTANT_COLOR0_G_MASK 0x0000ff00 +#define NV20TCL_RC_CONSTANT_COLOR0_R_SHIFT 16 +#define NV20TCL_RC_CONSTANT_COLOR0_R_MASK 0x00ff0000 +#define NV20TCL_RC_CONSTANT_COLOR0_A_SHIFT 24 +#define NV20TCL_RC_CONSTANT_COLOR0_A_MASK 0xff000000 +#define NV20TCL_RC_CONSTANT_COLOR1(x) (0x00000a80+((x)*4)) +#define NV20TCL_RC_CONSTANT_COLOR1__SIZE 0x00000008 +#define NV20TCL_RC_CONSTANT_COLOR1_B_SHIFT 0 +#define NV20TCL_RC_CONSTANT_COLOR1_B_MASK 0x000000ff +#define NV20TCL_RC_CONSTANT_COLOR1_G_SHIFT 8 +#define NV20TCL_RC_CONSTANT_COLOR1_G_MASK 0x0000ff00 +#define NV20TCL_RC_CONSTANT_COLOR1_R_SHIFT 16 +#define NV20TCL_RC_CONSTANT_COLOR1_R_MASK 0x00ff0000 +#define NV20TCL_RC_CONSTANT_COLOR1_A_SHIFT 24 +#define NV20TCL_RC_CONSTANT_COLOR1_A_MASK 0xff000000 +#define NV20TCL_RC_OUT_ALPHA(x) (0x00000aa0+((x)*4)) +#define NV20TCL_RC_OUT_ALPHA__SIZE 0x00000008 +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_SHIFT 0 +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_MASK 0x0000000f +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_ZERO 0x00000000 +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_FOG 0x00000003 +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_TEXTURE0_ARB 0x00000008 +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_TEXTURE1_ARB 0x00000009 +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_SPARE0_NV 0x0000000c +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_SPARE1_NV 0x0000000d +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV20TCL_RC_OUT_ALPHA_CD_OUTPUT_E_TIMES_F_NV 0x0000000f +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_SHIFT 4 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_MASK 0x000000f0 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_ZERO 0x00000000 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_CONSTANT_COLOR0_NV 0x00000010 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_CONSTANT_COLOR1_NV 0x00000020 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_FOG 0x00000030 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_PRIMARY_COLOR_NV 0x00000040 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_SECONDARY_COLOR_NV 0x00000050 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_TEXTURE0_ARB 0x00000080 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_TEXTURE1_ARB 0x00000090 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_SPARE0_NV 0x000000c0 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_SPARE1_NV 0x000000d0 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000000e0 +#define NV20TCL_RC_OUT_ALPHA_AB_OUTPUT_E_TIMES_F_NV 0x000000f0 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_SHIFT 8 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_MASK 0x00000f00 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_ZERO 0x00000000 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_FOG 0x00000300 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_TEXTURE0_ARB 0x00000800 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_TEXTURE1_ARB 0x00000900 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_SPARE0_NV 0x00000c00 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_SPARE1_NV 0x00000d00 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV20TCL_RC_OUT_ALPHA_SUM_OUTPUT_E_TIMES_F_NV 0x00000f00 +#define NV20TCL_RC_OUT_ALPHA_CD_DOT_PRODUCT (1 << 12) +#define NV20TCL_RC_OUT_ALPHA_AB_DOT_PRODUCT (1 << 13) +#define NV20TCL_RC_OUT_ALPHA_MUX_SUM (1 << 14) +#define NV20TCL_RC_OUT_ALPHA_BIAS (1 << 15) +#define NV20TCL_RC_OUT_ALPHA_BIAS_NONE 0x00000000 +#define NV20TCL_RC_OUT_ALPHA_BIAS_BIAS_BY_NEGATIVE_ONE_HALF_NV 0x00008000 +#define NV20TCL_RC_OUT_ALPHA_SCALE_SHIFT 17 +#define NV20TCL_RC_OUT_ALPHA_SCALE_MASK 0x00000000 +#define NV20TCL_RC_OUT_ALPHA_SCALE_NONE 0x00000000 +#define NV20TCL_RC_OUT_ALPHA_SCALE_SCALE_BY_TWO_NV 0x00020000 +#define NV20TCL_RC_OUT_ALPHA_SCALE_SCALE_BY_FOUR_NV 0x00040000 +#define NV20TCL_RC_OUT_ALPHA_SCALE_SCALE_BY_ONE_HALF_NV 0x00060000 +#define NV20TCL_RC_IN_RGB(x) (0x00000ac0+((x)*4)) +#define NV20TCL_RC_IN_RGB__SIZE 0x00000008 +#define NV20TCL_RC_IN_RGB_D_INPUT_SHIFT 0 +#define NV20TCL_RC_IN_RGB_D_INPUT_MASK 0x0000000f +#define NV20TCL_RC_IN_RGB_D_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_IN_RGB_D_INPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV20TCL_RC_IN_RGB_D_INPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV20TCL_RC_IN_RGB_D_INPUT_FOG 0x00000003 +#define NV20TCL_RC_IN_RGB_D_INPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV20TCL_RC_IN_RGB_D_INPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV20TCL_RC_IN_RGB_D_INPUT_TEXTURE0_ARB 0x00000008 +#define NV20TCL_RC_IN_RGB_D_INPUT_TEXTURE1_ARB 0x00000009 +#define NV20TCL_RC_IN_RGB_D_INPUT_SPARE0_NV 0x0000000c +#define NV20TCL_RC_IN_RGB_D_INPUT_SPARE1_NV 0x0000000d +#define NV20TCL_RC_IN_RGB_D_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV20TCL_RC_IN_RGB_D_INPUT_E_TIMES_F_NV 0x0000000f +#define NV20TCL_RC_IN_RGB_D_COMPONENT_USAGE (1 << 4) +#define NV20TCL_RC_IN_RGB_D_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_IN_RGB_D_COMPONENT_USAGE_ALPHA 0x00000010 +#define NV20TCL_RC_IN_RGB_D_MAPPING_SHIFT 5 +#define NV20TCL_RC_IN_RGB_D_MAPPING_MASK 0x000000e0 +#define NV20TCL_RC_IN_RGB_D_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_IN_RGB_D_MAPPING_UNSIGNED_INVERT_NV 0x00000020 +#define NV20TCL_RC_IN_RGB_D_MAPPING_EXPAND_NORMAL_NV 0x00000040 +#define NV20TCL_RC_IN_RGB_D_MAPPING_EXPAND_NEGATE_NV 0x00000060 +#define NV20TCL_RC_IN_RGB_D_MAPPING_HALF_BIAS_NORMAL_NV 0x00000080 +#define NV20TCL_RC_IN_RGB_D_MAPPING_HALF_BIAS_NEGATE_NV 0x000000a0 +#define NV20TCL_RC_IN_RGB_D_MAPPING_SIGNED_IDENTITY_NV 0x000000c0 +#define NV20TCL_RC_IN_RGB_D_MAPPING_SIGNED_NEGATE_NV 0x000000e0 +#define NV20TCL_RC_IN_RGB_C_INPUT_SHIFT 8 +#define NV20TCL_RC_IN_RGB_C_INPUT_MASK 0x00000f00 +#define NV20TCL_RC_IN_RGB_C_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_IN_RGB_C_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV20TCL_RC_IN_RGB_C_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV20TCL_RC_IN_RGB_C_INPUT_FOG 0x00000300 +#define NV20TCL_RC_IN_RGB_C_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV20TCL_RC_IN_RGB_C_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV20TCL_RC_IN_RGB_C_INPUT_TEXTURE0_ARB 0x00000800 +#define NV20TCL_RC_IN_RGB_C_INPUT_TEXTURE1_ARB 0x00000900 +#define NV20TCL_RC_IN_RGB_C_INPUT_SPARE0_NV 0x00000c00 +#define NV20TCL_RC_IN_RGB_C_INPUT_SPARE1_NV 0x00000d00 +#define NV20TCL_RC_IN_RGB_C_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV20TCL_RC_IN_RGB_C_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV20TCL_RC_IN_RGB_C_COMPONENT_USAGE (1 << 12) +#define NV20TCL_RC_IN_RGB_C_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_IN_RGB_C_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV20TCL_RC_IN_RGB_C_MAPPING_SHIFT 13 +#define NV20TCL_RC_IN_RGB_C_MAPPING_MASK 0x0000e000 +#define NV20TCL_RC_IN_RGB_C_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_IN_RGB_C_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV20TCL_RC_IN_RGB_C_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV20TCL_RC_IN_RGB_C_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV20TCL_RC_IN_RGB_C_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV20TCL_RC_IN_RGB_C_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV20TCL_RC_IN_RGB_C_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV20TCL_RC_IN_RGB_C_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV20TCL_RC_IN_RGB_B_INPUT_SHIFT 16 +#define NV20TCL_RC_IN_RGB_B_INPUT_MASK 0x000f0000 +#define NV20TCL_RC_IN_RGB_B_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_IN_RGB_B_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV20TCL_RC_IN_RGB_B_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV20TCL_RC_IN_RGB_B_INPUT_FOG 0x00030000 +#define NV20TCL_RC_IN_RGB_B_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV20TCL_RC_IN_RGB_B_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV20TCL_RC_IN_RGB_B_INPUT_TEXTURE0_ARB 0x00080000 +#define NV20TCL_RC_IN_RGB_B_INPUT_TEXTURE1_ARB 0x00090000 +#define NV20TCL_RC_IN_RGB_B_INPUT_SPARE0_NV 0x000c0000 +#define NV20TCL_RC_IN_RGB_B_INPUT_SPARE1_NV 0x000d0000 +#define NV20TCL_RC_IN_RGB_B_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV20TCL_RC_IN_RGB_B_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV20TCL_RC_IN_RGB_B_COMPONENT_USAGE (1 << 20) +#define NV20TCL_RC_IN_RGB_B_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_IN_RGB_B_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV20TCL_RC_IN_RGB_B_MAPPING_SHIFT 21 +#define NV20TCL_RC_IN_RGB_B_MAPPING_MASK 0x00e00000 +#define NV20TCL_RC_IN_RGB_B_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_IN_RGB_B_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV20TCL_RC_IN_RGB_B_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV20TCL_RC_IN_RGB_B_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV20TCL_RC_IN_RGB_B_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV20TCL_RC_IN_RGB_B_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV20TCL_RC_IN_RGB_B_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV20TCL_RC_IN_RGB_B_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV20TCL_RC_IN_RGB_A_INPUT_SHIFT 24 +#define NV20TCL_RC_IN_RGB_A_INPUT_MASK 0x0f000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_ZERO 0x00000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_FOG 0x03000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_TEXTURE0_ARB 0x08000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_TEXTURE1_ARB 0x09000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_SPARE0_NV 0x0c000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_SPARE1_NV 0x0d000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV20TCL_RC_IN_RGB_A_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV20TCL_RC_IN_RGB_A_COMPONENT_USAGE (1 << 28) +#define NV20TCL_RC_IN_RGB_A_COMPONENT_USAGE_RGB 0x00000000 +#define NV20TCL_RC_IN_RGB_A_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV20TCL_RC_IN_RGB_A_MAPPING_SHIFT 29 +#define NV20TCL_RC_IN_RGB_A_MAPPING_MASK 0xe0000000 +#define NV20TCL_RC_IN_RGB_A_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV20TCL_RC_IN_RGB_A_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV20TCL_RC_IN_RGB_A_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV20TCL_RC_IN_RGB_A_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV20TCL_RC_IN_RGB_A_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV20TCL_RC_IN_RGB_A_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV20TCL_RC_IN_RGB_A_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV20TCL_RC_IN_RGB_A_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV20TCL_VIEWPORT_SCALE1_X 0x00000af0 +#define NV20TCL_VIEWPORT_SCALE1_Y 0x00000af4 +#define NV20TCL_VIEWPORT_SCALE1_Z 0x00000af8 +#define NV20TCL_VIEWPORT_SCALE1_W 0x00000afc +#define NV20TCL_VP_UPLOAD_INST(x) (0x00000b00+((x)*4)) +#define NV20TCL_VP_UPLOAD_INST__SIZE 0x00000004 +#define NV20TCL_VP_UPLOAD_CONST(x) (0x00000b80+((x)*4)) +#define NV20TCL_VP_UPLOAD_CONST__SIZE 0x00000004 +#define NV20TCL_LIGHT_BACK_SIDE_PRODUCT_AMBIENT_R(x) (0x00000c00+((x)*64)) +#define NV20TCL_LIGHT_BACK_SIDE_PRODUCT_AMBIENT_R__SIZE 0x00000008 +#define NV20TCL_LIGHT_BACK_SIDE_PRODUCT_AMBIENT_G(x) (0x00000c04+((x)*64)) +#define NV20TCL_LIGHT_BACK_SIDE_PRODUCT_AMBIENT_G__SIZE 0x00000008 +#define NV20TCL_LIGHT_BACK_SIDE_PRODUCT_AMBIENT_B(x) (0x00000c08+((x)*64)) +#define NV20TCL_LIGHT_BACK_SIDE_PRODUCT_AMBIENT_B__SIZE 0x00000008 +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R(x) (0x00001000+((x)*128)) +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R__SIZE 0x00000008 +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_G(x) (0x00001004+((x)*128)) +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_G__SIZE 0x00000008 +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_B(x) (0x00001008+((x)*128)) +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_B__SIZE 0x00000008 +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R(x) (0x0000100c+((x)*128)) +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R__SIZE 0x00000008 +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_G(x) (0x00001010+((x)*128)) +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_G__SIZE 0x00000008 +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_B(x) (0x00001014+((x)*128)) +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_B__SIZE 0x00000008 +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R(x) (0x00001018+((x)*128)) +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R__SIZE 0x00000008 +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_G(x) (0x0000101c+((x)*128)) +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_G__SIZE 0x00000008 +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_B(x) (0x00001020+((x)*128)) +#define NV20TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_B__SIZE 0x00000008 +#define NV20TCL_LIGHT_HALF_VECTOR_X(x) (0x00001028+((x)*128)) +#define NV20TCL_LIGHT_HALF_VECTOR_X__SIZE 0x00000008 +#define NV20TCL_LIGHT_HALF_VECTOR_Y(x) (0x0000102c+((x)*128)) +#define NV20TCL_LIGHT_HALF_VECTOR_Y__SIZE 0x00000008 +#define NV20TCL_LIGHT_HALF_VECTOR_Z(x) (0x00001030+((x)*128)) +#define NV20TCL_LIGHT_HALF_VECTOR_Z__SIZE 0x00000008 +#define NV20TCL_LIGHT_DIRECTION_X(x) (0x00001034+((x)*128)) +#define NV20TCL_LIGHT_DIRECTION_X__SIZE 0x00000008 +#define NV20TCL_LIGHT_DIRECTION_Y(x) (0x00001038+((x)*128)) +#define NV20TCL_LIGHT_DIRECTION_Y__SIZE 0x00000008 +#define NV20TCL_LIGHT_DIRECTION_Z(x) (0x0000103c+((x)*128)) +#define NV20TCL_LIGHT_DIRECTION_Z__SIZE 0x00000008 +#define NV20TCL_LIGHT_POSITION_X(x) (0x0000105c+((x)*128)) +#define NV20TCL_LIGHT_POSITION_X__SIZE 0x00000008 +#define NV20TCL_LIGHT_POSITION_Y(x) (0x00001060+((x)*128)) +#define NV20TCL_LIGHT_POSITION_Y__SIZE 0x00000008 +#define NV20TCL_LIGHT_POSITION_Z(x) (0x00001064+((x)*128)) +#define NV20TCL_LIGHT_POSITION_Z__SIZE 0x00000008 +#define NV20TCL_LIGHT_CONSTANT_ATTENUATION(x) (0x00001068+((x)*128)) +#define NV20TCL_LIGHT_CONSTANT_ATTENUATION__SIZE 0x00000008 +#define NV20TCL_LIGHT_LINEAR_ATTENUATION(x) (0x0000106c+((x)*128)) +#define NV20TCL_LIGHT_LINEAR_ATTENUATION__SIZE 0x00000008 +#define NV20TCL_LIGHT_QUADRATIC_ATTENUATION(x) (0x00001070+((x)*128)) +#define NV20TCL_LIGHT_QUADRATIC_ATTENUATION__SIZE 0x00000008 +#define NV20TCL_POLYGON_STIPPLE_ENABLE 0x0000147c +#define NV20TCL_POLYGON_STIPPLE_PATTERN(x) (0x00001480+((x)*4)) +#define NV20TCL_POLYGON_STIPPLE_PATTERN__SIZE 0x00000020 +#define NV20TCL_VERTEX_POS_3F_X 0x00001500 +#define NV20TCL_VERTEX_POS_3F_Y 0x00001504 +#define NV20TCL_VERTEX_POS_3F_Z 0x00001508 +#define NV20TCL_VERTEX_POS_4F_X 0x00001518 +#define NV20TCL_VERTEX_POS_4F_Y 0x0000151c +#define NV20TCL_VERTEX_POS_4F_Z 0x00001520 +#define NV20TCL_VERTEX_POS_3I_XY 0x00001528 +#define NV20TCL_VERTEX_POS_3I_XY_X_SHIFT 0 +#define NV20TCL_VERTEX_POS_3I_XY_X_MASK 0x0000ffff +#define NV20TCL_VERTEX_POS_3I_XY_Y_SHIFT 16 +#define NV20TCL_VERTEX_POS_3I_XY_Y_MASK 0xffff0000 +#define NV20TCL_VERTEX_POS_3I_Z 0x0000152c +#define NV20TCL_VERTEX_POS_3I_Z_Z_SHIFT 0 +#define NV20TCL_VERTEX_POS_3I_Z_Z_MASK 0x0000ffff +#define NV20TCL_VERTEX_NOR_3F_X 0x00001530 +#define NV20TCL_VERTEX_NOR_3F_Y 0x00001534 +#define NV20TCL_VERTEX_NOR_3F_Z 0x00001538 +#define NV20TCL_VERTEX_NOR_3I_XY 0x00001540 +#define NV20TCL_VERTEX_NOR_3I_XY_X_SHIFT 0 +#define NV20TCL_VERTEX_NOR_3I_XY_X_MASK 0x0000ffff +#define NV20TCL_VERTEX_NOR_3I_XY_Y_SHIFT 16 +#define NV20TCL_VERTEX_NOR_3I_XY_Y_MASK 0xffff0000 +#define NV20TCL_VERTEX_NOR_3I_Z 0x00001544 +#define NV20TCL_VERTEX_NOR_3I_Z_Z_SHIFT 0 +#define NV20TCL_VERTEX_NOR_3I_Z_Z_MASK 0x0000ffff +#define NV20TCL_VERTEX_COL_4F_X 0x00001550 +#define NV20TCL_VERTEX_COL_4F_Y 0x00001554 +#define NV20TCL_VERTEX_COL_4F_Z 0x00001558 +#define NV20TCL_VERTEX_COL_4F_W 0x0000155c +#define NV20TCL_VERTEX_COL_3F_X 0x00001560 +#define NV20TCL_VERTEX_COL_3F_Y 0x00001564 +#define NV20TCL_VERTEX_COL_3F_Z 0x00001568 +#define NV20TCL_VERTEX_COL_4I 0x0000156c +#define NV20TCL_VERTEX_COL_4I_R_SHIFT 0 +#define NV20TCL_VERTEX_COL_4I_R_MASK 0x000000ff +#define NV20TCL_VERTEX_COL_4I_G_SHIFT 8 +#define NV20TCL_VERTEX_COL_4I_G_MASK 0x0000ff00 +#define NV20TCL_VERTEX_COL_4I_B_SHIFT 16 +#define NV20TCL_VERTEX_COL_4I_B_MASK 0x00ff0000 +#define NV20TCL_VERTEX_COL_4I_A_SHIFT 24 +#define NV20TCL_VERTEX_COL_4I_A_MASK 0xff000000 +#define NV20TCL_VERTEX_COL2_3F_X 0x00001580 +#define NV20TCL_VERTEX_COL2_3F_Y 0x00001584 +#define NV20TCL_VERTEX_COL2_3F_Z 0x00001588 +#define NV20TCL_VERTEX_COL2_4I 0x0000158c +#define NV20TCL_VERTEX_COL2_4I_R_SHIFT 0 +#define NV20TCL_VERTEX_COL2_4I_R_MASK 0x000000ff +#define NV20TCL_VERTEX_COL2_4I_G_SHIFT 8 +#define NV20TCL_VERTEX_COL2_4I_G_MASK 0x0000ff00 +#define NV20TCL_VERTEX_COL2_4I_B_SHIFT 16 +#define NV20TCL_VERTEX_COL2_4I_B_MASK 0x00ff0000 +#define NV20TCL_VERTEX_COL2_4I_A_SHIFT 24 +#define NV20TCL_VERTEX_COL2_4I_A_MASK 0xff000000 +#define NV20TCL_VERTEX_TX0_2F_S 0x00001590 +#define NV20TCL_VERTEX_TX0_2F_T 0x00001594 +#define NV20TCL_VERTEX_TX0_2I 0x00001598 +#define NV20TCL_VERTEX_TX0_2I_S_SHIFT 0 +#define NV20TCL_VERTEX_TX0_2I_S_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX0_2I_T_SHIFT 16 +#define NV20TCL_VERTEX_TX0_2I_T_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX0_4F_S 0x000015a0 +#define NV20TCL_VERTEX_TX0_4F_T 0x000015a4 +#define NV20TCL_VERTEX_TX0_4F_R 0x000015a8 +#define NV20TCL_VERTEX_TX0_4F_Q 0x000015ac +#define NV20TCL_VERTEX_TX0_4I_ST 0x000015b0 +#define NV20TCL_VERTEX_TX0_4I_ST_S_SHIFT 0 +#define NV20TCL_VERTEX_TX0_4I_ST_S_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX0_4I_ST_T_SHIFT 16 +#define NV20TCL_VERTEX_TX0_4I_ST_T_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX0_4I_RQ 0x000015b4 +#define NV20TCL_VERTEX_TX0_4I_RQ_R_SHIFT 0 +#define NV20TCL_VERTEX_TX0_4I_RQ_R_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX0_4I_RQ_Q_SHIFT 16 +#define NV20TCL_VERTEX_TX0_4I_RQ_Q_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX1_2F_S 0x000015b8 +#define NV20TCL_VERTEX_TX1_2F_T 0x000015bc +#define NV20TCL_VERTEX_TX1_2I 0x000015c0 +#define NV20TCL_VERTEX_TX1_2I_S_SHIFT 0 +#define NV20TCL_VERTEX_TX1_2I_S_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX1_2I_T_SHIFT 16 +#define NV20TCL_VERTEX_TX1_2I_T_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX1_4F_S 0x000015c8 +#define NV20TCL_VERTEX_TX1_4F_T 0x000015cc +#define NV20TCL_VERTEX_TX1_4F_R 0x000015d0 +#define NV20TCL_VERTEX_TX1_4F_Q 0x000015d4 +#define NV20TCL_VERTEX_TX1_4I_ST 0x000015d8 +#define NV20TCL_VERTEX_TX1_4I_ST_S_SHIFT 0 +#define NV20TCL_VERTEX_TX1_4I_ST_S_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX1_4I_ST_T_SHIFT 16 +#define NV20TCL_VERTEX_TX1_4I_ST_T_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX1_4I_RQ 0x000015dc +#define NV20TCL_VERTEX_TX1_4I_RQ_R_SHIFT 0 +#define NV20TCL_VERTEX_TX1_4I_RQ_R_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX1_4I_RQ_Q_SHIFT 16 +#define NV20TCL_VERTEX_TX1_4I_RQ_Q_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX2_2F_S 0x000015e0 +#define NV20TCL_VERTEX_TX2_2F_T 0x000015e4 +#define NV20TCL_VERTEX_TX2_2I 0x000015e8 +#define NV20TCL_VERTEX_TX2_2I_S_SHIFT 0 +#define NV20TCL_VERTEX_TX2_2I_S_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX2_2I_T_SHIFT 16 +#define NV20TCL_VERTEX_TX2_2I_T_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX2_4F_S 0x000015f0 +#define NV20TCL_VERTEX_TX2_4F_T 0x000015f4 +#define NV20TCL_VERTEX_TX2_4F_R 0x000015f8 +#define NV20TCL_VERTEX_TX2_4F_Q 0x000015fc +#define NV20TCL_VERTEX_TX2_4I_ST 0x00001600 +#define NV20TCL_VERTEX_TX2_4I_ST_S_SHIFT 0 +#define NV20TCL_VERTEX_TX2_4I_ST_S_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX2_4I_ST_T_SHIFT 16 +#define NV20TCL_VERTEX_TX2_4I_ST_T_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX2_4I_RQ 0x00001604 +#define NV20TCL_VERTEX_TX2_4I_RQ_R_SHIFT 0 +#define NV20TCL_VERTEX_TX2_4I_RQ_R_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX2_4I_RQ_Q_SHIFT 16 +#define NV20TCL_VERTEX_TX2_4I_RQ_Q_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX3_2F_S 0x00001608 +#define NV20TCL_VERTEX_TX3_2F_T 0x0000160c +#define NV20TCL_VERTEX_TX3_2I 0x00001610 +#define NV20TCL_VERTEX_TX3_2I_S_SHIFT 0 +#define NV20TCL_VERTEX_TX3_2I_S_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX3_2I_T_SHIFT 16 +#define NV20TCL_VERTEX_TX3_2I_T_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX3_4F_S 0x00001620 +#define NV20TCL_VERTEX_TX3_4F_T 0x00001624 +#define NV20TCL_VERTEX_TX3_4F_R 0x00001628 +#define NV20TCL_VERTEX_TX3_4F_Q 0x0000162c +#define NV20TCL_VERTEX_TX3_4I_ST 0x00001630 +#define NV20TCL_VERTEX_TX3_4I_ST_S_SHIFT 0 +#define NV20TCL_VERTEX_TX3_4I_ST_S_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX3_4I_ST_T_SHIFT 16 +#define NV20TCL_VERTEX_TX3_4I_ST_T_MASK 0xffff0000 +#define NV20TCL_VERTEX_TX3_4I_RQ 0x00001634 +#define NV20TCL_VERTEX_TX3_4I_RQ_R_SHIFT 0 +#define NV20TCL_VERTEX_TX3_4I_RQ_R_MASK 0x0000ffff +#define NV20TCL_VERTEX_TX3_4I_RQ_Q_SHIFT 16 +#define NV20TCL_VERTEX_TX3_4I_RQ_Q_MASK 0xffff0000 +#define NV20TCL_VERTEX_FOG_1F 0x00001698 +#define NV20TCL_EDGEFLAG_ENABLE 0x000016bc +#define NV20TCL_VTXBUF_ADDRESS(x) (0x00001720+((x)*4)) +#define NV20TCL_VTXBUF_ADDRESS__SIZE 0x00000010 +#define NV20TCL_VTXBUF_ADDRESS_DMA1 (1 << 31) +#define NV20TCL_VTXBUF_ADDRESS_OFFSET_SHIFT 0 +#define NV20TCL_VTXBUF_ADDRESS_OFFSET_MASK 0x0fffffff +#define NV20TCL_VTXFMT(x) (0x00001760+((x)*4)) +#define NV20TCL_VTXFMT__SIZE 0x00000010 +#define NV20TCL_VTXFMT_TYPE_SHIFT 0 +#define NV20TCL_VTXFMT_TYPE_MASK 0x0000000f +#define NV20TCL_VTXFMT_TYPE_FLOAT 0x00000002 +#define NV20TCL_VTXFMT_TYPE_UBYTE 0x00000004 +#define NV20TCL_VTXFMT_TYPE_USHORT 0x00000005 +#define NV20TCL_VTXFMT_SIZE_SHIFT 4 +#define NV20TCL_VTXFMT_SIZE_MASK 0x000000f0 +#define NV20TCL_VTXFMT_STRIDE_SHIFT 8 +#define NV20TCL_VTXFMT_STRIDE_MASK 0x0000ff00 +#define NV20TCL_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_R 0x000017a0 +#define NV20TCL_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_G 0x000017a4 +#define NV20TCL_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_B 0x000017a8 +#define NV20TCL_COLOR_MATERIAL_BACK_A 0x000017ac +#define NV20TCL_COLOR_MATERIAL_BACK_R 0x000017b0 +#define NV20TCL_COLOR_MATERIAL_BACK_G 0x000017b4 +#define NV20TCL_COLOR_MATERIAL_BACK_B 0x000017b8 +#define NV20TCL_COLOR_LOGIC_OP_ENABLE 0x000017bc +#define NV20TCL_COLOR_LOGIC_OP_OP 0x000017c0 +#define NV20TCL_COLOR_LOGIC_OP_OP_CLEAR 0x00001500 +#define NV20TCL_COLOR_LOGIC_OP_OP_AND 0x00001501 +#define NV20TCL_COLOR_LOGIC_OP_OP_AND_REVERSE 0x00001502 +#define NV20TCL_COLOR_LOGIC_OP_OP_COPY 0x00001503 +#define NV20TCL_COLOR_LOGIC_OP_OP_AND_INVERTED 0x00001504 +#define NV20TCL_COLOR_LOGIC_OP_OP_NOOP 0x00001505 +#define NV20TCL_COLOR_LOGIC_OP_OP_XOR 0x00001506 +#define NV20TCL_COLOR_LOGIC_OP_OP_OR 0x00001507 +#define NV20TCL_COLOR_LOGIC_OP_OP_NOR 0x00001508 +#define NV20TCL_COLOR_LOGIC_OP_OP_EQUIV 0x00001509 +#define NV20TCL_COLOR_LOGIC_OP_OP_INVERT 0x0000150a +#define NV20TCL_COLOR_LOGIC_OP_OP_OR_REVERSE 0x0000150b +#define NV20TCL_COLOR_LOGIC_OP_OP_COPY_INVERTED 0x0000150c +#define NV20TCL_COLOR_LOGIC_OP_OP_OR_INVERTED 0x0000150d +#define NV20TCL_COLOR_LOGIC_OP_OP_NAND 0x0000150e +#define NV20TCL_COLOR_LOGIC_OP_OP_SET 0x0000150f +#define NV20TCL_LIGHT_MODEL_TWO_SIDE_ENABLE 0x000017c4 +#define NV20TCL_TX_SHADER_CULL_MODE 0x000017f8 +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_S (1 << 0) +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_S_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_S_LESS 0x00000001 +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_T (1 << 1) +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_T_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_T_LESS 0x00000002 +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_R (1 << 2) +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_R_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_R_LESS 0x00000004 +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_Q (1 << 3) +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_Q_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX0_Q_LESS 0x00000008 +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_S (1 << 4) +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_S_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_S_LESS 0x00000010 +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_T (1 << 5) +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_T_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_T_LESS 0x00000020 +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_R (1 << 6) +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_R_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_R_LESS 0x00000040 +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_Q (1 << 7) +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_Q_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX1_Q_LESS 0x00000080 +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_S (1 << 8) +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_S_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_S_LESS 0x00000100 +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_T (1 << 9) +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_T_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_T_LESS 0x00000200 +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_R (1 << 10) +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_R_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_R_LESS 0x00000400 +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_Q (1 << 11) +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_Q_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX2_Q_LESS 0x00000800 +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_S (1 << 12) +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_S_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_S_LESS 0x00001000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_T (1 << 13) +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_T_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_T_LESS 0x00002000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_R (1 << 14) +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_R_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_R_LESS 0x00004000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_Q (1 << 15) +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_Q_GEQUAL 0x00000000 +#define NV20TCL_TX_SHADER_CULL_MODE_TX3_Q_LESS 0x00008000 +#define NV20TCL_VERTEX_BEGIN_END 0x000017fc +#define NV20TCL_VERTEX_BEGIN_END_STOP 0x00000000 +#define NV20TCL_VERTEX_BEGIN_END_POINTS 0x00000001 +#define NV20TCL_VERTEX_BEGIN_END_LINES 0x00000002 +#define NV20TCL_VERTEX_BEGIN_END_LINE_LOOP 0x00000003 +#define NV20TCL_VERTEX_BEGIN_END_LINE_STRIP 0x00000004 +#define NV20TCL_VERTEX_BEGIN_END_TRIANGLES 0x00000005 +#define NV20TCL_VERTEX_BEGIN_END_TRIANGLE_STRIP 0x00000006 +#define NV20TCL_VERTEX_BEGIN_END_TRIANGLE_FAN 0x00000007 +#define NV20TCL_VERTEX_BEGIN_END_QUADS 0x00000008 +#define NV20TCL_VERTEX_BEGIN_END_QUAD_STRIP 0x00000009 +#define NV20TCL_VERTEX_BEGIN_END_POLYGON 0x0000000a +#define NV20TCL_VB_ELEMENT_U16 0x00001800 +#define NV20TCL_VB_ELEMENT_U16_I0_SHIFT 0 +#define NV20TCL_VB_ELEMENT_U16_I0_MASK 0x0000ffff +#define NV20TCL_VB_ELEMENT_U16_I1_SHIFT 16 +#define NV20TCL_VB_ELEMENT_U16_I1_MASK 0xffff0000 +#define NV20TCL_VB_VERTEX_BATCH 0x00001810 +#define NV20TCL_VB_VERTEX_BATCH_OFFSET_SHIFT 0 +#define NV20TCL_VB_VERTEX_BATCH_OFFSET_MASK 0x00ffffff +#define NV20TCL_VB_VERTEX_BATCH_COUNT_SHIFT 24 +#define NV20TCL_VB_VERTEX_BATCH_COUNT_MASK 0xff000000 +#define NV20TCL_VERTEX_DATA 0x00001818 +#define NV20TCL_TX_SHADER_CONST_EYE_X 0x0000181c +#define NV20TCL_TX_SHADER_CONST_EYE_Y 0x00001820 +#define NV20TCL_TX_SHADER_CONST_EYE_Z 0x00001824 +#define NV20TCL_VTX_ATTR_4F_X(x) (0x00001a00+((x)*16)) +#define NV20TCL_VTX_ATTR_4F_X__SIZE 0x00000010 +#define NV20TCL_VTX_ATTR_4F_Y(x) (0x00001a04+((x)*16)) +#define NV20TCL_VTX_ATTR_4F_Y__SIZE 0x00000010 +#define NV20TCL_VTX_ATTR_4F_Z(x) (0x00001a08+((x)*16)) +#define NV20TCL_VTX_ATTR_4F_Z__SIZE 0x00000010 +#define NV20TCL_VTX_ATTR_4F_W(x) (0x00001a0c+((x)*16)) +#define NV20TCL_VTX_ATTR_4F_W__SIZE 0x00000010 +#define NV20TCL_TX_OFFSET(x) (0x00001b00+((x)*64)) +#define NV20TCL_TX_OFFSET__SIZE 0x00000004 +#define NV20TCL_TX_FORMAT(x) (0x00001b04+((x)*64)) +#define NV20TCL_TX_FORMAT__SIZE 0x00000004 +#define NV20TCL_TX_FORMAT_DMA0 (1 << 0) +#define NV20TCL_TX_FORMAT_DMA1 (1 << 1) +#define NV20TCL_TX_FORMAT_CUBIC (1 << 2) +#define NV20TCL_TX_FORMAT_NO_BORDER (1 << 3) +#define NV20TCL_TX_FORMAT_DIMS_SHIFT 4 +#define NV20TCL_TX_FORMAT_DIMS_MASK 0x000000f0 +#define NV20TCL_TX_FORMAT_DIMS_1D 0x00000010 +#define NV20TCL_TX_FORMAT_DIMS_2D 0x00000020 +#define NV20TCL_TX_FORMAT_DIMS_3D 0x00000030 +#define NV20TCL_TX_FORMAT_FORMAT_SHIFT 8 +#define NV20TCL_TX_FORMAT_FORMAT_MASK 0x0000ff00 +#define NV20TCL_TX_FORMAT_FORMAT_L8 0x00000000 +#define NV20TCL_TX_FORMAT_FORMAT_A8 0x00000100 +#define NV20TCL_TX_FORMAT_FORMAT_A1R5G5B5 0x00000200 +#define NV20TCL_TX_FORMAT_FORMAT_A8_RECT 0x00000300 +#define NV20TCL_TX_FORMAT_FORMAT_A4R4G4B4 0x00000400 +#define NV20TCL_TX_FORMAT_FORMAT_R5G6B5 0x00000500 +#define NV20TCL_TX_FORMAT_FORMAT_A8R8G8B8 0x00000600 +#define NV20TCL_TX_FORMAT_FORMAT_X8R8G8B8 0x00000700 +#define NV20TCL_TX_FORMAT_FORMAT_INDEX8 0x00000b00 +#define NV20TCL_TX_FORMAT_FORMAT_DXT1 0x00000c00 +#define NV20TCL_TX_FORMAT_FORMAT_DXT3 0x00000e00 +#define NV20TCL_TX_FORMAT_FORMAT_DXT5 0x00000f00 +#define NV20TCL_TX_FORMAT_FORMAT_A1R5G5B5_RECT 0x00001000 +#define NV20TCL_TX_FORMAT_FORMAT_R5G6B5_RECT 0x00001100 +#define NV20TCL_TX_FORMAT_FORMAT_A8R8G8B8_RECT 0x00001200 +#define NV20TCL_TX_FORMAT_FORMAT_L8_RECT 0x00001300 +#define NV20TCL_TX_FORMAT_FORMAT_A8L8 0x00001a00 +#define NV20TCL_TX_FORMAT_FORMAT_A8_RECT2 0x00001b00 +#define NV20TCL_TX_FORMAT_FORMAT_A4R4G4B4_RECT 0x00001d00 +#define NV20TCL_TX_FORMAT_FORMAT_R8G8B8_RECT 0x00001e00 +#define NV20TCL_TX_FORMAT_FORMAT_L8A8_RECT 0x00002000 +#define NV20TCL_TX_FORMAT_FORMAT_DSDT 0x00002800 +#define NV20TCL_TX_FORMAT_FORMAT_A16 0x00003200 +#define NV20TCL_TX_FORMAT_FORMAT_HILO16 0x00003300 +#define NV20TCL_TX_FORMAT_FORMAT_A16_RECT 0x00003500 +#define NV20TCL_TX_FORMAT_FORMAT_HILO16_RECT 0x00003600 +#define NV20TCL_TX_FORMAT_FORMAT_HILO8 0x00004400 +#define NV20TCL_TX_FORMAT_FORMAT_SIGNED_HILO8 0x00004500 +#define NV20TCL_TX_FORMAT_FORMAT_HILO8_RECT 0x00004600 +#define NV20TCL_TX_FORMAT_FORMAT_SIGNED_HILO8_RECT 0x00004700 +#define NV20TCL_TX_FORMAT_FORMAT_FLOAT_RGBA16_NV 0x00004a00 +#define NV20TCL_TX_FORMAT_FORMAT_FLOAT_RGBA32_NV 0x00004b00 +#define NV20TCL_TX_FORMAT_FORMAT_FLOAT_R32_NV 0x00004c00 +#define NV20TCL_TX_FORMAT_MIPMAP (1 << 19) +#define NV20TCL_TX_FORMAT_BASE_SIZE_U_SHIFT 20 +#define NV20TCL_TX_FORMAT_BASE_SIZE_U_MASK 0x00f00000 +#define NV20TCL_TX_FORMAT_BASE_SIZE_V_SHIFT 24 +#define NV20TCL_TX_FORMAT_BASE_SIZE_V_MASK 0x0f000000 +#define NV20TCL_TX_FORMAT_BASE_SIZE_W_SHIFT 28 +#define NV20TCL_TX_FORMAT_BASE_SIZE_W_MASK 0xf0000000 +#define NV20TCL_TX_WRAP(x) (0x00001b08+((x)*64)) +#define NV20TCL_TX_WRAP__SIZE 0x00000004 +#define NV20TCL_TX_WRAP_S_SHIFT 0 +#define NV20TCL_TX_WRAP_S_MASK 0x000000ff +#define NV20TCL_TX_WRAP_S_REPEAT 0x00000001 +#define NV20TCL_TX_WRAP_S_MIRRORED_REPEAT 0x00000002 +#define NV20TCL_TX_WRAP_S_CLAMP_TO_EDGE 0x00000003 +#define NV20TCL_TX_WRAP_S_CLAMP_TO_BORDER 0x00000004 +#define NV20TCL_TX_WRAP_S_CLAMP 0x00000005 +#define NV20TCL_TX_WRAP_T_SHIFT 8 +#define NV20TCL_TX_WRAP_T_MASK 0x00000f00 +#define NV20TCL_TX_WRAP_T_REPEAT 0x00000100 +#define NV20TCL_TX_WRAP_T_MIRRORED_REPEAT 0x00000200 +#define NV20TCL_TX_WRAP_T_CLAMP_TO_EDGE 0x00000300 +#define NV20TCL_TX_WRAP_T_CLAMP_TO_BORDER 0x00000400 +#define NV20TCL_TX_WRAP_T_CLAMP 0x00000500 +#define NV20TCL_TX_WRAP_R_SHIFT 16 +#define NV20TCL_TX_WRAP_R_MASK 0x000f0000 +#define NV20TCL_TX_WRAP_R_REPEAT 0x00010000 +#define NV20TCL_TX_WRAP_R_MIRRORED_REPEAT 0x00020000 +#define NV20TCL_TX_WRAP_R_CLAMP_TO_EDGE 0x00030000 +#define NV20TCL_TX_WRAP_R_CLAMP_TO_BORDER 0x00040000 +#define NV20TCL_TX_WRAP_R_CLAMP 0x00050000 +#define NV20TCL_TX_ENABLE(x) (0x00001b0c+((x)*64)) +#define NV20TCL_TX_ENABLE__SIZE 0x00000004 +#define NV20TCL_TX_ENABLE_ANISO_SHIFT 4 +#define NV20TCL_TX_ENABLE_ANISO_MASK 0x00000030 +#define NV20TCL_TX_ENABLE_ANISO_NONE 0x00000000 +#define NV20TCL_TX_ENABLE_ANISO_2X 0x00000010 +#define NV20TCL_TX_ENABLE_ANISO_4X 0x00000020 +#define NV20TCL_TX_ENABLE_ANISO_8X 0x00000030 +#define NV20TCL_TX_ENABLE_MIPMAP_MAX_LOD_SHIFT 14 +#define NV20TCL_TX_ENABLE_MIPMAP_MAX_LOD_MASK 0x0003c000 +#define NV20TCL_TX_ENABLE_MIPMAP_MIN_LOD_SHIFT 26 +#define NV20TCL_TX_ENABLE_MIPMAP_MIN_LOD_MASK 0x3c000000 +#define NV20TCL_TX_ENABLE_ENABLE (1 << 30) +#define NV20TCL_TX_SWIZZLE(x) (0x00001b10+((x)*64)) +#define NV20TCL_TX_SWIZZLE__SIZE 0x00000004 +#define NV20TCL_TX_SWIZZLE_RECT_PITCH_SHIFT 16 +#define NV20TCL_TX_SWIZZLE_RECT_PITCH_MASK 0xffff0000 +#define NV20TCL_TX_FILTER(x) (0x00001b14+((x)*64)) +#define NV20TCL_TX_FILTER__SIZE 0x00000004 +#define NV20TCL_TX_FILTER_LOD_BIAS_SHIFT 8 +#define NV20TCL_TX_FILTER_LOD_BIAS_MASK 0x00000f00 +#define NV20TCL_TX_FILTER_MINIFY_SHIFT 16 +#define NV20TCL_TX_FILTER_MINIFY_MASK 0x000f0000 +#define NV20TCL_TX_FILTER_MINIFY_NEAREST 0x00010000 +#define NV20TCL_TX_FILTER_MINIFY_LINEAR 0x00020000 +#define NV20TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST 0x00030000 +#define NV20TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST 0x00040000 +#define NV20TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR 0x00050000 +#define NV20TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR 0x00060000 +#define NV20TCL_TX_FILTER_MAGNIFY_SHIFT 24 +#define NV20TCL_TX_FILTER_MAGNIFY_MASK 0x0f000000 +#define NV20TCL_TX_FILTER_MAGNIFY_NEAREST 0x01000000 +#define NV20TCL_TX_FILTER_MAGNIFY_LINEAR 0x02000000 +#define NV20TCL_TX_NPOT_SIZE(x) (0x00001b1c+((x)*64)) +#define NV20TCL_TX_NPOT_SIZE__SIZE 0x00000004 +#define NV20TCL_TX_NPOT_SIZE_H_SHIFT 0 +#define NV20TCL_TX_NPOT_SIZE_H_MASK 0x0000ffff +#define NV20TCL_TX_NPOT_SIZE_W_SHIFT 16 +#define NV20TCL_TX_NPOT_SIZE_W_MASK 0xffff0000 +#define NV20TCL_TX_PALETTE_OFFSET(x) (0x00001b20+((x)*64)) +#define NV20TCL_TX_PALETTE_OFFSET__SIZE 0x00000004 +#define NV20TCL_TX_BORDER_COLOR(x) (0x00001b24+((x)*64)) +#define NV20TCL_TX_BORDER_COLOR__SIZE 0x00000004 +#define NV20TCL_TX_BORDER_COLOR_B_SHIFT 0 +#define NV20TCL_TX_BORDER_COLOR_B_MASK 0x000000ff +#define NV20TCL_TX_BORDER_COLOR_G_SHIFT 8 +#define NV20TCL_TX_BORDER_COLOR_G_MASK 0x0000ff00 +#define NV20TCL_TX_BORDER_COLOR_R_SHIFT 16 +#define NV20TCL_TX_BORDER_COLOR_R_MASK 0x00ff0000 +#define NV20TCL_TX_BORDER_COLOR_A_SHIFT 24 +#define NV20TCL_TX_BORDER_COLOR_A_MASK 0xff000000 +#define NV20TCL_TX_SHADER_OFFSET_MATRIX00(x) (0x00001b28+((x)*64)) +#define NV20TCL_TX_SHADER_OFFSET_MATRIX00__SIZE 0x00000004 +#define NV20TCL_TX_SHADER_OFFSET_MATRIX01(x) (0x00001b2c+((x)*64)) +#define NV20TCL_TX_SHADER_OFFSET_MATRIX01__SIZE 0x00000004 +#define NV20TCL_TX_SHADER_OFFSET_MATRIX11(x) (0x00001b30+((x)*64)) +#define NV20TCL_TX_SHADER_OFFSET_MATRIX11__SIZE 0x00000004 +#define NV20TCL_TX_SHADER_OFFSET_MATRIX10(x) (0x00001b34+((x)*64)) +#define NV20TCL_TX_SHADER_OFFSET_MATRIX10__SIZE 0x00000004 +#define NV20TCL_DEPTH_UNK17D8 0x00001d78 +#define NV20TCL_DEPTH_UNK17D8_CLAMP_SHIFT 4 +#define NV20TCL_DEPTH_UNK17D8_CLAMP_MASK 0x000000f0 +#define NV20TCL_MULTISAMPLE_CONTROL 0x00001d7c +#define NV20TCL_CLEAR_DEPTH_VALUE 0x00001d8c +#define NV20TCL_CLEAR_VALUE 0x00001d90 +#define NV20TCL_CLEAR_BUFFERS 0x00001d94 +#define NV20TCL_CLEAR_BUFFERS_COLOR_A (1 << 7) +#define NV20TCL_CLEAR_BUFFERS_COLOR_B (1 << 6) +#define NV20TCL_CLEAR_BUFFERS_COLOR_G (1 << 5) +#define NV20TCL_CLEAR_BUFFERS_COLOR_R (1 << 4) +#define NV20TCL_CLEAR_BUFFERS_STENCIL (1 << 1) +#define NV20TCL_CLEAR_BUFFERS_DEPTH (1 << 0) +#define NV20TCL_RC_COLOR0 0x00001e20 +#define NV20TCL_RC_COLOR0_B_SHIFT 0 +#define NV20TCL_RC_COLOR0_B_MASK 0x000000ff +#define NV20TCL_RC_COLOR0_G_SHIFT 8 +#define NV20TCL_RC_COLOR0_G_MASK 0x0000ff00 +#define NV20TCL_RC_COLOR0_R_SHIFT 16 +#define NV20TCL_RC_COLOR0_R_MASK 0x00ff0000 +#define NV20TCL_RC_COLOR0_A_SHIFT 24 +#define NV20TCL_RC_COLOR0_A_MASK 0xff000000 +#define NV20TCL_RC_COLOR1 0x00001e24 +#define NV20TCL_RC_COLOR1_B_SHIFT 0 +#define NV20TCL_RC_COLOR1_B_MASK 0x000000ff +#define NV20TCL_RC_COLOR1_G_SHIFT 8 +#define NV20TCL_RC_COLOR1_G_MASK 0x0000ff00 +#define NV20TCL_RC_COLOR1_R_SHIFT 16 +#define NV20TCL_RC_COLOR1_R_MASK 0x00ff0000 +#define NV20TCL_RC_COLOR1_A_SHIFT 24 +#define NV20TCL_RC_COLOR1_A_MASK 0xff000000 +#define NV20TCL_BACK_MATERIAL_SHININESS(x) (0x00001e28+((x)*4)) +#define NV20TCL_BACK_MATERIAL_SHININESS__SIZE 0x00000006 +#define NV20TCL_RC_OUT_RGB(x) (0x00001e40+((x)*4)) +#define NV20TCL_RC_OUT_RGB__SIZE 0x00000008 +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_SHIFT 0 +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_MASK 0x0000000f +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_ZERO 0x00000000 +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_FOG 0x00000003 +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_TEXTURE0_ARB 0x00000008 +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_TEXTURE1_ARB 0x00000009 +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_SPARE0_NV 0x0000000c +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_SPARE1_NV 0x0000000d +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV20TCL_RC_OUT_RGB_CD_OUTPUT_E_TIMES_F_NV 0x0000000f +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_SHIFT 4 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_MASK 0x000000f0 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_ZERO 0x00000000 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_CONSTANT_COLOR0_NV 0x00000010 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_CONSTANT_COLOR1_NV 0x00000020 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_FOG 0x00000030 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_PRIMARY_COLOR_NV 0x00000040 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_SECONDARY_COLOR_NV 0x00000050 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_TEXTURE0_ARB 0x00000080 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_TEXTURE1_ARB 0x00000090 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_SPARE0_NV 0x000000c0 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_SPARE1_NV 0x000000d0 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000000e0 +#define NV20TCL_RC_OUT_RGB_AB_OUTPUT_E_TIMES_F_NV 0x000000f0 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_SHIFT 8 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_MASK 0x00000f00 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_ZERO 0x00000000 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_FOG 0x00000300 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_TEXTURE0_ARB 0x00000800 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_TEXTURE1_ARB 0x00000900 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_SPARE0_NV 0x00000c00 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_SPARE1_NV 0x00000d00 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV20TCL_RC_OUT_RGB_SUM_OUTPUT_E_TIMES_F_NV 0x00000f00 +#define NV20TCL_RC_OUT_RGB_CD_DOT_PRODUCT (1 << 12) +#define NV20TCL_RC_OUT_RGB_AB_DOT_PRODUCT (1 << 13) +#define NV20TCL_RC_OUT_RGB_MUX_SUM (1 << 14) +#define NV20TCL_RC_OUT_RGB_BIAS (1 << 15) +#define NV20TCL_RC_OUT_RGB_BIAS_NONE 0x00000000 +#define NV20TCL_RC_OUT_RGB_BIAS_BIAS_BY_NEGATIVE_ONE_HALF_NV 0x00008000 +#define NV20TCL_RC_OUT_RGB_SCALE_SHIFT 17 +#define NV20TCL_RC_OUT_RGB_SCALE_MASK 0x00000000 +#define NV20TCL_RC_OUT_RGB_SCALE_NONE 0x00000000 +#define NV20TCL_RC_OUT_RGB_SCALE_SCALE_BY_TWO_NV 0x00020000 +#define NV20TCL_RC_OUT_RGB_SCALE_SCALE_BY_FOUR_NV 0x00040000 +#define NV20TCL_RC_OUT_RGB_SCALE_SCALE_BY_ONE_HALF_NV 0x00060000 +#define NV20TCL_RC_ENABLE 0x00001e60 +#define NV20TCL_RC_ENABLE_NUM_COMBINERS_SHIFT 0 +#define NV20TCL_RC_ENABLE_NUM_COMBINERS_MASK 0x0000000f +#define NV20TCL_TX_RCOMP 0x00001e6c +#define NV20TCL_TX_RCOMP_NEVER 0x00000000 +#define NV20TCL_TX_RCOMP_GREATER 0x00000001 +#define NV20TCL_TX_RCOMP_EQUAL 0x00000002 +#define NV20TCL_TX_RCOMP_GEQUAL 0x00000003 +#define NV20TCL_TX_RCOMP_LESS 0x00000004 +#define NV20TCL_TX_RCOMP_NOTEQUAL 0x00000005 +#define NV20TCL_TX_RCOMP_LEQUAL 0x00000006 +#define NV20TCL_TX_RCOMP_ALWAYS 0x00000007 +#define NV20TCL_TX_SHADER_OP 0x00001e70 +#define NV20TCL_TX_SHADER_OP_TX0_SHIFT 0 +#define NV20TCL_TX_SHADER_OP_TX0_MASK 0x0000001f +#define NV20TCL_TX_SHADER_OP_TX0_NONE 0x00000000 +#define NV20TCL_TX_SHADER_OP_TX0_TEXTURE_2D 0x00000001 +#define NV20TCL_TX_SHADER_OP_TX0_PASS_THROUGH 0x00000004 +#define NV20TCL_TX_SHADER_OP_TX0_CULL_FRAGMENT 0x00000005 +#define NV20TCL_TX_SHADER_OP_TX0_OFFSET_TEXTURE_2D 0x00000006 +#define NV20TCL_TX_SHADER_OP_TX0_DOT_PRODUCT_TEXTURE_2D 0x00000009 +#define NV20TCL_TX_SHADER_OP_TX0_DOT_PRODUCT_DEPTH_REPLACE 0x0000000a +#define NV20TCL_TX_SHADER_OP_TX0_DEPENDANT_AR_TEXTURE_2D 0x0000000f +#define NV20TCL_TX_SHADER_OP_TX0_DEPENDANT_GB_TEXTURE_2D 0x00000010 +#define NV20TCL_TX_SHADER_OP_TX0_DOT_PRODUCT 0x00000011 +#define NV20TCL_TX_SHADER_OP_TX1_SHIFT 5 +#define NV20TCL_TX_SHADER_OP_TX1_MASK 0x000003e0 +#define NV20TCL_TX_SHADER_OP_TX1_NONE 0x00000000 +#define NV20TCL_TX_SHADER_OP_TX1_TEXTURE_2D 0x00000020 +#define NV20TCL_TX_SHADER_OP_TX1_PASS_THROUGH 0x00000080 +#define NV20TCL_TX_SHADER_OP_TX1_CULL_FRAGMENT 0x000000a0 +#define NV20TCL_TX_SHADER_OP_TX1_OFFSET_TEXTURE_2D 0x000000c0 +#define NV20TCL_TX_SHADER_OP_TX1_DOT_PRODUCT_TEXTURE_2D 0x00000120 +#define NV20TCL_TX_SHADER_OP_TX1_DOT_PRODUCT_DEPTH_REPLACE 0x00000140 +#define NV20TCL_TX_SHADER_OP_TX1_DEPENDANT_AR_TEXTURE_2D 0x000001e0 +#define NV20TCL_TX_SHADER_OP_TX1_DEPENDANT_GB_TEXTURE_2D 0x00000200 +#define NV20TCL_TX_SHADER_OP_TX1_DOT_PRODUCT 0x00000220 +#define NV20TCL_TX_SHADER_OP_TX2_SHIFT 10 +#define NV20TCL_TX_SHADER_OP_TX2_MASK 0x00007c00 +#define NV20TCL_TX_SHADER_OP_TX2_NONE 0x00000000 +#define NV20TCL_TX_SHADER_OP_TX2_TEXTURE_2D 0x00000400 +#define NV20TCL_TX_SHADER_OP_TX2_PASS_THROUGH 0x00001000 +#define NV20TCL_TX_SHADER_OP_TX2_CULL_FRAGMENT 0x00001400 +#define NV20TCL_TX_SHADER_OP_TX2_OFFSET_TEXTURE_2D 0x00001800 +#define NV20TCL_TX_SHADER_OP_TX2_DOT_PRODUCT_TEXTURE_2D 0x00002400 +#define NV20TCL_TX_SHADER_OP_TX2_DOT_PRODUCT_DEPTH_REPLACE 0x00002800 +#define NV20TCL_TX_SHADER_OP_TX2_DEPENDANT_AR_TEXTURE_2D 0x00003c00 +#define NV20TCL_TX_SHADER_OP_TX2_DEPENDANT_GB_TEXTURE_2D 0x00004000 +#define NV20TCL_TX_SHADER_OP_TX2_DOT_PRODUCT 0x00004400 +#define NV20TCL_TX_SHADER_OP_TX3_SHIFT 15 +#define NV20TCL_TX_SHADER_OP_TX3_MASK 0x000f8000 +#define NV20TCL_TX_SHADER_OP_TX3_NONE 0x00000000 +#define NV20TCL_TX_SHADER_OP_TX3_TEXTURE_2D 0x00008000 +#define NV20TCL_TX_SHADER_OP_TX3_PASS_THROUGH 0x00020000 +#define NV20TCL_TX_SHADER_OP_TX3_CULL_FRAGMENT 0x00028000 +#define NV20TCL_TX_SHADER_OP_TX3_OFFSET_TEXTURE_2D 0x00030000 +#define NV20TCL_TX_SHADER_OP_TX3_DOT_PRODUCT_TEXTURE_2D 0x00048000 +#define NV20TCL_TX_SHADER_OP_TX3_DOT_PRODUCT_DEPTH_REPLACE 0x00050000 +#define NV20TCL_TX_SHADER_OP_TX3_DEPENDANT_AR_TEXTURE_2D 0x00078000 +#define NV20TCL_TX_SHADER_OP_TX3_DEPENDANT_GB_TEXTURE_2D 0x00080000 +#define NV20TCL_TX_SHADER_OP_TX3_DOT_PRODUCT 0x00088000 +#define NV20TCL_TX_SHADER_DOTMAPPING 0x00001e74 +#define NV20TCL_TX_SHADER_DOTMAPPING_TX0_SHIFT 0 +#define NV20TCL_TX_SHADER_DOTMAPPING_TX0_MASK 0x0000000f +#define NV20TCL_TX_SHADER_DOTMAPPING_TX1_SHIFT 4 +#define NV20TCL_TX_SHADER_DOTMAPPING_TX1_MASK 0x000000f0 +#define NV20TCL_TX_SHADER_DOTMAPPING_TX2_SHIFT 8 +#define NV20TCL_TX_SHADER_DOTMAPPING_TX2_MASK 0x00000f00 +#define NV20TCL_TX_SHADER_DOTMAPPING_TX3_SHIFT 12 +#define NV20TCL_TX_SHADER_DOTMAPPING_TX3_MASK 0x0000f000 +#define NV20TCL_TX_SHADER_PREVIOUS 0x00001e78 +#define NV20TCL_TX_SHADER_PREVIOUS_TX0_SHIFT 8 +#define NV20TCL_TX_SHADER_PREVIOUS_TX0_MASK 0x00000f00 +#define NV20TCL_TX_SHADER_PREVIOUS_TX1_SHIFT 12 +#define NV20TCL_TX_SHADER_PREVIOUS_TX1_MASK 0x0000f000 +#define NV20TCL_TX_SHADER_PREVIOUS_TX2_SHIFT 16 +#define NV20TCL_TX_SHADER_PREVIOUS_TX2_MASK 0x00030000 +#define NV20TCL_TX_SHADER_PREVIOUS_TX3_SHIFT 20 +#define NV20TCL_TX_SHADER_PREVIOUS_TX3_MASK 0x00300000 +#define NV20TCL_ENGINE 0x00001e94 +#define NV20TCL_ENGINE_VP (1 << 1) +#define NV20TCL_ENGINE_FIXED (1 << 2) +#define NV20TCL_VP_UPLOAD_FROM_ID 0x00001e9c +#define NV20TCL_VP_START_FROM_ID 0x00001ea0 +#define NV20TCL_VP_UPLOAD_CONST_ID 0x00001ea4 +#define NV20TCL_VIEWPORT_TRANSLATE_X 0x00001f00 +#define NV20TCL_VIEWPORT_TRANSLATE_Y 0x00001f04 +#define NV20TCL_VIEWPORT_TRANSLATE_Z 0x00001f08 +#define NV20TCL_VIEWPORT_TRANSLATE_W 0x00001f0c + + +#define NV17TCL 0x00000099 + +#define NV17TCL_DMA_IN_MEMORY4 0x000001ac +#define NV17TCL_DMA_IN_MEMORY5 0x000001b0 +#define NV17TCL_COLOR_MASK_ENABLE 0x000002bc +#define NV17TCL_LMA_DEPTH_BUFFER_PITCH 0x00000d5c +#define NV17TCL_LMA_DEPTH_BUFFER_OFFSET 0x00000d60 +#define NV17TCL_LMA_DEPTH_FILL_VALUE 0x00000d68 +#define NV17TCL_LMA_DEPTH_BUFFER_CLEAR 0x00000d6c +#define NV17TCL_LMA_DEPTH_ENABLE 0x00001658 + + +#define NV20_SWIZZLED_SURFACE 0x0000009e + + + +#define NV12_IMAGE_BLIT 0x0000009f + + + +#define NV30_CONTEXT_SURFACES_2D 0x00000362 + + + +#define NV30_STRETCHED_IMAGE_FROM_CPU 0x00000366 + + + +#define NV30_TEXTURE_FROM_CPU 0x0000037b + + + +#define NV30_SCALED_IMAGE_FROM_MEMORY 0x00000389 + + + +#define NV30_IMAGE_FROM_CPU 0x0000038a + + + +#define NV30TCL 0x00000397 + + + +#define NV30_SWIZZLED_SURFACE 0x0000039e + + + +#define NV35TCL 0x00000497 + + + +#define NV25TCL 0x00000597 + +#define NV25TCL_DMA_IN_MEMORY4 0x0000019c +#define NV25TCL_DMA_IN_MEMORY5 0x000001a0 +#define NV25TCL_DMA_IN_MEMORY8 0x000001ac +#define NV25TCL_DMA_IN_MEMORY9 0x000001b0 + + +#define NV34TCL 0x00000697 + +#define NV34TCL_NOP 0x00000100 +#define NV34TCL_NOTIFY 0x00000104 +#define NV34TCL_DMA_NOTIFY 0x00000180 +#define NV34TCL_DMA_TEXTURE0 0x00000184 +#define NV34TCL_DMA_TEXTURE1 0x00000188 +#define NV34TCL_DMA_COLOR1 0x0000018c +#define NV34TCL_DMA_COLOR0 0x00000194 +#define NV34TCL_DMA_ZETA 0x00000198 +#define NV34TCL_DMA_VTXBUF0 0x0000019c +#define NV34TCL_DMA_VTXBUF1 0x000001a0 +#define NV34TCL_DMA_FENCE 0x000001a4 +#define NV34TCL_DMA_QUERY 0x000001a8 +#define NV34TCL_DMA_IN_MEMORY7 0x000001ac +#define NV34TCL_DMA_IN_MEMORY8 0x000001b0 +#define NV34TCL_RT_HORIZ 0x00000200 +#define NV34TCL_RT_HORIZ_X_SHIFT 0 +#define NV34TCL_RT_HORIZ_X_MASK 0x0000ffff +#define NV34TCL_RT_HORIZ_W_SHIFT 16 +#define NV34TCL_RT_HORIZ_W_MASK 0xffff0000 +#define NV34TCL_RT_VERT 0x00000204 +#define NV34TCL_RT_VERT_Y_SHIFT 0 +#define NV34TCL_RT_VERT_Y_MASK 0x0000ffff +#define NV34TCL_RT_VERT_H_SHIFT 16 +#define NV34TCL_RT_VERT_H_MASK 0xffff0000 +#define NV34TCL_RT_FORMAT 0x00000208 +#define NV34TCL_RT_FORMAT_TYPE_SHIFT 8 +#define NV34TCL_RT_FORMAT_TYPE_MASK 0x00000f00 +#define NV34TCL_RT_FORMAT_TYPE_LINEAR 0x00000100 +#define NV34TCL_RT_FORMAT_TYPE_SWIZZLED 0x00000200 +#define NV34TCL_RT_FORMAT_ZETA_SHIFT 5 +#define NV34TCL_RT_FORMAT_ZETA_MASK 0x000000e0 +#define NV34TCL_RT_FORMAT_ZETA_Z16 0x00000020 +#define NV34TCL_RT_FORMAT_ZETA_Z24S8 0x00000040 +#define NV34TCL_RT_FORMAT_COLOR_SHIFT 0 +#define NV34TCL_RT_FORMAT_COLOR_MASK 0x0000001f +#define NV34TCL_RT_FORMAT_COLOR_R5G6B5 0x00000003 +#define NV34TCL_RT_FORMAT_COLOR_X8R8G8B8 0x00000005 +#define NV34TCL_RT_FORMAT_COLOR_A8R8G8B8 0x00000008 +#define NV34TCL_RT_FORMAT_COLOR_B8 0x00000009 +#define NV34TCL_RT_FORMAT_COLOR_UNKNOWN 0x0000000d +#define NV34TCL_RT_FORMAT_COLOR_X8B8G8R8 0x0000000f +#define NV34TCL_RT_FORMAT_COLOR_A8B8G8R8 0x00000010 +#define NV34TCL_COLOR0_PITCH 0x0000020c +#define NV34TCL_COLOR0_PITCH_COLOR0_SHIFT 0 +#define NV34TCL_COLOR0_PITCH_COLOR0_MASK 0x0000ffff +#define NV34TCL_COLOR0_PITCH_ZETA_SHIFT 16 +#define NV34TCL_COLOR0_PITCH_ZETA_MASK 0xffff0000 +#define NV34TCL_COLOR0_OFFSET 0x00000210 +#define NV34TCL_ZETA_OFFSET 0x00000214 +#define NV34TCL_COLOR1_OFFSET 0x00000218 +#define NV34TCL_COLOR1_PITCH 0x0000021c +#define NV34TCL_RT_ENABLE 0x00000220 +#define NV34TCL_RT_ENABLE_MRT (1 << 4) +#define NV34TCL_RT_ENABLE_COLOR1 (1 << 1) +#define NV34TCL_RT_ENABLE_COLOR0 (1 << 0) +#define NV34TCL_LMA_DEPTH_PITCH 0x0000022c +#define NV34TCL_LMA_DEPTH_OFFSET 0x00000230 +#define NV34TCL_TX_UNITS_ENABLE 0x0000023c +#define NV34TCL_TX_UNITS_ENABLE_TX0 (1 << 0) +#define NV34TCL_TX_UNITS_ENABLE_TX1 (1 << 1) +#define NV34TCL_TX_UNITS_ENABLE_TX2 (1 << 2) +#define NV34TCL_TX_UNITS_ENABLE_TX3 (1 << 3) +#define NV34TCL_TX_UNITS_ENABLE_TX4 (1 << 4) +#define NV34TCL_TX_UNITS_ENABLE_TX5 (1 << 5) +#define NV34TCL_TX_UNITS_ENABLE_TX6 (1 << 6) +#define NV34TCL_TX_UNITS_ENABLE_TX7 (1 << 7) +#define NV34TCL_TX_MATRIX_ENABLE(x) (0x00000240+((x)*4)) +#define NV34TCL_TX_MATRIX_ENABLE__SIZE 0x00000008 +#define NV34TCL_VIEWPORT_TX_ORIGIN 0x000002b8 +#define NV34TCL_VIEWPORT_TX_ORIGIN_X_SHIFT 0 +#define NV34TCL_VIEWPORT_TX_ORIGIN_X_MASK 0x0000ffff +#define NV34TCL_VIEWPORT_TX_ORIGIN_Y_SHIFT 16 +#define NV34TCL_VIEWPORT_TX_ORIGIN_Y_MASK 0xffff0000 +#define NV34TCL_VIEWPORT_CLIP_MODE 0x000002bc +#define NV34TCL_VIEWPORT_CLIP_HORIZ(x) (0x000002c0+((x)*8)) +#define NV34TCL_VIEWPORT_CLIP_HORIZ__SIZE 0x00000008 +#define NV34TCL_VIEWPORT_CLIP_HORIZ_L_SHIFT 0 +#define NV34TCL_VIEWPORT_CLIP_HORIZ_L_MASK 0x0000ffff +#define NV34TCL_VIEWPORT_CLIP_HORIZ_R_SHIFT 16 +#define NV34TCL_VIEWPORT_CLIP_HORIZ_R_MASK 0xffff0000 +#define NV34TCL_VIEWPORT_CLIP_VERT(x) (0x000002c4+((x)*8)) +#define NV34TCL_VIEWPORT_CLIP_VERT__SIZE 0x00000008 +#define NV34TCL_VIEWPORT_CLIP_VERT_T_SHIFT 0 +#define NV34TCL_VIEWPORT_CLIP_VERT_T_MASK 0x0000ffff +#define NV34TCL_VIEWPORT_CLIP_VERT_D_SHIFT 16 +#define NV34TCL_VIEWPORT_CLIP_VERT_D_MASK 0xffff0000 +#define NV34TCL_DITHER_ENABLE 0x00000300 +#define NV34TCL_ALPHA_FUNC_ENABLE 0x00000304 +#define NV34TCL_ALPHA_FUNC_FUNC 0x00000308 +#define NV34TCL_ALPHA_FUNC_FUNC_NEVER 0x00000200 +#define NV34TCL_ALPHA_FUNC_FUNC_LESS 0x00000201 +#define NV34TCL_ALPHA_FUNC_FUNC_EQUAL 0x00000202 +#define NV34TCL_ALPHA_FUNC_FUNC_LEQUAL 0x00000203 +#define NV34TCL_ALPHA_FUNC_FUNC_GREATER 0x00000204 +#define NV34TCL_ALPHA_FUNC_FUNC_GREATER 0x00000204 +#define NV34TCL_ALPHA_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV34TCL_ALPHA_FUNC_FUNC_GEQUAL 0x00000206 +#define NV34TCL_ALPHA_FUNC_FUNC_ALWAYS 0x00000207 +#define NV34TCL_ALPHA_FUNC_REF 0x0000030c +#define NV34TCL_BLEND_FUNC_ENABLE 0x00000310 +#define NV34TCL_BLEND_FUNC_SRC 0x00000314 +#define NV34TCL_BLEND_FUNC_SRC_RGB_SHIFT 0 +#define NV34TCL_BLEND_FUNC_SRC_RGB_MASK 0x0000ffff +#define NV34TCL_BLEND_FUNC_SRC_RGB_ZERO 0x00000000 +#define NV34TCL_BLEND_FUNC_SRC_RGB_ONE 0x00000001 +#define NV34TCL_BLEND_FUNC_SRC_RGB_SRC_COLOR 0x00000300 +#define NV34TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV34TCL_BLEND_FUNC_SRC_RGB_SRC_ALPHA 0x00000302 +#define NV34TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV34TCL_BLEND_FUNC_SRC_RGB_DST_ALPHA 0x00000304 +#define NV34TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV34TCL_BLEND_FUNC_SRC_RGB_DST_COLOR 0x00000306 +#define NV34TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_DST_COLOR 0x00000307 +#define NV34TCL_BLEND_FUNC_SRC_RGB_SRC_ALPHA_SATURATE 0x00000308 +#define NV34TCL_BLEND_FUNC_SRC_RGB_CONSTANT_COLOR 0x00008001 +#define NV34TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV34TCL_BLEND_FUNC_SRC_RGB_CONSTANT_ALPHA 0x00008003 +#define NV34TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_SHIFT 16 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_MASK 0xffff0000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_ZERO 0x00000000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_ONE 0x00010000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_SRC_COLOR 0x03000000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_SRC_COLOR 0x03010000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_SRC_ALPHA 0x03020000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_SRC_ALPHA 0x03030000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_DST_ALPHA 0x03040000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_DST_ALPHA 0x03050000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_DST_COLOR 0x03060000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_DST_COLOR 0x03070000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_SRC_ALPHA_SATURATE 0x03080000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_CONSTANT_COLOR 0x80010000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_CONSTANT_COLOR 0x80020000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_CONSTANT_ALPHA 0x80030000 +#define NV34TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_CONSTANT_ALPHA 0x80040000 +#define NV34TCL_BLEND_FUNC_DST 0x00000318 +#define NV34TCL_BLEND_FUNC_DST_RGB_SHIFT 0 +#define NV34TCL_BLEND_FUNC_DST_RGB_MASK 0x0000ffff +#define NV34TCL_BLEND_FUNC_DST_RGB_ZERO 0x00000000 +#define NV34TCL_BLEND_FUNC_DST_RGB_ONE 0x00000001 +#define NV34TCL_BLEND_FUNC_DST_RGB_SRC_COLOR 0x00000300 +#define NV34TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV34TCL_BLEND_FUNC_DST_RGB_SRC_ALPHA 0x00000302 +#define NV34TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV34TCL_BLEND_FUNC_DST_RGB_DST_ALPHA 0x00000304 +#define NV34TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV34TCL_BLEND_FUNC_DST_RGB_DST_COLOR 0x00000306 +#define NV34TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_DST_COLOR 0x00000307 +#define NV34TCL_BLEND_FUNC_DST_RGB_SRC_ALPHA_SATURATE 0x00000308 +#define NV34TCL_BLEND_FUNC_DST_RGB_CONSTANT_COLOR 0x00008001 +#define NV34TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV34TCL_BLEND_FUNC_DST_RGB_CONSTANT_ALPHA 0x00008003 +#define NV34TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_SHIFT 16 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_MASK 0xffff0000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_ZERO 0x00000000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_ONE 0x00010000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_SRC_COLOR 0x03000000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_SRC_COLOR 0x03010000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_SRC_ALPHA 0x03020000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_SRC_ALPHA 0x03030000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_DST_ALPHA 0x03040000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_DST_ALPHA 0x03050000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_DST_COLOR 0x03060000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_DST_COLOR 0x03070000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_SRC_ALPHA_SATURATE 0x03080000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_CONSTANT_COLOR 0x80010000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_CONSTANT_COLOR 0x80020000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_CONSTANT_ALPHA 0x80030000 +#define NV34TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_CONSTANT_ALPHA 0x80040000 +#define NV34TCL_BLEND_COLOR 0x0000031c +#define NV34TCL_BLEND_COLOR_B_SHIFT 0 +#define NV34TCL_BLEND_COLOR_B_MASK 0x000000ff +#define NV34TCL_BLEND_COLOR_G_SHIFT 8 +#define NV34TCL_BLEND_COLOR_G_MASK 0x0000ff00 +#define NV34TCL_BLEND_COLOR_R_SHIFT 16 +#define NV34TCL_BLEND_COLOR_R_MASK 0x00ff0000 +#define NV34TCL_BLEND_COLOR_A_SHIFT 24 +#define NV34TCL_BLEND_COLOR_A_MASK 0xff000000 +#define NV34TCL_BLEND_EQUATION 0x00000320 +#define NV34TCL_BLEND_EQUATION_FUNC_ADD 0x00008006 +#define NV34TCL_BLEND_EQUATION_MIN 0x00008007 +#define NV34TCL_BLEND_EQUATION_MAX 0x00008008 +#define NV34TCL_BLEND_EQUATION_FUNC_SUBTRACT 0x0000800a +#define NV34TCL_BLEND_EQUATION_FUNC_REVERSE_SUBTRACT 0x0000800b +#define NV34TCL_COLOR_MASK 0x00000324 +#define NV34TCL_COLOR_MASK_B_SHIFT 0 +#define NV34TCL_COLOR_MASK_B_MASK 0x000000ff +#define NV34TCL_COLOR_MASK_G_SHIFT 8 +#define NV34TCL_COLOR_MASK_G_MASK 0x0000ff00 +#define NV34TCL_COLOR_MASK_R_SHIFT 16 +#define NV34TCL_COLOR_MASK_R_MASK 0x00ff0000 +#define NV34TCL_COLOR_MASK_A_SHIFT 24 +#define NV34TCL_COLOR_MASK_A_MASK 0xff000000 +#define NV34TCL_STENCIL_BACK_ENABLE 0x00000328 +#define NV34TCL_STENCIL_BACK_MASK 0x0000032c +#define NV34TCL_STENCIL_BACK_FUNC_FUNC 0x00000330 +#define NV34TCL_STENCIL_BACK_FUNC_FUNC_NEVER 0x00000200 +#define NV34TCL_STENCIL_BACK_FUNC_FUNC_LESS 0x00000201 +#define NV34TCL_STENCIL_BACK_FUNC_FUNC_EQUAL 0x00000202 +#define NV34TCL_STENCIL_BACK_FUNC_FUNC_LEQUAL 0x00000203 +#define NV34TCL_STENCIL_BACK_FUNC_FUNC_GREATER 0x00000204 +#define NV34TCL_STENCIL_BACK_FUNC_FUNC_GREATER 0x00000204 +#define NV34TCL_STENCIL_BACK_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV34TCL_STENCIL_BACK_FUNC_FUNC_GEQUAL 0x00000206 +#define NV34TCL_STENCIL_BACK_FUNC_FUNC_ALWAYS 0x00000207 +#define NV34TCL_STENCIL_BACK_FUNC_REF 0x00000334 +#define NV34TCL_STENCIL_BACK_FUNC_MASK 0x00000338 +#define NV34TCL_STENCIL_BACK_OP_FAIL 0x0000033c +#define NV34TCL_STENCIL_BACK_OP_FAIL_ZERO 0x00000000 +#define NV34TCL_STENCIL_BACK_OP_FAIL_INVERT 0x0000150a +#define NV34TCL_STENCIL_BACK_OP_FAIL_KEEP 0x00001e00 +#define NV34TCL_STENCIL_BACK_OP_FAIL_REPLACE 0x00001e01 +#define NV34TCL_STENCIL_BACK_OP_FAIL_INCR 0x00001e02 +#define NV34TCL_STENCIL_BACK_OP_FAIL_DECR 0x00001e03 +#define NV34TCL_STENCIL_BACK_OP_FAIL_INCR_WRAP 0x00008507 +#define NV34TCL_STENCIL_BACK_OP_FAIL_DECR_WRAP 0x00008508 +#define NV34TCL_STENCIL_BACK_OP_ZFAIL 0x00000340 +#define NV34TCL_STENCIL_BACK_OP_ZFAIL_ZERO 0x00000000 +#define NV34TCL_STENCIL_BACK_OP_ZFAIL_INVERT 0x0000150a +#define NV34TCL_STENCIL_BACK_OP_ZFAIL_KEEP 0x00001e00 +#define NV34TCL_STENCIL_BACK_OP_ZFAIL_REPLACE 0x00001e01 +#define NV34TCL_STENCIL_BACK_OP_ZFAIL_INCR 0x00001e02 +#define NV34TCL_STENCIL_BACK_OP_ZFAIL_DECR 0x00001e03 +#define NV34TCL_STENCIL_BACK_OP_ZFAIL_INCR_WRAP 0x00008507 +#define NV34TCL_STENCIL_BACK_OP_ZFAIL_DECR_WRAP 0x00008508 +#define NV34TCL_STENCIL_BACK_OP_ZPASS 0x00000344 +#define NV34TCL_STENCIL_BACK_OP_ZPASS_ZERO 0x00000000 +#define NV34TCL_STENCIL_BACK_OP_ZPASS_INVERT 0x0000150a +#define NV34TCL_STENCIL_BACK_OP_ZPASS_KEEP 0x00001e00 +#define NV34TCL_STENCIL_BACK_OP_ZPASS_REPLACE 0x00001e01 +#define NV34TCL_STENCIL_BACK_OP_ZPASS_INCR 0x00001e02 +#define NV34TCL_STENCIL_BACK_OP_ZPASS_DECR 0x00001e03 +#define NV34TCL_STENCIL_BACK_OP_ZPASS_INCR_WRAP 0x00008507 +#define NV34TCL_STENCIL_BACK_OP_ZPASS_DECR_WRAP 0x00008508 +#define NV34TCL_STENCIL_FRONT_ENABLE 0x00000348 +#define NV34TCL_STENCIL_FRONT_MASK 0x0000034c +#define NV34TCL_STENCIL_FRONT_FUNC_FUNC 0x00000350 +#define NV34TCL_STENCIL_FRONT_FUNC_FUNC_NEVER 0x00000200 +#define NV34TCL_STENCIL_FRONT_FUNC_FUNC_LESS 0x00000201 +#define NV34TCL_STENCIL_FRONT_FUNC_FUNC_EQUAL 0x00000202 +#define NV34TCL_STENCIL_FRONT_FUNC_FUNC_LEQUAL 0x00000203 +#define NV34TCL_STENCIL_FRONT_FUNC_FUNC_GREATER 0x00000204 +#define NV34TCL_STENCIL_FRONT_FUNC_FUNC_GREATER 0x00000204 +#define NV34TCL_STENCIL_FRONT_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV34TCL_STENCIL_FRONT_FUNC_FUNC_GEQUAL 0x00000206 +#define NV34TCL_STENCIL_FRONT_FUNC_FUNC_ALWAYS 0x00000207 +#define NV34TCL_STENCIL_FRONT_FUNC_REF 0x00000354 +#define NV34TCL_STENCIL_FRONT_FUNC_MASK 0x00000358 +#define NV34TCL_STENCIL_FRONT_OP_FAIL 0x0000035c +#define NV34TCL_STENCIL_FRONT_OP_FAIL_ZERO 0x00000000 +#define NV34TCL_STENCIL_FRONT_OP_FAIL_INVERT 0x0000150a +#define NV34TCL_STENCIL_FRONT_OP_FAIL_KEEP 0x00001e00 +#define NV34TCL_STENCIL_FRONT_OP_FAIL_REPLACE 0x00001e01 +#define NV34TCL_STENCIL_FRONT_OP_FAIL_INCR 0x00001e02 +#define NV34TCL_STENCIL_FRONT_OP_FAIL_DECR 0x00001e03 +#define NV34TCL_STENCIL_FRONT_OP_FAIL_INCR_WRAP 0x00008507 +#define NV34TCL_STENCIL_FRONT_OP_FAIL_DECR_WRAP 0x00008508 +#define NV34TCL_STENCIL_FRONT_OP_ZFAIL 0x00000360 +#define NV34TCL_STENCIL_FRONT_OP_ZFAIL_ZERO 0x00000000 +#define NV34TCL_STENCIL_FRONT_OP_ZFAIL_INVERT 0x0000150a +#define NV34TCL_STENCIL_FRONT_OP_ZFAIL_KEEP 0x00001e00 +#define NV34TCL_STENCIL_FRONT_OP_ZFAIL_REPLACE 0x00001e01 +#define NV34TCL_STENCIL_FRONT_OP_ZFAIL_INCR 0x00001e02 +#define NV34TCL_STENCIL_FRONT_OP_ZFAIL_DECR 0x00001e03 +#define NV34TCL_STENCIL_FRONT_OP_ZFAIL_INCR_WRAP 0x00008507 +#define NV34TCL_STENCIL_FRONT_OP_ZFAIL_DECR_WRAP 0x00008508 +#define NV34TCL_STENCIL_FRONT_OP_ZPASS 0x00000364 +#define NV34TCL_STENCIL_FRONT_OP_ZPASS_ZERO 0x00000000 +#define NV34TCL_STENCIL_FRONT_OP_ZPASS_INVERT 0x0000150a +#define NV34TCL_STENCIL_FRONT_OP_ZPASS_KEEP 0x00001e00 +#define NV34TCL_STENCIL_FRONT_OP_ZPASS_REPLACE 0x00001e01 +#define NV34TCL_STENCIL_FRONT_OP_ZPASS_INCR 0x00001e02 +#define NV34TCL_STENCIL_FRONT_OP_ZPASS_DECR 0x00001e03 +#define NV34TCL_STENCIL_FRONT_OP_ZPASS_INCR_WRAP 0x00008507 +#define NV34TCL_STENCIL_FRONT_OP_ZPASS_DECR_WRAP 0x00008508 +#define NV34TCL_SHADE_MODEL 0x00000368 +#define NV34TCL_SHADE_MODEL_FLAT 0x00001d00 +#define NV34TCL_SHADE_MODEL_SMOOTH 0x00001d01 +#define NV34TCL_FOG_ENABLE 0x0000036c +#define NV34TCL_FOG_COLOR 0x00000370 +#define NV34TCL_FOG_COLOR_R_SHIFT 0 +#define NV34TCL_FOG_COLOR_R_MASK 0x000000ff +#define NV34TCL_FOG_COLOR_G_SHIFT 8 +#define NV34TCL_FOG_COLOR_G_MASK 0x0000ff00 +#define NV34TCL_FOG_COLOR_B_SHIFT 16 +#define NV34TCL_FOG_COLOR_B_MASK 0x00ff0000 +#define NV34TCL_FOG_COLOR_A_SHIFT 24 +#define NV34TCL_FOG_COLOR_A_MASK 0xff000000 +#define NV34TCL_COLOR_LOGIC_OP_ENABLE 0x00000374 +#define NV34TCL_COLOR_LOGIC_OP_OP 0x00000378 +#define NV34TCL_COLOR_LOGIC_OP_OP_CLEAR 0x00001500 +#define NV34TCL_COLOR_LOGIC_OP_OP_AND 0x00001501 +#define NV34TCL_COLOR_LOGIC_OP_OP_AND_REVERSE 0x00001502 +#define NV34TCL_COLOR_LOGIC_OP_OP_COPY 0x00001503 +#define NV34TCL_COLOR_LOGIC_OP_OP_AND_INVERTED 0x00001504 +#define NV34TCL_COLOR_LOGIC_OP_OP_NOOP 0x00001505 +#define NV34TCL_COLOR_LOGIC_OP_OP_XOR 0x00001506 +#define NV34TCL_COLOR_LOGIC_OP_OP_OR 0x00001507 +#define NV34TCL_COLOR_LOGIC_OP_OP_NOR 0x00001508 +#define NV34TCL_COLOR_LOGIC_OP_OP_EQUIV 0x00001509 +#define NV34TCL_COLOR_LOGIC_OP_OP_INVERT 0x0000150a +#define NV34TCL_COLOR_LOGIC_OP_OP_OR_REVERSE 0x0000150b +#define NV34TCL_COLOR_LOGIC_OP_OP_COPY_INVERTED 0x0000150c +#define NV34TCL_COLOR_LOGIC_OP_OP_OR_INVERTED 0x0000150d +#define NV34TCL_COLOR_LOGIC_OP_OP_NAND 0x0000150e +#define NV34TCL_COLOR_LOGIC_OP_OP_SET 0x0000150f +#define NV34TCL_NORMALIZE_ENABLE 0x0000037c +#define NV34TCL_COLOR_MATERIAL 0x00000390 +#define NV34TCL_COLOR_MATERIAL_FRONT_EMISSION_ENABLE (1 << 0) +#define NV34TCL_COLOR_MATERIAL_FRONT_AMBIENT_ENABLE (1 << 2) +#define NV34TCL_COLOR_MATERIAL_FRONT_DIFFUSE_ENABLE (1 << 4) +#define NV34TCL_COLOR_MATERIAL_FRONT_SPECULAR_ENABLE (1 << 6) +#define NV34TCL_COLOR_MATERIAL_BACK_EMISSION_ENABLE (1 << 8) +#define NV34TCL_COLOR_MATERIAL_BACK_AMBIENT_ENABLE (1 << 10) +#define NV34TCL_COLOR_MATERIAL_BACK_DIFFUSE_ENABLE (1 << 12) +#define NV34TCL_COLOR_MATERIAL_BACK_SPECULAR_ENABLE (1 << 14) +#define NV34TCL_DEPTH_RANGE_NEAR 0x00000394 +#define NV34TCL_DEPTH_RANGE_FAR 0x00000398 +#define NV34TCL_COLOR_MATERIAL_FRONT_R 0x000003a0 +#define NV34TCL_COLOR_MATERIAL_FRONT_G 0x000003a4 +#define NV34TCL_COLOR_MATERIAL_FRONT_B 0x000003a8 +#define NV34TCL_COLOR_MATERIAL_FRONT_A 0x000003b4 +#define NV34TCL_LINE_WIDTH 0x000003b8 +#define NV34TCL_LINE_SMOOTH_ENABLE 0x000003bc +#define NV34TCL_TX_GEN_S(x) (0x00000400+((x)*16)) +#define NV34TCL_TX_GEN_S__SIZE 0x00000008 +#define NV34TCL_TX_GEN_S_FALSE 0x00000000 +#define NV34TCL_TX_GEN_S_EYE_LINEAR 0x00002400 +#define NV34TCL_TX_GEN_S_OBJECT_LINEAR 0x00002401 +#define NV34TCL_TX_GEN_S_SPHERE_MAP 0x00002402 +#define NV34TCL_TX_GEN_S_NORMAL_MAP 0x00008511 +#define NV34TCL_TX_GEN_S_REFLECTION_MAP 0x00008512 +#define NV34TCL_TX_GEN_T(x) (0x00000404+((x)*16)) +#define NV34TCL_TX_GEN_T__SIZE 0x00000008 +#define NV34TCL_TX_GEN_T_FALSE 0x00000000 +#define NV34TCL_TX_GEN_T_EYE_LINEAR 0x00002400 +#define NV34TCL_TX_GEN_T_OBJECT_LINEAR 0x00002401 +#define NV34TCL_TX_GEN_T_SPHERE_MAP 0x00002402 +#define NV34TCL_TX_GEN_T_NORMAL_MAP 0x00008511 +#define NV34TCL_TX_GEN_T_REFLECTION_MAP 0x00008512 +#define NV34TCL_TX_GEN_R(x) (0x00000408+((x)*16)) +#define NV34TCL_TX_GEN_R__SIZE 0x00000008 +#define NV34TCL_TX_GEN_R_FALSE 0x00000000 +#define NV34TCL_TX_GEN_R_EYE_LINEAR 0x00002400 +#define NV34TCL_TX_GEN_R_OBJECT_LINEAR 0x00002401 +#define NV34TCL_TX_GEN_R_SPHERE_MAP 0x00002402 +#define NV34TCL_TX_GEN_R_NORMAL_MAP 0x00008511 +#define NV34TCL_TX_GEN_R_REFLECTION_MAP 0x00008512 +#define NV34TCL_TX_GEN_Q(x) (0x0000040c+((x)*16)) +#define NV34TCL_TX_GEN_Q__SIZE 0x00000008 +#define NV34TCL_TX_GEN_Q_FALSE 0x00000000 +#define NV34TCL_TX_GEN_Q_EYE_LINEAR 0x00002400 +#define NV34TCL_TX_GEN_Q_OBJECT_LINEAR 0x00002401 +#define NV34TCL_TX_GEN_Q_SPHERE_MAP 0x00002402 +#define NV34TCL_TX_GEN_Q_NORMAL_MAP 0x00008511 +#define NV34TCL_TX_GEN_Q_REFLECTION_MAP 0x00008512 +#define NV34TCL_MODELVIEW_MATRIX(x) (0x00000480+((x)*4)) +#define NV34TCL_MODELVIEW_MATRIX__SIZE 0x00000010 +#define NV34TCL_INVERSE_MODELVIEW_MATRIX(x) (0x00000580+((x)*4)) +#define NV34TCL_INVERSE_MODELVIEW_MATRIX__SIZE 0x0000000c +#define NV34TCL_PROJECTION_MATRIX(x) (0x00000680+((x)*4)) +#define NV34TCL_PROJECTION_MATRIX__SIZE 0x00000010 +#define NV34TCL_TX0_MATRIX(x) (0x000006c0+((x)*4)) +#define NV34TCL_TX0_MATRIX__SIZE 0x00000010 +#define NV34TCL_TX1_MATRIX(x) (0x00000700+((x)*4)) +#define NV34TCL_TX1_MATRIX__SIZE 0x00000010 +#define NV34TCL_TX2_MATRIX(x) (0x00000740+((x)*4)) +#define NV34TCL_TX2_MATRIX__SIZE 0x00000010 +#define NV34TCL_TX3_MATRIX(x) (0x00000780+((x)*4)) +#define NV34TCL_TX3_MATRIX__SIZE 0x00000010 +#define NV34TCL_TX4_MATRIX(x) (0x000007c0+((x)*4)) +#define NV34TCL_TX4_MATRIX__SIZE 0x00000010 +#define NV34TCL_TX5_MATRIX(x) (0x00000800+((x)*4)) +#define NV34TCL_TX5_MATRIX__SIZE 0x00000010 +#define NV34TCL_TX6_MATRIX(x) (0x00000840+((x)*4)) +#define NV34TCL_TX6_MATRIX__SIZE 0x00000010 +#define NV34TCL_TX7_MATRIX(x) (0x00000880+((x)*4)) +#define NV34TCL_TX7_MATRIX__SIZE 0x00000010 +#define NV34TCL_SCISSOR_HORIZ 0x000008c0 +#define NV34TCL_SCISSOR_HORIZ_X_SHIFT 0 +#define NV34TCL_SCISSOR_HORIZ_X_MASK 0x0000ffff +#define NV34TCL_SCISSOR_HORIZ_W_SHIFT 16 +#define NV34TCL_SCISSOR_HORIZ_W_MASK 0xffff0000 +#define NV34TCL_SCISSOR_VERT 0x000008c4 +#define NV34TCL_SCISSOR_VERT_Y_SHIFT 0 +#define NV34TCL_SCISSOR_VERT_Y_MASK 0x0000ffff +#define NV34TCL_SCISSOR_VERT_H_SHIFT 16 +#define NV34TCL_SCISSOR_VERT_H_MASK 0xffff0000 +#define NV34TCL_FOG_COORD_DIST 0x000008c8 +#define NV34TCL_FOG_COORD_DIST_COORD_FALSE 0x00000000 +#define NV34TCL_FOG_COORD_DIST_COORD_FRAGMENT_DEPTH_DISTANCE_EYE_RADIAL_NV 0x00000001 +#define NV34TCL_FOG_COORD_DIST_COORD_FRAGMENT_DEPTH_DISTANCE_EYE_PLANE_ABSOLUTE_NV 0x00000002 +#define NV34TCL_FOG_COORD_DIST_COORD_FOG 0x00000003 +#define NV34TCL_FOG_MODE 0x000008cc +#define NV34TCL_FOG_MODE_EXP 0x00000800 +#define NV34TCL_FOG_MODE_EXP_2 0x00000802 +#define NV34TCL_FOG_MODE_EXP2 0x00000803 +#define NV34TCL_FOG_MODE_LINEAR 0x00000804 +#define NV34TCL_FOG_MODE_LINEAR_2 0x00002601 +#define NV34TCL_FOG_EQUATION_CONSTANT 0x000008d0 +#define NV34TCL_FOG_EQUATION_LINEAR 0x000008d4 +#define NV34TCL_FOG_EQUATION_QUADRATIC 0x000008d8 +#define NV34TCL_FP_ACTIVE_PROGRAM 0x000008e4 +#define NV34TCL_FP_ACTIVE_PROGRAM_DMA0 (1 << 0) +#define NV34TCL_FP_ACTIVE_PROGRAM_DMA1 (1 << 1) +#define NV34TCL_FP_ACTIVE_PROGRAM_OFFSET_SHIFT 2 +#define NV34TCL_FP_ACTIVE_PROGRAM_OFFSET_MASK 0xfffffffc +#define NV34TCL_RC_COLOR0 0x000008ec +#define NV34TCL_RC_COLOR0_B_SHIFT 0 +#define NV34TCL_RC_COLOR0_B_MASK 0x000000ff +#define NV34TCL_RC_COLOR0_G_SHIFT 8 +#define NV34TCL_RC_COLOR0_G_MASK 0x0000ff00 +#define NV34TCL_RC_COLOR0_R_SHIFT 16 +#define NV34TCL_RC_COLOR0_R_MASK 0x00ff0000 +#define NV34TCL_RC_COLOR0_A_SHIFT 24 +#define NV34TCL_RC_COLOR0_A_MASK 0xff000000 +#define NV34TCL_RC_COLOR1 0x000008f0 +#define NV34TCL_RC_COLOR1_B_SHIFT 0 +#define NV34TCL_RC_COLOR1_B_MASK 0x000000ff +#define NV34TCL_RC_COLOR1_G_SHIFT 8 +#define NV34TCL_RC_COLOR1_G_MASK 0x0000ff00 +#define NV34TCL_RC_COLOR1_R_SHIFT 16 +#define NV34TCL_RC_COLOR1_R_MASK 0x00ff0000 +#define NV34TCL_RC_COLOR1_A_SHIFT 24 +#define NV34TCL_RC_COLOR1_A_MASK 0xff000000 +#define NV34TCL_RC_FINAL0 0x000008f4 +#define NV34TCL_RC_FINAL0_D_INPUT_SHIFT 0 +#define NV34TCL_RC_FINAL0_D_INPUT_MASK 0x0000000f +#define NV34TCL_RC_FINAL0_D_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_FINAL0_D_INPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV34TCL_RC_FINAL0_D_INPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV34TCL_RC_FINAL0_D_INPUT_FOG 0x00000003 +#define NV34TCL_RC_FINAL0_D_INPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV34TCL_RC_FINAL0_D_INPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV34TCL_RC_FINAL0_D_INPUT_TEXTURE0_ARB 0x00000008 +#define NV34TCL_RC_FINAL0_D_INPUT_TEXTURE1_ARB 0x00000009 +#define NV34TCL_RC_FINAL0_D_INPUT_SPARE0_NV 0x0000000c +#define NV34TCL_RC_FINAL0_D_INPUT_SPARE1_NV 0x0000000d +#define NV34TCL_RC_FINAL0_D_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV34TCL_RC_FINAL0_D_INPUT_E_TIMES_F_NV 0x0000000f +#define NV34TCL_RC_FINAL0_D_COMPONENT_USAGE (1 << 4) +#define NV34TCL_RC_FINAL0_D_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_FINAL0_D_COMPONENT_USAGE_ALPHA 0x00000010 +#define NV34TCL_RC_FINAL0_D_MAPPING_SHIFT 5 +#define NV34TCL_RC_FINAL0_D_MAPPING_MASK 0x000000e0 +#define NV34TCL_RC_FINAL0_D_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_FINAL0_D_MAPPING_UNSIGNED_INVERT_NV 0x00000020 +#define NV34TCL_RC_FINAL0_D_MAPPING_EXPAND_NORMAL_NV 0x00000040 +#define NV34TCL_RC_FINAL0_D_MAPPING_EXPAND_NEGATE_NV 0x00000060 +#define NV34TCL_RC_FINAL0_D_MAPPING_HALF_BIAS_NORMAL_NV 0x00000080 +#define NV34TCL_RC_FINAL0_D_MAPPING_HALF_BIAS_NEGATE_NV 0x000000a0 +#define NV34TCL_RC_FINAL0_D_MAPPING_SIGNED_IDENTITY_NV 0x000000c0 +#define NV34TCL_RC_FINAL0_D_MAPPING_SIGNED_NEGATE_NV 0x000000e0 +#define NV34TCL_RC_FINAL0_C_INPUT_SHIFT 8 +#define NV34TCL_RC_FINAL0_C_INPUT_MASK 0x00000f00 +#define NV34TCL_RC_FINAL0_C_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_FINAL0_C_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV34TCL_RC_FINAL0_C_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV34TCL_RC_FINAL0_C_INPUT_FOG 0x00000300 +#define NV34TCL_RC_FINAL0_C_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV34TCL_RC_FINAL0_C_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV34TCL_RC_FINAL0_C_INPUT_TEXTURE0_ARB 0x00000800 +#define NV34TCL_RC_FINAL0_C_INPUT_TEXTURE1_ARB 0x00000900 +#define NV34TCL_RC_FINAL0_C_INPUT_SPARE0_NV 0x00000c00 +#define NV34TCL_RC_FINAL0_C_INPUT_SPARE1_NV 0x00000d00 +#define NV34TCL_RC_FINAL0_C_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV34TCL_RC_FINAL0_C_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV34TCL_RC_FINAL0_C_COMPONENT_USAGE (1 << 12) +#define NV34TCL_RC_FINAL0_C_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_FINAL0_C_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV34TCL_RC_FINAL0_C_MAPPING_SHIFT 13 +#define NV34TCL_RC_FINAL0_C_MAPPING_MASK 0x0000e000 +#define NV34TCL_RC_FINAL0_C_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_FINAL0_C_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV34TCL_RC_FINAL0_C_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV34TCL_RC_FINAL0_C_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV34TCL_RC_FINAL0_C_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV34TCL_RC_FINAL0_C_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV34TCL_RC_FINAL0_C_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV34TCL_RC_FINAL0_C_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV34TCL_RC_FINAL0_B_INPUT_SHIFT 16 +#define NV34TCL_RC_FINAL0_B_INPUT_MASK 0x000f0000 +#define NV34TCL_RC_FINAL0_B_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_FINAL0_B_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV34TCL_RC_FINAL0_B_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV34TCL_RC_FINAL0_B_INPUT_FOG 0x00030000 +#define NV34TCL_RC_FINAL0_B_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV34TCL_RC_FINAL0_B_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV34TCL_RC_FINAL0_B_INPUT_TEXTURE0_ARB 0x00080000 +#define NV34TCL_RC_FINAL0_B_INPUT_TEXTURE1_ARB 0x00090000 +#define NV34TCL_RC_FINAL0_B_INPUT_SPARE0_NV 0x000c0000 +#define NV34TCL_RC_FINAL0_B_INPUT_SPARE1_NV 0x000d0000 +#define NV34TCL_RC_FINAL0_B_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV34TCL_RC_FINAL0_B_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV34TCL_RC_FINAL0_B_COMPONENT_USAGE (1 << 20) +#define NV34TCL_RC_FINAL0_B_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_FINAL0_B_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV34TCL_RC_FINAL0_B_MAPPING_SHIFT 21 +#define NV34TCL_RC_FINAL0_B_MAPPING_MASK 0x00e00000 +#define NV34TCL_RC_FINAL0_B_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_FINAL0_B_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV34TCL_RC_FINAL0_B_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV34TCL_RC_FINAL0_B_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV34TCL_RC_FINAL0_B_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV34TCL_RC_FINAL0_B_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV34TCL_RC_FINAL0_B_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV34TCL_RC_FINAL0_B_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV34TCL_RC_FINAL0_A_INPUT_SHIFT 24 +#define NV34TCL_RC_FINAL0_A_INPUT_MASK 0x0f000000 +#define NV34TCL_RC_FINAL0_A_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_FINAL0_A_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV34TCL_RC_FINAL0_A_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV34TCL_RC_FINAL0_A_INPUT_FOG 0x03000000 +#define NV34TCL_RC_FINAL0_A_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV34TCL_RC_FINAL0_A_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV34TCL_RC_FINAL0_A_INPUT_TEXTURE0_ARB 0x08000000 +#define NV34TCL_RC_FINAL0_A_INPUT_TEXTURE1_ARB 0x09000000 +#define NV34TCL_RC_FINAL0_A_INPUT_SPARE0_NV 0x0c000000 +#define NV34TCL_RC_FINAL0_A_INPUT_SPARE1_NV 0x0d000000 +#define NV34TCL_RC_FINAL0_A_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV34TCL_RC_FINAL0_A_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV34TCL_RC_FINAL0_A_COMPONENT_USAGE (1 << 28) +#define NV34TCL_RC_FINAL0_A_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_FINAL0_A_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV34TCL_RC_FINAL0_A_MAPPING_SHIFT 29 +#define NV34TCL_RC_FINAL0_A_MAPPING_MASK 0xe0000000 +#define NV34TCL_RC_FINAL0_A_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_FINAL0_A_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV34TCL_RC_FINAL0_A_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV34TCL_RC_FINAL0_A_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV34TCL_RC_FINAL0_A_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV34TCL_RC_FINAL0_A_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV34TCL_RC_FINAL0_A_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV34TCL_RC_FINAL0_A_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV34TCL_RC_FINAL1 0x000008f8 +#define NV34TCL_RC_FINAL1_COLOR_SUM_CLAMP (1 << 7) +#define NV34TCL_RC_FINAL1_G_INPUT_SHIFT 8 +#define NV34TCL_RC_FINAL1_G_INPUT_MASK 0x00000f00 +#define NV34TCL_RC_FINAL1_G_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_FINAL1_G_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV34TCL_RC_FINAL1_G_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV34TCL_RC_FINAL1_G_INPUT_FOG 0x00000300 +#define NV34TCL_RC_FINAL1_G_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV34TCL_RC_FINAL1_G_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV34TCL_RC_FINAL1_G_INPUT_TEXTURE0_ARB 0x00000800 +#define NV34TCL_RC_FINAL1_G_INPUT_TEXTURE1_ARB 0x00000900 +#define NV34TCL_RC_FINAL1_G_INPUT_SPARE0_NV 0x00000c00 +#define NV34TCL_RC_FINAL1_G_INPUT_SPARE1_NV 0x00000d00 +#define NV34TCL_RC_FINAL1_G_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV34TCL_RC_FINAL1_G_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV34TCL_RC_FINAL1_G_COMPONENT_USAGE (1 << 12) +#define NV34TCL_RC_FINAL1_G_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_FINAL1_G_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV34TCL_RC_FINAL1_G_MAPPING_SHIFT 13 +#define NV34TCL_RC_FINAL1_G_MAPPING_MASK 0x0000e000 +#define NV34TCL_RC_FINAL1_G_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_FINAL1_G_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV34TCL_RC_FINAL1_G_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV34TCL_RC_FINAL1_G_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV34TCL_RC_FINAL1_G_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV34TCL_RC_FINAL1_G_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV34TCL_RC_FINAL1_G_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV34TCL_RC_FINAL1_G_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV34TCL_RC_FINAL1_F_INPUT_SHIFT 16 +#define NV34TCL_RC_FINAL1_F_INPUT_MASK 0x000f0000 +#define NV34TCL_RC_FINAL1_F_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_FINAL1_F_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV34TCL_RC_FINAL1_F_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV34TCL_RC_FINAL1_F_INPUT_FOG 0x00030000 +#define NV34TCL_RC_FINAL1_F_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV34TCL_RC_FINAL1_F_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV34TCL_RC_FINAL1_F_INPUT_TEXTURE0_ARB 0x00080000 +#define NV34TCL_RC_FINAL1_F_INPUT_TEXTURE1_ARB 0x00090000 +#define NV34TCL_RC_FINAL1_F_INPUT_SPARE0_NV 0x000c0000 +#define NV34TCL_RC_FINAL1_F_INPUT_SPARE1_NV 0x000d0000 +#define NV34TCL_RC_FINAL1_F_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV34TCL_RC_FINAL1_F_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV34TCL_RC_FINAL1_F_COMPONENT_USAGE (1 << 20) +#define NV34TCL_RC_FINAL1_F_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_FINAL1_F_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV34TCL_RC_FINAL1_F_MAPPING_SHIFT 21 +#define NV34TCL_RC_FINAL1_F_MAPPING_MASK 0x00e00000 +#define NV34TCL_RC_FINAL1_F_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_FINAL1_F_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV34TCL_RC_FINAL1_F_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV34TCL_RC_FINAL1_F_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV34TCL_RC_FINAL1_F_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV34TCL_RC_FINAL1_F_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV34TCL_RC_FINAL1_F_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV34TCL_RC_FINAL1_F_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV34TCL_RC_FINAL1_E_INPUT_SHIFT 24 +#define NV34TCL_RC_FINAL1_E_INPUT_MASK 0x0f000000 +#define NV34TCL_RC_FINAL1_E_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_FINAL1_E_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV34TCL_RC_FINAL1_E_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV34TCL_RC_FINAL1_E_INPUT_FOG 0x03000000 +#define NV34TCL_RC_FINAL1_E_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV34TCL_RC_FINAL1_E_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV34TCL_RC_FINAL1_E_INPUT_TEXTURE0_ARB 0x08000000 +#define NV34TCL_RC_FINAL1_E_INPUT_TEXTURE1_ARB 0x09000000 +#define NV34TCL_RC_FINAL1_E_INPUT_SPARE0_NV 0x0c000000 +#define NV34TCL_RC_FINAL1_E_INPUT_SPARE1_NV 0x0d000000 +#define NV34TCL_RC_FINAL1_E_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV34TCL_RC_FINAL1_E_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV34TCL_RC_FINAL1_E_COMPONENT_USAGE (1 << 28) +#define NV34TCL_RC_FINAL1_E_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_FINAL1_E_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV34TCL_RC_FINAL1_E_MAPPING_SHIFT 29 +#define NV34TCL_RC_FINAL1_E_MAPPING_MASK 0xe0000000 +#define NV34TCL_RC_FINAL1_E_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_FINAL1_E_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV34TCL_RC_FINAL1_E_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV34TCL_RC_FINAL1_E_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV34TCL_RC_FINAL1_E_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV34TCL_RC_FINAL1_E_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV34TCL_RC_FINAL1_E_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV34TCL_RC_FINAL1_E_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV34TCL_RC_ENABLE 0x000008fc +#define NV34TCL_RC_ENABLE_NUM_COMBINERS_SHIFT 0 +#define NV34TCL_RC_ENABLE_NUM_COMBINERS_MASK 0x0000000f +#define NV34TCL_RC_ENABLE_STAGE_CONSTANT_COLOR0_SHIFT 12 +#define NV34TCL_RC_ENABLE_STAGE_CONSTANT_COLOR0_MASK 0x0000f000 +#define NV34TCL_RC_ENABLE_STAGE_CONSTANT_COLOR1_SHIFT 16 +#define NV34TCL_RC_ENABLE_STAGE_CONSTANT_COLOR1_MASK 0x000f0000 +#define NV34TCL_RC_IN_ALPHA(x) (0x00000900+((x)*32)) +#define NV34TCL_RC_IN_ALPHA__SIZE 0x00000008 +#define NV34TCL_RC_IN_ALPHA_D_INPUT_SHIFT 0 +#define NV34TCL_RC_IN_ALPHA_D_INPUT_MASK 0x0000000f +#define NV34TCL_RC_IN_ALPHA_D_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_IN_ALPHA_D_INPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV34TCL_RC_IN_ALPHA_D_INPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV34TCL_RC_IN_ALPHA_D_INPUT_FOG 0x00000003 +#define NV34TCL_RC_IN_ALPHA_D_INPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV34TCL_RC_IN_ALPHA_D_INPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV34TCL_RC_IN_ALPHA_D_INPUT_TEXTURE0_ARB 0x00000008 +#define NV34TCL_RC_IN_ALPHA_D_INPUT_TEXTURE1_ARB 0x00000009 +#define NV34TCL_RC_IN_ALPHA_D_INPUT_SPARE0_NV 0x0000000c +#define NV34TCL_RC_IN_ALPHA_D_INPUT_SPARE1_NV 0x0000000d +#define NV34TCL_RC_IN_ALPHA_D_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV34TCL_RC_IN_ALPHA_D_INPUT_E_TIMES_F_NV 0x0000000f +#define NV34TCL_RC_IN_ALPHA_D_COMPONENT_USAGE (1 << 4) +#define NV34TCL_RC_IN_ALPHA_D_COMPONENT_USAGE_BLUE 0x00000000 +#define NV34TCL_RC_IN_ALPHA_D_COMPONENT_USAGE_ALPHA 0x00000010 +#define NV34TCL_RC_IN_ALPHA_D_MAPPING_SHIFT 5 +#define NV34TCL_RC_IN_ALPHA_D_MAPPING_MASK 0x000000e0 +#define NV34TCL_RC_IN_ALPHA_D_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_IN_ALPHA_D_MAPPING_UNSIGNED_INVERT_NV 0x00000020 +#define NV34TCL_RC_IN_ALPHA_D_MAPPING_EXPAND_NORMAL_NV 0x00000040 +#define NV34TCL_RC_IN_ALPHA_D_MAPPING_EXPAND_NEGATE_NV 0x00000060 +#define NV34TCL_RC_IN_ALPHA_D_MAPPING_HALF_BIAS_NORMAL_NV 0x00000080 +#define NV34TCL_RC_IN_ALPHA_D_MAPPING_HALF_BIAS_NEGATE_NV 0x000000a0 +#define NV34TCL_RC_IN_ALPHA_D_MAPPING_SIGNED_IDENTITY_NV 0x000000c0 +#define NV34TCL_RC_IN_ALPHA_D_MAPPING_SIGNED_NEGATE_NV 0x000000e0 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_SHIFT 8 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_MASK 0x00000f00 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_FOG 0x00000300 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_TEXTURE0_ARB 0x00000800 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_TEXTURE1_ARB 0x00000900 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_SPARE0_NV 0x00000c00 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_SPARE1_NV 0x00000d00 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV34TCL_RC_IN_ALPHA_C_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV34TCL_RC_IN_ALPHA_C_COMPONENT_USAGE (1 << 12) +#define NV34TCL_RC_IN_ALPHA_C_COMPONENT_USAGE_BLUE 0x00000000 +#define NV34TCL_RC_IN_ALPHA_C_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV34TCL_RC_IN_ALPHA_C_MAPPING_SHIFT 13 +#define NV34TCL_RC_IN_ALPHA_C_MAPPING_MASK 0x0000e000 +#define NV34TCL_RC_IN_ALPHA_C_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_IN_ALPHA_C_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV34TCL_RC_IN_ALPHA_C_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV34TCL_RC_IN_ALPHA_C_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV34TCL_RC_IN_ALPHA_C_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV34TCL_RC_IN_ALPHA_C_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV34TCL_RC_IN_ALPHA_C_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV34TCL_RC_IN_ALPHA_C_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_SHIFT 16 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_MASK 0x000f0000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_FOG 0x00030000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_TEXTURE0_ARB 0x00080000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_TEXTURE1_ARB 0x00090000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_SPARE0_NV 0x000c0000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_SPARE1_NV 0x000d0000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV34TCL_RC_IN_ALPHA_B_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV34TCL_RC_IN_ALPHA_B_COMPONENT_USAGE (1 << 20) +#define NV34TCL_RC_IN_ALPHA_B_COMPONENT_USAGE_BLUE 0x00000000 +#define NV34TCL_RC_IN_ALPHA_B_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV34TCL_RC_IN_ALPHA_B_MAPPING_SHIFT 21 +#define NV34TCL_RC_IN_ALPHA_B_MAPPING_MASK 0x00e00000 +#define NV34TCL_RC_IN_ALPHA_B_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_IN_ALPHA_B_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV34TCL_RC_IN_ALPHA_B_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV34TCL_RC_IN_ALPHA_B_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV34TCL_RC_IN_ALPHA_B_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV34TCL_RC_IN_ALPHA_B_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV34TCL_RC_IN_ALPHA_B_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV34TCL_RC_IN_ALPHA_B_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_SHIFT 24 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_MASK 0x0f000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_FOG 0x03000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_TEXTURE0_ARB 0x08000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_TEXTURE1_ARB 0x09000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_SPARE0_NV 0x0c000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_SPARE1_NV 0x0d000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV34TCL_RC_IN_ALPHA_A_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV34TCL_RC_IN_ALPHA_A_COMPONENT_USAGE (1 << 28) +#define NV34TCL_RC_IN_ALPHA_A_COMPONENT_USAGE_BLUE 0x00000000 +#define NV34TCL_RC_IN_ALPHA_A_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV34TCL_RC_IN_ALPHA_A_MAPPING_SHIFT 29 +#define NV34TCL_RC_IN_ALPHA_A_MAPPING_MASK 0xe0000000 +#define NV34TCL_RC_IN_ALPHA_A_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_IN_ALPHA_A_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV34TCL_RC_IN_ALPHA_A_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV34TCL_RC_IN_ALPHA_A_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV34TCL_RC_IN_ALPHA_A_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV34TCL_RC_IN_ALPHA_A_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV34TCL_RC_IN_ALPHA_A_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV34TCL_RC_IN_ALPHA_A_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV34TCL_RC_IN_RGB(x) (0x00000904+((x)*32)) +#define NV34TCL_RC_IN_RGB__SIZE 0x00000008 +#define NV34TCL_RC_IN_RGB_D_INPUT_SHIFT 0 +#define NV34TCL_RC_IN_RGB_D_INPUT_MASK 0x0000000f +#define NV34TCL_RC_IN_RGB_D_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_IN_RGB_D_INPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV34TCL_RC_IN_RGB_D_INPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV34TCL_RC_IN_RGB_D_INPUT_FOG 0x00000003 +#define NV34TCL_RC_IN_RGB_D_INPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV34TCL_RC_IN_RGB_D_INPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV34TCL_RC_IN_RGB_D_INPUT_TEXTURE0_ARB 0x00000008 +#define NV34TCL_RC_IN_RGB_D_INPUT_TEXTURE1_ARB 0x00000009 +#define NV34TCL_RC_IN_RGB_D_INPUT_SPARE0_NV 0x0000000c +#define NV34TCL_RC_IN_RGB_D_INPUT_SPARE1_NV 0x0000000d +#define NV34TCL_RC_IN_RGB_D_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV34TCL_RC_IN_RGB_D_INPUT_E_TIMES_F_NV 0x0000000f +#define NV34TCL_RC_IN_RGB_D_COMPONENT_USAGE (1 << 4) +#define NV34TCL_RC_IN_RGB_D_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_IN_RGB_D_COMPONENT_USAGE_ALPHA 0x00000010 +#define NV34TCL_RC_IN_RGB_D_MAPPING_SHIFT 5 +#define NV34TCL_RC_IN_RGB_D_MAPPING_MASK 0x000000e0 +#define NV34TCL_RC_IN_RGB_D_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_IN_RGB_D_MAPPING_UNSIGNED_INVERT_NV 0x00000020 +#define NV34TCL_RC_IN_RGB_D_MAPPING_EXPAND_NORMAL_NV 0x00000040 +#define NV34TCL_RC_IN_RGB_D_MAPPING_EXPAND_NEGATE_NV 0x00000060 +#define NV34TCL_RC_IN_RGB_D_MAPPING_HALF_BIAS_NORMAL_NV 0x00000080 +#define NV34TCL_RC_IN_RGB_D_MAPPING_HALF_BIAS_NEGATE_NV 0x000000a0 +#define NV34TCL_RC_IN_RGB_D_MAPPING_SIGNED_IDENTITY_NV 0x000000c0 +#define NV34TCL_RC_IN_RGB_D_MAPPING_SIGNED_NEGATE_NV 0x000000e0 +#define NV34TCL_RC_IN_RGB_C_INPUT_SHIFT 8 +#define NV34TCL_RC_IN_RGB_C_INPUT_MASK 0x00000f00 +#define NV34TCL_RC_IN_RGB_C_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_IN_RGB_C_INPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV34TCL_RC_IN_RGB_C_INPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV34TCL_RC_IN_RGB_C_INPUT_FOG 0x00000300 +#define NV34TCL_RC_IN_RGB_C_INPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV34TCL_RC_IN_RGB_C_INPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV34TCL_RC_IN_RGB_C_INPUT_TEXTURE0_ARB 0x00000800 +#define NV34TCL_RC_IN_RGB_C_INPUT_TEXTURE1_ARB 0x00000900 +#define NV34TCL_RC_IN_RGB_C_INPUT_SPARE0_NV 0x00000c00 +#define NV34TCL_RC_IN_RGB_C_INPUT_SPARE1_NV 0x00000d00 +#define NV34TCL_RC_IN_RGB_C_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV34TCL_RC_IN_RGB_C_INPUT_E_TIMES_F_NV 0x00000f00 +#define NV34TCL_RC_IN_RGB_C_COMPONENT_USAGE (1 << 12) +#define NV34TCL_RC_IN_RGB_C_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_IN_RGB_C_COMPONENT_USAGE_ALPHA 0x00001000 +#define NV34TCL_RC_IN_RGB_C_MAPPING_SHIFT 13 +#define NV34TCL_RC_IN_RGB_C_MAPPING_MASK 0x0000e000 +#define NV34TCL_RC_IN_RGB_C_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_IN_RGB_C_MAPPING_UNSIGNED_INVERT_NV 0x00002000 +#define NV34TCL_RC_IN_RGB_C_MAPPING_EXPAND_NORMAL_NV 0x00004000 +#define NV34TCL_RC_IN_RGB_C_MAPPING_EXPAND_NEGATE_NV 0x00006000 +#define NV34TCL_RC_IN_RGB_C_MAPPING_HALF_BIAS_NORMAL_NV 0x00008000 +#define NV34TCL_RC_IN_RGB_C_MAPPING_HALF_BIAS_NEGATE_NV 0x0000a000 +#define NV34TCL_RC_IN_RGB_C_MAPPING_SIGNED_IDENTITY_NV 0x0000c000 +#define NV34TCL_RC_IN_RGB_C_MAPPING_SIGNED_NEGATE_NV 0x0000e000 +#define NV34TCL_RC_IN_RGB_B_INPUT_SHIFT 16 +#define NV34TCL_RC_IN_RGB_B_INPUT_MASK 0x000f0000 +#define NV34TCL_RC_IN_RGB_B_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_IN_RGB_B_INPUT_CONSTANT_COLOR0_NV 0x00010000 +#define NV34TCL_RC_IN_RGB_B_INPUT_CONSTANT_COLOR1_NV 0x00020000 +#define NV34TCL_RC_IN_RGB_B_INPUT_FOG 0x00030000 +#define NV34TCL_RC_IN_RGB_B_INPUT_PRIMARY_COLOR_NV 0x00040000 +#define NV34TCL_RC_IN_RGB_B_INPUT_SECONDARY_COLOR_NV 0x00050000 +#define NV34TCL_RC_IN_RGB_B_INPUT_TEXTURE0_ARB 0x00080000 +#define NV34TCL_RC_IN_RGB_B_INPUT_TEXTURE1_ARB 0x00090000 +#define NV34TCL_RC_IN_RGB_B_INPUT_SPARE0_NV 0x000c0000 +#define NV34TCL_RC_IN_RGB_B_INPUT_SPARE1_NV 0x000d0000 +#define NV34TCL_RC_IN_RGB_B_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000e0000 +#define NV34TCL_RC_IN_RGB_B_INPUT_E_TIMES_F_NV 0x000f0000 +#define NV34TCL_RC_IN_RGB_B_COMPONENT_USAGE (1 << 20) +#define NV34TCL_RC_IN_RGB_B_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_IN_RGB_B_COMPONENT_USAGE_ALPHA 0x00100000 +#define NV34TCL_RC_IN_RGB_B_MAPPING_SHIFT 21 +#define NV34TCL_RC_IN_RGB_B_MAPPING_MASK 0x00e00000 +#define NV34TCL_RC_IN_RGB_B_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_IN_RGB_B_MAPPING_UNSIGNED_INVERT_NV 0x00200000 +#define NV34TCL_RC_IN_RGB_B_MAPPING_EXPAND_NORMAL_NV 0x00400000 +#define NV34TCL_RC_IN_RGB_B_MAPPING_EXPAND_NEGATE_NV 0x00600000 +#define NV34TCL_RC_IN_RGB_B_MAPPING_HALF_BIAS_NORMAL_NV 0x00800000 +#define NV34TCL_RC_IN_RGB_B_MAPPING_HALF_BIAS_NEGATE_NV 0x00a00000 +#define NV34TCL_RC_IN_RGB_B_MAPPING_SIGNED_IDENTITY_NV 0x00c00000 +#define NV34TCL_RC_IN_RGB_B_MAPPING_SIGNED_NEGATE_NV 0x00e00000 +#define NV34TCL_RC_IN_RGB_A_INPUT_SHIFT 24 +#define NV34TCL_RC_IN_RGB_A_INPUT_MASK 0x0f000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_ZERO 0x00000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_CONSTANT_COLOR0_NV 0x01000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_CONSTANT_COLOR1_NV 0x02000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_FOG 0x03000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_PRIMARY_COLOR_NV 0x04000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_SECONDARY_COLOR_NV 0x05000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_TEXTURE0_ARB 0x08000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_TEXTURE1_ARB 0x09000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_SPARE0_NV 0x0c000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_SPARE1_NV 0x0d000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0e000000 +#define NV34TCL_RC_IN_RGB_A_INPUT_E_TIMES_F_NV 0x0f000000 +#define NV34TCL_RC_IN_RGB_A_COMPONENT_USAGE (1 << 28) +#define NV34TCL_RC_IN_RGB_A_COMPONENT_USAGE_RGB 0x00000000 +#define NV34TCL_RC_IN_RGB_A_COMPONENT_USAGE_ALPHA 0x10000000 +#define NV34TCL_RC_IN_RGB_A_MAPPING_SHIFT 29 +#define NV34TCL_RC_IN_RGB_A_MAPPING_MASK 0xe0000000 +#define NV34TCL_RC_IN_RGB_A_MAPPING_UNSIGNED_IDENTITY_NV 0x00000000 +#define NV34TCL_RC_IN_RGB_A_MAPPING_UNSIGNED_INVERT_NV 0x20000000 +#define NV34TCL_RC_IN_RGB_A_MAPPING_EXPAND_NORMAL_NV 0x40000000 +#define NV34TCL_RC_IN_RGB_A_MAPPING_EXPAND_NEGATE_NV 0x60000000 +#define NV34TCL_RC_IN_RGB_A_MAPPING_HALF_BIAS_NORMAL_NV 0x80000000 +#define NV34TCL_RC_IN_RGB_A_MAPPING_HALF_BIAS_NEGATE_NV 0xa0000000 +#define NV34TCL_RC_IN_RGB_A_MAPPING_SIGNED_IDENTITY_NV 0xc0000000 +#define NV34TCL_RC_IN_RGB_A_MAPPING_SIGNED_NEGATE_NV 0xe0000000 +#define NV34TCL_RC_CONSTANT_COLOR0(x) (0x00000908+((x)*32)) +#define NV34TCL_RC_CONSTANT_COLOR0__SIZE 0x00000008 +#define NV34TCL_RC_CONSTANT_COLOR0_B_SHIFT 0 +#define NV34TCL_RC_CONSTANT_COLOR0_B_MASK 0x000000ff +#define NV34TCL_RC_CONSTANT_COLOR0_G_SHIFT 8 +#define NV34TCL_RC_CONSTANT_COLOR0_G_MASK 0x0000ff00 +#define NV34TCL_RC_CONSTANT_COLOR0_R_SHIFT 16 +#define NV34TCL_RC_CONSTANT_COLOR0_R_MASK 0x00ff0000 +#define NV34TCL_RC_CONSTANT_COLOR0_A_SHIFT 24 +#define NV34TCL_RC_CONSTANT_COLOR0_A_MASK 0xff000000 +#define NV34TCL_RC_CONSTANT_COLOR1(x) (0x0000090c+((x)*32)) +#define NV34TCL_RC_CONSTANT_COLOR1__SIZE 0x00000008 +#define NV34TCL_RC_CONSTANT_COLOR1_B_SHIFT 0 +#define NV34TCL_RC_CONSTANT_COLOR1_B_MASK 0x000000ff +#define NV34TCL_RC_CONSTANT_COLOR1_G_SHIFT 8 +#define NV34TCL_RC_CONSTANT_COLOR1_G_MASK 0x0000ff00 +#define NV34TCL_RC_CONSTANT_COLOR1_R_SHIFT 16 +#define NV34TCL_RC_CONSTANT_COLOR1_R_MASK 0x00ff0000 +#define NV34TCL_RC_CONSTANT_COLOR1_A_SHIFT 24 +#define NV34TCL_RC_CONSTANT_COLOR1_A_MASK 0xff000000 +#define NV34TCL_RC_OUT_ALPHA(x) (0x00000910+((x)*32)) +#define NV34TCL_RC_OUT_ALPHA__SIZE 0x00000008 +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_SHIFT 0 +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_MASK 0x0000000f +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_ZERO 0x00000000 +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_FOG 0x00000003 +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_TEXTURE0_ARB 0x00000008 +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_TEXTURE1_ARB 0x00000009 +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_SPARE0_NV 0x0000000c +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_SPARE1_NV 0x0000000d +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV34TCL_RC_OUT_ALPHA_CD_OUTPUT_E_TIMES_F_NV 0x0000000f +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_SHIFT 4 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_MASK 0x000000f0 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_ZERO 0x00000000 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_CONSTANT_COLOR0_NV 0x00000010 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_CONSTANT_COLOR1_NV 0x00000020 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_FOG 0x00000030 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_PRIMARY_COLOR_NV 0x00000040 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_SECONDARY_COLOR_NV 0x00000050 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_TEXTURE0_ARB 0x00000080 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_TEXTURE1_ARB 0x00000090 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_SPARE0_NV 0x000000c0 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_SPARE1_NV 0x000000d0 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000000e0 +#define NV34TCL_RC_OUT_ALPHA_AB_OUTPUT_E_TIMES_F_NV 0x000000f0 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_SHIFT 8 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_MASK 0x00000f00 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_ZERO 0x00000000 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_FOG 0x00000300 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_TEXTURE0_ARB 0x00000800 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_TEXTURE1_ARB 0x00000900 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_SPARE0_NV 0x00000c00 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_SPARE1_NV 0x00000d00 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV34TCL_RC_OUT_ALPHA_SUM_OUTPUT_E_TIMES_F_NV 0x00000f00 +#define NV34TCL_RC_OUT_ALPHA_CD_DOT_PRODUCT (1 << 12) +#define NV34TCL_RC_OUT_ALPHA_AB_DOT_PRODUCT (1 << 13) +#define NV34TCL_RC_OUT_ALPHA_MUX_SUM (1 << 14) +#define NV34TCL_RC_OUT_ALPHA_BIAS (1 << 15) +#define NV34TCL_RC_OUT_ALPHA_BIAS_NONE 0x00000000 +#define NV34TCL_RC_OUT_ALPHA_BIAS_BIAS_BY_NEGATIVE_ONE_HALF_NV 0x00008000 +#define NV34TCL_RC_OUT_ALPHA_SCALE_SHIFT 17 +#define NV34TCL_RC_OUT_ALPHA_SCALE_MASK 0x00000000 +#define NV34TCL_RC_OUT_ALPHA_SCALE_NONE 0x00000000 +#define NV34TCL_RC_OUT_ALPHA_SCALE_SCALE_BY_TWO_NV 0x00020000 +#define NV34TCL_RC_OUT_ALPHA_SCALE_SCALE_BY_FOUR_NV 0x00040000 +#define NV34TCL_RC_OUT_ALPHA_SCALE_SCALE_BY_ONE_HALF_NV 0x00060000 +#define NV34TCL_RC_OUT_RGB(x) (0x00000914+((x)*32)) +#define NV34TCL_RC_OUT_RGB__SIZE 0x00000008 +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_SHIFT 0 +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_MASK 0x0000000f +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_ZERO 0x00000000 +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_CONSTANT_COLOR0_NV 0x00000001 +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_CONSTANT_COLOR1_NV 0x00000002 +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_FOG 0x00000003 +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_PRIMARY_COLOR_NV 0x00000004 +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_SECONDARY_COLOR_NV 0x00000005 +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_TEXTURE0_ARB 0x00000008 +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_TEXTURE1_ARB 0x00000009 +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_SPARE0_NV 0x0000000c +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_SPARE1_NV 0x0000000d +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x0000000e +#define NV34TCL_RC_OUT_RGB_CD_OUTPUT_E_TIMES_F_NV 0x0000000f +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_SHIFT 4 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_MASK 0x000000f0 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_ZERO 0x00000000 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_CONSTANT_COLOR0_NV 0x00000010 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_CONSTANT_COLOR1_NV 0x00000020 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_FOG 0x00000030 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_PRIMARY_COLOR_NV 0x00000040 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_SECONDARY_COLOR_NV 0x00000050 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_TEXTURE0_ARB 0x00000080 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_TEXTURE1_ARB 0x00000090 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_SPARE0_NV 0x000000c0 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_SPARE1_NV 0x000000d0 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x000000e0 +#define NV34TCL_RC_OUT_RGB_AB_OUTPUT_E_TIMES_F_NV 0x000000f0 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_SHIFT 8 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_MASK 0x00000f00 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_ZERO 0x00000000 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_CONSTANT_COLOR0_NV 0x00000100 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_CONSTANT_COLOR1_NV 0x00000200 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_FOG 0x00000300 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_PRIMARY_COLOR_NV 0x00000400 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_SECONDARY_COLOR_NV 0x00000500 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_TEXTURE0_ARB 0x00000800 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_TEXTURE1_ARB 0x00000900 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_SPARE0_NV 0x00000c00 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_SPARE1_NV 0x00000d00 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_SPARE0_PLUS_SECONDARY_COLOR_NV 0x00000e00 +#define NV34TCL_RC_OUT_RGB_SUM_OUTPUT_E_TIMES_F_NV 0x00000f00 +#define NV34TCL_RC_OUT_RGB_CD_DOT_PRODUCT (1 << 12) +#define NV34TCL_RC_OUT_RGB_AB_DOT_PRODUCT (1 << 13) +#define NV34TCL_RC_OUT_RGB_MUX_SUM (1 << 14) +#define NV34TCL_RC_OUT_RGB_BIAS (1 << 15) +#define NV34TCL_RC_OUT_RGB_BIAS_NONE 0x00000000 +#define NV34TCL_RC_OUT_RGB_BIAS_BIAS_BY_NEGATIVE_ONE_HALF_NV 0x00008000 +#define NV34TCL_RC_OUT_RGB_SCALE_SHIFT 17 +#define NV34TCL_RC_OUT_RGB_SCALE_MASK 0x00000000 +#define NV34TCL_RC_OUT_RGB_SCALE_NONE 0x00000000 +#define NV34TCL_RC_OUT_RGB_SCALE_SCALE_BY_TWO_NV 0x00020000 +#define NV34TCL_RC_OUT_RGB_SCALE_SCALE_BY_FOUR_NV 0x00040000 +#define NV34TCL_RC_OUT_RGB_SCALE_SCALE_BY_ONE_HALF_NV 0x00060000 +#define NV34TCL_VIEWPORT_HORIZ 0x00000a00 +#define NV34TCL_VIEWPORT_HORIZ_X_SHIFT 0 +#define NV34TCL_VIEWPORT_HORIZ_X_MASK 0x0000ffff +#define NV34TCL_VIEWPORT_HORIZ_W_SHIFT 16 +#define NV34TCL_VIEWPORT_HORIZ_W_MASK 0xffff0000 +#define NV34TCL_VIEWPORT_VERT 0x00000a04 +#define NV34TCL_VIEWPORT_VERT_Y_SHIFT 0 +#define NV34TCL_VIEWPORT_VERT_Y_MASK 0x0000ffff +#define NV34TCL_VIEWPORT_VERT_H_SHIFT 16 +#define NV34TCL_VIEWPORT_VERT_H_MASK 0xffff0000 +#define NV34TCL_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_R 0x00000a10 +#define NV34TCL_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_G 0x00000a14 +#define NV34TCL_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_B 0x00000a18 +#define NV34TCL_VIEWPORT_TRANSLATE_X 0x00000a20 +#define NV34TCL_VIEWPORT_TRANSLATE_Y 0x00000a24 +#define NV34TCL_VIEWPORT_TRANSLATE_Z 0x00000a28 +#define NV34TCL_VIEWPORT_TRANSLATE_W 0x00000a2c +#define NV34TCL_VIEWPORT_SCALE_X 0x00000a30 +#define NV34TCL_VIEWPORT_SCALE_Y 0x00000a34 +#define NV34TCL_VIEWPORT_SCALE_Z 0x00000a38 +#define NV34TCL_VIEWPORT_SCALE_W 0x00000a3c +#define NV34TCL_POLYGON_OFFSET_POINT_ENABLE 0x00000a60 +#define NV34TCL_POLYGON_OFFSET_LINE_ENABLE 0x00000a64 +#define NV34TCL_POLYGON_OFFSET_FILL_ENABLE 0x00000a68 +#define NV34TCL_DEPTH_FUNC 0x00000a6c +#define NV34TCL_DEPTH_FUNC_NEVER 0x00000200 +#define NV34TCL_DEPTH_FUNC_LESS 0x00000201 +#define NV34TCL_DEPTH_FUNC_EQUAL 0x00000202 +#define NV34TCL_DEPTH_FUNC_LEQUAL 0x00000203 +#define NV34TCL_DEPTH_FUNC_GREATER 0x00000204 +#define NV34TCL_DEPTH_FUNC_GREATER 0x00000204 +#define NV34TCL_DEPTH_FUNC_NOTEQUAL 0x00000205 +#define NV34TCL_DEPTH_FUNC_GEQUAL 0x00000206 +#define NV34TCL_DEPTH_FUNC_ALWAYS 0x00000207 +#define NV34TCL_DEPTH_WRITE_ENABLE 0x00000a70 +#define NV34TCL_DEPTH_TEST_ENABLE 0x00000a74 +#define NV34TCL_POLYGON_OFFSET_FACTOR 0x00000a78 +#define NV34TCL_POLYGON_OFFSET_UNITS 0x00000a7c +#define NV34TCL_VTX_ATTR_3I_XY(x) (0x00000a80+((x)*8)) +#define NV34TCL_VTX_ATTR_3I_XY__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_3I_XY_X_SHIFT 0 +#define NV34TCL_VTX_ATTR_3I_XY_X_MASK 0x0000ffff +#define NV34TCL_VTX_ATTR_3I_XY_Y_SHIFT 16 +#define NV34TCL_VTX_ATTR_3I_XY_Y_MASK 0xffff0000 +#define NV34TCL_VTX_ATTR_3I_Z(x) (0x00000a84+((x)*8)) +#define NV34TCL_VTX_ATTR_3I_Z__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_3I_Z_Z_SHIFT 0 +#define NV34TCL_VTX_ATTR_3I_Z_Z_MASK 0x0000ffff +#define NV34TCL_VP_UPLOAD_INST(x) (0x00000b80+((x)*4)) +#define NV34TCL_VP_UPLOAD_INST__SIZE 0x00000004 +#define NV34TCL_TX0_CLIP_PLANE_A(x) (0x00000e00+((x)*16)) +#define NV34TCL_TX0_CLIP_PLANE_A__SIZE 0x00000004 +#define NV34TCL_TX0_CLIP_PLANE_B(x) (0x00000e04+((x)*16)) +#define NV34TCL_TX0_CLIP_PLANE_B__SIZE 0x00000004 +#define NV34TCL_TX0_CLIP_PLANE_C(x) (0x00000e08+((x)*16)) +#define NV34TCL_TX0_CLIP_PLANE_C__SIZE 0x00000004 +#define NV34TCL_TX0_CLIP_PLANE_D(x) (0x00000e0c+((x)*16)) +#define NV34TCL_TX0_CLIP_PLANE_D__SIZE 0x00000004 +#define NV34TCL_TX1_CLIP_PLANE_A(x) (0x00000e40+((x)*16)) +#define NV34TCL_TX1_CLIP_PLANE_A__SIZE 0x00000004 +#define NV34TCL_TX1_CLIP_PLANE_B(x) (0x00000e44+((x)*16)) +#define NV34TCL_TX1_CLIP_PLANE_B__SIZE 0x00000004 +#define NV34TCL_TX1_CLIP_PLANE_C(x) (0x00000e48+((x)*16)) +#define NV34TCL_TX1_CLIP_PLANE_C__SIZE 0x00000004 +#define NV34TCL_TX1_CLIP_PLANE_D(x) (0x00000e4c+((x)*16)) +#define NV34TCL_TX1_CLIP_PLANE_D__SIZE 0x00000004 +#define NV34TCL_TX2_CLIP_PLANE_A(x) (0x00000e80+((x)*16)) +#define NV34TCL_TX2_CLIP_PLANE_A__SIZE 0x00000004 +#define NV34TCL_TX2_CLIP_PLANE_B(x) (0x00000e84+((x)*16)) +#define NV34TCL_TX2_CLIP_PLANE_B__SIZE 0x00000004 +#define NV34TCL_TX2_CLIP_PLANE_C(x) (0x00000e88+((x)*16)) +#define NV34TCL_TX2_CLIP_PLANE_C__SIZE 0x00000004 +#define NV34TCL_TX2_CLIP_PLANE_D(x) (0x00000e8c+((x)*16)) +#define NV34TCL_TX2_CLIP_PLANE_D__SIZE 0x00000004 +#define NV34TCL_TX3_CLIP_PLANE_A(x) (0x00000ec0+((x)*16)) +#define NV34TCL_TX3_CLIP_PLANE_A__SIZE 0x00000004 +#define NV34TCL_TX3_CLIP_PLANE_B(x) (0x00000ec4+((x)*16)) +#define NV34TCL_TX3_CLIP_PLANE_B__SIZE 0x00000004 +#define NV34TCL_TX3_CLIP_PLANE_C(x) (0x00000ec8+((x)*16)) +#define NV34TCL_TX3_CLIP_PLANE_C__SIZE 0x00000004 +#define NV34TCL_TX3_CLIP_PLANE_D(x) (0x00000ecc+((x)*16)) +#define NV34TCL_TX3_CLIP_PLANE_D__SIZE 0x00000004 +#define NV34TCL_TX4_CLIP_PLANE_A(x) (0x00000f00+((x)*16)) +#define NV34TCL_TX4_CLIP_PLANE_A__SIZE 0x00000004 +#define NV34TCL_TX4_CLIP_PLANE_B(x) (0x00000f04+((x)*16)) +#define NV34TCL_TX4_CLIP_PLANE_B__SIZE 0x00000004 +#define NV34TCL_TX4_CLIP_PLANE_C(x) (0x00000f08+((x)*16)) +#define NV34TCL_TX4_CLIP_PLANE_C__SIZE 0x00000004 +#define NV34TCL_TX4_CLIP_PLANE_D(x) (0x00000f0c+((x)*16)) +#define NV34TCL_TX4_CLIP_PLANE_D__SIZE 0x00000004 +#define NV34TCL_TX5_CLIP_PLANE_A(x) (0x00000f40+((x)*16)) +#define NV34TCL_TX5_CLIP_PLANE_A__SIZE 0x00000004 +#define NV34TCL_TX5_CLIP_PLANE_B(x) (0x00000f44+((x)*16)) +#define NV34TCL_TX5_CLIP_PLANE_B__SIZE 0x00000004 +#define NV34TCL_TX5_CLIP_PLANE_C(x) (0x00000f48+((x)*16)) +#define NV34TCL_TX5_CLIP_PLANE_C__SIZE 0x00000004 +#define NV34TCL_TX5_CLIP_PLANE_D(x) (0x00000f4c+((x)*16)) +#define NV34TCL_TX5_CLIP_PLANE_D__SIZE 0x00000004 +#define NV34TCL_TX6_CLIP_PLANE_A(x) (0x00000f80+((x)*16)) +#define NV34TCL_TX6_CLIP_PLANE_A__SIZE 0x00000004 +#define NV34TCL_TX6_CLIP_PLANE_B(x) (0x00000f84+((x)*16)) +#define NV34TCL_TX6_CLIP_PLANE_B__SIZE 0x00000004 +#define NV34TCL_TX6_CLIP_PLANE_C(x) (0x00000f88+((x)*16)) +#define NV34TCL_TX6_CLIP_PLANE_C__SIZE 0x00000004 +#define NV34TCL_TX6_CLIP_PLANE_D(x) (0x00000f8c+((x)*16)) +#define NV34TCL_TX6_CLIP_PLANE_D__SIZE 0x00000004 +#define NV34TCL_TX7_CLIP_PLANE_A(x) (0x00000fc0+((x)*16)) +#define NV34TCL_TX7_CLIP_PLANE_A__SIZE 0x00000004 +#define NV34TCL_TX7_CLIP_PLANE_B(x) (0x00000fc4+((x)*16)) +#define NV34TCL_TX7_CLIP_PLANE_B__SIZE 0x00000004 +#define NV34TCL_TX7_CLIP_PLANE_C(x) (0x00000fc8+((x)*16)) +#define NV34TCL_TX7_CLIP_PLANE_C__SIZE 0x00000004 +#define NV34TCL_TX7_CLIP_PLANE_D(x) (0x00000fcc+((x)*16)) +#define NV34TCL_TX7_CLIP_PLANE_D__SIZE 0x00000004 +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R(x) (0x00001000+((x)*64)) +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R__SIZE 0x00000008 +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_G(x) (0x00001004+((x)*64)) +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_G__SIZE 0x00000008 +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_B(x) (0x00001008+((x)*64)) +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_B__SIZE 0x00000008 +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R(x) (0x0000100c+((x)*64)) +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R__SIZE 0x00000008 +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_G(x) (0x00001010+((x)*64)) +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_G__SIZE 0x00000008 +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_B(x) (0x00001014+((x)*64)) +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_B__SIZE 0x00000008 +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R(x) (0x00001018+((x)*64)) +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R__SIZE 0x00000008 +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_G(x) (0x0000101c+((x)*64)) +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_G__SIZE 0x00000008 +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_B(x) (0x00001020+((x)*64)) +#define NV34TCL_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_B__SIZE 0x00000008 +#define NV34TCL_LIGHT_HALF_VECTOR_X(x) (0x00001028+((x)*64)) +#define NV34TCL_LIGHT_HALF_VECTOR_X__SIZE 0x00000008 +#define NV34TCL_LIGHT_HALF_VECTOR_Y(x) (0x0000102c+((x)*64)) +#define NV34TCL_LIGHT_HALF_VECTOR_Y__SIZE 0x00000008 +#define NV34TCL_LIGHT_HALF_VECTOR_Z(x) (0x00001030+((x)*64)) +#define NV34TCL_LIGHT_HALF_VECTOR_Z__SIZE 0x00000008 +#define NV34TCL_LIGHT_DIRECTION_X(x) (0x00001034+((x)*64)) +#define NV34TCL_LIGHT_DIRECTION_X__SIZE 0x00000008 +#define NV34TCL_LIGHT_DIRECTION_Y(x) (0x00001038+((x)*64)) +#define NV34TCL_LIGHT_DIRECTION_Y__SIZE 0x00000008 +#define NV34TCL_LIGHT_DIRECTION_Z(x) (0x0000103c+((x)*64)) +#define NV34TCL_LIGHT_DIRECTION_Z__SIZE 0x00000008 +#define NV34TCL_LIGHT_SPOT_CUTOFF_A(x) (0x00001200+((x)*64)) +#define NV34TCL_LIGHT_SPOT_CUTOFF_A__SIZE 0x00000008 +#define NV34TCL_LIGHT_SPOT_CUTOFF_B(x) (0x00001204+((x)*64)) +#define NV34TCL_LIGHT_SPOT_CUTOFF_B__SIZE 0x00000008 +#define NV34TCL_LIGHT_SPOT_CUTOFF_C(x) (0x00001208+((x)*64)) +#define NV34TCL_LIGHT_SPOT_CUTOFF_C__SIZE 0x00000008 +#define NV34TCL_LIGHT_SPOT_DIR_X(x) (0x0000120c+((x)*64)) +#define NV34TCL_LIGHT_SPOT_DIR_X__SIZE 0x00000008 +#define NV34TCL_LIGHT_SPOT_DIR_Y(x) (0x00001210+((x)*64)) +#define NV34TCL_LIGHT_SPOT_DIR_Y__SIZE 0x00000008 +#define NV34TCL_LIGHT_SPOT_DIR_Z(x) (0x00001214+((x)*64)) +#define NV34TCL_LIGHT_SPOT_DIR_Z__SIZE 0x00000008 +#define NV34TCL_LIGHT_SPOT_CUTOFF_D(x) (0x00001218+((x)*64)) +#define NV34TCL_LIGHT_SPOT_CUTOFF_D__SIZE 0x00000008 +#define NV34TCL_LIGHT_POSITION_X(x) (0x0000121c+((x)*64)) +#define NV34TCL_LIGHT_POSITION_X__SIZE 0x00000008 +#define NV34TCL_LIGHT_POSITION_Y(x) (0x00001220+((x)*64)) +#define NV34TCL_LIGHT_POSITION_Y__SIZE 0x00000008 +#define NV34TCL_LIGHT_POSITION_Z(x) (0x00001224+((x)*64)) +#define NV34TCL_LIGHT_POSITION_Z__SIZE 0x00000008 +#define NV34TCL_LIGHT_ATTENUATION_CONSTANT(x) (0x00001228+((x)*64)) +#define NV34TCL_LIGHT_ATTENUATION_CONSTANT__SIZE 0x00000008 +#define NV34TCL_LIGHT_ATTENUATION_LINEAR(x) (0x0000122c+((x)*64)) +#define NV34TCL_LIGHT_ATTENUATION_LINEAR__SIZE 0x00000008 +#define NV34TCL_LIGHT_ATTENUATION_QUADRATIC(x) (0x00001230+((x)*64)) +#define NV34TCL_LIGHT_ATTENUATION_QUADRATIC__SIZE 0x00000008 +#define NV34TCL_FRONT_MATERIAL_SHININESS(x) (0x00001400+((x)*4)) +#define NV34TCL_FRONT_MATERIAL_SHININESS__SIZE 0x00000006 +#define NV34TCL_ENABLED_LIGHTS 0x00001420 +#define NV34TCL_FP_REG_CONTROL 0x00001450 +#define NV34TCL_FP_REG_CONTROL_UNK1_SHIFT 16 +#define NV34TCL_FP_REG_CONTROL_UNK1_MASK 0xffff0000 +#define NV34TCL_FP_REG_CONTROL_UNK0_SHIFT 0 +#define NV34TCL_FP_REG_CONTROL_UNK0_MASK 0x0000ffff +#define NV34TCL_VP_CLIP_PLANES_ENABLE 0x00001478 +#define NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE0 (1 << 1) +#define NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE1 (1 << 5) +#define NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE2 (1 << 9) +#define NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE3 (1 << 13) +#define NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE4 (1 << 17) +#define NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE5 (1 << 21) +#define NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE6 (1 << 25) +#define NV34TCL_VP_CLIP_PLANES_ENABLE_PLANE7 (1 << 29) +#define NV34TCL_POLYGON_STIPPLE_ENABLE 0x0000147c +#define NV34TCL_POLYGON_STIPPLE_PATTERN(x) (0x00001480+((x)*4)) +#define NV34TCL_POLYGON_STIPPLE_PATTERN__SIZE 0x00000020 +#define NV34TCL_VTX_ATTR_3F_X(x) (0x00001500+((x)*16)) +#define NV34TCL_VTX_ATTR_3F_X__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_3F_Y(x) (0x00001504+((x)*16)) +#define NV34TCL_VTX_ATTR_3F_Y__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_3F_Z(x) (0x00001508+((x)*16)) +#define NV34TCL_VTX_ATTR_3F_Z__SIZE 0x00000010 +#define NV34TCL_VP_CLIP_PLANE_A(x) (0x00001600+((x)*16)) +#define NV34TCL_VP_CLIP_PLANE_A__SIZE 0x00000006 +#define NV34TCL_VP_CLIP_PLANE_B(x) (0x00001604+((x)*16)) +#define NV34TCL_VP_CLIP_PLANE_B__SIZE 0x00000006 +#define NV34TCL_VP_CLIP_PLANE_C(x) (0x00001608+((x)*16)) +#define NV34TCL_VP_CLIP_PLANE_C__SIZE 0x00000006 +#define NV34TCL_VP_CLIP_PLANE_D(x) (0x0000160c+((x)*16)) +#define NV34TCL_VP_CLIP_PLANE_D__SIZE 0x00000006 +#define NV34TCL_VTXBUF_ADDRESS(x) (0x00001680+((x)*4)) +#define NV34TCL_VTXBUF_ADDRESS__SIZE 0x00000010 +#define NV34TCL_VTXBUF_ADDRESS_DMA1 (1 << 31) +#define NV34TCL_VTXBUF_ADDRESS_OFFSET_SHIFT 0 +#define NV34TCL_VTXBUF_ADDRESS_OFFSET_MASK 0x0fffffff +#define NV34TCL_VTXFMT(x) (0x00001740+((x)*4)) +#define NV34TCL_VTXFMT__SIZE 0x00000010 +#define NV34TCL_VTXFMT_TYPE_SHIFT 0 +#define NV34TCL_VTXFMT_TYPE_MASK 0x0000000f +#define NV34TCL_VTXFMT_TYPE_FLOAT 0x00000002 +#define NV34TCL_VTXFMT_TYPE_UBYTE 0x00000004 +#define NV34TCL_VTXFMT_TYPE_USHORT 0x00000005 +#define NV34TCL_VTXFMT_SIZE_SHIFT 4 +#define NV34TCL_VTXFMT_SIZE_MASK 0x000000f0 +#define NV34TCL_VTXFMT_STRIDE_SHIFT 8 +#define NV34TCL_VTXFMT_STRIDE_MASK 0x0000ff00 +#define NV34TCL_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_R 0x000017a0 +#define NV34TCL_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_G 0x000017a4 +#define NV34TCL_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_B 0x000017a8 +#define NV34TCL_COLOR_MATERIAL_BACK_R 0x000017b0 +#define NV34TCL_COLOR_MATERIAL_BACK_G 0x000017b4 +#define NV34TCL_COLOR_MATERIAL_BACK_B 0x000017b8 +#define NV34TCL_COLOR_MATERIAL_BACK_A 0x000017c0 +#define NV34TCL_QUERY_RESET 0x000017c8 +#define NV34TCL_QUERY_UNK17CC 0x000017cc +#define NV34TCL_QUERY_GET 0x00001800 +#define NV34TCL_QUERY_GET_UNK24_SHIFT 24 +#define NV34TCL_QUERY_GET_UNK24_MASK 0xff000000 +#define NV34TCL_QUERY_GET_OFFSET_SHIFT 0 +#define NV34TCL_QUERY_GET_OFFSET_MASK 0x00ffffff +#define NV34TCL_VERTEX_BEGIN_END 0x00001808 +#define NV34TCL_VERTEX_BEGIN_END_STOP 0x00000000 +#define NV34TCL_VERTEX_BEGIN_END_POINTS 0x00000001 +#define NV34TCL_VERTEX_BEGIN_END_LINES 0x00000002 +#define NV34TCL_VERTEX_BEGIN_END_LINE_LOOP 0x00000003 +#define NV34TCL_VERTEX_BEGIN_END_LINE_STRIP 0x00000004 +#define NV34TCL_VERTEX_BEGIN_END_TRIANGLES 0x00000005 +#define NV34TCL_VERTEX_BEGIN_END_TRIANGLE_STRIP 0x00000006 +#define NV34TCL_VERTEX_BEGIN_END_TRIANGLE_FAN 0x00000007 +#define NV34TCL_VERTEX_BEGIN_END_QUADS 0x00000008 +#define NV34TCL_VERTEX_BEGIN_END_QUAD_STRIP 0x00000009 +#define NV34TCL_VERTEX_BEGIN_END_POLYGON 0x0000000a +#define NV34TCL_VB_ELEMENT_U16 0x0000180c +#define NV34TCL_VB_ELEMENT_U16_I0_SHIFT 0 +#define NV34TCL_VB_ELEMENT_U16_I0_MASK 0x0000ffff +#define NV34TCL_VB_ELEMENT_U16_I1_SHIFT 16 +#define NV34TCL_VB_ELEMENT_U16_I1_MASK 0xffff0000 +#define NV34TCL_VB_ELEMENT_U32 0x00001810 +#define NV34TCL_VB_VERTEX_BATCH 0x00001814 +#define NV34TCL_VB_VERTEX_BATCH_OFFSET_SHIFT 0 +#define NV34TCL_VB_VERTEX_BATCH_OFFSET_MASK 0x00ffffff +#define NV34TCL_VB_VERTEX_BATCH_COUNT_SHIFT 24 +#define NV34TCL_VB_VERTEX_BATCH_COUNT_MASK 0xff000000 +#define NV34TCL_VERTEX_DATA 0x00001818 +#define NV34TCL_IDXBUF_ADDRESS 0x0000181c +#define NV34TCL_IDXBUF_FORMAT 0x00001820 +#define NV34TCL_IDXBUF_FORMAT_TYPE_SHIFT 4 +#define NV34TCL_IDXBUF_FORMAT_TYPE_MASK 0x000000f0 +#define NV34TCL_IDXBUF_FORMAT_TYPE_U32 0x00000000 +#define NV34TCL_IDXBUF_FORMAT_TYPE_U16 0x00000010 +#define NV34TCL_IDXBUF_FORMAT_DMA1 (1 << 0) +#define NV34TCL_VB_INDEX_BATCH 0x00001824 +#define NV34TCL_VB_INDEX_BATCH_COUNT_SHIFT 24 +#define NV34TCL_VB_INDEX_BATCH_COUNT_MASK 0xff000000 +#define NV34TCL_VB_INDEX_BATCH_START_SHIFT 0 +#define NV34TCL_VB_INDEX_BATCH_START_MASK 0x00ffffff +#define NV34TCL_POLYGON_MODE_FRONT 0x00001828 +#define NV34TCL_POLYGON_MODE_FRONT_POINT 0x00001b00 +#define NV34TCL_POLYGON_MODE_FRONT_LINE 0x00001b01 +#define NV34TCL_POLYGON_MODE_FRONT_FILL 0x00001b02 +#define NV34TCL_POLYGON_MODE_BACK 0x0000182c +#define NV34TCL_POLYGON_MODE_BACK_POINT 0x00001b00 +#define NV34TCL_POLYGON_MODE_BACK_LINE 0x00001b01 +#define NV34TCL_POLYGON_MODE_BACK_FILL 0x00001b02 +#define NV34TCL_CULL_FACE 0x00001830 +#define NV34TCL_CULL_FACE_FRONT 0x00000404 +#define NV34TCL_CULL_FACE_BACK 0x00000405 +#define NV34TCL_CULL_FACE_FRONT_AND_BACK 0x00000408 +#define NV34TCL_FRONT_FACE 0x00001834 +#define NV34TCL_FRONT_FACE_CW 0x00000900 +#define NV34TCL_FRONT_FACE_CCW 0x00000901 +#define NV34TCL_POLYGON_SMOOTH_ENABLE 0x00001838 +#define NV34TCL_CULL_FACE_ENABLE 0x0000183c +#define NV34TCL_TX_PALETTE_OFFSET(x) (0x00001840+((x)*4)) +#define NV34TCL_TX_PALETTE_OFFSET__SIZE 0x00000004 +#define NV34TCL_VTX_ATTR_2F_X(x) (0x00001880+((x)*8)) +#define NV34TCL_VTX_ATTR_2F_X__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_2F_Y(x) (0x00001884+((x)*8)) +#define NV34TCL_VTX_ATTR_2F_Y__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_2I(x) (0x00001900+((x)*4)) +#define NV34TCL_VTX_ATTR_2I__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_2I_X_SHIFT 0 +#define NV34TCL_VTX_ATTR_2I_X_MASK 0x0000ffff +#define NV34TCL_VTX_ATTR_2I_Y_SHIFT 16 +#define NV34TCL_VTX_ATTR_2I_Y_MASK 0xffff0000 +#define NV34TCL_VTX_ATTR_4UB(x) (0x00001940+((x)*4)) +#define NV34TCL_VTX_ATTR_4UB__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_4UB_X_SHIFT 0 +#define NV34TCL_VTX_ATTR_4UB_X_MASK 0x000000ff +#define NV34TCL_VTX_ATTR_4UB_Y_SHIFT 8 +#define NV34TCL_VTX_ATTR_4UB_Y_MASK 0x0000ff00 +#define NV34TCL_VTX_ATTR_4UB_Z_SHIFT 16 +#define NV34TCL_VTX_ATTR_4UB_Z_MASK 0x00ff0000 +#define NV34TCL_VTX_ATTR_4UB_W_SHIFT 24 +#define NV34TCL_VTX_ATTR_4UB_W_MASK 0xff000000 +#define NV34TCL_VTX_ATTR_4I_XY(x) (0x00001980+((x)*8)) +#define NV34TCL_VTX_ATTR_4I_XY__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_4I_XY_X_SHIFT 0 +#define NV34TCL_VTX_ATTR_4I_XY_X_MASK 0x0000ffff +#define NV34TCL_VTX_ATTR_4I_XY_Y_SHIFT 16 +#define NV34TCL_VTX_ATTR_4I_XY_Y_MASK 0xffff0000 +#define NV34TCL_VTX_ATTR_4I_ZW(x) (0x00001984+((x)*8)) +#define NV34TCL_VTX_ATTR_4I_ZW__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_4I_ZW_Z_SHIFT 0 +#define NV34TCL_VTX_ATTR_4I_ZW_Z_MASK 0x0000ffff +#define NV34TCL_VTX_ATTR_4I_ZW_W_SHIFT 16 +#define NV34TCL_VTX_ATTR_4I_ZW_W_MASK 0xffff0000 +#define NV34TCL_TX_OFFSET(x) (0x00001a00+((x)*32)) +#define NV34TCL_TX_OFFSET__SIZE 0x00000004 +#define NV34TCL_TX_FORMAT(x) (0x00001a04+((x)*32)) +#define NV34TCL_TX_FORMAT__SIZE 0x00000004 +#define NV34TCL_TX_FORMAT_DMA0 (1 << 0) +#define NV34TCL_TX_FORMAT_DMA1 (1 << 1) +#define NV34TCL_TX_FORMAT_CUBIC (1 << 2) +#define NV34TCL_TX_FORMAT_NO_BORDER (1 << 3) +#define NV34TCL_TX_FORMAT_DIMS_SHIFT 4 +#define NV34TCL_TX_FORMAT_DIMS_MASK 0x000000f0 +#define NV34TCL_TX_FORMAT_DIMS_1D 0x00000010 +#define NV34TCL_TX_FORMAT_DIMS_2D 0x00000020 +#define NV34TCL_TX_FORMAT_DIMS_3D 0x00000030 +#define NV34TCL_TX_FORMAT_FORMAT_SHIFT 8 +#define NV34TCL_TX_FORMAT_FORMAT_MASK 0x0000ff00 +#define NV34TCL_TX_FORMAT_FORMAT_L8 0x00000000 +#define NV34TCL_TX_FORMAT_FORMAT_A8 0x00000100 +#define NV34TCL_TX_FORMAT_FORMAT_A1R5G5B5 0x00000200 +#define NV34TCL_TX_FORMAT_FORMAT_A8_RECT 0x00000300 +#define NV34TCL_TX_FORMAT_FORMAT_A4R4G4B4 0x00000400 +#define NV34TCL_TX_FORMAT_FORMAT_R5G6B5 0x00000500 +#define NV34TCL_TX_FORMAT_FORMAT_A8R8G8B8 0x00000600 +#define NV34TCL_TX_FORMAT_FORMAT_X8R8G8B8 0x00000700 +#define NV34TCL_TX_FORMAT_FORMAT_INDEX8 0x00000b00 +#define NV34TCL_TX_FORMAT_FORMAT_DXT1 0x00000c00 +#define NV34TCL_TX_FORMAT_FORMAT_DXT3 0x00000e00 +#define NV34TCL_TX_FORMAT_FORMAT_DXT5 0x00000f00 +#define NV34TCL_TX_FORMAT_FORMAT_A1R5G5B5_RECT 0x00001000 +#define NV34TCL_TX_FORMAT_FORMAT_R5G6B5_RECT 0x00001100 +#define NV34TCL_TX_FORMAT_FORMAT_A8R8G8B8_RECT 0x00001200 +#define NV34TCL_TX_FORMAT_FORMAT_L8_RECT 0x00001300 +#define NV34TCL_TX_FORMAT_FORMAT_A8L8 0x00001a00 +#define NV34TCL_TX_FORMAT_FORMAT_A8_RECT2 0x00001b00 +#define NV34TCL_TX_FORMAT_FORMAT_A4R4G4B4_RECT 0x00001d00 +#define NV34TCL_TX_FORMAT_FORMAT_R8G8B8_RECT 0x00001e00 +#define NV34TCL_TX_FORMAT_FORMAT_L8A8_RECT 0x00002000 +#define NV34TCL_TX_FORMAT_FORMAT_DSDT 0x00002800 +#define NV34TCL_TX_FORMAT_FORMAT_A16 0x00003200 +#define NV34TCL_TX_FORMAT_FORMAT_HILO16 0x00003300 +#define NV34TCL_TX_FORMAT_FORMAT_A16_RECT 0x00003500 +#define NV34TCL_TX_FORMAT_FORMAT_HILO16_RECT 0x00003600 +#define NV34TCL_TX_FORMAT_FORMAT_HILO8 0x00004400 +#define NV34TCL_TX_FORMAT_FORMAT_SIGNED_HILO8 0x00004500 +#define NV34TCL_TX_FORMAT_FORMAT_HILO8_RECT 0x00004600 +#define NV34TCL_TX_FORMAT_FORMAT_SIGNED_HILO8_RECT 0x00004700 +#define NV34TCL_TX_FORMAT_FORMAT_FLOAT_RGBA16_NV 0x00004a00 +#define NV34TCL_TX_FORMAT_FORMAT_FLOAT_RGBA32_NV 0x00004b00 +#define NV34TCL_TX_FORMAT_FORMAT_FLOAT_R32_NV 0x00004c00 +#define NV34TCL_TX_FORMAT_MIPMAP (1 << 19) +#define NV34TCL_TX_FORMAT_BASE_SIZE_U_SHIFT 20 +#define NV34TCL_TX_FORMAT_BASE_SIZE_U_MASK 0x00f00000 +#define NV34TCL_TX_FORMAT_BASE_SIZE_V_SHIFT 24 +#define NV34TCL_TX_FORMAT_BASE_SIZE_V_MASK 0x0f000000 +#define NV34TCL_TX_FORMAT_BASE_SIZE_W_SHIFT 28 +#define NV34TCL_TX_FORMAT_BASE_SIZE_W_MASK 0xf0000000 +#define NV34TCL_TX_WRAP(x) (0x00001a08+((x)*32)) +#define NV34TCL_TX_WRAP__SIZE 0x00000004 +#define NV34TCL_TX_WRAP_S_SHIFT 0 +#define NV34TCL_TX_WRAP_S_MASK 0x000000ff +#define NV34TCL_TX_WRAP_S_REPEAT 0x00000001 +#define NV34TCL_TX_WRAP_S_MIRRORED_REPEAT 0x00000002 +#define NV34TCL_TX_WRAP_S_CLAMP_TO_EDGE 0x00000003 +#define NV34TCL_TX_WRAP_S_CLAMP_TO_BORDER 0x00000004 +#define NV34TCL_TX_WRAP_S_CLAMP 0x00000005 +#define NV34TCL_TX_WRAP_T_SHIFT 8 +#define NV34TCL_TX_WRAP_T_MASK 0x00000f00 +#define NV34TCL_TX_WRAP_T_REPEAT 0x00000100 +#define NV34TCL_TX_WRAP_T_MIRRORED_REPEAT 0x00000200 +#define NV34TCL_TX_WRAP_T_CLAMP_TO_EDGE 0x00000300 +#define NV34TCL_TX_WRAP_T_CLAMP_TO_BORDER 0x00000400 +#define NV34TCL_TX_WRAP_T_CLAMP 0x00000500 +#define NV34TCL_TX_WRAP_EXPAND_NORMAL_SHIFT 12 +#define NV34TCL_TX_WRAP_EXPAND_NORMAL_MASK 0x0000f000 +#define NV34TCL_TX_WRAP_R_SHIFT 16 +#define NV34TCL_TX_WRAP_R_MASK 0x000f0000 +#define NV34TCL_TX_WRAP_R_REPEAT 0x00010000 +#define NV34TCL_TX_WRAP_R_MIRRORED_REPEAT 0x00020000 +#define NV34TCL_TX_WRAP_R_CLAMP_TO_EDGE 0x00030000 +#define NV34TCL_TX_WRAP_R_CLAMP_TO_BORDER 0x00040000 +#define NV34TCL_TX_WRAP_R_CLAMP 0x00050000 +#define NV34TCL_TX_WRAP_RCOMP_SHIFT 28 +#define NV34TCL_TX_WRAP_RCOMP_MASK 0xf0000000 +#define NV34TCL_TX_WRAP_RCOMP_NEVER 0x00000000 +#define NV34TCL_TX_WRAP_RCOMP_GREATER 0x10000000 +#define NV34TCL_TX_WRAP_RCOMP_EQUAL 0x20000000 +#define NV34TCL_TX_WRAP_RCOMP_GEQUAL 0x30000000 +#define NV34TCL_TX_WRAP_RCOMP_LESS 0x40000000 +#define NV34TCL_TX_WRAP_RCOMP_NOTEQUAL 0x50000000 +#define NV34TCL_TX_WRAP_RCOMP_LEQUAL 0x60000000 +#define NV34TCL_TX_WRAP_RCOMP_ALWAYS 0x70000000 +#define NV34TCL_TX_ENABLE(x) (0x00001a0c+((x)*32)) +#define NV34TCL_TX_ENABLE__SIZE 0x00000004 +#define NV34TCL_TX_ENABLE_ANISO_SHIFT 4 +#define NV34TCL_TX_ENABLE_ANISO_MASK 0x00000030 +#define NV34TCL_TX_ENABLE_ANISO_NONE 0x00000000 +#define NV34TCL_TX_ENABLE_ANISO_2X 0x00000010 +#define NV34TCL_TX_ENABLE_ANISO_4X 0x00000020 +#define NV34TCL_TX_ENABLE_ANISO_8X 0x00000030 +#define NV34TCL_TX_ENABLE_MIPMAP_MAX_LOD_SHIFT 14 +#define NV34TCL_TX_ENABLE_MIPMAP_MAX_LOD_MASK 0x0003c000 +#define NV34TCL_TX_ENABLE_MIPMAP_MIN_LOD_SHIFT 26 +#define NV34TCL_TX_ENABLE_MIPMAP_MIN_LOD_MASK 0x3c000000 +#define NV34TCL_TX_ENABLE_ENABLE (1 << 30) +#define NV34TCL_TX_SWIZZLE(x) (0x00001a10+((x)*32)) +#define NV34TCL_TX_SWIZZLE__SIZE 0x00000004 +#define NV34TCL_TX_SWIZZLE_S0_X_SHIFT 14 +#define NV34TCL_TX_SWIZZLE_S0_X_MASK 0x0000c000 +#define NV34TCL_TX_SWIZZLE_S0_X_ZERO 0x00000000 +#define NV34TCL_TX_SWIZZLE_S0_X_ONE 0x00004000 +#define NV34TCL_TX_SWIZZLE_S0_X_S1 0x00008000 +#define NV34TCL_TX_SWIZZLE_S0_Y_SHIFT 12 +#define NV34TCL_TX_SWIZZLE_S0_Y_MASK 0x00003000 +#define NV34TCL_TX_SWIZZLE_S0_Y_ZERO 0x00000000 +#define NV34TCL_TX_SWIZZLE_S0_Y_ONE 0x00001000 +#define NV34TCL_TX_SWIZZLE_S0_Y_S1 0x00002000 +#define NV34TCL_TX_SWIZZLE_S0_Z_SHIFT 10 +#define NV34TCL_TX_SWIZZLE_S0_Z_MASK 0x00000c00 +#define NV34TCL_TX_SWIZZLE_S0_Z_ZERO 0x00000000 +#define NV34TCL_TX_SWIZZLE_S0_Z_ONE 0x00000400 +#define NV34TCL_TX_SWIZZLE_S0_Z_S1 0x00000800 +#define NV34TCL_TX_SWIZZLE_S0_W_SHIFT 8 +#define NV34TCL_TX_SWIZZLE_S0_W_MASK 0x00000300 +#define NV34TCL_TX_SWIZZLE_S0_W_ZERO 0x00000000 +#define NV34TCL_TX_SWIZZLE_S0_W_ONE 0x00000100 +#define NV34TCL_TX_SWIZZLE_S0_W_S1 0x00000200 +#define NV34TCL_TX_SWIZZLE_S1_X_SHIFT 6 +#define NV34TCL_TX_SWIZZLE_S1_X_MASK 0x000000c0 +#define NV34TCL_TX_SWIZZLE_S1_X_W 0x00000000 +#define NV34TCL_TX_SWIZZLE_S1_X_Z 0x00000040 +#define NV34TCL_TX_SWIZZLE_S1_X_Y 0x00000080 +#define NV34TCL_TX_SWIZZLE_S1_X_X 0x000000c0 +#define NV34TCL_TX_SWIZZLE_S1_Y_SHIFT 4 +#define NV34TCL_TX_SWIZZLE_S1_Y_MASK 0x00000030 +#define NV34TCL_TX_SWIZZLE_S1_Y_W 0x00000000 +#define NV34TCL_TX_SWIZZLE_S1_Y_Z 0x00000010 +#define NV34TCL_TX_SWIZZLE_S1_Y_Y 0x00000020 +#define NV34TCL_TX_SWIZZLE_S1_Y_X 0x00000030 +#define NV34TCL_TX_SWIZZLE_S1_Z_SHIFT 2 +#define NV34TCL_TX_SWIZZLE_S1_Z_MASK 0x0000000c +#define NV34TCL_TX_SWIZZLE_S1_Z_W 0x00000000 +#define NV34TCL_TX_SWIZZLE_S1_Z_Z 0x00000004 +#define NV34TCL_TX_SWIZZLE_S1_Z_Y 0x00000008 +#define NV34TCL_TX_SWIZZLE_S1_Z_X 0x0000000c +#define NV34TCL_TX_SWIZZLE_S1_W_SHIFT 0 +#define NV34TCL_TX_SWIZZLE_S1_W_MASK 0x00000003 +#define NV34TCL_TX_SWIZZLE_S1_W_W 0x00000000 +#define NV34TCL_TX_SWIZZLE_S1_W_Z 0x00000001 +#define NV34TCL_TX_SWIZZLE_S1_W_Y 0x00000002 +#define NV34TCL_TX_SWIZZLE_S1_W_X 0x00000003 +#define NV34TCL_TX_SWIZZLE_RECT_PITCH_SHIFT 16 +#define NV34TCL_TX_SWIZZLE_RECT_PITCH_MASK 0xffff0000 +#define NV34TCL_TX_FILTER(x) (0x00001a14+((x)*32)) +#define NV34TCL_TX_FILTER__SIZE 0x00000004 +#define NV34TCL_TX_FILTER_LOD_BIAS_SHIFT 8 +#define NV34TCL_TX_FILTER_LOD_BIAS_MASK 0x00000f00 +#define NV34TCL_TX_FILTER_MINIFY_SHIFT 16 +#define NV34TCL_TX_FILTER_MINIFY_MASK 0x000f0000 +#define NV34TCL_TX_FILTER_MINIFY_NEAREST 0x00010000 +#define NV34TCL_TX_FILTER_MINIFY_LINEAR 0x00020000 +#define NV34TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST 0x00030000 +#define NV34TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST 0x00040000 +#define NV34TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR 0x00050000 +#define NV34TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR 0x00060000 +#define NV34TCL_TX_FILTER_MAGNIFY_SHIFT 24 +#define NV34TCL_TX_FILTER_MAGNIFY_MASK 0x0f000000 +#define NV34TCL_TX_FILTER_MAGNIFY_NEAREST 0x01000000 +#define NV34TCL_TX_FILTER_MAGNIFY_LINEAR 0x02000000 +#define NV34TCL_TX_FILTER_SIGNED_BLUE (1 << 28) +#define NV34TCL_TX_FILTER_SIGNED_GREEN (1 << 29) +#define NV34TCL_TX_FILTER_SIGNED_RED (1 << 30) +#define NV34TCL_TX_FILTER_SIGNED_ALPHA (1 << 31) +#define NV34TCL_TX_NPOT_SIZE(x) (0x00001a18+((x)*32)) +#define NV34TCL_TX_NPOT_SIZE__SIZE 0x00000004 +#define NV34TCL_TX_NPOT_SIZE_H_SHIFT 0 +#define NV34TCL_TX_NPOT_SIZE_H_MASK 0x0000ffff +#define NV34TCL_TX_NPOT_SIZE_W_SHIFT 16 +#define NV34TCL_TX_NPOT_SIZE_W_MASK 0xffff0000 +#define NV34TCL_TX_BORDER_COLOR(x) (0x00001a1c+((x)*32)) +#define NV34TCL_TX_BORDER_COLOR__SIZE 0x00000004 +#define NV34TCL_TX_BORDER_COLOR_B_SHIFT 0 +#define NV34TCL_TX_BORDER_COLOR_B_MASK 0x000000ff +#define NV34TCL_TX_BORDER_COLOR_G_SHIFT 8 +#define NV34TCL_TX_BORDER_COLOR_G_MASK 0x0000ff00 +#define NV34TCL_TX_BORDER_COLOR_R_SHIFT 16 +#define NV34TCL_TX_BORDER_COLOR_R_MASK 0x00ff0000 +#define NV34TCL_TX_BORDER_COLOR_A_SHIFT 24 +#define NV34TCL_TX_BORDER_COLOR_A_MASK 0xff000000 +#define NV34TCL_VTX_ATTR_4F_X(x) (0x00001c00+((x)*16)) +#define NV34TCL_VTX_ATTR_4F_X__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_4F_Y(x) (0x00001c04+((x)*16)) +#define NV34TCL_VTX_ATTR_4F_Y__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_4F_Z(x) (0x00001c08+((x)*16)) +#define NV34TCL_VTX_ATTR_4F_Z__SIZE 0x00000010 +#define NV34TCL_VTX_ATTR_4F_W(x) (0x00001c0c+((x)*16)) +#define NV34TCL_VTX_ATTR_4F_W__SIZE 0x00000010 +#define NV34TCL_FP_CONTROL 0x00001d60 +#define NV34TCL_FP_CONTROL_USES_KIL (1 << 7) +#define NV34TCL_FP_CONTROL_USED_REGS_MINUS1_DIV2_SHIFT 0 +#define NV34TCL_FP_CONTROL_USED_REGS_MINUS1_DIV2_MASK 0x0000000f +#define NV34TCL_DEPTH_UNK17D8 0x00001d78 +#define NV34TCL_DEPTH_UNK17D8_CLAMP_SHIFT 4 +#define NV34TCL_DEPTH_UNK17D8_CLAMP_MASK 0x000000f0 +#define NV34TCL_MULTISAMPLE_CONTROL 0x00001d7c +#define NV34TCL_MULTISAMPLE_CONTROL_ENABLE (1 << 0) +#define NV34TCL_MULTISAMPLE_CONTROL_SAMPLE_ALPHA_TO_COVERAGE (1 << 4) +#define NV34TCL_MULTISAMPLE_CONTROL_SAMPLE_ALPHA_TO_ONE (1 << 8) +#define NV34TCL_MULTISAMPLE_CONTROL_SAMPLE_COVERAGE_SHIFT 16 +#define NV34TCL_MULTISAMPLE_CONTROL_SAMPLE_COVERAGE_MASK 0xffff0000 +#define NV34TCL_CLEAR_DEPTH_VALUE 0x00001d8c +#define NV34TCL_CLEAR_COLOR_VALUE 0x00001d90 +#define NV34TCL_CLEAR_COLOR_VALUE_B_SHIFT 0 +#define NV34TCL_CLEAR_COLOR_VALUE_B_MASK 0x000000ff +#define NV34TCL_CLEAR_COLOR_VALUE_G_SHIFT 8 +#define NV34TCL_CLEAR_COLOR_VALUE_G_MASK 0x0000ff00 +#define NV34TCL_CLEAR_COLOR_VALUE_R_SHIFT 16 +#define NV34TCL_CLEAR_COLOR_VALUE_R_MASK 0x00ff0000 +#define NV34TCL_CLEAR_COLOR_VALUE_A_SHIFT 24 +#define NV34TCL_CLEAR_COLOR_VALUE_A_MASK 0xff000000 +#define NV34TCL_CLEAR_BUFFERS 0x00001d94 +#define NV34TCL_CLEAR_BUFFERS_COLOR_A (1 << 7) +#define NV34TCL_CLEAR_BUFFERS_COLOR_B (1 << 6) +#define NV34TCL_CLEAR_BUFFERS_COLOR_G (1 << 5) +#define NV34TCL_CLEAR_BUFFERS_COLOR_R (1 << 4) +#define NV34TCL_CLEAR_BUFFERS_STENCIL (1 << 1) +#define NV34TCL_CLEAR_BUFFERS_DEPTH (1 << 0) +#define NV34TCL_DO_VERTICES 0x00001dac +#define NV34TCL_LINE_STIPPLE_ENABLE 0x00001db4 +#define NV34TCL_LINE_STIPPLE_PATTERN 0x00001db8 +#define NV34TCL_LINE_STIPPLE_PATTERN_FACTOR_SHIFT 0 +#define NV34TCL_LINE_STIPPLE_PATTERN_FACTOR_MASK 0x0000ffff +#define NV34TCL_LINE_STIPPLE_PATTERN_PATTERN_SHIFT 16 +#define NV34TCL_LINE_STIPPLE_PATTERN_PATTERN_MASK 0xffff0000 +#define NV34TCL_BACK_MATERIAL_SHININESS(x) (0x00001e20+((x)*4)) +#define NV34TCL_BACK_MATERIAL_SHININESS__SIZE 0x00000006 +#define NV34TCL_VTX_ATTR_1F(x) (0x00001e40+((x)*4)) +#define NV34TCL_VTX_ATTR_1F__SIZE 0x00000010 +#define NV34TCL_ENGINE 0x00001e94 +#define NV34TCL_ENGINE_FP (1 << 0) +#define NV34TCL_ENGINE_VP (1 << 1) +#define NV34TCL_ENGINE_FIXED (1 << 2) +#define NV34TCL_VP_UPLOAD_FROM_ID 0x00001e9c +#define NV34TCL_VP_START_FROM_ID 0x00001ea0 +#define NV34TCL_POINT_PARAMETERS(x) (0x00001ec0+((x)*4)) +#define NV34TCL_POINT_PARAMETERS__SIZE 0x00000008 +#define NV34TCL_POINT_SIZE 0x00001ee0 +#define NV34TCL_POINT_PARAMETERS_ENABLE 0x00001ee4 +#define NV34TCL_POINT_SPRITE 0x00001ee8 +#define NV34TCL_POINT_SPRITE_ENABLE (1 << 0) +#define NV34TCL_POINT_SPRITE_R_MODE_SHIFT 1 +#define NV34TCL_POINT_SPRITE_R_MODE_MASK 0x00000006 +#define NV34TCL_POINT_SPRITE_R_MODE_ZERO 0x00000000 +#define NV34TCL_POINT_SPRITE_R_MODE_R 0x00000002 +#define NV34TCL_POINT_SPRITE_R_MODE_S 0x00000004 +#define NV34TCL_POINT_SPRITE_COORD_REPLACE (1 << 11) +#define NV34TCL_VP_UPLOAD_CONST_ID 0x00001efc +#define NV34TCL_VP_UPLOAD_CONST_X(x) (0x00001f00+((x)*16)) +#define NV34TCL_VP_UPLOAD_CONST_X__SIZE 0x00000004 +#define NV34TCL_VP_UPLOAD_CONST_Y(x) (0x00001f04+((x)*16)) +#define NV34TCL_VP_UPLOAD_CONST_Y__SIZE 0x00000004 +#define NV34TCL_VP_UPLOAD_CONST_Z(x) (0x00001f08+((x)*16)) +#define NV34TCL_VP_UPLOAD_CONST_Z__SIZE 0x00000004 +#define NV34TCL_VP_UPLOAD_CONST_W(x) (0x00001f0c+((x)*16)) +#define NV34TCL_VP_UPLOAD_CONST_W__SIZE 0x00000004 +#define NV34TCL_UNK1f80(x) (0x00001f80+((x)*4)) +#define NV34TCL_UNK1f80__SIZE 0x00000010 + + +#define NV40_CONTEXT_SURFACES_2D 0x00003062 + + + +#define NV40_STRETCHED_IMAGE_FROM_CPU 0x00003066 + + + +#define NV40_TEXTURE_FROM_CPU 0x0000307b + + + +#define NV40_SCALED_IMAGE_FROM_MEMORY 0x00003089 + + + +#define NV40_IMAGE_FROM_CPU 0x0000308a + + + +#define NV40_SWIZZLED_SURFACE 0x0000309e + + + +#define NV40TCL 0x00004097 + +#define NV40TCL_REF_CNT 0x00000050 +#define NV40TCL_NOP 0x00000100 +#define NV40TCL_NOTIFY 0x00000104 +#define NV40TCL_DMA_NOTIFY 0x00000180 +#define NV40TCL_DMA_TEXTURE0 0x00000184 +#define NV40TCL_DMA_TEXTURE1 0x00000188 +#define NV40TCL_DMA_COLOR1 0x0000018c +#define NV40TCL_DMA_COLOR0 0x00000194 +#define NV40TCL_DMA_ZETA 0x00000198 +#define NV40TCL_DMA_VTXBUF0 0x0000019c +#define NV40TCL_DMA_VTXBUF1 0x000001a0 +#define NV40TCL_DMA_FENCE 0x000001a4 +#define NV40TCL_DMA_QUERY 0x000001a8 +#define NV40TCL_DMA_UNK01AC 0x000001ac +#define NV40TCL_DMA_UNK01B0 0x000001b0 +#define NV40TCL_DMA_COLOR2 0x000001b4 +#define NV40TCL_DMA_COLOR3 0x000001b8 +#define NV40TCL_RT_HORIZ 0x00000200 +#define NV40TCL_RT_HORIZ_W_SHIFT 16 +#define NV40TCL_RT_HORIZ_W_MASK 0xffff0000 +#define NV40TCL_RT_HORIZ_X_SHIFT 0 +#define NV40TCL_RT_HORIZ_X_MASK 0x0000ffff +#define NV40TCL_RT_VERT 0x00000204 +#define NV40TCL_RT_VERT_H_SHIFT 16 +#define NV40TCL_RT_VERT_H_MASK 0xffff0000 +#define NV40TCL_RT_VERT_Y_SHIFT 0 +#define NV40TCL_RT_VERT_Y_MASK 0x0000ffff +#define NV40TCL_RT_FORMAT 0x00000208 +#define NV40TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT 24 +#define NV40TCL_RT_FORMAT_LOG2_HEIGHT_MASK 0xff000000 +#define NV40TCL_RT_FORMAT_LOG2_WIDTH_SHIFT 16 +#define NV40TCL_RT_FORMAT_LOG2_WIDTH_MASK 0x00ff0000 +#define NV40TCL_RT_FORMAT_TYPE_SHIFT 8 +#define NV40TCL_RT_FORMAT_TYPE_MASK 0x00000f00 +#define NV40TCL_RT_FORMAT_TYPE_LINEAR 0x00000100 +#define NV40TCL_RT_FORMAT_TYPE_SWIZZLED 0x00000200 +#define NV40TCL_RT_FORMAT_ZETA_SHIFT 5 +#define NV40TCL_RT_FORMAT_ZETA_MASK 0x000000e0 +#define NV40TCL_RT_FORMAT_ZETA_Z16 0x00000020 +#define NV40TCL_RT_FORMAT_ZETA_Z24S8 0x00000040 +#define NV40TCL_RT_FORMAT_COLOR_SHIFT 0 +#define NV40TCL_RT_FORMAT_COLOR_MASK 0x0000001f +#define NV40TCL_RT_FORMAT_COLOR_R5G6B5 0x00000003 +#define NV40TCL_RT_FORMAT_COLOR_X8R8G8B8 0x00000005 +#define NV40TCL_RT_FORMAT_COLOR_A8R8G8B8 0x00000008 +#define NV40TCL_RT_FORMAT_COLOR_B8 0x00000009 +#define NV40TCL_RT_FORMAT_COLOR_UNKNOWN 0x0000000d +#define NV40TCL_RT_FORMAT_COLOR_X8B8G8R8 0x0000000f +#define NV40TCL_RT_FORMAT_COLOR_A8B8G8R8 0x00000010 +#define NV40TCL_COLOR0_PITCH 0x0000020c +#define NV40TCL_COLOR0_OFFSET 0x00000210 +#define NV40TCL_ZETA_OFFSET 0x00000214 +#define NV40TCL_COLOR1_OFFSET 0x00000218 +#define NV40TCL_COLOR1_PITCH 0x0000021c +#define NV40TCL_RT_ENABLE 0x00000220 +#define NV40TCL_RT_ENABLE_MRT (1 << 4) +#define NV40TCL_RT_ENABLE_COLOR3 (1 << 3) +#define NV40TCL_RT_ENABLE_COLOR2 (1 << 2) +#define NV40TCL_RT_ENABLE_COLOR1 (1 << 1) +#define NV40TCL_RT_ENABLE_COLOR0 (1 << 0) +#define NV40TCL_ZETA_PITCH 0x0000022c +#define NV40TCL_COLOR2_PITCH 0x00000280 +#define NV40TCL_COLOR3_PITCH 0x00000284 +#define NV40TCL_COLOR2_OFFSET 0x00000288 +#define NV40TCL_COLOR3_OFFSET 0x0000028c +#define NV40TCL_VIEWPORT_CLIP_HORIZ(x) (0x000002c0+((x)*8)) +#define NV40TCL_VIEWPORT_CLIP_HORIZ__SIZE 0x00000008 +#define NV40TCL_VIEWPORT_CLIP_VERT(x) (0x000002c4+((x)*8)) +#define NV40TCL_VIEWPORT_CLIP_VERT__SIZE 0x00000008 +#define NV40TCL_DITHER_ENABLE 0x00000300 +#define NV40TCL_ALPHA_TEST_ENABLE 0x00000304 +#define NV40TCL_ALPHA_TEST_FUNC 0x00000308 +#define NV40TCL_ALPHA_TEST_FUNC_NEVER 0x00000200 +#define NV40TCL_ALPHA_TEST_FUNC_LESS 0x00000201 +#define NV40TCL_ALPHA_TEST_FUNC_EQUAL 0x00000202 +#define NV40TCL_ALPHA_TEST_FUNC_LEQUAL 0x00000203 +#define NV40TCL_ALPHA_TEST_FUNC_GREATER 0x00000204 +#define NV40TCL_ALPHA_TEST_FUNC_GREATER 0x00000204 +#define NV40TCL_ALPHA_TEST_FUNC_NOTEQUAL 0x00000205 +#define NV40TCL_ALPHA_TEST_FUNC_GEQUAL 0x00000206 +#define NV40TCL_ALPHA_TEST_FUNC_ALWAYS 0x00000207 +#define NV40TCL_ALPHA_TEST_REF 0x0000030c +#define NV40TCL_BLEND_ENABLE 0x00000310 +#define NV40TCL_BLEND_FUNC_SRC 0x00000314 +#define NV40TCL_BLEND_FUNC_SRC_RGB_SHIFT 0 +#define NV40TCL_BLEND_FUNC_SRC_RGB_MASK 0x0000ffff +#define NV40TCL_BLEND_FUNC_SRC_RGB_ZERO 0x00000000 +#define NV40TCL_BLEND_FUNC_SRC_RGB_ONE 0x00000001 +#define NV40TCL_BLEND_FUNC_SRC_RGB_SRC_COLOR 0x00000300 +#define NV40TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV40TCL_BLEND_FUNC_SRC_RGB_SRC_ALPHA 0x00000302 +#define NV40TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV40TCL_BLEND_FUNC_SRC_RGB_DST_ALPHA 0x00000304 +#define NV40TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV40TCL_BLEND_FUNC_SRC_RGB_DST_COLOR 0x00000306 +#define NV40TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_DST_COLOR 0x00000307 +#define NV40TCL_BLEND_FUNC_SRC_RGB_SRC_ALPHA_SATURATE 0x00000308 +#define NV40TCL_BLEND_FUNC_SRC_RGB_CONSTANT_COLOR 0x00008001 +#define NV40TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV40TCL_BLEND_FUNC_SRC_RGB_CONSTANT_ALPHA 0x00008003 +#define NV40TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_SHIFT 16 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_MASK 0xffff0000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_ZERO 0x00000000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_ONE 0x00010000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_SRC_COLOR 0x03000000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_SRC_COLOR 0x03010000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_SRC_ALPHA 0x03020000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_SRC_ALPHA 0x03030000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_DST_ALPHA 0x03040000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_DST_ALPHA 0x03050000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_DST_COLOR 0x03060000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_DST_COLOR 0x03070000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_SRC_ALPHA_SATURATE 0x03080000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_CONSTANT_COLOR 0x80010000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_CONSTANT_COLOR 0x80020000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_CONSTANT_ALPHA 0x80030000 +#define NV40TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_CONSTANT_ALPHA 0x80040000 +#define NV40TCL_BLEND_FUNC_DST 0x00000318 +#define NV40TCL_BLEND_FUNC_DST_RGB_SHIFT 0 +#define NV40TCL_BLEND_FUNC_DST_RGB_MASK 0x0000ffff +#define NV40TCL_BLEND_FUNC_DST_RGB_ZERO 0x00000000 +#define NV40TCL_BLEND_FUNC_DST_RGB_ONE 0x00000001 +#define NV40TCL_BLEND_FUNC_DST_RGB_SRC_COLOR 0x00000300 +#define NV40TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV40TCL_BLEND_FUNC_DST_RGB_SRC_ALPHA 0x00000302 +#define NV40TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV40TCL_BLEND_FUNC_DST_RGB_DST_ALPHA 0x00000304 +#define NV40TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV40TCL_BLEND_FUNC_DST_RGB_DST_COLOR 0x00000306 +#define NV40TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_DST_COLOR 0x00000307 +#define NV40TCL_BLEND_FUNC_DST_RGB_SRC_ALPHA_SATURATE 0x00000308 +#define NV40TCL_BLEND_FUNC_DST_RGB_CONSTANT_COLOR 0x00008001 +#define NV40TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV40TCL_BLEND_FUNC_DST_RGB_CONSTANT_ALPHA 0x00008003 +#define NV40TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_SHIFT 16 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_MASK 0xffff0000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_ZERO 0x00000000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_ONE 0x00010000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_SRC_COLOR 0x03000000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_SRC_COLOR 0x03010000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_SRC_ALPHA 0x03020000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_SRC_ALPHA 0x03030000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_DST_ALPHA 0x03040000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_DST_ALPHA 0x03050000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_DST_COLOR 0x03060000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_DST_COLOR 0x03070000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_SRC_ALPHA_SATURATE 0x03080000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_CONSTANT_COLOR 0x80010000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_CONSTANT_COLOR 0x80020000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_CONSTANT_ALPHA 0x80030000 +#define NV40TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_CONSTANT_ALPHA 0x80040000 +#define NV40TCL_BLEND_COLOR 0x0000031c +#define NV40TCL_BLEND_EQUATION 0x00000320 +#define NV40TCL_BLEND_EQUATION_RGB_SHIFT 0 +#define NV40TCL_BLEND_EQUATION_RGB_MASK 0x0000ffff +#define NV40TCL_BLEND_EQUATION_RGB_FUNC_ADD 0x00008006 +#define NV40TCL_BLEND_EQUATION_RGB_MIN 0x00008007 +#define NV40TCL_BLEND_EQUATION_RGB_MAX 0x00008008 +#define NV40TCL_BLEND_EQUATION_RGB_FUNC_SUBTRACT 0x0000800a +#define NV40TCL_BLEND_EQUATION_RGB_FUNC_REVERSE_SUBTRACT 0x0000800b +#define NV40TCL_BLEND_EQUATION_ALPHA_SHIFT 16 +#define NV40TCL_BLEND_EQUATION_ALPHA_MASK 0xffff0000 +#define NV40TCL_BLEND_EQUATION_ALPHA_FUNC_ADD 0x80060000 +#define NV40TCL_BLEND_EQUATION_ALPHA_MIN 0x80070000 +#define NV40TCL_BLEND_EQUATION_ALPHA_MAX 0x80080000 +#define NV40TCL_BLEND_EQUATION_ALPHA_FUNC_SUBTRACT 0x800a0000 +#define NV40TCL_BLEND_EQUATION_ALPHA_FUNC_REVERSE_SUBTRACT 0x800b0000 +#define NV40TCL_COLOR_MASK 0x00000324 +#define NV40TCL_COLOR_MASK_BUFFER0_B_SHIFT 0 +#define NV40TCL_COLOR_MASK_BUFFER0_B_MASK 0x000000ff +#define NV40TCL_COLOR_MASK_BUFFER0_G_SHIFT 8 +#define NV40TCL_COLOR_MASK_BUFFER0_G_MASK 0x0000ff00 +#define NV40TCL_COLOR_MASK_BUFFER0_R_SHIFT 16 +#define NV40TCL_COLOR_MASK_BUFFER0_R_MASK 0x00ff0000 +#define NV40TCL_COLOR_MASK_BUFFER0_A_SHIFT 24 +#define NV40TCL_COLOR_MASK_BUFFER0_A_MASK 0xff000000 +#define NV40TCL_STENCIL_FRONT_ENABLE 0x00000328 +#define NV40TCL_STENCIL_FRONT_MASK 0x0000032c +#define NV40TCL_STENCIL_FRONT_FUNC_FUNC 0x00000330 +#define NV40TCL_STENCIL_FRONT_FUNC_FUNC_NEVER 0x00000200 +#define NV40TCL_STENCIL_FRONT_FUNC_FUNC_LESS 0x00000201 +#define NV40TCL_STENCIL_FRONT_FUNC_FUNC_EQUAL 0x00000202 +#define NV40TCL_STENCIL_FRONT_FUNC_FUNC_LEQUAL 0x00000203 +#define NV40TCL_STENCIL_FRONT_FUNC_FUNC_GREATER 0x00000204 +#define NV40TCL_STENCIL_FRONT_FUNC_FUNC_GREATER 0x00000204 +#define NV40TCL_STENCIL_FRONT_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV40TCL_STENCIL_FRONT_FUNC_FUNC_GEQUAL 0x00000206 +#define NV40TCL_STENCIL_FRONT_FUNC_FUNC_ALWAYS 0x00000207 +#define NV40TCL_STENCIL_FRONT_FUNC_REF 0x00000334 +#define NV40TCL_STENCIL_FRONT_FUNC_MASK 0x00000338 +#define NV40TCL_STENCIL_FRONT_OP_FAIL 0x0000033c +#define NV40TCL_STENCIL_FRONT_OP_FAIL_ZERO 0x00000000 +#define NV40TCL_STENCIL_FRONT_OP_FAIL_INVERT 0x0000150a +#define NV40TCL_STENCIL_FRONT_OP_FAIL_KEEP 0x00001e00 +#define NV40TCL_STENCIL_FRONT_OP_FAIL_REPLACE 0x00001e01 +#define NV40TCL_STENCIL_FRONT_OP_FAIL_INCR 0x00001e02 +#define NV40TCL_STENCIL_FRONT_OP_FAIL_DECR 0x00001e03 +#define NV40TCL_STENCIL_FRONT_OP_FAIL_INCR_WRAP 0x00008507 +#define NV40TCL_STENCIL_FRONT_OP_FAIL_DECR_WRAP 0x00008508 +#define NV40TCL_STENCIL_FRONT_OP_ZFAIL 0x00000340 +#define NV40TCL_STENCIL_FRONT_OP_ZFAIL_ZERO 0x00000000 +#define NV40TCL_STENCIL_FRONT_OP_ZFAIL_INVERT 0x0000150a +#define NV40TCL_STENCIL_FRONT_OP_ZFAIL_KEEP 0x00001e00 +#define NV40TCL_STENCIL_FRONT_OP_ZFAIL_REPLACE 0x00001e01 +#define NV40TCL_STENCIL_FRONT_OP_ZFAIL_INCR 0x00001e02 +#define NV40TCL_STENCIL_FRONT_OP_ZFAIL_DECR 0x00001e03 +#define NV40TCL_STENCIL_FRONT_OP_ZFAIL_INCR_WRAP 0x00008507 +#define NV40TCL_STENCIL_FRONT_OP_ZFAIL_DECR_WRAP 0x00008508 +#define NV40TCL_STENCIL_FRONT_OP_ZPASS 0x00000344 +#define NV40TCL_STENCIL_FRONT_OP_ZPASS_ZERO 0x00000000 +#define NV40TCL_STENCIL_FRONT_OP_ZPASS_INVERT 0x0000150a +#define NV40TCL_STENCIL_FRONT_OP_ZPASS_KEEP 0x00001e00 +#define NV40TCL_STENCIL_FRONT_OP_ZPASS_REPLACE 0x00001e01 +#define NV40TCL_STENCIL_FRONT_OP_ZPASS_INCR 0x00001e02 +#define NV40TCL_STENCIL_FRONT_OP_ZPASS_DECR 0x00001e03 +#define NV40TCL_STENCIL_FRONT_OP_ZPASS_INCR_WRAP 0x00008507 +#define NV40TCL_STENCIL_FRONT_OP_ZPASS_DECR_WRAP 0x00008508 +#define NV40TCL_STENCIL_BACK_ENABLE 0x00000348 +#define NV40TCL_STENCIL_BACK_MASK 0x0000034c +#define NV40TCL_STENCIL_BACK_FUNC_FUNC 0x00000350 +#define NV40TCL_STENCIL_BACK_FUNC_FUNC_NEVER 0x00000200 +#define NV40TCL_STENCIL_BACK_FUNC_FUNC_LESS 0x00000201 +#define NV40TCL_STENCIL_BACK_FUNC_FUNC_EQUAL 0x00000202 +#define NV40TCL_STENCIL_BACK_FUNC_FUNC_LEQUAL 0x00000203 +#define NV40TCL_STENCIL_BACK_FUNC_FUNC_GREATER 0x00000204 +#define NV40TCL_STENCIL_BACK_FUNC_FUNC_GREATER 0x00000204 +#define NV40TCL_STENCIL_BACK_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV40TCL_STENCIL_BACK_FUNC_FUNC_GEQUAL 0x00000206 +#define NV40TCL_STENCIL_BACK_FUNC_FUNC_ALWAYS 0x00000207 +#define NV40TCL_STENCIL_BACK_FUNC_REF 0x00000354 +#define NV40TCL_STENCIL_BACK_FUNC_MASK 0x00000358 +#define NV40TCL_STENCIL_BACK_OP_FAIL 0x0000035c +#define NV40TCL_STENCIL_BACK_OP_FAIL_ZERO 0x00000000 +#define NV40TCL_STENCIL_BACK_OP_FAIL_INVERT 0x0000150a +#define NV40TCL_STENCIL_BACK_OP_FAIL_KEEP 0x00001e00 +#define NV40TCL_STENCIL_BACK_OP_FAIL_REPLACE 0x00001e01 +#define NV40TCL_STENCIL_BACK_OP_FAIL_INCR 0x00001e02 +#define NV40TCL_STENCIL_BACK_OP_FAIL_DECR 0x00001e03 +#define NV40TCL_STENCIL_BACK_OP_FAIL_INCR_WRAP 0x00008507 +#define NV40TCL_STENCIL_BACK_OP_FAIL_DECR_WRAP 0x00008508 +#define NV40TCL_STENCIL_BACK_OP_ZFAIL 0x00000360 +#define NV40TCL_STENCIL_BACK_OP_ZFAIL_ZERO 0x00000000 +#define NV40TCL_STENCIL_BACK_OP_ZFAIL_INVERT 0x0000150a +#define NV40TCL_STENCIL_BACK_OP_ZFAIL_KEEP 0x00001e00 +#define NV40TCL_STENCIL_BACK_OP_ZFAIL_REPLACE 0x00001e01 +#define NV40TCL_STENCIL_BACK_OP_ZFAIL_INCR 0x00001e02 +#define NV40TCL_STENCIL_BACK_OP_ZFAIL_DECR 0x00001e03 +#define NV40TCL_STENCIL_BACK_OP_ZFAIL_INCR_WRAP 0x00008507 +#define NV40TCL_STENCIL_BACK_OP_ZFAIL_DECR_WRAP 0x00008508 +#define NV40TCL_STENCIL_BACK_OP_ZPASS 0x00000364 +#define NV40TCL_STENCIL_BACK_OP_ZPASS_ZERO 0x00000000 +#define NV40TCL_STENCIL_BACK_OP_ZPASS_INVERT 0x0000150a +#define NV40TCL_STENCIL_BACK_OP_ZPASS_KEEP 0x00001e00 +#define NV40TCL_STENCIL_BACK_OP_ZPASS_REPLACE 0x00001e01 +#define NV40TCL_STENCIL_BACK_OP_ZPASS_INCR 0x00001e02 +#define NV40TCL_STENCIL_BACK_OP_ZPASS_DECR 0x00001e03 +#define NV40TCL_STENCIL_BACK_OP_ZPASS_INCR_WRAP 0x00008507 +#define NV40TCL_STENCIL_BACK_OP_ZPASS_DECR_WRAP 0x00008508 +#define NV40TCL_SHADE_MODEL 0x00000368 +#define NV40TCL_SHADE_MODEL_FLAT 0x00001d00 +#define NV40TCL_SHADE_MODEL_SMOOTH 0x00001d01 +#define NV40TCL_MRT_COLOR_MASK 0x00000370 +#define NV40TCL_MRT_COLOR_MASK_BUFFER1_A (1 << 4) +#define NV40TCL_MRT_COLOR_MASK_BUFFER1_R (1 << 5) +#define NV40TCL_MRT_COLOR_MASK_BUFFER1_G (1 << 6) +#define NV40TCL_MRT_COLOR_MASK_BUFFER1_B (1 << 7) +#define NV40TCL_MRT_COLOR_MASK_BUFFER2_A (1 << 8) +#define NV40TCL_MRT_COLOR_MASK_BUFFER2_R (1 << 9) +#define NV40TCL_MRT_COLOR_MASK_BUFFER2_G (1 << 10) +#define NV40TCL_MRT_COLOR_MASK_BUFFER2_B (1 << 11) +#define NV40TCL_MRT_COLOR_MASK_BUFFER3_A (1 << 12) +#define NV40TCL_MRT_COLOR_MASK_BUFFER3_R (1 << 13) +#define NV40TCL_MRT_COLOR_MASK_BUFFER3_G (1 << 14) +#define NV40TCL_MRT_COLOR_MASK_BUFFER3_B (1 << 15) +#define NV40TCL_COLOR_LOGIC_OP_ENABLE 0x00000374 +#define NV40TCL_COLOR_LOGIC_OP 0x00000378 +#define NV40TCL_COLOR_LOGIC_OP_CLEAR 0x00001500 +#define NV40TCL_COLOR_LOGIC_OP_AND 0x00001501 +#define NV40TCL_COLOR_LOGIC_OP_AND_REVERSE 0x00001502 +#define NV40TCL_COLOR_LOGIC_OP_COPY 0x00001503 +#define NV40TCL_COLOR_LOGIC_OP_AND_INVERTED 0x00001504 +#define NV40TCL_COLOR_LOGIC_OP_NOOP 0x00001505 +#define NV40TCL_COLOR_LOGIC_OP_XOR 0x00001506 +#define NV40TCL_COLOR_LOGIC_OP_OR 0x00001507 +#define NV40TCL_COLOR_LOGIC_OP_NOR 0x00001508 +#define NV40TCL_COLOR_LOGIC_OP_EQUIV 0x00001509 +#define NV40TCL_COLOR_LOGIC_OP_INVERT 0x0000150a +#define NV40TCL_COLOR_LOGIC_OP_OR_REVERSE 0x0000150b +#define NV40TCL_COLOR_LOGIC_OP_COPY_INVERTED 0x0000150c +#define NV40TCL_COLOR_LOGIC_OP_OR_INVERTED 0x0000150d +#define NV40TCL_COLOR_LOGIC_OP_NAND 0x0000150e +#define NV40TCL_COLOR_LOGIC_OP_SET 0x0000150f +#define NV40TCL_DEPTH_RANGE_NEAR 0x00000394 +#define NV40TCL_DEPTH_RANGE_FAR 0x00000398 +#define NV40TCL_LINE_WIDTH 0x000003b8 +#define NV40TCL_LINE_SMOOTH_ENABLE 0x000003bc +#define NV40TCL_UNK03C0(x) (0x000003c0+((x)*4)) +#define NV40TCL_UNK03C0__SIZE 0x00000010 +#define NV40TCL_UNK0400(x) (0x00000400+((x)*4)) +#define NV40TCL_UNK0400__SIZE 0x00000010 +#define NV40TCL_UNK0440(x) (0x00000440+((x)*4)) +#define NV40TCL_UNK0440__SIZE 0x00000020 +#define NV40TCL_SCISSOR_HORIZ 0x000008c0 +#define NV40TCL_SCISSOR_HORIZ_X_SHIFT 0 +#define NV40TCL_SCISSOR_HORIZ_X_MASK 0x0000ffff +#define NV40TCL_SCISSOR_HORIZ_W_SHIFT 16 +#define NV40TCL_SCISSOR_HORIZ_W_MASK 0xffff0000 +#define NV40TCL_SCISSOR_VERT 0x000008c4 +#define NV40TCL_SCISSOR_VERT_Y_SHIFT 0 +#define NV40TCL_SCISSOR_VERT_Y_MASK 0x0000ffff +#define NV40TCL_SCISSOR_VERT_H_SHIFT 16 +#define NV40TCL_SCISSOR_VERT_H_MASK 0xffff0000 +#define NV40TCL_FOG_MODE 0x000008cc +#define NV40TCL_FOG_EQUATION_CONSTANT 0x000008d0 +#define NV40TCL_FOG_EQUATION_LINEAR 0x000008d4 +#define NV40TCL_FOG_EQUATION_QUADRATIC 0x000008d8 +#define NV40TCL_FP_ADDRESS 0x000008e4 +#define NV40TCL_FP_ADDRESS_OFFSET_SHIFT 8 +#define NV40TCL_FP_ADDRESS_OFFSET_MASK 0xffffff00 +#define NV40TCL_FP_ADDRESS_DMA1 (1 << 1) +#define NV40TCL_FP_ADDRESS_DMA0 (1 << 0) +#define NV40TCL_VIEWPORT_HORIZ 0x00000a00 +#define NV40TCL_VIEWPORT_HORIZ_W_SHIFT 16 +#define NV40TCL_VIEWPORT_HORIZ_W_MASK 0xffff0000 +#define NV40TCL_VIEWPORT_HORIZ_X_SHIFT 0 +#define NV40TCL_VIEWPORT_HORIZ_X_MASK 0x0000ffff +#define NV40TCL_VIEWPORT_VERT 0x00000a04 +#define NV40TCL_VIEWPORT_VERT_H_SHIFT 16 +#define NV40TCL_VIEWPORT_VERT_H_MASK 0xffff0000 +#define NV40TCL_VIEWPORT_VERT_Y_SHIFT 0 +#define NV40TCL_VIEWPORT_VERT_Y_MASK 0x0000ffff +#define NV40TCL_VIEWPORT_TRANSLATE_X 0x00000a20 +#define NV40TCL_VIEWPORT_TRANSLATE_Y 0x00000a24 +#define NV40TCL_VIEWPORT_TRANSLATE_Z 0x00000a28 +#define NV40TCL_VIEWPORT_TRANSLATE_W 0x00000a2c +#define NV40TCL_VIEWPORT_SCALE_X 0x00000a30 +#define NV40TCL_VIEWPORT_SCALE_Y 0x00000a34 +#define NV40TCL_VIEWPORT_SCALE_Z 0x00000a38 +#define NV40TCL_VIEWPORT_SCALE_W 0x00000a3c +#define NV40TCL_POLYGON_OFFSET_POINT_ENABLE 0x00000a60 +#define NV40TCL_POLYGON_OFFSET_LINE_ENABLE 0x00000a64 +#define NV40TCL_POLYGON_OFFSET_FILL_ENABLE 0x00000a68 +#define NV40TCL_DEPTH_FUNC 0x00000a6c +#define NV40TCL_DEPTH_FUNC_NEVER 0x00000200 +#define NV40TCL_DEPTH_FUNC_LESS 0x00000201 +#define NV40TCL_DEPTH_FUNC_EQUAL 0x00000202 +#define NV40TCL_DEPTH_FUNC_LEQUAL 0x00000203 +#define NV40TCL_DEPTH_FUNC_GREATER 0x00000204 +#define NV40TCL_DEPTH_FUNC_GREATER 0x00000204 +#define NV40TCL_DEPTH_FUNC_NOTEQUAL 0x00000205 +#define NV40TCL_DEPTH_FUNC_GEQUAL 0x00000206 +#define NV40TCL_DEPTH_FUNC_ALWAYS 0x00000207 +#define NV40TCL_DEPTH_WRITE_ENABLE 0x00000a70 +#define NV40TCL_DEPTH_TEST_ENABLE 0x00000a74 +#define NV40TCL_POLYGON_OFFSET_FACTOR 0x00000a78 +#define NV40TCL_POLYGON_OFFSET_UNITS 0x00000a7c +#define NV40TCL_VTX_ATTR_3I_XY(x) (0x00000a80+((x)*8)) +#define NV40TCL_VTX_ATTR_3I_XY__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_3I_XY_X_SHIFT 0 +#define NV40TCL_VTX_ATTR_3I_XY_X_MASK 0x0000ffff +#define NV40TCL_VTX_ATTR_3I_XY_Y_SHIFT 16 +#define NV40TCL_VTX_ATTR_3I_XY_Y_MASK 0xffff0000 +#define NV40TCL_VTX_ATTR_3I_Z(x) (0x00000a84+((x)*8)) +#define NV40TCL_VTX_ATTR_3I_Z__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_3I_Z_Z_SHIFT 0 +#define NV40TCL_VTX_ATTR_3I_Z_Z_MASK 0x0000ffff +#define NV40TCL_UNK0B40(x) (0x00000b40+((x)*4)) +#define NV40TCL_UNK0B40__SIZE 0x00000008 +#define NV40TCL_VP_UPLOAD_INST(x) (0x00000b80+((x)*4)) +#define NV40TCL_VP_UPLOAD_INST__SIZE 0x00000004 +#define NV40TCL_CLIP_PLANE_ENABLE 0x00001478 +#define NV40TCL_CLIP_PLANE_ENABLE_PLANE0 (1 << 1) +#define NV40TCL_CLIP_PLANE_ENABLE_PLANE1 (1 << 5) +#define NV40TCL_CLIP_PLANE_ENABLE_PLANE2 (1 << 9) +#define NV40TCL_CLIP_PLANE_ENABLE_PLANE3 (1 << 13) +#define NV40TCL_CLIP_PLANE_ENABLE_PLANE4 (1 << 17) +#define NV40TCL_CLIP_PLANE_ENABLE_PLANE5 (1 << 21) +#define NV40TCL_POLYGON_STIPPLE_ENABLE 0x0000147c +#define NV40TCL_POLYGON_STIPPLE_PATTERN(x) (0x00001480+((x)*4)) +#define NV40TCL_POLYGON_STIPPLE_PATTERN__SIZE 0x00000020 +#define NV40TCL_VTX_ATTR_3F_X(x) (0x00001500+((x)*16)) +#define NV40TCL_VTX_ATTR_3F_X__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_3F_Y(x) (0x00001504+((x)*16)) +#define NV40TCL_VTX_ATTR_3F_Y__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_3F_Z(x) (0x00001508+((x)*16)) +#define NV40TCL_VTX_ATTR_3F_Z__SIZE 0x00000010 +#define NV40TCL_VTXBUF_ADDRESS(x) (0x00001680+((x)*4)) +#define NV40TCL_VTXBUF_ADDRESS__SIZE 0x00000010 +#define NV40TCL_VTXBUF_ADDRESS_DMA1 (1 << 31) +#define NV40TCL_VTXBUF_ADDRESS_OFFSET_SHIFT 0 +#define NV40TCL_VTXBUF_ADDRESS_OFFSET_MASK 0x0fffffff +#define NV40TCL_VTX_CACHE_INVALIDATE 0x00001714 +#define NV40TCL_VTXFMT(x) (0x00001740+((x)*4)) +#define NV40TCL_VTXFMT__SIZE 0x00000010 +#define NV40TCL_VTXFMT_TYPE_SHIFT 0 +#define NV40TCL_VTXFMT_TYPE_MASK 0x0000000f +#define NV40TCL_VTXFMT_TYPE_FLOAT 0x00000002 +#define NV40TCL_VTXFMT_TYPE_UBYTE 0x00000004 +#define NV40TCL_VTXFMT_TYPE_USHORT 0x00000005 +#define NV40TCL_VTXFMT_SIZE_SHIFT 4 +#define NV40TCL_VTXFMT_SIZE_MASK 0x000000f0 +#define NV40TCL_VTXFMT_STRIDE_SHIFT 8 +#define NV40TCL_VTXFMT_STRIDE_MASK 0x0000ff00 +#define NV40TCL_QUERY_RESET 0x000017c8 +#define NV40TCL_QUERY_UNK17CC 0x000017cc +#define NV40TCL_QUERY_GET 0x00001800 +#define NV40TCL_QUERY_GET_UNK24_SHIFT 24 +#define NV40TCL_QUERY_GET_UNK24_MASK 0xff000000 +#define NV40TCL_QUERY_GET_OFFSET_SHIFT 0 +#define NV40TCL_QUERY_GET_OFFSET_MASK 0x00ffffff +#define NV40TCL_BEGIN_END 0x00001808 +#define NV40TCL_BEGIN_END_STOP 0x00000000 +#define NV40TCL_BEGIN_END_POINTS 0x00000001 +#define NV40TCL_BEGIN_END_LINES 0x00000002 +#define NV40TCL_BEGIN_END_LINE_LOOP 0x00000003 +#define NV40TCL_BEGIN_END_LINE_STRIP 0x00000004 +#define NV40TCL_BEGIN_END_TRIANGLES 0x00000005 +#define NV40TCL_BEGIN_END_TRIANGLE_STRIP 0x00000006 +#define NV40TCL_BEGIN_END_TRIANGLE_FAN 0x00000007 +#define NV40TCL_BEGIN_END_QUADS 0x00000008 +#define NV40TCL_BEGIN_END_QUAD_STRIP 0x00000009 +#define NV40TCL_BEGIN_END_POLYGON 0x0000000a +#define NV40TCL_VB_ELEMENT_U16 0x0000180c +#define NV40TCL_VB_ELEMENT_U16_1_SHIFT 16 +#define NV40TCL_VB_ELEMENT_U16_1_MASK 0xffff0000 +#define NV40TCL_VB_ELEMENT_U16_0_SHIFT 0 +#define NV40TCL_VB_ELEMENT_U16_0_MASK 0x0000ffff +#define NV40TCL_VB_ELEMENT_U32 0x00001810 +#define NV40TCL_VB_VERTEX_BATCH 0x00001814 +#define NV40TCL_VB_VERTEX_BATCH_COUNT_SHIFT 24 +#define NV40TCL_VB_VERTEX_BATCH_COUNT_MASK 0xff000000 +#define NV40TCL_VB_VERTEX_BATCH_START_SHIFT 0 +#define NV40TCL_VB_VERTEX_BATCH_START_MASK 0x00ffffff +#define NV40TCL_VERTEX_DATA 0x00001818 +#define NV40TCL_IDXBUF_ADDRESS 0x0000181c +#define NV40TCL_IDXBUF_FORMAT 0x00001820 +#define NV40TCL_IDXBUF_FORMAT_TYPE_SHIFT 4 +#define NV40TCL_IDXBUF_FORMAT_TYPE_MASK 0x000000f0 +#define NV40TCL_IDXBUF_FORMAT_TYPE_U32 0x00000000 +#define NV40TCL_IDXBUF_FORMAT_TYPE_U16 0x00000010 +#define NV40TCL_IDXBUF_FORMAT_DMA1 (1 << 0) +#define NV40TCL_VB_INDEX_BATCH 0x00001824 +#define NV40TCL_VB_INDEX_BATCH_COUNT_SHIFT 24 +#define NV40TCL_VB_INDEX_BATCH_COUNT_MASK 0xff000000 +#define NV40TCL_VB_INDEX_BATCH_START_SHIFT 0 +#define NV40TCL_VB_INDEX_BATCH_START_MASK 0x00ffffff +#define NV40TCL_POLYGON_MODE_FRONT 0x00001828 +#define NV40TCL_POLYGON_MODE_FRONT_POINT 0x00001b00 +#define NV40TCL_POLYGON_MODE_FRONT_LINE 0x00001b01 +#define NV40TCL_POLYGON_MODE_FRONT_FILL 0x00001b02 +#define NV40TCL_POLYGON_MODE_BACK 0x0000182c +#define NV40TCL_POLYGON_MODE_BACK_POINT 0x00001b00 +#define NV40TCL_POLYGON_MODE_BACK_LINE 0x00001b01 +#define NV40TCL_POLYGON_MODE_BACK_FILL 0x00001b02 +#define NV40TCL_CULL_FACE 0x00001830 +#define NV40TCL_CULL_FACE_FRONT 0x00000404 +#define NV40TCL_CULL_FACE_BACK 0x00000405 +#define NV40TCL_CULL_FACE_FRONT_AND_BACK 0x00000408 +#define NV40TCL_FRONT_FACE 0x00001834 +#define NV40TCL_FRONT_FACE_CW 0x00000900 +#define NV40TCL_FRONT_FACE_CCW 0x00000901 +#define NV40TCL_POLYGON_SMOOTH_ENABLE 0x00001838 +#define NV40TCL_CULL_FACE_ENABLE 0x0000183c +#define NV40TCL_TEX_SIZE1(x) (0x00001840+((x)*4)) +#define NV40TCL_TEX_SIZE1__SIZE 0x00000008 +#define NV40TCL_TEX_SIZE1_DEPTH_SHIFT 20 +#define NV40TCL_TEX_SIZE1_DEPTH_MASK 0xfff00000 +#define NV40TCL_TEX_SIZE1_PITCH_SHIFT 0 +#define NV40TCL_TEX_SIZE1_PITCH_MASK 0x0000ffff +#define NV40TCL_VTX_ATTR_2F_X(x) (0x00001880+((x)*8)) +#define NV40TCL_VTX_ATTR_2F_X__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_2F_Y(x) (0x00001884+((x)*8)) +#define NV40TCL_VTX_ATTR_2F_Y__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_2I(x) (0x00001900+((x)*4)) +#define NV40TCL_VTX_ATTR_2I__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_2I_X_SHIFT 0 +#define NV40TCL_VTX_ATTR_2I_X_MASK 0x0000ffff +#define NV40TCL_VTX_ATTR_2I_Y_SHIFT 16 +#define NV40TCL_VTX_ATTR_2I_Y_MASK 0xffff0000 +#define NV40TCL_VTX_ATTR_4UB(x) (0x00001940+((x)*4)) +#define NV40TCL_VTX_ATTR_4UB__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_4UB_X_SHIFT 0 +#define NV40TCL_VTX_ATTR_4UB_X_MASK 0x000000ff +#define NV40TCL_VTX_ATTR_4UB_Y_SHIFT 8 +#define NV40TCL_VTX_ATTR_4UB_Y_MASK 0x0000ff00 +#define NV40TCL_VTX_ATTR_4UB_Z_SHIFT 16 +#define NV40TCL_VTX_ATTR_4UB_Z_MASK 0x00ff0000 +#define NV40TCL_VTX_ATTR_4UB_W_SHIFT 24 +#define NV40TCL_VTX_ATTR_4UB_W_MASK 0xff000000 +#define NV40TCL_VTX_ATTR_4I_XY(x) (0x00001980+((x)*8)) +#define NV40TCL_VTX_ATTR_4I_XY__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_4I_XY_X_SHIFT 0 +#define NV40TCL_VTX_ATTR_4I_XY_X_MASK 0x0000ffff +#define NV40TCL_VTX_ATTR_4I_XY_Y_SHIFT 16 +#define NV40TCL_VTX_ATTR_4I_XY_Y_MASK 0xffff0000 +#define NV40TCL_VTX_ATTR_4I_ZW(x) (0x00001984+((x)*8)) +#define NV40TCL_VTX_ATTR_4I_ZW__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_4I_ZW_Z_SHIFT 0 +#define NV40TCL_VTX_ATTR_4I_ZW_Z_MASK 0x0000ffff +#define NV40TCL_VTX_ATTR_4I_ZW_W_SHIFT 16 +#define NV40TCL_VTX_ATTR_4I_ZW_W_MASK 0xffff0000 +#define NV40TCL_TEX_OFFSET(x) (0x00001a00+((x)*32)) +#define NV40TCL_TEX_OFFSET__SIZE 0x00000010 +#define NV40TCL_TEX_FORMAT(x) (0x00001a04+((x)*32)) +#define NV40TCL_TEX_FORMAT__SIZE 0x00000010 +#define NV40TCL_TEX_FORMAT_MIPMAP_COUNT_SHIFT 16 +#define NV40TCL_TEX_FORMAT_MIPMAP_COUNT_MASK 0x000f0000 +#define NV40TCL_TEX_FORMAT_RECT (1 << 14) +#define NV40TCL_TEX_FORMAT_LINEAR (1 << 13) +#define NV40TCL_TEX_FORMAT_FORMAT_SHIFT 8 +#define NV40TCL_TEX_FORMAT_FORMAT_MASK 0x00001f00 +#define NV40TCL_TEX_FORMAT_FORMAT_L8 0x00000100 +#define NV40TCL_TEX_FORMAT_FORMAT_A1R5G5B5 0x00000200 +#define NV40TCL_TEX_FORMAT_FORMAT_A4R4G4B4 0x00000300 +#define NV40TCL_TEX_FORMAT_FORMAT_R5G6B5 0x00000400 +#define NV40TCL_TEX_FORMAT_FORMAT_A8R8G8B8 0x00000500 +#define NV40TCL_TEX_FORMAT_FORMAT_DXT1 0x00000600 +#define NV40TCL_TEX_FORMAT_FORMAT_DXT3 0x00000700 +#define NV40TCL_TEX_FORMAT_FORMAT_DXT5 0x00000800 +#define NV40TCL_TEX_FORMAT_FORMAT_A8L8 0x00000b00 +#define NV40TCL_TEX_FORMAT_FORMAT_Z24 0x00001000 +#define NV40TCL_TEX_FORMAT_FORMAT_Z16 0x00001200 +#define NV40TCL_TEX_FORMAT_FORMAT_A16 0x00001400 +#define NV40TCL_TEX_FORMAT_FORMAT_A16L16 0x00001500 +#define NV40TCL_TEX_FORMAT_FORMAT_HILO8 0x00001800 +#define NV40TCL_TEX_FORMAT_FORMAT_RGBA16F 0x00001a00 +#define NV40TCL_TEX_FORMAT_FORMAT_RGBA32F 0x00001b00 +#define NV40TCL_TEX_FORMAT_DIMS_SHIFT 4 +#define NV40TCL_TEX_FORMAT_DIMS_MASK 0x000000f0 +#define NV40TCL_TEX_FORMAT_DIMS_1D 0x00000010 +#define NV40TCL_TEX_FORMAT_DIMS_2D 0x00000020 +#define NV40TCL_TEX_FORMAT_DIMS_3D 0x00000030 +#define NV40TCL_TEX_FORMAT_NO_BORDER (1 << 3) +#define NV40TCL_TEX_FORMAT_CUBIC (1 << 2) +#define NV40TCL_TEX_FORMAT_DMA1 (1 << 1) +#define NV40TCL_TEX_FORMAT_DMA0 (1 << 0) +#define NV40TCL_TEX_WRAP(x) (0x00001a08+((x)*32)) +#define NV40TCL_TEX_WRAP__SIZE 0x00000010 +#define NV40TCL_TEX_WRAP_S_SHIFT 0 +#define NV40TCL_TEX_WRAP_S_MASK 0x000000ff +#define NV40TCL_TEX_WRAP_S_REPEAT 0x00000001 +#define NV40TCL_TEX_WRAP_S_MIRRORED_REPEAT 0x00000002 +#define NV40TCL_TEX_WRAP_S_CLAMP_TO_EDGE 0x00000003 +#define NV40TCL_TEX_WRAP_S_CLAMP_TO_BORDER 0x00000004 +#define NV40TCL_TEX_WRAP_S_CLAMP 0x00000005 +#define NV40TCL_TEX_WRAP_S_MIRROR_CLAMP_TO_EDGE 0x00000006 +#define NV40TCL_TEX_WRAP_S_MIRROR_CLAMP_TO_BORDER 0x00000007 +#define NV40TCL_TEX_WRAP_S_MIRROR_CLAMP 0x00000008 +#define NV40TCL_TEX_WRAP_T_SHIFT 8 +#define NV40TCL_TEX_WRAP_T_MASK 0x00000f00 +#define NV40TCL_TEX_WRAP_T_REPEAT 0x00000100 +#define NV40TCL_TEX_WRAP_T_MIRRORED_REPEAT 0x00000200 +#define NV40TCL_TEX_WRAP_T_CLAMP_TO_EDGE 0x00000300 +#define NV40TCL_TEX_WRAP_T_CLAMP_TO_BORDER 0x00000400 +#define NV40TCL_TEX_WRAP_T_CLAMP 0x00000500 +#define NV40TCL_TEX_WRAP_T_MIRROR_CLAMP_TO_EDGE 0x00000600 +#define NV40TCL_TEX_WRAP_T_MIRROR_CLAMP_TO_BORDER 0x00000700 +#define NV40TCL_TEX_WRAP_T_MIRROR_CLAMP 0x00000800 +#define NV40TCL_TEX_WRAP_EXPAND_NORMAL_SHIFT 12 +#define NV40TCL_TEX_WRAP_EXPAND_NORMAL_MASK 0x0000f000 +#define NV40TCL_TEX_WRAP_R_SHIFT 16 +#define NV40TCL_TEX_WRAP_R_MASK 0x00ff0000 +#define NV40TCL_TEX_WRAP_R_REPEAT 0x00010000 +#define NV40TCL_TEX_WRAP_R_MIRRORED_REPEAT 0x00020000 +#define NV40TCL_TEX_WRAP_R_CLAMP_TO_EDGE 0x00030000 +#define NV40TCL_TEX_WRAP_R_CLAMP_TO_BORDER 0x00040000 +#define NV40TCL_TEX_WRAP_R_CLAMP 0x00050000 +#define NV40TCL_TEX_WRAP_R_MIRROR_CLAMP_TO_EDGE 0x00060000 +#define NV40TCL_TEX_WRAP_R_MIRROR_CLAMP_TO_BORDER 0x00070000 +#define NV40TCL_TEX_WRAP_R_MIRROR_CLAMP 0x00080000 +#define NV40TCL_TEX_WRAP_RCOMP_SHIFT 28 +#define NV40TCL_TEX_WRAP_RCOMP_MASK 0xf0000000 +#define NV40TCL_TEX_WRAP_RCOMP_NEVER 0x00000000 +#define NV40TCL_TEX_WRAP_RCOMP_GREATER 0x10000000 +#define NV40TCL_TEX_WRAP_RCOMP_EQUAL 0x20000000 +#define NV40TCL_TEX_WRAP_RCOMP_GEQUAL 0x30000000 +#define NV40TCL_TEX_WRAP_RCOMP_LESS 0x40000000 +#define NV40TCL_TEX_WRAP_RCOMP_NOTEQUAL 0x50000000 +#define NV40TCL_TEX_WRAP_RCOMP_LEQUAL 0x60000000 +#define NV40TCL_TEX_WRAP_RCOMP_ALWAYS 0x70000000 +#define NV40TCL_TEX_ENABLE(x) (0x00001a0c+((x)*32)) +#define NV40TCL_TEX_ENABLE__SIZE 0x00000010 +#define NV40TCL_TEX_ENABLE_ENABLE (1 << 31) +#define NV40TCL_TEX_ENABLE_MIPMAP_MIN_LOD_SHIFT 27 +#define NV40TCL_TEX_ENABLE_MIPMAP_MIN_LOD_MASK 0x38000000 +#define NV40TCL_TEX_ENABLE_MIPMAP_MAX_LOD_SHIFT 15 +#define NV40TCL_TEX_ENABLE_MIPMAP_MAX_LOD_MASK 0x00038000 +#define NV40TCL_TEX_ENABLE_ANISO_SHIFT 4 +#define NV40TCL_TEX_ENABLE_ANISO_MASK 0x000000f0 +#define NV40TCL_TEX_ENABLE_ANISO_NONE 0x00000000 +#define NV40TCL_TEX_ENABLE_ANISO_2X 0x00000010 +#define NV40TCL_TEX_ENABLE_ANISO_4X 0x00000020 +#define NV40TCL_TEX_ENABLE_ANISO_6X 0x00000030 +#define NV40TCL_TEX_ENABLE_ANISO_8X 0x00000040 +#define NV40TCL_TEX_ENABLE_ANISO_10X 0x00000050 +#define NV40TCL_TEX_ENABLE_ANISO_12X 0x00000060 +#define NV40TCL_TEX_ENABLE_ANISO_16X 0x00000070 +#define NV40TCL_TEX_SWIZZLE(x) (0x00001a10+((x)*32)) +#define NV40TCL_TEX_SWIZZLE__SIZE 0x00000010 +#define NV40TCL_TEX_SWIZZLE_S0_X_SHIFT 14 +#define NV40TCL_TEX_SWIZZLE_S0_X_MASK 0x0000c000 +#define NV40TCL_TEX_SWIZZLE_S0_X_ZERO 0x00000000 +#define NV40TCL_TEX_SWIZZLE_S0_X_ONE 0x00004000 +#define NV40TCL_TEX_SWIZZLE_S0_X_S1 0x00008000 +#define NV40TCL_TEX_SWIZZLE_S0_Y_SHIFT 12 +#define NV40TCL_TEX_SWIZZLE_S0_Y_MASK 0x00003000 +#define NV40TCL_TEX_SWIZZLE_S0_Y_ZERO 0x00000000 +#define NV40TCL_TEX_SWIZZLE_S0_Y_ONE 0x00001000 +#define NV40TCL_TEX_SWIZZLE_S0_Y_S1 0x00002000 +#define NV40TCL_TEX_SWIZZLE_S0_Z_SHIFT 10 +#define NV40TCL_TEX_SWIZZLE_S0_Z_MASK 0x00000c00 +#define NV40TCL_TEX_SWIZZLE_S0_Z_ZERO 0x00000000 +#define NV40TCL_TEX_SWIZZLE_S0_Z_ONE 0x00000400 +#define NV40TCL_TEX_SWIZZLE_S0_Z_S1 0x00000800 +#define NV40TCL_TEX_SWIZZLE_S0_W_SHIFT 8 +#define NV40TCL_TEX_SWIZZLE_S0_W_MASK 0x00000300 +#define NV40TCL_TEX_SWIZZLE_S0_W_ZERO 0x00000000 +#define NV40TCL_TEX_SWIZZLE_S0_W_ONE 0x00000100 +#define NV40TCL_TEX_SWIZZLE_S0_W_S1 0x00000200 +#define NV40TCL_TEX_SWIZZLE_S1_X_SHIFT 6 +#define NV40TCL_TEX_SWIZZLE_S1_X_MASK 0x000000c0 +#define NV40TCL_TEX_SWIZZLE_S1_X_W 0x00000000 +#define NV40TCL_TEX_SWIZZLE_S1_X_Z 0x00000040 +#define NV40TCL_TEX_SWIZZLE_S1_X_Y 0x00000080 +#define NV40TCL_TEX_SWIZZLE_S1_X_X 0x000000c0 +#define NV40TCL_TEX_SWIZZLE_S1_Y_SHIFT 4 +#define NV40TCL_TEX_SWIZZLE_S1_Y_MASK 0x00000030 +#define NV40TCL_TEX_SWIZZLE_S1_Y_W 0x00000000 +#define NV40TCL_TEX_SWIZZLE_S1_Y_Z 0x00000010 +#define NV40TCL_TEX_SWIZZLE_S1_Y_Y 0x00000020 +#define NV40TCL_TEX_SWIZZLE_S1_Y_X 0x00000030 +#define NV40TCL_TEX_SWIZZLE_S1_Z_SHIFT 2 +#define NV40TCL_TEX_SWIZZLE_S1_Z_MASK 0x0000000c +#define NV40TCL_TEX_SWIZZLE_S1_Z_W 0x00000000 +#define NV40TCL_TEX_SWIZZLE_S1_Z_Z 0x00000004 +#define NV40TCL_TEX_SWIZZLE_S1_Z_Y 0x00000008 +#define NV40TCL_TEX_SWIZZLE_S1_Z_X 0x0000000c +#define NV40TCL_TEX_SWIZZLE_S1_W_SHIFT 0 +#define NV40TCL_TEX_SWIZZLE_S1_W_MASK 0x00000003 +#define NV40TCL_TEX_SWIZZLE_S1_W_W 0x00000000 +#define NV40TCL_TEX_SWIZZLE_S1_W_Z 0x00000001 +#define NV40TCL_TEX_SWIZZLE_S1_W_Y 0x00000002 +#define NV40TCL_TEX_SWIZZLE_S1_W_X 0x00000003 +#define NV40TCL_TEX_FILTER(x) (0x00001a14+((x)*32)) +#define NV40TCL_TEX_FILTER__SIZE 0x00000010 +#define NV40TCL_TEX_FILTER_SIGNED_ALPHA (1 << 31) +#define NV40TCL_TEX_FILTER_SIGNED_RED (1 << 30) +#define NV40TCL_TEX_FILTER_SIGNED_GREEN (1 << 29) +#define NV40TCL_TEX_FILTER_SIGNED_BLUE (1 << 28) +#define NV40TCL_TEX_FILTER_MIN_SHIFT 16 +#define NV40TCL_TEX_FILTER_MIN_MASK 0x000f0000 +#define NV40TCL_TEX_FILTER_MIN_NEAREST 0x00010000 +#define NV40TCL_TEX_FILTER_MIN_LINEAR 0x00020000 +#define NV40TCL_TEX_FILTER_MIN_NEAREST_MIPMAP_NEAREST 0x00030000 +#define NV40TCL_TEX_FILTER_MIN_LINEAR_MIPMAP_NEAREST 0x00040000 +#define NV40TCL_TEX_FILTER_MIN_NEAREST_MIPMAP_LINEAR 0x00050000 +#define NV40TCL_TEX_FILTER_MIN_LINEAR_MIPMAP_LINEAR 0x00060000 +#define NV40TCL_TEX_FILTER_MAG_SHIFT 24 +#define NV40TCL_TEX_FILTER_MAG_MASK 0x0f000000 +#define NV40TCL_TEX_FILTER_MAG_NEAREST 0x01000000 +#define NV40TCL_TEX_FILTER_MAG_LINEAR 0x02000000 +#define NV40TCL_TEX_SIZE0(x) (0x00001a18+((x)*32)) +#define NV40TCL_TEX_SIZE0__SIZE 0x00000010 +#define NV40TCL_TEX_SIZE0_H_SHIFT 0 +#define NV40TCL_TEX_SIZE0_H_MASK 0x0000ffff +#define NV40TCL_TEX_SIZE0_W_SHIFT 16 +#define NV40TCL_TEX_SIZE0_W_MASK 0xffff0000 +#define NV40TCL_TEX_BORDER_COLOR(x) (0x00001a1c+((x)*32)) +#define NV40TCL_TEX_BORDER_COLOR__SIZE 0x00000010 +#define NV40TCL_TEX_BORDER_COLOR_B_SHIFT 0 +#define NV40TCL_TEX_BORDER_COLOR_B_MASK 0x000000ff +#define NV40TCL_TEX_BORDER_COLOR_G_SHIFT 8 +#define NV40TCL_TEX_BORDER_COLOR_G_MASK 0x0000ff00 +#define NV40TCL_TEX_BORDER_COLOR_R_SHIFT 16 +#define NV40TCL_TEX_BORDER_COLOR_R_MASK 0x00ff0000 +#define NV40TCL_TEX_BORDER_COLOR_A_SHIFT 24 +#define NV40TCL_TEX_BORDER_COLOR_A_MASK 0xff000000 +#define NV40TCL_VTX_ATTR_4F_X(x) (0x00001c00+((x)*16)) +#define NV40TCL_VTX_ATTR_4F_X__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_4F_Y(x) (0x00001c04+((x)*16)) +#define NV40TCL_VTX_ATTR_4F_Y__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_4F_Z(x) (0x00001c08+((x)*16)) +#define NV40TCL_VTX_ATTR_4F_Z__SIZE 0x00000010 +#define NV40TCL_VTX_ATTR_4F_W(x) (0x00001c0c+((x)*16)) +#define NV40TCL_VTX_ATTR_4F_W__SIZE 0x00000010 +#define NV40TCL_FP_CONTROL 0x00001d60 +#define NV40TCL_FP_CONTROL_TEMP_COUNT_SHIFT 24 +#define NV40TCL_FP_CONTROL_TEMP_COUNT_MASK 0xff000000 +#define NV40TCL_FP_CONTROL_KIL (1 << 7) +#define NV40TCL_MULTISAMPLE_CONTROL 0x00001d7c +#define NV40TCL_CLEAR_VALUE_DEPTH 0x00001d8c +#define NV40TCL_CLEAR_VALUE_COLOR 0x00001d90 +#define NV40TCL_CLEAR_BUFFERS 0x00001d94 +#define NV40TCL_CLEAR_BUFFERS_COLOR_A (1 << 7) +#define NV40TCL_CLEAR_BUFFERS_COLOR_B (1 << 6) +#define NV40TCL_CLEAR_BUFFERS_COLOR_G (1 << 5) +#define NV40TCL_CLEAR_BUFFERS_COLOR_R (1 << 4) +#define NV40TCL_CLEAR_BUFFERS_STENCIL (1 << 1) +#define NV40TCL_CLEAR_BUFFERS_DEPTH (1 << 0) +#define NV40TCL_LINE_STIPPLE_ENABLE 0x00001db4 +#define NV40TCL_LINE_STIPPLE_PATTERN 0x00001db8 +#define NV40TCL_LINE_STIPPLE_PATTERN_FACTOR_SHIFT 0 +#define NV40TCL_LINE_STIPPLE_PATTERN_FACTOR_MASK 0x0000ffff +#define NV40TCL_LINE_STIPPLE_PATTERN_PATTERN_SHIFT 16 +#define NV40TCL_LINE_STIPPLE_PATTERN_PATTERN_MASK 0xffff0000 +#define NV40TCL_VTX_ATTR_1F(x) (0x00001e40+((x)*4)) +#define NV40TCL_VTX_ATTR_1F__SIZE 0x00000010 +#define NV40TCL_VP_UPLOAD_FROM_ID 0x00001e9c +#define NV40TCL_VP_START_FROM_ID 0x00001ea0 +#define NV40TCL_POINT_SIZE 0x00001ee0 +#define NV40TCL_POINT_SPRITE 0x00001ee8 +#define NV40TCL_VP_UPLOAD_CONST_ID 0x00001efc +#define NV40TCL_VP_UPLOAD_CONST_X(x) (0x00001f00+((x)*16)) +#define NV40TCL_VP_UPLOAD_CONST_X__SIZE 0x00000004 +#define NV40TCL_VP_UPLOAD_CONST_Y(x) (0x00001f04+((x)*16)) +#define NV40TCL_VP_UPLOAD_CONST_Y__SIZE 0x00000004 +#define NV40TCL_VP_UPLOAD_CONST_Z(x) (0x00001f08+((x)*16)) +#define NV40TCL_VP_UPLOAD_CONST_Z__SIZE 0x00000004 +#define NV40TCL_VP_UPLOAD_CONST_W(x) (0x00001f0c+((x)*16)) +#define NV40TCL_VP_UPLOAD_CONST_W__SIZE 0x00000004 +#define NV40TCL_TEX_CACHE_CTL 0x00001fd8 +#define NV40TCL_VP_ATTRIB_EN 0x00001ff0 +#define NV40TCL_VP_RESULT_EN 0x00001ff4 + + +#define NV44TCL 0x00004497 + + + +#define NV50_2D 0x0000502d + +#define NV50_2D_NOP 0x00000100 +#define NV50_2D_NOTIFY 0x00000104 +#define NV50_2D_DMA_NOTIFY 0x00000180 +#define NV50_2D_DMA_IN_MEMORY0 0x00000184 +#define NV50_2D_DMA_IN_MEMORY1 0x00000188 +#define NV50_2D_DMA_IN_MEMORY2 0x0000018c +#define NV50_2D_DST_FORMAT 0x00000200 +#define NV50_2D_DST_FORMAT_32BPP 0x000000cf +#define NV50_2D_DST_FORMAT_24BPP 0x000000e6 +#define NV50_2D_DST_FORMAT_16BPP 0x000000e8 +#define NV50_2D_DST_FORMAT_8BPP 0x000000f3 +#define NV50_2D_DST_FORMAT_15BPP 0x000000f8 +#define NV50_2D_DST_PITCH 0x00000214 +#define NV50_2D_DST_WIDTH 0x00000218 +#define NV50_2D_DST_HEIGHT 0x0000021c +#define NV50_2D_DST_ADDRESS_HIGH 0x00000220 +#define NV50_2D_DST_ADDRESS_LOW 0x00000224 +#define NV50_2D_SRC_FORMAT 0x00000230 +#define NV50_2D_SRC_FORMAT_32BPP 0x000000cf +#define NV50_2D_SRC_FORMAT_24BPP 0x000000e6 +#define NV50_2D_SRC_FORMAT_16BPP 0x000000e8 +#define NV50_2D_SRC_FORMAT_8BPP 0x000000f3 +#define NV50_2D_SRC_FORMAT_15BPP 0x000000f8 +#define NV50_2D_SRC_PITCH 0x00000244 +#define NV50_2D_SRC_WIDTH 0x00000248 +#define NV50_2D_SRC_HEIGHT 0x0000024c +#define NV50_2D_SRC_ADDRESS_HIGH 0x00000250 +#define NV50_2D_SRC_ADDRESS_LOW 0x00000254 +#define NV50_2D_CLIP_X 0x00000280 +#define NV50_2D_CLIP_Y 0x00000284 +#define NV50_2D_CLIP_Z 0x00000288 +#define NV50_2D_CLIP_W 0x0000028c +#define NV50_2D_ROP 0x000002a0 +#define NV50_2D_OPERATION 0x000002ac +#define NV50_2D_OPERATION_SRCCOPY_AND 0x00000000 +#define NV50_2D_OPERATION_ROP_AND 0x00000001 +#define NV50_2D_OPERATION_BLEND_AND 0x00000002 +#define NV50_2D_OPERATION_SRCCOPY 0x00000003 +#define NV50_2D_OPERATION_SRCCOPY_PREMULT 0x00000004 +#define NV50_2D_OPERATION_BLEND_PREMULT 0x00000005 +#define NV50_2D_PATTERN_FORMAT 0x000002e8 +#define NV50_2D_PATTERN_FORMAT_16BPP 0x00000000 +#define NV50_2D_PATTERN_FORMAT_15BPP 0x00000001 +#define NV50_2D_PATTERN_FORMAT_32BPP 0x00000002 +#define NV50_2D_PATTERN_FORMAT_8BPP 0x00000003 +#define NV50_2D_PATTERN_COLOR(x) (0x000002f0+((x)*4)) +#define NV50_2D_PATTERN_COLOR__SIZE 0x00000002 +#define NV50_2D_PATTERN_BITMAP(x) (0x000002f8+((x)*4)) +#define NV50_2D_PATTERN_BITMAP__SIZE 0x00000002 +#define NV50_2D_RECT_FORMAT 0x00000584 +#define NV50_2D_RECT_FORMAT_32BPP 0x000000cf +#define NV50_2D_RECT_FORMAT_24BPP 0x000000e6 +#define NV50_2D_RECT_FORMAT_16BPP 0x000000e8 +#define NV50_2D_RECT_FORMAT_8BPP 0x000000f3 +#define NV50_2D_RECT_FORMAT_15BPP 0x000000f8 +#define NV50_2D_RECT_COLOR 0x00000588 +#define NV50_2D_RECT_X1 0x00000600 +#define NV50_2D_RECT_Y1 0x00000604 +#define NV50_2D_RECT_X2 0x00000608 +#define NV50_2D_RECT_Y2 0x0000060c +#define NV50_2D_SIFC_UNK0800 0x00000800 +#define NV50_2D_SIFC_FORMAT 0x00000804 +#define NV50_2D_SIFC_FORMAT_32BPP 0x000000cf +#define NV50_2D_SIFC_FORMAT_24BPP 0x000000e6 +#define NV50_2D_SIFC_FORMAT_16BPP 0x000000e8 +#define NV50_2D_SIFC_FORMAT_8BPP 0x000000f3 +#define NV50_2D_SIFC_FORMAT_15BPP 0x000000f8 +#define NV50_2D_SIFC_WIDTH 0x00000838 +#define NV50_2D_SIFC_HEIGHT 0x0000083c +#define NV50_2D_SIFC_SCALE_UNK0840 0x00000840 +#define NV50_2D_SIFC_SCALE_UNK0844 0x00000844 +#define NV50_2D_SIFC_SCALE_UNK0848 0x00000848 +#define NV50_2D_SIFC_SCALE_UNK084C 0x0000084c +#define NV50_2D_SIFC_UNK0850 0x00000850 +#define NV50_2D_SIFC_DST_X 0x00000854 +#define NV50_2D_SIFC_UNK0858 0x00000858 +#define NV50_2D_SIFC_DST_Y 0x0000085c +#define NV50_2D_SIFC_DATA 0x00000860 +#define NV50_2D_BLIT_DST_X 0x000008b0 +#define NV50_2D_BLIT_DST_Y 0x000008b4 +#define NV50_2D_BLIT_DST_W 0x000008b8 +#define NV50_2D_BLIT_DST_H 0x000008bc +#define NV50_2D_BLIT_SRC_X 0x000008d4 +#define NV50_2D_BLIT_SRC_Y 0x000008dc + + +#define NV50_MEMORY_TO_MEMORY_FORMAT 0x00005039 + +#define NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN_HIGH 0x00000238 +#define NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_OUT_HIGH 0x0000023c + + +#define NV50TCL 0x00005097 + +#define NV50TCL_NOP 0x00000100 +#define NV50TCL_NOTIFY 0x00000104 +#define NV50TCL_DMA_NOTIFY 0x00000180 +#define NV50TCL_DMA_UNK0(x) (0x00000184+((x)*4)) +#define NV50TCL_DMA_UNK0__SIZE 0x0000000b +#define NV50TCL_DMA_UNK1(x) (0x000001c0+((x)*4)) +#define NV50TCL_DMA_UNK1__SIZE 0x00000008 +#define NV50TCL_RT_ADDRESS_HIGH(x) (0x00000200+((x)*32)) +#define NV50TCL_RT_ADDRESS_HIGH__SIZE 0x00000008 +#define NV50TCL_RT_ADDRESS_LOW(x) (0x00000204+((x)*32)) +#define NV50TCL_RT_ADDRESS_LOW__SIZE 0x00000008 +#define NV50TCL_RT_FORMAT(x) (0x00000208+((x)*32)) +#define NV50TCL_RT_FORMAT__SIZE 0x00000008 +#define NV50TCL_RT_FORMAT_32BPP 0x000000cf +#define NV50TCL_RT_FORMAT_24BPP 0x000000e6 +#define NV50TCL_RT_FORMAT_16BPP 0x000000e8 +#define NV50TCL_RT_FORMAT_8BPP 0x000000f3 +#define NV50TCL_RT_FORMAT_15BPP 0x000000f8 +#define NV50TCL_RT_TILE_UNK(x) (0x0000020c+((x)*32)) +#define NV50TCL_RT_TILE_UNK__SIZE 0x00000008 +#define NV50TCL_RT_UNK4(x) (0x00000210+((x)*32)) +#define NV50TCL_RT_UNK4__SIZE 0x00000008 +#define NV50TCL_VTX_ATTR_1F(x) (0x00000300+((x)*4)) +#define NV50TCL_VTX_ATTR_1F__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_2F_X(x) (0x00000380+((x)*8)) +#define NV50TCL_VTX_ATTR_2F_X__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_2F_Y(x) (0x00000384+((x)*8)) +#define NV50TCL_VTX_ATTR_2F_Y__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_3F_X(x) (0x00000400+((x)*16)) +#define NV50TCL_VTX_ATTR_3F_X__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_3F_Y(x) (0x00000404+((x)*16)) +#define NV50TCL_VTX_ATTR_3F_Y__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_3F_Z(x) (0x00000408+((x)*16)) +#define NV50TCL_VTX_ATTR_3F_Z__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_3F_W(x) (0x0000040c+((x)*16)) +#define NV50TCL_VTX_ATTR_3F_W__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_4F_X(x) (0x00000500+((x)*16)) +#define NV50TCL_VTX_ATTR_4F_X__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_4F_Y(x) (0x00000504+((x)*16)) +#define NV50TCL_VTX_ATTR_4F_Y__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_4F_Z(x) (0x00000508+((x)*16)) +#define NV50TCL_VTX_ATTR_4F_Z__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_4F_W(x) (0x0000050c+((x)*16)) +#define NV50TCL_VTX_ATTR_4F_W__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_2I(x) (0x00000680+((x)*4)) +#define NV50TCL_VTX_ATTR_2I__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_2I_X_SHIFT 0 +#define NV50TCL_VTX_ATTR_2I_X_MASK 0x0000ffff +#define NV50TCL_VTX_ATTR_2I_Y_SHIFT 16 +#define NV50TCL_VTX_ATTR_2I_Y_MASK 0xffff0000 +#define NV50TCL_VTX_ATTR_4I_0(x) (0x00000700+((x)*8)) +#define NV50TCL_VTX_ATTR_4I_0__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_4I_0_X_SHIFT 0 +#define NV50TCL_VTX_ATTR_4I_0_X_MASK 0x0000ffff +#define NV50TCL_VTX_ATTR_4I_0_Y_SHIFT 16 +#define NV50TCL_VTX_ATTR_4I_0_Y_MASK 0xffff0000 +#define NV50TCL_VTX_ATTR_4I_1(x) (0x00000704+((x)*8)) +#define NV50TCL_VTX_ATTR_4I_1__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_4I_1_Z_SHIFT 0 +#define NV50TCL_VTX_ATTR_4I_1_Z_MASK 0x0000ffff +#define NV50TCL_VTX_ATTR_4I_1_W_SHIFT 16 +#define NV50TCL_VTX_ATTR_4I_1_W_MASK 0xffff0000 +#define NV50TCL_VTX_ATTR_4NI_0(x) (0x00000780+((x)*8)) +#define NV50TCL_VTX_ATTR_4NI_0__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_4NI_0_X_SHIFT 0 +#define NV50TCL_VTX_ATTR_4NI_0_X_MASK 0x0000ffff +#define NV50TCL_VTX_ATTR_4NI_0_Y_SHIFT 16 +#define NV50TCL_VTX_ATTR_4NI_0_Y_MASK 0xffff0000 +#define NV50TCL_VTX_ATTR_4NI_1(x) (0x00000784+((x)*8)) +#define NV50TCL_VTX_ATTR_4NI_1__SIZE 0x00000010 +#define NV50TCL_VTX_ATTR_4NI_1_Z_SHIFT 0 +#define NV50TCL_VTX_ATTR_4NI_1_Z_MASK 0x0000ffff +#define NV50TCL_VTX_ATTR_4NI_1_W_SHIFT 16 +#define NV50TCL_VTX_ATTR_4NI_1_W_MASK 0xffff0000 +#define NV50TCL_VERTEX_ARRAY_FORMAT(x) (0x00000900+((x)*16)) +#define NV50TCL_VERTEX_ARRAY_FORMAT__SIZE 0x00000010 +#define NV50TCL_VIEWPORT_UNK0(x) (0x00000a00+((x)*4)) +#define NV50TCL_VIEWPORT_UNK0__SIZE 0x00000003 +#define NV50TCL_VIEWPORT_UNK1(x) (0x00000a0c+((x)*4)) +#define NV50TCL_VIEWPORT_UNK1__SIZE 0x00000003 +#define NV50TCL_VIEWPORT_HORIZ 0x00000c00 +#define NV50TCL_VIEWPORT_HORIZ_X_SHIFT 0 +#define NV50TCL_VIEWPORT_HORIZ_X_MASK 0x0000ffff +#define NV50TCL_VIEWPORT_HORIZ_W_SHIFT 16 +#define NV50TCL_VIEWPORT_HORIZ_W_MASK 0xffff0000 +#define NV50TCL_VIEWPORT_VERT 0x00000c04 +#define NV50TCL_VIEWPORT_VERT_Y_SHIFT 0 +#define NV50TCL_VIEWPORT_VERT_Y_MASK 0x0000ffff +#define NV50TCL_VIEWPORT_VERT_H_SHIFT 16 +#define NV50TCL_VIEWPORT_VERT_H_MASK 0xffff0000 +#define NV50TCL_DEPTH_RANGE_NEAR 0x00000c08 +#define NV50TCL_DEPTH_RANGE_FAR 0x00000c0c +#define NV50TCL_VIEWPORT_CLIP_HORIZ(x) (0x00000d00+((x)*8)) +#define NV50TCL_VIEWPORT_CLIP_HORIZ__SIZE 0x00000008 +#define NV50TCL_VIEWPORT_CLIP_VERT(x) (0x00000d04+((x)*8)) +#define NV50TCL_VIEWPORT_CLIP_VERT__SIZE 0x00000008 +#define NV50TCL_VERTEX_BUFFER_FIRST 0x00000d74 +#define NV50TCL_VERTEX_BUFFER_COUNT 0x00000d78 +#define NV50TCL_CLEAR_COLOR(x) (0x00000d80+((x)*4)) +#define NV50TCL_CLEAR_COLOR__SIZE 0x00000004 +#define NV50TCL_CLEAR_DEPTH 0x00000d90 +#define NV50TCL_CLEAR_STENCIL 0x00000da0 +#define NV50TCL_POLYGON_MODE_FRONT 0x00000dac +#define NV50TCL_POLYGON_MODE_FRONT_POINT 0x00001b00 +#define NV50TCL_POLYGON_MODE_FRONT_LINE 0x00001b01 +#define NV50TCL_POLYGON_MODE_FRONT_FILL 0x00001b02 +#define NV50TCL_POLYGON_MODE_BACK 0x00000db0 +#define NV50TCL_POLYGON_MODE_BACK_POINT 0x00001b00 +#define NV50TCL_POLYGON_MODE_BACK_LINE 0x00001b01 +#define NV50TCL_POLYGON_MODE_BACK_FILL 0x00001b02 +#define NV50TCL_POLYGON_SMOOTH_ENABLE 0x00000db4 +#define NV50TCL_POLYGON_OFFSET_POINT_ENABLE 0x00000dc0 +#define NV50TCL_POLYGON_OFFSET_LINE_ENABLE 0x00000dc4 +#define NV50TCL_POLYGON_OFFSET_FILL_ENABLE 0x00000dc8 +#define NV50TCL_SCISSOR_HORIZ 0x00000e04 +#define NV50TCL_SCISSOR_HORIZ_L_SHIFT 0 +#define NV50TCL_SCISSOR_HORIZ_L_MASK 0x0000ffff +#define NV50TCL_SCISSOR_HORIZ_R_SHIFT 16 +#define NV50TCL_SCISSOR_HORIZ_R_MASK 0xffff0000 +#define NV50TCL_SCISSOR_VERT 0x00000e08 +#define NV50TCL_SCISSOR_VERT_T_SHIFT 0 +#define NV50TCL_SCISSOR_VERT_T_MASK 0x0000ffff +#define NV50TCL_SCISSOR_VERT_B_SHIFT 16 +#define NV50TCL_SCISSOR_VERT_B_MASK 0xffff0000 +#define NV50TCL_CB_ADDR 0x00000f00 +#define NV50TCL_CB_ADDR_ID_SHIFT 8 +#define NV50TCL_CB_ADDR_ID_MASK 0xffffff00 +#define NV50TCL_CB_ADDR_BUFFER_SHIFT 0 +#define NV50TCL_CB_ADDR_BUFFER_MASK 0x000000ff +#define NV50TCL_CB_DATA(x) (0x00000f04+((x)*4)) +#define NV50TCL_CB_DATA__SIZE 0x00000010 +#define NV50TCL_STENCIL_FRONT_FUNC_REF 0x00000f54 +#define NV50TCL_STENCIL_FRONT_MASK 0x00000f58 +#define NV50TCL_STENCIL_FRONT_FUNC_MASK 0x00000f5c +#define NV50TCL_GP_ADDRESS_HIGH 0x00000f70 +#define NV50TCL_GP_ADDRESS_LOW 0x00000f74 +#define NV50TCL_VP_ADDRESS_HIGH 0x00000f7c +#define NV50TCL_VP_ADDRESS_LOW 0x00000f80 +#define NV50TCL_FP_ADDRESS_HIGH 0x00000fa4 +#define NV50TCL_FP_ADDRESS_LOW 0x00000fa8 +#define NV50TCL_ZETA_ADDRESS_HIGH 0x00000fe0 +#define NV50TCL_ZETA_ADDRESS_LOW 0x00000fe4 +#define NV50TCL_UNKFF4 0x00000ff4 +#define NV50TCL_UNKFF4_W_SHIFT 16 +#define NV50TCL_UNKFF4_W_MASK 0xffff0000 +#define NV50TCL_UNKFF8 0x00000ff8 +#define NV50TCL_UNKFF8_H_SHIFT 16 +#define NV50TCL_UNKFF8_H_MASK 0xffff0000 +#define NV50TCL_RT_HORIZ(x) (0x00001240+((x)*8)) +#define NV50TCL_RT_HORIZ__SIZE 0x00000008 +#define NV50TCL_RT_VERT(x) (0x00001244+((x)*8)) +#define NV50TCL_RT_VERT__SIZE 0x00000008 +#define NV50TCL_CB_DEF_ADDRESS_HIGH 0x00001280 +#define NV50TCL_CB_DEF_ADDRESS_LOW 0x00001284 +#define NV50TCL_CB_DEF_SET 0x00001288 +#define NV50TCL_CB_DEF_SET_SIZE_SHIFT 0 +#define NV50TCL_CB_DEF_SET_SIZE_MASK 0x0000ffff +#define NV50TCL_CB_DEF_SET_BUFFER_SHIFT 16 +#define NV50TCL_CB_DEF_SET_BUFFER_MASK 0xffff0000 +#define NV50TCL_DEPTH_TEST_ENABLE 0x000012cc +#define NV50TCL_SHADE_MODEL 0x000012d4 +#define NV50TCL_SHADE_MODEL_FLAT 0x00001d00 +#define NV50TCL_SHADE_MODEL_SMOOTH 0x00001d01 +#define NV50TCL_DEPTH_WRITE_ENABLE 0x000012e8 +#define NV50TCL_ALPHA_TEST_ENABLE 0x000012ec +#define NV50TCL_DEPTH_TEST_FUNC 0x0000130c +#define NV50TCL_DEPTH_TEST_FUNC_NEVER 0x00000200 +#define NV50TCL_DEPTH_TEST_FUNC_LESS 0x00000201 +#define NV50TCL_DEPTH_TEST_FUNC_EQUAL 0x00000202 +#define NV50TCL_DEPTH_TEST_FUNC_LEQUAL 0x00000203 +#define NV50TCL_DEPTH_TEST_FUNC_GREATER 0x00000204 +#define NV50TCL_DEPTH_TEST_FUNC_GREATER 0x00000204 +#define NV50TCL_DEPTH_TEST_FUNC_NOTEQUAL 0x00000205 +#define NV50TCL_DEPTH_TEST_FUNC_GEQUAL 0x00000206 +#define NV50TCL_DEPTH_TEST_FUNC_ALWAYS 0x00000207 +#define NV50TCL_ALPHA_TEST_REF 0x00001310 +#define NV50TCL_ALPHA_TEST_FUNC 0x00001314 +#define NV50TCL_ALPHA_TEST_FUNC_NEVER 0x00000200 +#define NV50TCL_ALPHA_TEST_FUNC_LESS 0x00000201 +#define NV50TCL_ALPHA_TEST_FUNC_EQUAL 0x00000202 +#define NV50TCL_ALPHA_TEST_FUNC_LEQUAL 0x00000203 +#define NV50TCL_ALPHA_TEST_FUNC_GREATER 0x00000204 +#define NV50TCL_ALPHA_TEST_FUNC_GREATER 0x00000204 +#define NV50TCL_ALPHA_TEST_FUNC_NOTEQUAL 0x00000205 +#define NV50TCL_ALPHA_TEST_FUNC_GEQUAL 0x00000206 +#define NV50TCL_ALPHA_TEST_FUNC_ALWAYS 0x00000207 +#define NV50TCL_BLEND_COLOR(x) (0x0000131c+((x)*4)) +#define NV50TCL_BLEND_COLOR__SIZE 0x00000004 +#define NV50TCL_BLEND_EQUATION_RGB 0x00001340 +#define NV50TCL_BLEND_EQUATION_RGB_FUNC_ADD 0x00008006 +#define NV50TCL_BLEND_EQUATION_RGB_MIN 0x00008007 +#define NV50TCL_BLEND_EQUATION_RGB_MAX 0x00008008 +#define NV50TCL_BLEND_EQUATION_RGB_FUNC_SUBTRACT 0x0000800a +#define NV50TCL_BLEND_EQUATION_RGB_FUNC_REVERSE_SUBTRACT 0x0000800b +#define NV50TCL_BLEND_FUNC_SRC_RGB 0x00001344 +#define NV50TCL_BLEND_FUNC_SRC_RGB_ZERO 0x00000000 +#define NV50TCL_BLEND_FUNC_SRC_RGB_ONE 0x00000001 +#define NV50TCL_BLEND_FUNC_SRC_RGB_SRC_COLOR 0x00000300 +#define NV50TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV50TCL_BLEND_FUNC_SRC_RGB_SRC_ALPHA 0x00000302 +#define NV50TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV50TCL_BLEND_FUNC_SRC_RGB_DST_ALPHA 0x00000304 +#define NV50TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV50TCL_BLEND_FUNC_SRC_RGB_DST_COLOR 0x00000306 +#define NV50TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_DST_COLOR 0x00000307 +#define NV50TCL_BLEND_FUNC_SRC_RGB_SRC_ALPHA_SATURATE 0x00000308 +#define NV50TCL_BLEND_FUNC_SRC_RGB_CONSTANT_COLOR 0x00008001 +#define NV50TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV50TCL_BLEND_FUNC_SRC_RGB_CONSTANT_ALPHA 0x00008003 +#define NV50TCL_BLEND_FUNC_SRC_RGB_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV50TCL_BLEND_FUNC_DST_RGB 0x00001348 +#define NV50TCL_BLEND_FUNC_DST_RGB_ZERO 0x00000000 +#define NV50TCL_BLEND_FUNC_DST_RGB_ONE 0x00000001 +#define NV50TCL_BLEND_FUNC_DST_RGB_SRC_COLOR 0x00000300 +#define NV50TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV50TCL_BLEND_FUNC_DST_RGB_SRC_ALPHA 0x00000302 +#define NV50TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV50TCL_BLEND_FUNC_DST_RGB_DST_ALPHA 0x00000304 +#define NV50TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV50TCL_BLEND_FUNC_DST_RGB_DST_COLOR 0x00000306 +#define NV50TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_DST_COLOR 0x00000307 +#define NV50TCL_BLEND_FUNC_DST_RGB_SRC_ALPHA_SATURATE 0x00000308 +#define NV50TCL_BLEND_FUNC_DST_RGB_CONSTANT_COLOR 0x00008001 +#define NV50TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV50TCL_BLEND_FUNC_DST_RGB_CONSTANT_ALPHA 0x00008003 +#define NV50TCL_BLEND_FUNC_DST_RGB_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV50TCL_BLEND_EQUATION_ALPHA 0x0000134c +#define NV50TCL_BLEND_EQUATION_ALPHA_FUNC_ADD 0x00008006 +#define NV50TCL_BLEND_EQUATION_ALPHA_MIN 0x00008007 +#define NV50TCL_BLEND_EQUATION_ALPHA_MAX 0x00008008 +#define NV50TCL_BLEND_EQUATION_ALPHA_FUNC_SUBTRACT 0x0000800a +#define NV50TCL_BLEND_EQUATION_ALPHA_FUNC_REVERSE_SUBTRACT 0x0000800b +#define NV50TCL_BLEND_FUNC_SRC_ALPHA 0x00001350 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_ZERO 0x00000000 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_ONE 0x00000001 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_SRC_COLOR 0x00000300 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_SRC_ALPHA 0x00000302 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_DST_ALPHA 0x00000304 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_DST_COLOR 0x00000306 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_DST_COLOR 0x00000307 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_SRC_ALPHA_SATURATE 0x00000308 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_CONSTANT_COLOR 0x00008001 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_CONSTANT_ALPHA 0x00008003 +#define NV50TCL_BLEND_FUNC_SRC_ALPHA_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV50TCL_BLEND_FUNC_DST_ALPHA 0x00001358 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_ZERO 0x00000000 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_ONE 0x00000001 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_SRC_COLOR 0x00000300 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_SRC_COLOR 0x00000301 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_SRC_ALPHA 0x00000302 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_SRC_ALPHA 0x00000303 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_DST_ALPHA 0x00000304 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_DST_ALPHA 0x00000305 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_DST_COLOR 0x00000306 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_DST_COLOR 0x00000307 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_SRC_ALPHA_SATURATE 0x00000308 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_CONSTANT_COLOR 0x00008001 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_CONSTANT_COLOR 0x00008002 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_CONSTANT_ALPHA 0x00008003 +#define NV50TCL_BLEND_FUNC_DST_ALPHA_ONE_MINUS_CONSTANT_ALPHA 0x00008004 +#define NV50TCL_BLEND_ENABLE(x) (0x00001360+((x)*4)) +#define NV50TCL_BLEND_ENABLE__SIZE 0x00000008 +#define NV50TCL_STENCIL_BACK_ENABLE 0x00001380 +#define NV50TCL_STENCIL_BACK_OP_FAIL 0x00001384 +#define NV50TCL_STENCIL_BACK_OP_FAIL_ZERO 0x00000000 +#define NV50TCL_STENCIL_BACK_OP_FAIL_INVERT 0x0000150a +#define NV50TCL_STENCIL_BACK_OP_FAIL_KEEP 0x00001e00 +#define NV50TCL_STENCIL_BACK_OP_FAIL_REPLACE 0x00001e01 +#define NV50TCL_STENCIL_BACK_OP_FAIL_INCR 0x00001e02 +#define NV50TCL_STENCIL_BACK_OP_FAIL_DECR 0x00001e03 +#define NV50TCL_STENCIL_BACK_OP_FAIL_INCR_WRAP 0x00008507 +#define NV50TCL_STENCIL_BACK_OP_FAIL_DECR_WRAP 0x00008508 +#define NV50TCL_STENCIL_BACK_OP_ZFAIL 0x00001388 +#define NV50TCL_STENCIL_BACK_OP_ZFAIL_ZERO 0x00000000 +#define NV50TCL_STENCIL_BACK_OP_ZFAIL_INVERT 0x0000150a +#define NV50TCL_STENCIL_BACK_OP_ZFAIL_KEEP 0x00001e00 +#define NV50TCL_STENCIL_BACK_OP_ZFAIL_REPLACE 0x00001e01 +#define NV50TCL_STENCIL_BACK_OP_ZFAIL_INCR 0x00001e02 +#define NV50TCL_STENCIL_BACK_OP_ZFAIL_DECR 0x00001e03 +#define NV50TCL_STENCIL_BACK_OP_ZFAIL_INCR_WRAP 0x00008507 +#define NV50TCL_STENCIL_BACK_OP_ZFAIL_DECR_WRAP 0x00008508 +#define NV50TCL_STENCIL_BACK_OP_ZPASS 0x0000138c +#define NV50TCL_STENCIL_BACK_OP_ZPASS_ZERO 0x00000000 +#define NV50TCL_STENCIL_BACK_OP_ZPASS_INVERT 0x0000150a +#define NV50TCL_STENCIL_BACK_OP_ZPASS_KEEP 0x00001e00 +#define NV50TCL_STENCIL_BACK_OP_ZPASS_REPLACE 0x00001e01 +#define NV50TCL_STENCIL_BACK_OP_ZPASS_INCR 0x00001e02 +#define NV50TCL_STENCIL_BACK_OP_ZPASS_DECR 0x00001e03 +#define NV50TCL_STENCIL_BACK_OP_ZPASS_INCR_WRAP 0x00008507 +#define NV50TCL_STENCIL_BACK_OP_ZPASS_DECR_WRAP 0x00008508 +#define NV50TCL_STENCIL_BACK_FUNC_FUNC 0x00001390 +#define NV50TCL_STENCIL_BACK_FUNC_FUNC_NEVER 0x00000200 +#define NV50TCL_STENCIL_BACK_FUNC_FUNC_LESS 0x00000201 +#define NV50TCL_STENCIL_BACK_FUNC_FUNC_EQUAL 0x00000202 +#define NV50TCL_STENCIL_BACK_FUNC_FUNC_LEQUAL 0x00000203 +#define NV50TCL_STENCIL_BACK_FUNC_FUNC_GREATER 0x00000204 +#define NV50TCL_STENCIL_BACK_FUNC_FUNC_GREATER 0x00000204 +#define NV50TCL_STENCIL_BACK_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV50TCL_STENCIL_BACK_FUNC_FUNC_GEQUAL 0x00000206 +#define NV50TCL_STENCIL_BACK_FUNC_FUNC_ALWAYS 0x00000207 +#define NV50TCL_STENCIL_BACK_FUNC_REF 0x00001394 +#define NV50TCL_STENCIL_BACK_MASK 0x00001398 +#define NV50TCL_STENCIL_BACK_FUNC_MASK 0x0000139c +#define NV50TCL_LINE_WIDTH 0x000013b0 +#define NV50TCL_VP_START_ID 0x0000140c +#define NV50TCL_GP_START_ID 0x00001410 +#define NV50TCL_FP_START_ID 0x00001414 +#define NV50TCL_POINT_SIZE 0x00001518 +#define NV50TCL_TSC_ADDRESS_HIGH 0x0000155c +#define NV50TCL_TSC_ADDRESS_LOW 0x00001560 +#define NV50TCL_POLYGON_OFFSET_FACTOR 0x0000156c +#define NV50TCL_LINE_SMOOTH_ENABLE 0x00001570 +#define NV50TCL_TIC_ADDRESS_HIGH 0x00001574 +#define NV50TCL_TIC_ADDRESS_LOW 0x00001578 +#define NV50TCL_STENCIL_FRONT_ENABLE 0x00001594 +#define NV50TCL_STENCIL_FRONT_OP_FAIL 0x00001598 +#define NV50TCL_STENCIL_FRONT_OP_FAIL_ZERO 0x00000000 +#define NV50TCL_STENCIL_FRONT_OP_FAIL_INVERT 0x0000150a +#define NV50TCL_STENCIL_FRONT_OP_FAIL_KEEP 0x00001e00 +#define NV50TCL_STENCIL_FRONT_OP_FAIL_REPLACE 0x00001e01 +#define NV50TCL_STENCIL_FRONT_OP_FAIL_INCR 0x00001e02 +#define NV50TCL_STENCIL_FRONT_OP_FAIL_DECR 0x00001e03 +#define NV50TCL_STENCIL_FRONT_OP_FAIL_INCR_WRAP 0x00008507 +#define NV50TCL_STENCIL_FRONT_OP_FAIL_DECR_WRAP 0x00008508 +#define NV50TCL_STENCIL_FRONT_OP_ZFAIL 0x0000159c +#define NV50TCL_STENCIL_FRONT_OP_ZFAIL_ZERO 0x00000000 +#define NV50TCL_STENCIL_FRONT_OP_ZFAIL_INVERT 0x0000150a +#define NV50TCL_STENCIL_FRONT_OP_ZFAIL_KEEP 0x00001e00 +#define NV50TCL_STENCIL_FRONT_OP_ZFAIL_REPLACE 0x00001e01 +#define NV50TCL_STENCIL_FRONT_OP_ZFAIL_INCR 0x00001e02 +#define NV50TCL_STENCIL_FRONT_OP_ZFAIL_DECR 0x00001e03 +#define NV50TCL_STENCIL_FRONT_OP_ZFAIL_INCR_WRAP 0x00008507 +#define NV50TCL_STENCIL_FRONT_OP_ZFAIL_DECR_WRAP 0x00008508 +#define NV50TCL_STENCIL_FRONT_OP_ZPASS 0x000015a0 +#define NV50TCL_STENCIL_FRONT_OP_ZPASS_ZERO 0x00000000 +#define NV50TCL_STENCIL_FRONT_OP_ZPASS_INVERT 0x0000150a +#define NV50TCL_STENCIL_FRONT_OP_ZPASS_KEEP 0x00001e00 +#define NV50TCL_STENCIL_FRONT_OP_ZPASS_REPLACE 0x00001e01 +#define NV50TCL_STENCIL_FRONT_OP_ZPASS_INCR 0x00001e02 +#define NV50TCL_STENCIL_FRONT_OP_ZPASS_DECR 0x00001e03 +#define NV50TCL_STENCIL_FRONT_OP_ZPASS_INCR_WRAP 0x00008507 +#define NV50TCL_STENCIL_FRONT_OP_ZPASS_DECR_WRAP 0x00008508 +#define NV50TCL_STENCIL_FRONT_FUNC_FUNC 0x000015a4 +#define NV50TCL_STENCIL_FRONT_FUNC_FUNC_NEVER 0x00000200 +#define NV50TCL_STENCIL_FRONT_FUNC_FUNC_LESS 0x00000201 +#define NV50TCL_STENCIL_FRONT_FUNC_FUNC_EQUAL 0x00000202 +#define NV50TCL_STENCIL_FRONT_FUNC_FUNC_LEQUAL 0x00000203 +#define NV50TCL_STENCIL_FRONT_FUNC_FUNC_GREATER 0x00000204 +#define NV50TCL_STENCIL_FRONT_FUNC_FUNC_GREATER 0x00000204 +#define NV50TCL_STENCIL_FRONT_FUNC_FUNC_NOTEQUAL 0x00000205 +#define NV50TCL_STENCIL_FRONT_FUNC_FUNC_GEQUAL 0x00000206 +#define NV50TCL_STENCIL_FRONT_FUNC_FUNC_ALWAYS 0x00000207 +#define NV50TCL_POLYGON_OFFSET_UNITS 0x000015bc +#define NV50TCL_VERTEX_BEGIN 0x000015dc +#define NV50TCL_VERTEX_BEGIN_POINTS 0x00000000 +#define NV50TCL_VERTEX_BEGIN_LINES 0x00000001 +#define NV50TCL_VERTEX_BEGIN_LINE_LOOP 0x00000002 +#define NV50TCL_VERTEX_BEGIN_LINE_STRIP 0x00000003 +#define NV50TCL_VERTEX_BEGIN_TRIANGLES 0x00000004 +#define NV50TCL_VERTEX_BEGIN_TRIANGLE_STRIP 0x00000005 +#define NV50TCL_VERTEX_BEGIN_TRIANGLE_FAN 0x00000006 +#define NV50TCL_VERTEX_BEGIN_QUADS 0x00000007 +#define NV50TCL_VERTEX_BEGIN_QUAD_STRIP 0x00000008 +#define NV50TCL_VERTEX_BEGIN_POLYGON 0x00000009 +#define NV50TCL_VERTEX_END 0x000015e0 +#define NV50TCL_VERTEX_DATA 0x00001640 +#define NV50TCL_VP_ATTR_EN_0 0x00001650 +#define NV50TCL_VP_ATTR_EN_0_7_SHIFT 28 +#define NV50TCL_VP_ATTR_EN_0_7_MASK 0xf0000000 +#define NV50TCL_VP_ATTR_EN_0_7_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_0_7_XNNN 0x10000000 +#define NV50TCL_VP_ATTR_EN_0_7_NYNN 0x20000000 +#define NV50TCL_VP_ATTR_EN_0_7_XYNN 0x30000000 +#define NV50TCL_VP_ATTR_EN_0_7_NNZN 0x40000000 +#define NV50TCL_VP_ATTR_EN_0_7_XNZN 0x50000000 +#define NV50TCL_VP_ATTR_EN_0_7_NYZN 0x60000000 +#define NV50TCL_VP_ATTR_EN_0_7_XYZN 0x70000000 +#define NV50TCL_VP_ATTR_EN_0_7_NNNW 0x80000000 +#define NV50TCL_VP_ATTR_EN_0_7_XNNW 0x90000000 +#define NV50TCL_VP_ATTR_EN_0_7_NYNW 0xa0000000 +#define NV50TCL_VP_ATTR_EN_0_7_XYNW 0xb0000000 +#define NV50TCL_VP_ATTR_EN_0_7_NNZW 0xc0000000 +#define NV50TCL_VP_ATTR_EN_0_7_XNZW 0xd0000000 +#define NV50TCL_VP_ATTR_EN_0_7_NYZW 0xe0000000 +#define NV50TCL_VP_ATTR_EN_0_7_XYZW 0xf0000000 +#define NV50TCL_VP_ATTR_EN_0_6_SHIFT 24 +#define NV50TCL_VP_ATTR_EN_0_6_MASK 0x0f000000 +#define NV50TCL_VP_ATTR_EN_0_6_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_0_6_XNNN 0x01000000 +#define NV50TCL_VP_ATTR_EN_0_6_NYNN 0x02000000 +#define NV50TCL_VP_ATTR_EN_0_6_XYNN 0x03000000 +#define NV50TCL_VP_ATTR_EN_0_6_NNZN 0x04000000 +#define NV50TCL_VP_ATTR_EN_0_6_XNZN 0x05000000 +#define NV50TCL_VP_ATTR_EN_0_6_NYZN 0x06000000 +#define NV50TCL_VP_ATTR_EN_0_6_XYZN 0x07000000 +#define NV50TCL_VP_ATTR_EN_0_6_NNNW 0x08000000 +#define NV50TCL_VP_ATTR_EN_0_6_XNNW 0x09000000 +#define NV50TCL_VP_ATTR_EN_0_6_NYNW 0x0a000000 +#define NV50TCL_VP_ATTR_EN_0_6_XYNW 0x0b000000 +#define NV50TCL_VP_ATTR_EN_0_6_NNZW 0x0c000000 +#define NV50TCL_VP_ATTR_EN_0_6_XNZW 0x0d000000 +#define NV50TCL_VP_ATTR_EN_0_6_NYZW 0x0e000000 +#define NV50TCL_VP_ATTR_EN_0_6_XYZW 0x0f000000 +#define NV50TCL_VP_ATTR_EN_0_5_SHIFT 20 +#define NV50TCL_VP_ATTR_EN_0_5_MASK 0x00f00000 +#define NV50TCL_VP_ATTR_EN_0_5_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_0_5_XNNN 0x00100000 +#define NV50TCL_VP_ATTR_EN_0_5_NYNN 0x00200000 +#define NV50TCL_VP_ATTR_EN_0_5_XYNN 0x00300000 +#define NV50TCL_VP_ATTR_EN_0_5_NNZN 0x00400000 +#define NV50TCL_VP_ATTR_EN_0_5_XNZN 0x00500000 +#define NV50TCL_VP_ATTR_EN_0_5_NYZN 0x00600000 +#define NV50TCL_VP_ATTR_EN_0_5_XYZN 0x00700000 +#define NV50TCL_VP_ATTR_EN_0_5_NNNW 0x00800000 +#define NV50TCL_VP_ATTR_EN_0_5_XNNW 0x00900000 +#define NV50TCL_VP_ATTR_EN_0_5_NYNW 0x00a00000 +#define NV50TCL_VP_ATTR_EN_0_5_XYNW 0x00b00000 +#define NV50TCL_VP_ATTR_EN_0_5_NNZW 0x00c00000 +#define NV50TCL_VP_ATTR_EN_0_5_XNZW 0x00d00000 +#define NV50TCL_VP_ATTR_EN_0_5_NYZW 0x00e00000 +#define NV50TCL_VP_ATTR_EN_0_5_XYZW 0x00f00000 +#define NV50TCL_VP_ATTR_EN_0_4_SHIFT 16 +#define NV50TCL_VP_ATTR_EN_0_4_MASK 0x000f0000 +#define NV50TCL_VP_ATTR_EN_0_4_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_0_4_XNNN 0x00010000 +#define NV50TCL_VP_ATTR_EN_0_4_NYNN 0x00020000 +#define NV50TCL_VP_ATTR_EN_0_4_XYNN 0x00030000 +#define NV50TCL_VP_ATTR_EN_0_4_NNZN 0x00040000 +#define NV50TCL_VP_ATTR_EN_0_4_XNZN 0x00050000 +#define NV50TCL_VP_ATTR_EN_0_4_NYZN 0x00060000 +#define NV50TCL_VP_ATTR_EN_0_4_XYZN 0x00070000 +#define NV50TCL_VP_ATTR_EN_0_4_NNNW 0x00080000 +#define NV50TCL_VP_ATTR_EN_0_4_XNNW 0x00090000 +#define NV50TCL_VP_ATTR_EN_0_4_NYNW 0x000a0000 +#define NV50TCL_VP_ATTR_EN_0_4_XYNW 0x000b0000 +#define NV50TCL_VP_ATTR_EN_0_4_NNZW 0x000c0000 +#define NV50TCL_VP_ATTR_EN_0_4_XNZW 0x000d0000 +#define NV50TCL_VP_ATTR_EN_0_4_NYZW 0x000e0000 +#define NV50TCL_VP_ATTR_EN_0_4_XYZW 0x000f0000 +#define NV50TCL_VP_ATTR_EN_0_3_SHIFT 12 +#define NV50TCL_VP_ATTR_EN_0_3_MASK 0x0000f000 +#define NV50TCL_VP_ATTR_EN_0_3_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_0_3_XNNN 0x00001000 +#define NV50TCL_VP_ATTR_EN_0_3_NYNN 0x00002000 +#define NV50TCL_VP_ATTR_EN_0_3_XYNN 0x00003000 +#define NV50TCL_VP_ATTR_EN_0_3_NNZN 0x00004000 +#define NV50TCL_VP_ATTR_EN_0_3_XNZN 0x00005000 +#define NV50TCL_VP_ATTR_EN_0_3_NYZN 0x00006000 +#define NV50TCL_VP_ATTR_EN_0_3_XYZN 0x00007000 +#define NV50TCL_VP_ATTR_EN_0_3_NNNW 0x00008000 +#define NV50TCL_VP_ATTR_EN_0_3_XNNW 0x00009000 +#define NV50TCL_VP_ATTR_EN_0_3_NYNW 0x0000a000 +#define NV50TCL_VP_ATTR_EN_0_3_XYNW 0x0000b000 +#define NV50TCL_VP_ATTR_EN_0_3_NNZW 0x0000c000 +#define NV50TCL_VP_ATTR_EN_0_3_XNZW 0x0000d000 +#define NV50TCL_VP_ATTR_EN_0_3_NYZW 0x0000e000 +#define NV50TCL_VP_ATTR_EN_0_3_XYZW 0x0000f000 +#define NV50TCL_VP_ATTR_EN_0_2_SHIFT 8 +#define NV50TCL_VP_ATTR_EN_0_2_MASK 0x00000f00 +#define NV50TCL_VP_ATTR_EN_0_2_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_0_2_XNNN 0x00000100 +#define NV50TCL_VP_ATTR_EN_0_2_NYNN 0x00000200 +#define NV50TCL_VP_ATTR_EN_0_2_XYNN 0x00000300 +#define NV50TCL_VP_ATTR_EN_0_2_NNZN 0x00000400 +#define NV50TCL_VP_ATTR_EN_0_2_XNZN 0x00000500 +#define NV50TCL_VP_ATTR_EN_0_2_NYZN 0x00000600 +#define NV50TCL_VP_ATTR_EN_0_2_XYZN 0x00000700 +#define NV50TCL_VP_ATTR_EN_0_2_NNNW 0x00000800 +#define NV50TCL_VP_ATTR_EN_0_2_XNNW 0x00000900 +#define NV50TCL_VP_ATTR_EN_0_2_NYNW 0x00000a00 +#define NV50TCL_VP_ATTR_EN_0_2_XYNW 0x00000b00 +#define NV50TCL_VP_ATTR_EN_0_2_NNZW 0x00000c00 +#define NV50TCL_VP_ATTR_EN_0_2_XNZW 0x00000d00 +#define NV50TCL_VP_ATTR_EN_0_2_NYZW 0x00000e00 +#define NV50TCL_VP_ATTR_EN_0_2_XYZW 0x00000f00 +#define NV50TCL_VP_ATTR_EN_0_1_SHIFT 4 +#define NV50TCL_VP_ATTR_EN_0_1_MASK 0x000000f0 +#define NV50TCL_VP_ATTR_EN_0_1_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_0_1_XNNN 0x00000010 +#define NV50TCL_VP_ATTR_EN_0_1_NYNN 0x00000020 +#define NV50TCL_VP_ATTR_EN_0_1_XYNN 0x00000030 +#define NV50TCL_VP_ATTR_EN_0_1_NNZN 0x00000040 +#define NV50TCL_VP_ATTR_EN_0_1_XNZN 0x00000050 +#define NV50TCL_VP_ATTR_EN_0_1_NYZN 0x00000060 +#define NV50TCL_VP_ATTR_EN_0_1_XYZN 0x00000070 +#define NV50TCL_VP_ATTR_EN_0_1_NNNW 0x00000080 +#define NV50TCL_VP_ATTR_EN_0_1_XNNW 0x00000090 +#define NV50TCL_VP_ATTR_EN_0_1_NYNW 0x000000a0 +#define NV50TCL_VP_ATTR_EN_0_1_XYNW 0x000000b0 +#define NV50TCL_VP_ATTR_EN_0_1_NNZW 0x000000c0 +#define NV50TCL_VP_ATTR_EN_0_1_XNZW 0x000000d0 +#define NV50TCL_VP_ATTR_EN_0_1_NYZW 0x000000e0 +#define NV50TCL_VP_ATTR_EN_0_1_XYZW 0x000000f0 +#define NV50TCL_VP_ATTR_EN_0_0_SHIFT 0 +#define NV50TCL_VP_ATTR_EN_0_0_MASK 0x0000000f +#define NV50TCL_VP_ATTR_EN_0_0_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_0_0_XNNN 0x00000001 +#define NV50TCL_VP_ATTR_EN_0_0_NYNN 0x00000002 +#define NV50TCL_VP_ATTR_EN_0_0_XYNN 0x00000003 +#define NV50TCL_VP_ATTR_EN_0_0_NNZN 0x00000004 +#define NV50TCL_VP_ATTR_EN_0_0_XNZN 0x00000005 +#define NV50TCL_VP_ATTR_EN_0_0_NYZN 0x00000006 +#define NV50TCL_VP_ATTR_EN_0_0_XYZN 0x00000007 +#define NV50TCL_VP_ATTR_EN_0_0_NNNW 0x00000008 +#define NV50TCL_VP_ATTR_EN_0_0_XNNW 0x00000009 +#define NV50TCL_VP_ATTR_EN_0_0_NYNW 0x0000000a +#define NV50TCL_VP_ATTR_EN_0_0_XYNW 0x0000000b +#define NV50TCL_VP_ATTR_EN_0_0_NNZW 0x0000000c +#define NV50TCL_VP_ATTR_EN_0_0_XNZW 0x0000000d +#define NV50TCL_VP_ATTR_EN_0_0_NYZW 0x0000000e +#define NV50TCL_VP_ATTR_EN_0_0_XYZW 0x0000000f +#define NV50TCL_VP_ATTR_EN_1 0x00001654 +#define NV50TCL_VP_ATTR_EN_1_15_SHIFT 28 +#define NV50TCL_VP_ATTR_EN_1_15_MASK 0xf0000000 +#define NV50TCL_VP_ATTR_EN_1_15_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_1_15_XNNN 0x10000000 +#define NV50TCL_VP_ATTR_EN_1_15_NYNN 0x20000000 +#define NV50TCL_VP_ATTR_EN_1_15_XYNN 0x30000000 +#define NV50TCL_VP_ATTR_EN_1_15_NNZN 0x40000000 +#define NV50TCL_VP_ATTR_EN_1_15_XNZN 0x50000000 +#define NV50TCL_VP_ATTR_EN_1_15_NYZN 0x60000000 +#define NV50TCL_VP_ATTR_EN_1_15_XYZN 0x70000000 +#define NV50TCL_VP_ATTR_EN_1_15_NNNW 0x80000000 +#define NV50TCL_VP_ATTR_EN_1_15_XNNW 0x90000000 +#define NV50TCL_VP_ATTR_EN_1_15_NYNW 0xa0000000 +#define NV50TCL_VP_ATTR_EN_1_15_XYNW 0xb0000000 +#define NV50TCL_VP_ATTR_EN_1_15_NNZW 0xc0000000 +#define NV50TCL_VP_ATTR_EN_1_15_XNZW 0xd0000000 +#define NV50TCL_VP_ATTR_EN_1_15_NYZW 0xe0000000 +#define NV50TCL_VP_ATTR_EN_1_15_XYZW 0xf0000000 +#define NV50TCL_VP_ATTR_EN_1_14_SHIFT 24 +#define NV50TCL_VP_ATTR_EN_1_14_MASK 0x0f000000 +#define NV50TCL_VP_ATTR_EN_1_14_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_1_14_XNNN 0x01000000 +#define NV50TCL_VP_ATTR_EN_1_14_NYNN 0x02000000 +#define NV50TCL_VP_ATTR_EN_1_14_XYNN 0x03000000 +#define NV50TCL_VP_ATTR_EN_1_14_NNZN 0x04000000 +#define NV50TCL_VP_ATTR_EN_1_14_XNZN 0x05000000 +#define NV50TCL_VP_ATTR_EN_1_14_NYZN 0x06000000 +#define NV50TCL_VP_ATTR_EN_1_14_XYZN 0x07000000 +#define NV50TCL_VP_ATTR_EN_1_14_NNNW 0x08000000 +#define NV50TCL_VP_ATTR_EN_1_14_XNNW 0x09000000 +#define NV50TCL_VP_ATTR_EN_1_14_NYNW 0x0a000000 +#define NV50TCL_VP_ATTR_EN_1_14_XYNW 0x0b000000 +#define NV50TCL_VP_ATTR_EN_1_14_NNZW 0x0c000000 +#define NV50TCL_VP_ATTR_EN_1_14_XNZW 0x0d000000 +#define NV50TCL_VP_ATTR_EN_1_14_NYZW 0x0e000000 +#define NV50TCL_VP_ATTR_EN_1_14_XYZW 0x0f000000 +#define NV50TCL_VP_ATTR_EN_1_13_SHIFT 20 +#define NV50TCL_VP_ATTR_EN_1_13_MASK 0x00f00000 +#define NV50TCL_VP_ATTR_EN_1_13_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_1_13_XNNN 0x00100000 +#define NV50TCL_VP_ATTR_EN_1_13_NYNN 0x00200000 +#define NV50TCL_VP_ATTR_EN_1_13_XYNN 0x00300000 +#define NV50TCL_VP_ATTR_EN_1_13_NNZN 0x00400000 +#define NV50TCL_VP_ATTR_EN_1_13_XNZN 0x00500000 +#define NV50TCL_VP_ATTR_EN_1_13_NYZN 0x00600000 +#define NV50TCL_VP_ATTR_EN_1_13_XYZN 0x00700000 +#define NV50TCL_VP_ATTR_EN_1_13_NNNW 0x00800000 +#define NV50TCL_VP_ATTR_EN_1_13_XNNW 0x00900000 +#define NV50TCL_VP_ATTR_EN_1_13_NYNW 0x00a00000 +#define NV50TCL_VP_ATTR_EN_1_13_XYNW 0x00b00000 +#define NV50TCL_VP_ATTR_EN_1_13_NNZW 0x00c00000 +#define NV50TCL_VP_ATTR_EN_1_13_XNZW 0x00d00000 +#define NV50TCL_VP_ATTR_EN_1_13_NYZW 0x00e00000 +#define NV50TCL_VP_ATTR_EN_1_13_XYZW 0x00f00000 +#define NV50TCL_VP_ATTR_EN_1_12_SHIFT 16 +#define NV50TCL_VP_ATTR_EN_1_12_MASK 0x000f0000 +#define NV50TCL_VP_ATTR_EN_1_12_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_1_12_XNNN 0x00010000 +#define NV50TCL_VP_ATTR_EN_1_12_NYNN 0x00020000 +#define NV50TCL_VP_ATTR_EN_1_12_XYNN 0x00030000 +#define NV50TCL_VP_ATTR_EN_1_12_NNZN 0x00040000 +#define NV50TCL_VP_ATTR_EN_1_12_XNZN 0x00050000 +#define NV50TCL_VP_ATTR_EN_1_12_NYZN 0x00060000 +#define NV50TCL_VP_ATTR_EN_1_12_XYZN 0x00070000 +#define NV50TCL_VP_ATTR_EN_1_12_NNNW 0x00080000 +#define NV50TCL_VP_ATTR_EN_1_12_XNNW 0x00090000 +#define NV50TCL_VP_ATTR_EN_1_12_NYNW 0x000a0000 +#define NV50TCL_VP_ATTR_EN_1_12_XYNW 0x000b0000 +#define NV50TCL_VP_ATTR_EN_1_12_NNZW 0x000c0000 +#define NV50TCL_VP_ATTR_EN_1_12_XNZW 0x000d0000 +#define NV50TCL_VP_ATTR_EN_1_12_NYZW 0x000e0000 +#define NV50TCL_VP_ATTR_EN_1_12_XYZW 0x000f0000 +#define NV50TCL_VP_ATTR_EN_1_11_SHIFT 12 +#define NV50TCL_VP_ATTR_EN_1_11_MASK 0x0000f000 +#define NV50TCL_VP_ATTR_EN_1_11_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_1_11_XNNN 0x00001000 +#define NV50TCL_VP_ATTR_EN_1_11_NYNN 0x00002000 +#define NV50TCL_VP_ATTR_EN_1_11_XYNN 0x00003000 +#define NV50TCL_VP_ATTR_EN_1_11_NNZN 0x00004000 +#define NV50TCL_VP_ATTR_EN_1_11_XNZN 0x00005000 +#define NV50TCL_VP_ATTR_EN_1_11_NYZN 0x00006000 +#define NV50TCL_VP_ATTR_EN_1_11_XYZN 0x00007000 +#define NV50TCL_VP_ATTR_EN_1_11_NNNW 0x00008000 +#define NV50TCL_VP_ATTR_EN_1_11_XNNW 0x00009000 +#define NV50TCL_VP_ATTR_EN_1_11_NYNW 0x0000a000 +#define NV50TCL_VP_ATTR_EN_1_11_XYNW 0x0000b000 +#define NV50TCL_VP_ATTR_EN_1_11_NNZW 0x0000c000 +#define NV50TCL_VP_ATTR_EN_1_11_XNZW 0x0000d000 +#define NV50TCL_VP_ATTR_EN_1_11_NYZW 0x0000e000 +#define NV50TCL_VP_ATTR_EN_1_11_XYZW 0x0000f000 +#define NV50TCL_VP_ATTR_EN_1_10_SHIFT 8 +#define NV50TCL_VP_ATTR_EN_1_10_MASK 0x00000f00 +#define NV50TCL_VP_ATTR_EN_1_10_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_1_10_XNNN 0x00000100 +#define NV50TCL_VP_ATTR_EN_1_10_NYNN 0x00000200 +#define NV50TCL_VP_ATTR_EN_1_10_XYNN 0x00000300 +#define NV50TCL_VP_ATTR_EN_1_10_NNZN 0x00000400 +#define NV50TCL_VP_ATTR_EN_1_10_XNZN 0x00000500 +#define NV50TCL_VP_ATTR_EN_1_10_NYZN 0x00000600 +#define NV50TCL_VP_ATTR_EN_1_10_XYZN 0x00000700 +#define NV50TCL_VP_ATTR_EN_1_10_NNNW 0x00000800 +#define NV50TCL_VP_ATTR_EN_1_10_XNNW 0x00000900 +#define NV50TCL_VP_ATTR_EN_1_10_NYNW 0x00000a00 +#define NV50TCL_VP_ATTR_EN_1_10_XYNW 0x00000b00 +#define NV50TCL_VP_ATTR_EN_1_10_NNZW 0x00000c00 +#define NV50TCL_VP_ATTR_EN_1_10_XNZW 0x00000d00 +#define NV50TCL_VP_ATTR_EN_1_10_NYZW 0x00000e00 +#define NV50TCL_VP_ATTR_EN_1_10_XYZW 0x00000f00 +#define NV50TCL_VP_ATTR_EN_1_9_SHIFT 4 +#define NV50TCL_VP_ATTR_EN_1_9_MASK 0x000000f0 +#define NV50TCL_VP_ATTR_EN_1_9_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_1_9_XNNN 0x00000010 +#define NV50TCL_VP_ATTR_EN_1_9_NYNN 0x00000020 +#define NV50TCL_VP_ATTR_EN_1_9_XYNN 0x00000030 +#define NV50TCL_VP_ATTR_EN_1_9_NNZN 0x00000040 +#define NV50TCL_VP_ATTR_EN_1_9_XNZN 0x00000050 +#define NV50TCL_VP_ATTR_EN_1_9_NYZN 0x00000060 +#define NV50TCL_VP_ATTR_EN_1_9_XYZN 0x00000070 +#define NV50TCL_VP_ATTR_EN_1_9_NNNW 0x00000080 +#define NV50TCL_VP_ATTR_EN_1_9_XNNW 0x00000090 +#define NV50TCL_VP_ATTR_EN_1_9_NYNW 0x000000a0 +#define NV50TCL_VP_ATTR_EN_1_9_XYNW 0x000000b0 +#define NV50TCL_VP_ATTR_EN_1_9_NNZW 0x000000c0 +#define NV50TCL_VP_ATTR_EN_1_9_XNZW 0x000000d0 +#define NV50TCL_VP_ATTR_EN_1_9_NYZW 0x000000e0 +#define NV50TCL_VP_ATTR_EN_1_9_XYZW 0x000000f0 +#define NV50TCL_VP_ATTR_EN_1_8_SHIFT 0 +#define NV50TCL_VP_ATTR_EN_1_8_MASK 0x0000000f +#define NV50TCL_VP_ATTR_EN_1_8_NONE 0x00000000 +#define NV50TCL_VP_ATTR_EN_1_8_XNNN 0x00000001 +#define NV50TCL_VP_ATTR_EN_1_8_NYNN 0x00000002 +#define NV50TCL_VP_ATTR_EN_1_8_XYNN 0x00000003 +#define NV50TCL_VP_ATTR_EN_1_8_NNZN 0x00000004 +#define NV50TCL_VP_ATTR_EN_1_8_XNZN 0x00000005 +#define NV50TCL_VP_ATTR_EN_1_8_NYZN 0x00000006 +#define NV50TCL_VP_ATTR_EN_1_8_XYZN 0x00000007 +#define NV50TCL_VP_ATTR_EN_1_8_NNNW 0x00000008 +#define NV50TCL_VP_ATTR_EN_1_8_XNNW 0x00000009 +#define NV50TCL_VP_ATTR_EN_1_8_NYNW 0x0000000a +#define NV50TCL_VP_ATTR_EN_1_8_XYNW 0x0000000b +#define NV50TCL_VP_ATTR_EN_1_8_NNZW 0x0000000c +#define NV50TCL_VP_ATTR_EN_1_8_XNZW 0x0000000d +#define NV50TCL_VP_ATTR_EN_1_8_NYZW 0x0000000e +#define NV50TCL_VP_ATTR_EN_1_8_XYZW 0x0000000f +#define NV50TCL_LINE_STIPPLE_ENABLE 0x0000166c +#define NV50TCL_LINE_STIPPLE_PATTERN 0x00001680 +#define NV50TCL_POLYGON_STIPPLE_ENABLE 0x0000168c +#define NV50TCL_VP_REG_HPOS 0x000016bc +#define NV50TCL_VP_REG_HPOS_X_SHIFT 0 +#define NV50TCL_VP_REG_HPOS_X_MASK 0x000000ff +#define NV50TCL_VP_REG_HPOS_Y_SHIFT 8 +#define NV50TCL_VP_REG_HPOS_Y_MASK 0x0000ff00 +#define NV50TCL_VP_REG_HPOS_Z_SHIFT 16 +#define NV50TCL_VP_REG_HPOS_Z_MASK 0x00ff0000 +#define NV50TCL_VP_REG_HPOS_W_SHIFT 24 +#define NV50TCL_VP_REG_HPOS_W_MASK 0xff000000 +#define NV50TCL_VP_REG_COL0 0x000016c0 +#define NV50TCL_VP_REG_COL0_X_SHIFT 0 +#define NV50TCL_VP_REG_COL0_X_MASK 0x000000ff +#define NV50TCL_VP_REG_COL0_Y_SHIFT 8 +#define NV50TCL_VP_REG_COL0_Y_MASK 0x0000ff00 +#define NV50TCL_VP_REG_COL0_Z_SHIFT 16 +#define NV50TCL_VP_REG_COL0_Z_MASK 0x00ff0000 +#define NV50TCL_VP_REG_COL0_W_SHIFT 24 +#define NV50TCL_VP_REG_COL0_W_MASK 0xff000000 +#define NV50TCL_POLYGON_STIPPLE_PATTERN(x) (0x00001700+((x)*4)) +#define NV50TCL_POLYGON_STIPPLE_PATTERN__SIZE 0x00000020 +#define NV50TCL_CULL_FACE_ENABLE 0x00001918 +#define NV50TCL_FRONT_FACE 0x0000191c +#define NV50TCL_FRONT_FACE_CW 0x00000900 +#define NV50TCL_FRONT_FACE_CCW 0x00000901 +#define NV50TCL_CULL_FACE 0x00001920 +#define NV50TCL_CULL_FACE_FRONT 0x00000404 +#define NV50TCL_CULL_FACE_BACK 0x00000405 +#define NV50TCL_CULL_FACE_FRONT_AND_BACK 0x00000408 +#define NV50TCL_LOGIC_OP_ENABLE 0x000019c4 +#define NV50TCL_LOGIC_OP 0x000019c8 +#define NV50TCL_LOGIC_OP_CLEAR 0x00001500 +#define NV50TCL_LOGIC_OP_AND 0x00001501 +#define NV50TCL_LOGIC_OP_AND_REVERSE 0x00001502 +#define NV50TCL_LOGIC_OP_COPY 0x00001503 +#define NV50TCL_LOGIC_OP_AND_INVERTED 0x00001504 +#define NV50TCL_LOGIC_OP_NOOP 0x00001505 +#define NV50TCL_LOGIC_OP_XOR 0x00001506 +#define NV50TCL_LOGIC_OP_OR 0x00001507 +#define NV50TCL_LOGIC_OP_NOR 0x00001508 +#define NV50TCL_LOGIC_OP_EQUIV 0x00001509 +#define NV50TCL_LOGIC_OP_INVERT 0x0000150a +#define NV50TCL_LOGIC_OP_OR_REVERSE 0x0000150b +#define NV50TCL_LOGIC_OP_COPY_INVERTED 0x0000150c +#define NV50TCL_LOGIC_OP_OR_INVERTED 0x0000150d +#define NV50TCL_LOGIC_OP_NAND 0x0000150e +#define NV50TCL_LOGIC_OP_SET 0x0000150f +#define NV50TCL_CLEAR_BUFFERS 0x000019d0 +#define NV50TCL_COLOR_MASK(x) (0x00001a00+((x)*4)) +#define NV50TCL_COLOR_MASK__SIZE 0x00000008 +#define NV50TCL_COLOR_MASK_R_SHIFT 0 +#define NV50TCL_COLOR_MASK_R_MASK 0x0000000f +#define NV50TCL_COLOR_MASK_G_SHIFT 4 +#define NV50TCL_COLOR_MASK_G_MASK 0x000000f0 +#define NV50TCL_COLOR_MASK_B_SHIFT 8 +#define NV50TCL_COLOR_MASK_B_MASK 0x00000f00 +#define NV50TCL_COLOR_MASK_A_SHIFT 12 +#define NV50TCL_COLOR_MASK_A_MASK 0x0000f000 + + +#define NV50_COMPUTE 0x000050c0 + +#define NV50_COMPUTE_DMA_UNK0 0x000001a0 +#define NV50_COMPUTE_DMA_STATUS 0x000001a4 +#define NV50_COMPUTE_DMA_UNK1 0x000001b8 +#define NV50_COMPUTE_DMA_UNK2 0x000001bc +#define NV50_COMPUTE_DMA_UNK3 0x000001c0 +#define NV50_COMPUTE_UNK4_HIGH 0x00000210 +#define NV50_COMPUTE_UNK4_LOW 0x00000214 +#define NV50_COMPUTE_UNK5_HIGH 0x00000218 +#define NV50_COMPUTE_UNK5_LOW 0x0000021c +#define NV50_COMPUTE_UNK6_HIGH 0x00000294 +#define NV50_COMPUTE_UNK6_LOW 0x00000298 +#define NV50_COMPUTE_CONST_BASE_HIGH 0x000002a4 +#define NV50_COMPUTE_CONST_BASE_LO 0x000002a8 +#define NV50_COMPUTE_CONST_SIZE_SEG 0x000002ac +#define NV50_COMPUTE_REG_COUNT 0x000002c0 +#define NV50_COMPUTE_STATUS_HIGH 0x00000310 +#define NV50_COMPUTE_STATUS_LOW 0x00000314 +#define NV50_COMPUTE_EXECUTE 0x0000031c +#define NV50_COMPUTE_USER_PARAM_COUNT 0x00000374 +#define NV50_COMPUTE_GRIDDIM_YX 0x000003a4 +#define NV50_COMPUTE_SHARED_SIZE 0x000003a8 +#define NV50_COMPUTE_BLOCKDIM_YX 0x000003ac +#define NV50_COMPUTE_BLOCKDIM_Z 0x000003b0 +#define NV50_COMPUTE_CALL_ADDRESS 0x000003b4 +#define NV50_COMPUTE_GLOBAL_BASE_HIGH(x) (0x00000400+((x)*32)) +#define NV50_COMPUTE_GLOBAL_BASE_HIGH__SIZE 0x00000010 +#define NV50_COMPUTE_GLOBAL_BASE_LOW(x) (0x00000404+((x)*32)) +#define NV50_COMPUTE_GLOBAL_BASE_LOW__SIZE 0x00000010 +#define NV50_COMPUTE_GLOBAL_LIMIT_HIGH(x) (0x00000408+((x)*32)) +#define NV50_COMPUTE_GLOBAL_LIMIT_HIGH__SIZE 0x00000010 +#define NV50_COMPUTE_GLOBAL_LIMIT_LOW(x) (0x0000040c+((x)*32)) +#define NV50_COMPUTE_GLOBAL_LIMIT_LOW__SIZE 0x00000010 +#define NV50_COMPUTE_GLOBAL_UNK(x) (0x00000410+((x)*32)) +#define NV50_COMPUTE_GLOBAL_UNK__SIZE 0x00000010 +#define NV50_COMPUTE_USER_PARAM(x) (0x00000600+((x)*4)) +#define NV50_COMPUTE_USER_PARAM__SIZE 0x00000040 + + +#define NV54TCL 0x00008297 + + + +#endif /* NOUVEAU_REG_H */ diff --git a/src/gallium/drivers/nouveau/nouveau_device.h b/src/gallium/drivers/nouveau/nouveau_device.h new file mode 100644 index 0000000000..e25e89fedd --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_device.h @@ -0,0 +1,30 @@ +/* + * Copyright 2007 Nouveau Project + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef __NOUVEAU_DEVICE_H__ +#define __NOUVEAU_DEVICE_H__ + +struct nouveau_device { + unsigned chipset; +}; + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_gldefs.h b/src/gallium/drivers/nouveau/nouveau_gldefs.h new file mode 100644 index 0000000000..ff97aaa9af --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_gldefs.h @@ -0,0 +1,196 @@ +#ifndef __NOUVEAU_GLDEFS_H__ +#define __NOUVEAU_GLDEFS_H__ + +static INLINE unsigned +nvgl_blend_func(unsigned factor) +{ + switch (factor) { + case PIPE_BLENDFACTOR_ZERO: + return 0x0000; + case PIPE_BLENDFACTOR_ONE: + return 0x0001; + case PIPE_BLENDFACTOR_SRC_COLOR: + return 0x0300; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + return 0x0301; + case PIPE_BLENDFACTOR_SRC_ALPHA: + return 0x0302; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + return 0x0303; + case PIPE_BLENDFACTOR_DST_ALPHA: + return 0x0304; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + return 0x0305; + case PIPE_BLENDFACTOR_DST_COLOR: + return 0x0306; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + return 0x0307; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + return 0x0308; + case PIPE_BLENDFACTOR_CONST_COLOR: + return 0x8001; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + return 0x8002; + case PIPE_BLENDFACTOR_CONST_ALPHA: + return 0x8003; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + return 0x8004; + default: + return 0x0000; + } +} + +static INLINE unsigned +nvgl_blend_eqn(unsigned func) +{ + switch (func) { + case PIPE_BLEND_ADD: + return 0x8006; + case PIPE_BLEND_MIN: + return 0x8007; + case PIPE_BLEND_MAX: + return 0x8008; + case PIPE_BLEND_SUBTRACT: + return 0x800a; + case PIPE_BLEND_REVERSE_SUBTRACT: + return 0x800b; + default: + return 0x8006; + } +} + +static INLINE unsigned +nvgl_logicop_func(unsigned func) +{ + switch (func) { + case PIPE_LOGICOP_CLEAR: + return 0x1500; + case PIPE_LOGICOP_NOR: + return 0x1508; + case PIPE_LOGICOP_AND_INVERTED: + return 0x1504; + case PIPE_LOGICOP_COPY_INVERTED: + return 0x150c; + case PIPE_LOGICOP_AND_REVERSE: + return 0x1502; + case PIPE_LOGICOP_INVERT: + return 0x150a; + case PIPE_LOGICOP_XOR: + return 0x1506; + case PIPE_LOGICOP_NAND: + return 0x150e; + case PIPE_LOGICOP_AND: + return 0x1501; + case PIPE_LOGICOP_EQUIV: + return 0x1509; + case PIPE_LOGICOP_NOOP: + return 0x1505; + case PIPE_LOGICOP_OR_INVERTED: + return 0x150d; + case PIPE_LOGICOP_COPY: + return 0x1503; + case PIPE_LOGICOP_OR_REVERSE: + return 0x150b; + case PIPE_LOGICOP_OR: + return 0x1507; + case PIPE_LOGICOP_SET: + return 0x150f; + default: + return 0x1505; + } +} + +static INLINE unsigned +nvgl_comparison_op(unsigned op) +{ + switch (op) { + case PIPE_FUNC_NEVER: + return 0x0200; + case PIPE_FUNC_LESS: + return 0x0201; + case PIPE_FUNC_EQUAL: + return 0x0202; + case PIPE_FUNC_LEQUAL: + return 0x0203; + case PIPE_FUNC_GREATER: + return 0x0204; + case PIPE_FUNC_NOTEQUAL: + return 0x0205; + case PIPE_FUNC_GEQUAL: + return 0x0206; + case PIPE_FUNC_ALWAYS: + return 0x0207; + default: + return 0x0207; + } +} + +static INLINE unsigned +nvgl_polygon_mode(unsigned mode) +{ + switch (mode) { + case PIPE_POLYGON_MODE_POINT: + return 0x1b00; + case PIPE_POLYGON_MODE_LINE: + return 0x1b01; + case PIPE_POLYGON_MODE_FILL: + return 0x1b02; + default: + return 0x1b02; + } +} + +static INLINE unsigned +nvgl_stencil_op(unsigned op) +{ + switch (op) { + case PIPE_STENCIL_OP_ZERO: + return 0x0000; + case PIPE_STENCIL_OP_INVERT: + return 0x150a; + case PIPE_STENCIL_OP_KEEP: + return 0x1e00; + case PIPE_STENCIL_OP_REPLACE: + return 0x1e01; + case PIPE_STENCIL_OP_INCR: + return 0x1e02; + case PIPE_STENCIL_OP_DECR: + return 0x1e03; + case PIPE_STENCIL_OP_INCR_WRAP: + return 0x8507; + case PIPE_STENCIL_OP_DECR_WRAP: + return 0x8508; + default: + return 0x1e00; + } +} + +static INLINE unsigned +nvgl_primitive(unsigned prim) { + switch (prim) { + case PIPE_PRIM_POINTS: + return 0x0001; + case PIPE_PRIM_LINES: + return 0x0002; + case PIPE_PRIM_LINE_LOOP: + return 0x0003; + case PIPE_PRIM_LINE_STRIP: + return 0x0004; + case PIPE_PRIM_TRIANGLES: + return 0x0005; + case PIPE_PRIM_TRIANGLE_STRIP: + return 0x0006; + case PIPE_PRIM_TRIANGLE_FAN: + return 0x0007; + case PIPE_PRIM_QUADS: + return 0x0008; + case PIPE_PRIM_QUAD_STRIP: + return 0x0009; + case PIPE_PRIM_POLYGON: + return 0x000a; + default: + return 0; + } +} + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_grobj.h b/src/gallium/drivers/nouveau/nouveau_grobj.h new file mode 100644 index 0000000000..8f5abf9051 --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_grobj.h @@ -0,0 +1,35 @@ +/* + * Copyright 2007 Nouveau Project + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef __NOUVEAU_GROBJ_H__ +#define __NOUVEAU_GROBJ_H__ + +#include "nouveau_channel.h" + +struct nouveau_grobj { + struct nouveau_channel *channel; + int grclass; + uint32_t handle; + int subc; +}; + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_notifier.h b/src/gallium/drivers/nouveau/nouveau_notifier.h new file mode 100644 index 0000000000..35adde1e32 --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_notifier.h @@ -0,0 +1,43 @@ +/* + * Copyright 2007 Nouveau Project + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef __NOUVEAU_NOTIFIER_H__ +#define __NOUVEAU_NOTIFIER_H__ + +#define NV_NOTIFIER_SIZE 32 +#define NV_NOTIFY_TIME_0 0x00000000 +#define NV_NOTIFY_TIME_1 0x00000004 +#define NV_NOTIFY_RETURN_VALUE 0x00000008 +#define NV_NOTIFY_STATE 0x0000000C +#define NV_NOTIFY_STATE_STATUS_MASK 0xFF000000 +#define NV_NOTIFY_STATE_STATUS_SHIFT 24 +#define NV_NOTIFY_STATE_STATUS_COMPLETED 0x00 +#define NV_NOTIFY_STATE_STATUS_IN_PROCESS 0x01 +#define NV_NOTIFY_STATE_ERROR_CODE_MASK 0x0000FFFF +#define NV_NOTIFY_STATE_ERROR_CODE_SHIFT 0 + +struct nouveau_notifier { + struct nouveau_channel *channel; + uint32_t handle; +}; + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_push.h b/src/gallium/drivers/nouveau/nouveau_push.h new file mode 100644 index 0000000000..54ef1c1291 --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_push.h @@ -0,0 +1,82 @@ +#ifndef __NOUVEAU_PUSH_H__ +#define __NOUVEAU_PUSH_H__ + +#include "nouveau/nouveau_winsys.h" + +#ifndef NOUVEAU_PUSH_CONTEXT +#error undefined push context +#endif + +#define OUT_RING(data) do { \ + NOUVEAU_PUSH_CONTEXT(pc); \ + (*pc->nvws->channel->pushbuf->cur++) = (data); \ +} while(0) + +#define OUT_RINGp(src,size) do { \ + NOUVEAU_PUSH_CONTEXT(pc); \ + memcpy(pc->nvws->channel->pushbuf->cur, (src), (size) * 4); \ + pc->nvws->channel->pushbuf->cur += (size); \ +} while(0) + +#define OUT_RINGf(data) do { \ + union { float v; uint32_t u; } c; \ + c.v = (data); \ + OUT_RING(c.u); \ +} while(0) + +#define BEGIN_RING(obj,mthd,size) do { \ + NOUVEAU_PUSH_CONTEXT(pc); \ + if (pc->nvws->channel->pushbuf->remaining < ((size) + 1)) \ + pc->nvws->push_flush(pc->nvws, ((size) + 1), NULL); \ + OUT_RING((pc->obj->subc << 13) | ((size) << 18) | (mthd)); \ + pc->nvws->channel->pushbuf->remaining -= ((size) + 1); \ +} while(0) + +#define BEGIN_RING_NI(obj,mthd,size) do { \ + BEGIN_RING(obj, (mthd) | 0x40000000, (size)); \ +} while(0) + +#define FIRE_RING(fence) do { \ + NOUVEAU_PUSH_CONTEXT(pc); \ + pc->nvws->push_flush(pc->nvws, 0, fence); \ +} while(0) + +#define OUT_RELOC(bo,data,flags,vor,tor) do { \ + NOUVEAU_PUSH_CONTEXT(pc); \ + pc->nvws->push_reloc(pc->nvws, pc->nvws->channel->pushbuf->cur++, \ + (bo), (data), (flags), (vor), (tor)); \ +} while(0) + +/* Raw data + flags depending on FB/TT buffer */ +#define OUT_RELOCd(bo,data,flags,vor,tor) do { \ + OUT_RELOC((bo), (data), (flags) | NOUVEAU_BO_OR, (vor), (tor)); \ +} while(0) + +/* FB/TT object handle */ +#define OUT_RELOCo(bo,flags) do { \ + OUT_RELOC((bo), 0, (flags) | NOUVEAU_BO_OR, \ + pc->nvws->channel->vram->handle, \ + pc->nvws->channel->gart->handle); \ +} while(0) + +/* Low 32-bits of offset */ +#define OUT_RELOCl(bo,delta,flags) do { \ + OUT_RELOC((bo), (delta), (flags) | NOUVEAU_BO_LOW, 0, 0); \ +} while(0) + +/* High 32-bits of offset */ +#define OUT_RELOCh(bo,delta,flags) do { \ + OUT_RELOC((bo), (delta), (flags) | NOUVEAU_BO_HIGH, 0, 0); \ +} while(0) + +/* A reloc which'll recombine into a NV_DMA_METHOD packet header */ +#define OUT_RELOCm(bo, flags, obj, mthd, size) do { \ + NOUVEAU_PUSH_CONTEXT(pc); \ + if (pc->nvws->channel->pushbuf->remaining < ((size) + 1)) \ + pc->nvws->push_flush(pc->nvws->channel, ((size) + 1), NULL); \ + OUT_RELOCd((bo), (pc->obj->subc << 13) | ((size) << 18) | (mthd), \ + (flags), 0, 0); \ + pc->nvws->channel->pushbuf->remaining -= ((size) + 1); \ +} while(0) + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_pushbuf.h b/src/gallium/drivers/nouveau/nouveau_pushbuf.h new file mode 100644 index 0000000000..1909765098 --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_pushbuf.h @@ -0,0 +1,32 @@ +/* + * Copyright 2007 Nouveau Project + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef __NOUVEAU_PUSHBUF_H__ +#define __NOUVEAU_PUSHBUF_H__ + +struct nouveau_pushbuf { + struct nouveau_channel *channel; + unsigned remaining; + uint32_t *cur; +}; + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_resource.h b/src/gallium/drivers/nouveau/nouveau_resource.h new file mode 100644 index 0000000000..1af7961d30 --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_resource.h @@ -0,0 +1,37 @@ +/* + * Copyright 2007 Nouveau Project + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef __NOUVEAU_RESOURCE_H__ +#define __NOUVEAU_RESOURCE_H__ + +struct nouveau_resource { + struct nouveau_resource *prev; + struct nouveau_resource *next; + + int in_use; + void *priv; + + unsigned int start; + unsigned int size; +}; + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_stateobj.h b/src/gallium/drivers/nouveau/nouveau_stateobj.h new file mode 100644 index 0000000000..729988b095 --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_stateobj.h @@ -0,0 +1,158 @@ +#ifndef __NOUVEAU_STATEOBJ_H__ +#define __NOUVEAU_STATEOBJ_H__ + +#include "pipe/p_debug.h" + +struct nouveau_stateobj_reloc { + struct pipe_buffer *bo; + + unsigned offset; + unsigned packet; + + unsigned data; + unsigned flags; + unsigned vor; + unsigned tor; +}; + +struct nouveau_stateobj { + int refcount; + + unsigned *push; + struct nouveau_stateobj_reloc *reloc; + + unsigned *cur; + unsigned cur_packet; + unsigned cur_reloc; +}; + +static INLINE struct nouveau_stateobj * +so_new(unsigned push, unsigned reloc) +{ + struct nouveau_stateobj *so; + + so = MALLOC(sizeof(struct nouveau_stateobj)); + so->refcount = 0; + so->push = MALLOC(sizeof(unsigned) * push); + so->reloc = MALLOC(sizeof(struct nouveau_stateobj_reloc) * reloc); + + so->cur = so->push; + so->cur_reloc = so->cur_packet = 0; + + return so; +} + +static INLINE void +so_ref(struct nouveau_stateobj *ref, struct nouveau_stateobj **pso) +{ + struct nouveau_stateobj *so = *pso; + + if (ref) { + ref->refcount++; + } + + if (so && --so->refcount <= 0) { + free(so->push); + free(so->reloc); + free(so); + } + + *pso = ref; +} + +static INLINE void +so_data(struct nouveau_stateobj *so, unsigned data) +{ + (*so->cur++) = (data); + so->cur_packet += 4; +} + +static INLINE void +so_datap(struct nouveau_stateobj *so, unsigned *data, unsigned size) +{ + so->cur_packet += (4 * size); + while (size--) + (*so->cur++) = (*data++); +} + +static INLINE void +so_method(struct nouveau_stateobj *so, struct nouveau_grobj *gr, + unsigned mthd, unsigned size) +{ + so->cur_packet = (gr->subc << 13) | (1 << 18) | (mthd - 4); + so_data(so, (gr->subc << 13) | (size << 18) | mthd); +} + +static INLINE void +so_reloc(struct nouveau_stateobj *so, struct pipe_buffer *bo, + unsigned data, unsigned flags, unsigned vor, unsigned tor) +{ + struct nouveau_stateobj_reloc *r = &so->reloc[so->cur_reloc++]; + + r->bo = bo; + r->offset = so->cur - so->push; + r->packet = so->cur_packet; + r->data = data; + r->flags = flags; + r->vor = vor; + r->tor = tor; + so_data(so, data); +} + +static INLINE void +so_dump(struct nouveau_stateobj *so) +{ + unsigned i, nr = so->cur - so->push; + + for (i = 0; i < nr; i++) + debug_printf("+0x%04x: 0x%08x\n", i, so->push[i]); +} + +static INLINE void +so_emit(struct nouveau_winsys *nvws, struct nouveau_stateobj *so) +{ + struct nouveau_pushbuf *pb = nvws->channel->pushbuf; + unsigned nr, i; + + nr = so->cur - so->push; + if (pb->remaining < nr) + nvws->push_flush(nvws, nr, NULL); + pb->remaining -= nr; + + memcpy(pb->cur, so->push, nr * 4); + for (i = 0; i < so->cur_reloc; i++) { + struct nouveau_stateobj_reloc *r = &so->reloc[i]; + + nvws->push_reloc(nvws, pb->cur + r->offset, r->bo, + r->data, r->flags, r->vor, r->tor); + } + pb->cur += nr; +} + +static INLINE void +so_emit_reloc_markers(struct nouveau_winsys *nvws, struct nouveau_stateobj *so) +{ + struct nouveau_pushbuf *pb = nvws->channel->pushbuf; + unsigned i; + + if (!so) + return; + + i = so->cur_reloc << 1; + if (nvws->channel->pushbuf->remaining < i) + nvws->push_flush(nvws, i, NULL); + nvws->channel->pushbuf->remaining -= i; + + for (i = 0; i < so->cur_reloc; i++) { + struct nouveau_stateobj_reloc *r = &so->reloc[i]; + + nvws->push_reloc(nvws, pb->cur++, r->bo, r->packet, + (r->flags & + (NOUVEAU_BO_VRAM | NOUVEAU_BO_GART)) | + NOUVEAU_BO_DUMMY, 0, 0); + nvws->push_reloc(nvws, pb->cur++, r->bo, r->data, + r->flags | NOUVEAU_BO_DUMMY, r->vor, r->tor); + } +} + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_util.h b/src/gallium/drivers/nouveau/nouveau_util.h new file mode 100644 index 0000000000..a10114beab --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_util.h @@ -0,0 +1,91 @@ +#ifndef __NOUVEAU_UTIL_H__ +#define __NOUVEAU_UTIL_H__ + +/* Determine how many vertices can be pushed into the command stream. + * Where the remaining space isn't large enough to represent all verices, + * split the buffer at primitive boundaries. + * + * Returns a count of vertices that can be rendered, and an index to + * restart drawing at after a flush. + */ +static INLINE unsigned +nouveau_vbuf_split(unsigned remaining, unsigned overhead, unsigned vpp, + unsigned mode, unsigned start, unsigned count, + unsigned *restart) +{ + int max, adj = 0; + + max = remaining - overhead; + if (max < 0) + return 0; + + max *= vpp; + if (max >= count) + return count; + + switch (mode) { + case PIPE_PRIM_POINTS: + break; + case PIPE_PRIM_LINES: + max = max & 1; + break; + case PIPE_PRIM_TRIANGLES: + max = max - (max % 3); + break; + case PIPE_PRIM_QUADS: + max = max & 3; + break; + case PIPE_PRIM_LINE_LOOP: + case PIPE_PRIM_LINE_STRIP: + if (max < 2) + max = 0; + adj = 1; + break; + case PIPE_PRIM_POLYGON: + case PIPE_PRIM_TRIANGLE_STRIP: + case PIPE_PRIM_TRIANGLE_FAN: + if (max < 3) + max = 0; + adj = 2; + break; + case PIPE_PRIM_QUAD_STRIP: + if (max < 4) + max = 0; + adj = 3; + break; + default: + assert(0); + } + + *restart = start + max - adj; + return max; +} + +/* Integer base-2 logarithm, rounded towards zero. */ +static INLINE unsigned log2i(unsigned i) +{ + unsigned r = 0; + + if (i & 0xffff0000) { + i >>= 16; + r += 16; + } + if (i & 0x0000ff00) { + i >>= 8; + r += 8; + } + if (i & 0x000000f0) { + i >>= 4; + r += 4; + } + if (i & 0x0000000c) { + i >>= 2; + r += 2; + } + if (i & 0x00000002) { + r += 1; + } + return r; +} + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_winsys.h b/src/gallium/drivers/nouveau/nouveau_winsys.h new file mode 100644 index 0000000000..5535ebb6a9 --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_winsys.h @@ -0,0 +1,98 @@ +#ifndef NOUVEAU_WINSYS_H +#define NOUVEAU_WINSYS_H + +#include <stdint.h> +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" + +#include "nouveau/nouveau_bo.h" +#include "nouveau/nouveau_channel.h" +#include "nouveau/nouveau_class.h" +#include "nouveau/nouveau_device.h" +#include "nouveau/nouveau_grobj.h" +#include "nouveau/nouveau_notifier.h" +#include "nouveau/nouveau_resource.h" +#include "nouveau/nouveau_pushbuf.h" + +#define NOUVEAU_CAP_HW_VTXBUF (0xbeef0000) +#define NOUVEAU_CAP_HW_IDXBUF (0xbeef0001) + +#define NOUVEAU_TEXTURE_USAGE_LINEAR (1 << 16) + +#define NOUVEAU_BUFFER_USAGE_TEXTURE (1 << 16) +#define NOUVEAU_BUFFER_USAGE_ZETA (1 << 17) + +struct nouveau_winsys { + struct nouveau_context *nv; + + struct nouveau_channel *channel; + + int (*res_init)(struct nouveau_resource **heap, unsigned start, + unsigned size); + int (*res_alloc)(struct nouveau_resource *heap, int size, void *priv, + struct nouveau_resource **); + void (*res_free)(struct nouveau_resource **); + + int (*push_reloc)(struct nouveau_winsys *, void *ptr, + struct pipe_buffer *, uint32_t data, + uint32_t flags, uint32_t vor, uint32_t tor); + int (*push_flush)(struct nouveau_winsys *, unsigned size, + struct pipe_fence_handle **fence); + + int (*grobj_alloc)(struct nouveau_winsys *, int grclass, + struct nouveau_grobj **); + void (*grobj_free)(struct nouveau_grobj **); + + int (*notifier_alloc)(struct nouveau_winsys *, int count, + struct nouveau_notifier **); + void (*notifier_free)(struct nouveau_notifier **); + void (*notifier_reset)(struct nouveau_notifier *, int id); + uint32_t (*notifier_status)(struct nouveau_notifier *, int id); + uint32_t (*notifier_retval)(struct nouveau_notifier *, int id); + int (*notifier_wait)(struct nouveau_notifier *, int id, + int status, int timeout); + + int (*surface_copy)(struct nouveau_winsys *, struct pipe_surface *, + unsigned, unsigned, struct pipe_surface *, + unsigned, unsigned, unsigned, unsigned); + int (*surface_fill)(struct nouveau_winsys *, struct pipe_surface *, + unsigned, unsigned, unsigned, unsigned, unsigned); +}; + +extern struct pipe_screen * +nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); + +extern struct pipe_context * +nv04_create(struct pipe_screen *, unsigned pctx_id); + +extern struct pipe_screen * +nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); + +extern struct pipe_context * +nv10_create(struct pipe_screen *, unsigned pctx_id); + +extern struct pipe_screen * +nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); + +extern struct pipe_context * +nv20_create(struct pipe_screen *, unsigned pctx_id); + +extern struct pipe_screen * +nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); + +extern struct pipe_context * +nv30_create(struct pipe_screen *, unsigned pctx_id); + +extern struct pipe_screen * +nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); + +extern struct pipe_context * +nv40_create(struct pipe_screen *, unsigned pctx_id); + +extern struct pipe_screen * +nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); + +extern struct pipe_context * +nv50_create(struct pipe_screen *, unsigned pctx_id); + +#endif diff --git a/src/gallium/drivers/nv04/Makefile b/src/gallium/drivers/nv04/Makefile new file mode 100644 index 0000000000..5ea51a2f42 --- /dev/null +++ b/src/gallium/drivers/nv04/Makefile @@ -0,0 +1,28 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = nv04 + +DRIVER_SOURCES = \ + nv04_clear.c \ + nv04_context.c \ + nv04_fragprog.c \ + nv04_fragtex.c \ + nv04_miptree.c \ + nv04_prim_vbuf.c \ + nv04_screen.c \ + nv04_state.c \ + nv04_state_emit.c \ + nv04_surface.c \ + nv04_vbo.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/drivers/nv04/nv04_clear.c b/src/gallium/drivers/nv04/nv04_clear.c new file mode 100644 index 0000000000..01cacd36fe --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_clear.c @@ -0,0 +1,12 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "nv04_context.h" + +void +nv04_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); +} diff --git a/src/gallium/drivers/nv04/nv04_context.c b/src/gallium/drivers/nv04/nv04_context.c new file mode 100644 index 0000000000..9f75253363 --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_context.c @@ -0,0 +1,106 @@ +#include "draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" + +#include "nv04_context.h" +#include "nv04_screen.h" + +static void +nv04_flush(struct pipe_context *pipe, unsigned flags, + struct pipe_fence_handle **fence) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + draw_flush(nv04->draw); + + FIRE_RING(fence); +} + +static void +nv04_destroy(struct pipe_context *pipe) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + if (nv04->draw) + draw_destroy(nv04->draw); + + FREE(nv04); +} + +static void +nv04_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield) +{ +} + +static boolean +nv04_init_hwctx(struct nv04_context *nv04) +{ + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_NOTIFY, 1); + OUT_RING(0); + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_NOP, 1); + OUT_RING(0); + + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_CONTROL, 1); + OUT_RING(0x40182800); +// OUT_RING(1<<20/*no cull*/); + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_BLEND, 1); +// OUT_RING(0x24|(1<<6)|(1<<8)); + OUT_RING(0x120001a4); + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_FORMAT, 1); + OUT_RING(0x332213a1); + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_FILTER, 1); + OUT_RING(0x11001010); + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_COLORKEY, 1); + OUT_RING(0x0); +// BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_OFFSET, 1); +// OUT_RING(SCREEN_OFFSET); + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR, 1); + OUT_RING(0xff000000); + + + + FIRE_RING (NULL); + return TRUE; +} + +struct pipe_context * +nv04_create(struct pipe_screen *pscreen, unsigned pctx_id) +{ + struct nv04_screen *screen = nv04_screen(pscreen); + struct pipe_winsys *ws = pscreen->winsys; + struct nv04_context *nv04; + struct nouveau_winsys *nvws = screen->nvws; + + nv04 = CALLOC(1, sizeof(struct nv04_context)); + if (!nv04) + return NULL; + nv04->screen = screen; + nv04->pctx_id = pctx_id; + + nv04->nvws = nvws; + + nv04->pipe.winsys = ws; + nv04->pipe.screen = pscreen; + nv04->pipe.destroy = nv04_destroy; + nv04->pipe.set_edgeflags = nv04_set_edgeflags; + nv04->pipe.draw_arrays = nv04_draw_arrays; + nv04->pipe.draw_elements = nv04_draw_elements; + nv04->pipe.clear = nv04_clear; + nv04->pipe.flush = nv04_flush; + + nv04_init_surface_functions(nv04); + nv04_init_state_functions(nv04); + + nv04->draw = draw_create(); + assert(nv04->draw); + draw_wide_point_threshold(nv04->draw, 0.0); + draw_wide_line_threshold(nv04->draw, 0.0); + draw_enable_line_stipple(nv04->draw, FALSE); + draw_enable_point_sprites(nv04->draw, FALSE); + draw_set_rasterize_stage(nv04->draw, nv04_draw_vbuf_stage(nv04)); + + nv04_init_hwctx(nv04); + + return &nv04->pipe; +} + diff --git a/src/gallium/drivers/nv04/nv04_context.h b/src/gallium/drivers/nv04/nv04_context.h new file mode 100644 index 0000000000..3e6a085270 --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_context.h @@ -0,0 +1,146 @@ +#ifndef __NV04_CONTEXT_H__ +#define __NV04_CONTEXT_H__ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_compiler.h" + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "draw/draw_vertex.h" + +#include "nouveau/nouveau_winsys.h" +#include "nouveau/nouveau_gldefs.h" + +#define NOUVEAU_PUSH_CONTEXT(ctx) \ + struct nv04_screen *ctx = nv04->screen +#include "nouveau/nouveau_push.h" + +#include "nv04_state.h" + +#define NOUVEAU_ERR(fmt, args...) \ + fprintf(stderr, "%s:%d - "fmt, __func__, __LINE__, ##args); +#define NOUVEAU_MSG(fmt, args...) \ + fprintf(stderr, "nouveau: "fmt, ##args); + +#include "nv04_screen.h" + +#define NV04_NEW_VERTPROG (1 << 1) +#define NV04_NEW_FRAGPROG (1 << 2) +#define NV04_NEW_BLEND (1 << 3) +#define NV04_NEW_RAST (1 << 4) +#define NV04_NEW_CONTROL (1 << 5) +#define NV04_NEW_VIEWPORT (1 << 6) +#define NV04_NEW_SAMPLER (1 << 7) + +struct nv04_context { + struct pipe_context pipe; + + struct nouveau_winsys *nvws; + struct nv04_screen *screen; + unsigned pctx_id; + + struct draw_context *draw; + + int chipset; + struct nouveau_notifier *sync; + + uint32_t dirty; + + struct nv04_blend_state *blend; + struct nv04_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + struct nv04_fragtex_state fragtex; + struct nv04_rasterizer_state *rast; + struct nv04_depth_stencil_alpha_state *dsa; + + struct nv04_miptree *tex_miptree[PIPE_MAX_SAMPLERS]; + unsigned dirty_samplers; + unsigned fp_samplers; + unsigned vp_samplers; + + uint32_t rt_enable; + struct pipe_buffer *rt[4]; + struct pipe_buffer *zeta; + + struct { + struct pipe_buffer *buffer; + uint32_t format; + } tex[16]; + + unsigned vb_enable; + struct { + struct pipe_buffer *buffer; + unsigned delta; + } vb[16]; + + struct vertex_info vertex_info; + struct { + + struct nouveau_resource *exec_heap; + struct nouveau_resource *data_heap; + + struct nv04_vertex_program *active; + + struct nv04_vertex_program *current; + struct pipe_buffer *constant_buf; + } vertprog; + + struct { + struct nv04_fragment_program *active; + + struct nv04_fragment_program *current; + struct pipe_buffer *constant_buf; + } fragprog; + + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + unsigned num_vertex_buffers; + unsigned num_vertex_elements; + + struct pipe_viewport_state viewport; +}; + +static INLINE struct nv04_context * +nv04_context(struct pipe_context *pipe) +{ + return (struct nv04_context *)pipe; +} + +extern void nv04_init_state_functions(struct nv04_context *nv04); +extern void nv04_init_surface_functions(struct nv04_context *nv04); +extern void nv04_init_miptree_functions(struct pipe_screen *screen); + +/* nv04_clear.c */ +extern void nv04_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue); + +/* nv04_draw.c */ +extern struct draw_stage *nv04_draw_render_stage(struct nv04_context *nv04); + +/* nv04_fragprog.c */ +extern void nv04_fragprog_bind(struct nv04_context *, + struct nv04_fragment_program *); +extern void nv04_fragprog_destroy(struct nv04_context *, + struct nv04_fragment_program *); + +/* nv04_fragtex.c */ +extern void nv04_fragtex_bind(struct nv04_context *); + +/* nv04_prim_vbuf.c */ +struct draw_stage *nv04_draw_vbuf_stage( struct nv04_context *nv04 ); + +/* nv04_state.c and friends */ +extern void nv04_emit_hw_state(struct nv04_context *nv04); +extern void nv04_state_tex_update(struct nv04_context *nv04); + +/* nv04_vbo.c */ +extern boolean nv04_draw_arrays(struct pipe_context *, unsigned mode, + unsigned start, unsigned count); +extern boolean nv04_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count); + + +#endif diff --git a/src/gallium/drivers/nv04/nv04_fragprog.c b/src/gallium/drivers/nv04/nv04_fragprog.c new file mode 100644 index 0000000000..8a2af41fe0 --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_fragprog.c @@ -0,0 +1,21 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" + +#include "nv04_context.h" + +void +nv04_fragprog_bind(struct nv04_context *nv04, struct nv04_fragment_program *fp) +{ +} + +void +nv04_fragprog_destroy(struct nv04_context *nv04, + struct nv04_fragment_program *fp) +{ +} + diff --git a/src/gallium/drivers/nv04/nv04_fragtex.c b/src/gallium/drivers/nv04/nv04_fragtex.c new file mode 100644 index 0000000000..21f990fd53 --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_fragtex.c @@ -0,0 +1,73 @@ +#include "nv04_context.h" +#include "nouveau/nouveau_util.h" + +#define _(m,tf) \ +{ \ + PIPE_FORMAT_##m, \ + NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_##tf, \ +} + +struct nv04_texture_format { + uint pipe; + int format; +}; + +static struct nv04_texture_format +nv04_texture_formats[] = { + _(A8R8G8B8_UNORM, A8R8G8B8), + _(X8R8G8B8_UNORM, X8R8G8B8), + _(A1R5G5B5_UNORM, A1R5G5B5), + _(A4R4G4B4_UNORM, A4R4G4B4), + _(L8_UNORM, Y8 ), + _(A8_UNORM, Y8 ), +}; + +static uint32_t +nv04_fragtex_format(uint pipe_format) +{ + struct nv04_texture_format *tf = nv04_texture_formats; + int i; + + for (i=0; i< sizeof(nv04_texture_formats)/sizeof(nv04_texture_formats[0]); i++) { + if (tf->pipe == pipe_format) + return tf->format; + tf++; + } + + NOUVEAU_ERR("unknown texture format %s\n", pf_name(pipe_format)); + return 0; +} + + +static void +nv04_fragtex_build(struct nv04_context *nv04, int unit) +{ + struct nv04_miptree *nv04mt = nv04->tex_miptree[unit]; + struct pipe_texture *pt = &nv04mt->base; + + switch (pt->target) { + case PIPE_TEXTURE_2D: + break; + default: + NOUVEAU_ERR("Unknown target %d\n", pt->target); + return; + } + + nv04->fragtex.format = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_ZOH_CORNER + | NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_FOH_CORNER + | nv04_fragtex_format(pt->format) + | ( (pt->last_level + 1) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_MIPMAP_LEVELS_SHIFT ) + | ( log2i(pt->width[0]) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_U_SHIFT ) + | ( log2i(pt->height[0]) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_V_SHIFT ) + | NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP_TO_EDGE + | NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_CLAMP_TO_EDGE + ; +} + + +void +nv04_fragtex_bind(struct nv04_context *nv04) +{ + nv04_fragtex_build(nv04, 0); +} + diff --git a/src/gallium/drivers/nv04/nv04_miptree.c b/src/gallium/drivers/nv04/nv04_miptree.c new file mode 100644 index 0000000000..0cbb91e187 --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_miptree.c @@ -0,0 +1,138 @@ +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "nv04_context.h" +#include "nv04_screen.h" + +static void +nv04_miptree_layout(struct nv04_miptree *nv04mt) +{ + struct pipe_texture *pt = &nv04mt->base; + uint width = pt->width[0], height = pt->height[0]; + uint offset = 0; + int nr_faces, l, f; + + nr_faces = 1; + + for (l = 0; l <= pt->last_level; l++) { + pt->width[l] = width; + pt->height[l] = height; + + pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); + pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); + + nv04mt->level[l].pitch = pt->width[0] * pt->block.size; + nv04mt->level[l].pitch = (nv04mt->level[l].pitch + 63) & ~63; + + nv04mt->level[l].image_offset = + CALLOC(nr_faces, sizeof(unsigned)); + + width = MAX2(1, width >> 1); + height = MAX2(1, height >> 1); + + } + + for (f = 0; f < nr_faces; f++) { + for (l = 0; l <= pt->last_level; l++) { + nv04mt->level[l].image_offset[f] = offset; + offset += nv04mt->level[l].pitch * pt->height[l]; + } + } + + nv04mt->total_size = offset; +} + +static struct pipe_texture * +nv04_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) +{ + struct pipe_winsys *ws = screen->winsys; + struct nv04_miptree *mt; + + mt = MALLOC(sizeof(struct nv04_miptree)); + if (!mt) + return NULL; + mt->base = *pt; + mt->base.refcount = 1; + mt->base.screen = screen; + + nv04_miptree_layout(mt); + + mt->buffer = ws->buffer_create(ws, 256, PIPE_BUFFER_USAGE_PIXEL, + mt->total_size); + if (!mt->buffer) { + FREE(mt); + return NULL; + } + + return &mt->base; +} + +static void +nv04_miptree_release(struct pipe_screen *screen, struct pipe_texture **pt) +{ + struct pipe_texture *mt = *pt; + + *pt = NULL; + if (--mt->refcount <= 0) { + struct nv04_miptree *nv04mt = (struct nv04_miptree *)mt; + int l; + + pipe_buffer_reference(screen, &nv04mt->buffer, NULL); + for (l = 0; l <= mt->last_level; l++) { + if (nv04mt->level[l].image_offset) + FREE(nv04mt->level[l].image_offset); + } + FREE(nv04mt); + } +} + +static struct pipe_surface * +nv04_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + unsigned flags) +{ + struct pipe_winsys *ws = pscreen->winsys; + struct nv04_miptree *nv04mt = (struct nv04_miptree *)pt; + struct pipe_surface *ps; + + ps = ws->surface_alloc(ws); + if (!ps) + return NULL; + pipe_buffer_reference(pscreen, &ps->buffer, nv04mt->buffer); + ps->format = pt->format; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->block = pt->block; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = nv04mt->level[level].pitch; + ps->refcount = 1; + ps->winsys = pscreen->winsys; + + if (pt->target == PIPE_TEXTURE_CUBE) { + ps->offset = nv04mt->level[level].image_offset[face]; + } else { + ps->offset = nv04mt->level[level].image_offset[0]; + } + + return ps; +} + +static void +nv04_miptree_surface_del(struct pipe_screen *pscreen, + struct pipe_surface **psurface) +{ +} + +void +nv04_init_miptree_functions(struct pipe_screen *pscreen) +{ + pscreen->texture_create = nv04_miptree_create; + pscreen->texture_release = nv04_miptree_release; + pscreen->get_tex_surface = nv04_miptree_surface_new; + pscreen->tex_surface_release = nv04_miptree_surface_del; +} + diff --git a/src/gallium/drivers/nv04/nv04_prim_vbuf.c b/src/gallium/drivers/nv04/nv04_prim_vbuf.c new file mode 100644 index 0000000000..19979fff79 --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_prim_vbuf.c @@ -0,0 +1,309 @@ + +#include "pipe/p_debug.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "pipe/p_compiler.h" + +#include "draw/draw_vbuf.h" + +#include "nv04_context.h" +#include "nv04_state.h" + +#define VERTEX_SIZE 40 +#define VERTEX_BUFFER_SIZE (4096*VERTEX_SIZE) // 4096 vertices of 40 bytes each + +/** + * Primitive renderer for nv04. + */ +struct nv04_vbuf_render { + struct vbuf_render base; + + struct nv04_context *nv04; + + /** Vertex buffer */ + unsigned char* buffer; + + /** Vertex size in bytes */ + unsigned vertex_size; + + /** Current primitive */ + unsigned prim; +}; + + +/** + * Basically a cast wrapper. + */ +static INLINE struct nv04_vbuf_render * +nv04_vbuf_render( struct vbuf_render *render ) +{ + assert(render); + return (struct nv04_vbuf_render *)render; +} + + +static const struct vertex_info * +nv04_vbuf_render_get_vertex_info( struct vbuf_render *render ) +{ + struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render); + struct nv04_context *nv04 = nv04_render->nv04; + return &nv04->vertex_info; +} + + +static void * +nv04_vbuf_render_allocate_vertices( struct vbuf_render *render, + ushort vertex_size, + ushort nr_vertices ) +{ + struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render); + + nv04_render->buffer = (unsigned char*) MALLOC(VERTEX_BUFFER_SIZE); + assert(!nv04_render->buffer); + + return nv04_render->buffer; +} + + +static boolean +nv04_vbuf_render_set_primitive( struct vbuf_render *render, + unsigned prim ) +{ + struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render); + + if (prim <= PIPE_PRIM_LINE_STRIP) + return FALSE; + + nv04_render->prim = prim; + return TRUE; +} + +static INLINE void nv04_2triangles(struct nv04_context* nv04, unsigned char* buffer, ushort v0, ushort v1, ushort v2, ushort v3, ushort v4, ushort v5) +{ + BEGIN_RING(fahrenheit,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0xA),49); + OUT_RINGp(buffer + VERTEX_SIZE * v0,8); + OUT_RINGp(buffer + VERTEX_SIZE * v1,8); + OUT_RINGp(buffer + VERTEX_SIZE * v2,8); + OUT_RINGp(buffer + VERTEX_SIZE * v3,8); + OUT_RINGp(buffer + VERTEX_SIZE * v4,8); + OUT_RINGp(buffer + VERTEX_SIZE * v5,8); + OUT_RING(0xFEDCBA); +} + +static INLINE void nv04_1triangle(struct nv04_context* nv04, unsigned char* buffer, ushort v0, ushort v1, ushort v2) +{ + BEGIN_RING(fahrenheit,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0xD),25); + OUT_RINGp(buffer + VERTEX_SIZE * v0,8); + OUT_RINGp(buffer + VERTEX_SIZE * v1,8); + OUT_RINGp(buffer + VERTEX_SIZE * v2,8); + OUT_RING(0xFED); +} + +static INLINE void nv04_1quad(struct nv04_context* nv04, unsigned char* buffer, ushort v0, ushort v1, ushort v2, ushort v3) +{ + BEGIN_RING(fahrenheit,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0xC),33); + OUT_RINGp(buffer + VERTEX_SIZE * v0,8); + OUT_RINGp(buffer + VERTEX_SIZE * v1,8); + OUT_RINGp(buffer + VERTEX_SIZE * v2,8); + OUT_RINGp(buffer + VERTEX_SIZE * v3,8); + OUT_RING(0xFECEDC); +} + +static void nv04_vbuf_render_triangles_elts(struct nv04_vbuf_render * render, const ushort * indices, uint nr_indices) +{ + unsigned char* buffer = render->buffer; + struct nv04_context* nv04 = render->nv04; + int i; + + for( i=0; i< nr_indices-5; i+=6) + nv04_2triangles(nv04, + buffer, + indices[i+0], + indices[i+1], + indices[i+2], + indices[i+3], + indices[i+4], + indices[i+5] + ); + if (i != nr_indices) + { + nv04_1triangle(nv04, + buffer, + indices[i+0], + indices[i+1], + indices[i+2] + ); + i+=3; + } + if (i != nr_indices) + NOUVEAU_ERR("Houston, we have lost some vertices\n"); +} + +static void nv04_vbuf_render_tri_strip_elts(struct nv04_vbuf_render* render, const ushort* indices, uint nr_indices) +{ + const uint32_t striptbl[]={0x321210,0x543432,0x765654,0x987876,0xBA9A98,0xDCBCBA,0xFEDEDC}; + unsigned char* buffer = render->buffer; + struct nv04_context* nv04 = render->nv04; + int i,j; + + for(i = 0; i<nr_indices; i+=14) + { + int numvert = MIN2(16, nr_indices - i); + int numtri = numvert - 2; + if (numvert<3) + break; + + BEGIN_RING( fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x0), numvert*8 ); + for(j = 0; j<numvert; j++) + OUT_RINGp( buffer + VERTEX_SIZE * indices [i+j], 8 ); + + BEGIN_RING_NI( fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE(0), (numtri+1)/2 ); + for(j = 0; j<numtri/2; j++ ) + OUT_RING(striptbl[j]); + if (numtri%2) + OUT_RING(striptbl[numtri/2]&0xFFF); + } +} + +static void nv04_vbuf_render_tri_fan_elts(struct nv04_vbuf_render* render, const ushort* indices, uint nr_indices) +{ + const uint32_t fantbl[]={0x320210,0x540430,0x760650,0x980870,0xBA0A90,0xDC0CB0,0xFE0ED0}; + unsigned char* buffer = render->buffer; + struct nv04_context* nv04 = render->nv04; + int i,j; + + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x0), 8); + OUT_RINGp(buffer + VERTEX_SIZE * indices[0], 8); + + for(i = 1; i<nr_indices; i+=14) + { + int numvert=MIN2(15, nr_indices - i); + int numtri=numvert-2; + if (numvert < 3) + break; + + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x1), numvert*8); + + for(j=0;j<numvert;j++) + OUT_RINGp( buffer + VERTEX_SIZE * indices[ i+j ], 8 ); + + BEGIN_RING_NI(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE(0), (numtri+1)/2); + for(j = 0; j<numtri/2; j++) + OUT_RING(fantbl[j]); + if (numtri%2) + OUT_RING(fantbl[numtri/2]&0xFFF); + } +} + +static void nv04_vbuf_render_quads_elts(struct nv04_vbuf_render* render, const ushort* indices, uint nr_indices) +{ + unsigned char* buffer = render->buffer; + struct nv04_context* nv04 = render->nv04; + int i; + + for(i = 0; i < nr_indices; i += 4) + nv04_1quad(nv04, + buffer, + indices[i+0], + indices[i+1], + indices[i+2], + indices[i+3] + ); +} + + +static void +nv04_vbuf_render_draw( struct vbuf_render *render, + const ushort *indices, + uint nr_indices) +{ + struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render); + + // emit the indices + switch( nv04_render->prim ) + { + case PIPE_PRIM_TRIANGLES: + nv04_vbuf_render_triangles_elts(nv04_render, indices, nr_indices); + break; + case PIPE_PRIM_QUAD_STRIP: + case PIPE_PRIM_TRIANGLE_STRIP: + nv04_vbuf_render_tri_strip_elts(nv04_render, indices, nr_indices); + break; + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + nv04_vbuf_render_tri_fan_elts(nv04_render, indices, nr_indices); + break; + case PIPE_PRIM_QUADS: + nv04_vbuf_render_quads_elts(nv04_render, indices, nr_indices); + break; + default: + NOUVEAU_ERR("You have to implement primitive %d, young padawan\n", nv04_render->prim); + break; + } +} + + +static void +nv04_vbuf_render_release_vertices( struct vbuf_render *render, + void *vertices, + unsigned vertex_size, + unsigned vertices_used ) +{ + struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render); + + free(nv04_render->buffer); + nv04_render->buffer = NULL; +} + + +static void +nv04_vbuf_render_destroy( struct vbuf_render *render ) +{ + struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render); + FREE(nv04_render); +} + + +/** + * Create a new primitive render. + */ +static struct vbuf_render * +nv04_vbuf_render_create( struct nv04_context *nv04 ) +{ + struct nv04_vbuf_render *nv04_render = CALLOC_STRUCT(nv04_vbuf_render); + + nv04_render->nv04 = nv04; + + nv04_render->base.max_vertex_buffer_bytes = VERTEX_BUFFER_SIZE; + nv04_render->base.max_indices = 65536; + nv04_render->base.get_vertex_info = nv04_vbuf_render_get_vertex_info; + nv04_render->base.allocate_vertices = nv04_vbuf_render_allocate_vertices; + nv04_render->base.set_primitive = nv04_vbuf_render_set_primitive; + nv04_render->base.draw = nv04_vbuf_render_draw; + nv04_render->base.release_vertices = nv04_vbuf_render_release_vertices; + nv04_render->base.destroy = nv04_vbuf_render_destroy; + + return &nv04_render->base; +} + + +/** + * Create a new primitive vbuf/render stage. + */ +struct draw_stage *nv04_draw_vbuf_stage( struct nv04_context *nv04 ) +{ + struct vbuf_render *render; + struct draw_stage *stage; + + render = nv04_vbuf_render_create(nv04); + if(!render) + return NULL; + + stage = draw_vbuf_stage( nv04->draw, render ); + if(!stage) { + render->destroy(render); + return NULL; + } + + return stage; +} diff --git a/src/gallium/drivers/nv04/nv04_screen.c b/src/gallium/drivers/nv04/nv04_screen.c new file mode 100644 index 0000000000..65eacde6b2 --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_screen.c @@ -0,0 +1,214 @@ +#include "pipe/p_screen.h" + +#include "nv04_context.h" +#include "nv04_screen.h" + +static const char * +nv04_screen_get_name(struct pipe_screen *screen) +{ + struct nv04_screen *nv04screen = nv04_screen(screen); + struct nouveau_device *dev = nv04screen->nvws->channel->device; + static char buffer[128]; + + snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); + return buffer; +} + +static const char * +nv04_screen_get_vendor(struct pipe_screen *screen) +{ + return "nouveau"; +} + +static int +nv04_screen_get_param(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 1; + case PIPE_CAP_NPOT_TEXTURES: + return 0; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 0; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 0; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 10; + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 0; + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 0; + case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: + return 0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0; + } +} + +static float +nv04_screen_get_paramf(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 0.0; + case PIPE_CAP_MAX_POINT_WIDTH: + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 0.0; + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 0.0; + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 0.0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0.0; + } +} + +static boolean +nv04_screen_is_format_supported(struct pipe_screen *screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, unsigned geom_flags) +{ + if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_Z16_UNORM: + return TRUE; + default: + break; + } + } else { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_X8R8G8B8_UNORM: + case PIPE_FORMAT_A1R5G5B5_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_A8_UNORM: + return TRUE; + default: + break; + } + } + + return FALSE; +} + +static void * +nv04_surface_map(struct pipe_screen *screen, struct pipe_surface *surface, + unsigned flags ) +{ + struct pipe_winsys *ws = screen->winsys; + void *map; + + map = ws->buffer_map(ws, surface->buffer, flags); + if (!map) + return NULL; + + return map + surface->offset; +} + +static void +nv04_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) +{ + struct pipe_winsys *ws = screen->winsys; + + ws->buffer_unmap(ws, surface->buffer); +} + +static void +nv04_screen_destroy(struct pipe_screen *pscreen) +{ + struct nv04_screen *screen = nv04_screen(pscreen); + struct nouveau_winsys *nvws = screen->nvws; + + nvws->notifier_free(&screen->sync); + nvws->grobj_free(&screen->fahrenheit); + + FREE(pscreen); +} + +struct pipe_screen * +nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +{ + struct nv04_screen *screen = CALLOC_STRUCT(nv04_screen); + unsigned fahrenheit_class = 0, sub3d_class = 0; + unsigned chipset = nvws->channel->device->chipset; + int ret; + + if (!screen) + return NULL; + screen->nvws = nvws; + + if (chipset>=0x20) { + fahrenheit_class = 0; + sub3d_class = 0; + } else if (chipset>=0x10) { + fahrenheit_class = NV10_DX5_TEXTURED_TRIANGLE; + sub3d_class = NV10_CONTEXT_SURFACES_3D; + } else { + fahrenheit_class=NV04_DX5_TEXTURED_TRIANGLE; + sub3d_class = NV04_CONTEXT_SURFACES_3D; + } + + if (!fahrenheit_class) { + NOUVEAU_ERR("Unknown nv04 chipset: nv%02x\n", chipset); + return NULL; + } + + /* 3D object */ + ret = nvws->grobj_alloc(nvws, fahrenheit_class, &screen->fahrenheit); + if (ret) { + NOUVEAU_ERR("Error creating 3D object: %d\n", ret); + return NULL; + } + + /* 3D surface object */ + ret = nvws->grobj_alloc(nvws, sub3d_class, &screen->context_surfaces_3d); + if (ret) { + NOUVEAU_ERR("Error creating 3D surface object: %d\n", ret); + return NULL; + } + + /* Notifier for sync purposes */ + ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + if (ret) { + NOUVEAU_ERR("Error creating notifier object: %d\n", ret); + nv04_screen_destroy(&screen->pipe); + return NULL; + } + + screen->pipe.winsys = ws; + screen->pipe.destroy = nv04_screen_destroy; + + screen->pipe.get_name = nv04_screen_get_name; + screen->pipe.get_vendor = nv04_screen_get_vendor; + screen->pipe.get_param = nv04_screen_get_param; + screen->pipe.get_paramf = nv04_screen_get_paramf; + + screen->pipe.is_format_supported = nv04_screen_is_format_supported; + + screen->pipe.surface_map = nv04_surface_map; + screen->pipe.surface_unmap = nv04_surface_unmap; + + nv04_screen_init_miptree_functions(&screen->pipe); + + return &screen->pipe; +} + diff --git a/src/gallium/drivers/nv04/nv04_screen.h b/src/gallium/drivers/nv04/nv04_screen.h new file mode 100644 index 0000000000..99a49cdf7a --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_screen.h @@ -0,0 +1,25 @@ +#ifndef __NV04_SCREEN_H__ +#define __NV04_SCREEN_H__ + +#include "pipe/p_screen.h" + +struct nv04_screen { + struct pipe_screen pipe; + + struct nouveau_winsys *nvws; + unsigned chipset; + + /* HW graphics objects */ + struct nouveau_grobj *fahrenheit; + struct nouveau_grobj *context_surfaces_3d; + struct nouveau_notifier *sync; + +}; + +static INLINE struct nv04_screen * +nv04_screen(struct pipe_screen *screen) +{ + return (struct nv04_screen *)screen; +} + +#endif diff --git a/src/gallium/drivers/nv04/nv04_state.c b/src/gallium/drivers/nv04/nv04_state.c new file mode 100644 index 0000000000..ff1933b550 --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_state.c @@ -0,0 +1,495 @@ +#include "draw/draw_context.h" +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/tgsi_parse.h" + +#include "nv04_context.h" +#include "nv04_state.h" + +static void * +nv04_blend_state_create(struct pipe_context *pipe, + const struct pipe_blend_state *cso) +{ + struct nv04_blend_state *cb; + + cb = MALLOC(sizeof(struct nv04_blend_state)); + + cb->b_enable = cso->blend_enable ? 1 : 0; + cb->b_src = ((nvgl_blend_func(cso->alpha_src_factor)<<16) | + (nvgl_blend_func(cso->rgb_src_factor))); + cb->b_dst = ((nvgl_blend_func(cso->alpha_dst_factor)<<16) | + (nvgl_blend_func(cso->rgb_dst_factor))); + + + return (void *)cb; +} + +static void +nv04_blend_state_bind(struct pipe_context *pipe, void *blend) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + nv04->blend = (struct nv04_blend_state*)blend; + + nv04->dirty |= NV04_NEW_BLEND; +} + +static void +nv04_blend_state_delete(struct pipe_context *pipe, void *hwcso) +{ + free(hwcso); +} + + +static INLINE unsigned +wrap_mode(unsigned wrap) { + unsigned ret; + + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_REPEAT; + break; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_MIRRORED_REPEAT; + break; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP_TO_EDGE; + break; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP_TO_BORDER; + break; + case PIPE_TEX_WRAP_CLAMP: + ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + case PIPE_TEX_WRAP_MIRROR_CLAMP: + default: + NOUVEAU_ERR("unknown wrap mode: %d\n", wrap); + ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP; + } + return ret >> NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_SHIFT; +} + +static void * +nv04_sampler_state_create(struct pipe_context *pipe, + const struct pipe_sampler_state *cso) +{ + + struct nv04_sampler_state *ss; + uint32_t filter = 0; + + ss = MALLOC(sizeof(struct nv04_sampler_state)); + + ss->format = ((wrap_mode(cso->wrap_s) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_SHIFT) | + (wrap_mode(cso->wrap_t) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_SHIFT)); + + if (cso->max_anisotropy > 1.0) { + filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_ANISOTROPIC_MINIFY_ENABLE | NV04_DX5_TEXTURED_TRIANGLE_FILTER_ANISOTROPIC_MAGNIFY_ENABLE; + } + + switch (cso->mag_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MAGNIFY_LINEAR; + break; + case PIPE_TEX_FILTER_NEAREST: + default: + filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MAGNIFY_NEAREST; + break; + } + + switch (cso->min_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_LINEAR; + break; + } + break; + case PIPE_TEX_FILTER_NEAREST: + default: + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_NEAREST; + break; + } + break; + } + + ss->filter = filter; + + return (void *)ss; +} + +static void +nv04_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler) +{ + struct nv04_context *nv04 = nv04_context(pipe); + unsigned unit; + + for (unit = 0; unit < nr; unit++) { + nv04->sampler[unit] = sampler[unit]; + nv04->dirty_samplers |= (1 << unit); + } +} + +static void +nv04_sampler_state_delete(struct pipe_context *pipe, void *hwcso) +{ + free(hwcso); +} + +static void +nv04_set_sampler_texture(struct pipe_context *pipe, unsigned nr, + struct pipe_texture **miptree) +{ + struct nv04_context *nv04 = nv04_context(pipe); + unsigned unit; + + for (unit = 0; unit < nr; unit++) { + nv04->tex_miptree[unit] = (struct nv04_miptree *)miptree[unit]; + nv04->dirty_samplers |= (1 << unit); + } +} + +static void * +nv04_rasterizer_state_create(struct pipe_context *pipe, + const struct pipe_rasterizer_state *cso) +{ + struct nv04_rasterizer_state *rs; + + /*XXX: ignored: + * scissor + * points/lines (no hw support, emulated with tris in gallium) + */ + rs = MALLOC(sizeof(struct nv04_rasterizer_state)); + + rs->blend = cso->flatshade ? NV04_DX5_TEXTURED_TRIANGLE_BLEND_SHADE_MODE_FLAT : NV04_DX5_TEXTURED_TRIANGLE_BLEND_SHADE_MODE_GOURAUD; + + return (void *)rs; +} + +static void +nv04_rasterizer_state_bind(struct pipe_context *pipe, void *rast) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + nv04->rast = (struct nv04_rasterizer_state*)rast; + + draw_set_rasterizer_state(nv04->draw, (nv04->rast ? nv04->rast->templ : NULL)); + + nv04->dirty |= NV04_NEW_RAST | NV04_NEW_BLEND; +} + +static void +nv04_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso) +{ + free(hwcso); +} + +static INLINE uint32_t nv04_compare_func(uint32_t f) +{ + switch ( f ) { + case PIPE_FUNC_NEVER: return 1; + case PIPE_FUNC_LESS: return 2; + case PIPE_FUNC_EQUAL: return 3; + case PIPE_FUNC_LEQUAL: return 4; + case PIPE_FUNC_GREATER: return 5; + case PIPE_FUNC_NOTEQUAL: return 6; + case PIPE_FUNC_GEQUAL: return 7; + case PIPE_FUNC_ALWAYS: return 8; + } + NOUVEAU_MSG("Unable to find the function\n"); + return 0; +} + +static void * +nv04_depth_stencil_alpha_state_create(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *cso) +{ + struct nv04_depth_stencil_alpha_state *hw; + + hw = MALLOC(sizeof(struct nv04_depth_stencil_alpha_state)); + + hw->control = float_to_ubyte(cso->alpha.ref); + hw->control |= ( nv04_compare_func(cso->alpha.func) << NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ALPHA_FUNC_SHIFT ); + hw->control |= cso->alpha.enabled ? NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ALPHA_TEST_ENABLE : 0; + hw->control |= NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ORIGIN; + hw->control |= cso->depth.enabled ? (1 << NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_ENABLE_SHIFT) : 0; + hw->control |= ( nv04_compare_func(cso->depth.func)<< NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_FUNC_SHIFT ); + hw->control |= 1 << NV04_DX5_TEXTURED_TRIANGLE_CONTROL_CULL_MODE_SHIFT; // no culling, handled by the draw module + hw->control |= NV04_DX5_TEXTURED_TRIANGLE_CONTROL_DITHER_ENABLE; + hw->control |= NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_PERSPECTIVE_ENABLE; + hw->control |= cso->depth.writemask ? (1 << NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_WRITE_ENABLE_SHIFT) : 0; + hw->control |= 1 << NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_FORMAT_SHIFT; // integer zbuffer format + + return (void *)hw; +} + +static void +nv04_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + nv04->dsa = hwcso; + nv04->dirty |= NV04_NEW_CONTROL; +} + +static void +nv04_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso) +{ + free(hwcso); +} + +static void * +nv04_vp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + return draw_create_vertex_shader(nv04->draw, templ); +} + +static void +nv04_vp_state_bind(struct pipe_context *pipe, void *shader) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + draw_bind_vertex_shader(nv04->draw, (struct draw_vertex_shader *) shader); + + nv04->dirty |= NV04_NEW_VERTPROG; +} + +static void +nv04_vp_state_delete(struct pipe_context *pipe, void *shader) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + draw_delete_vertex_shader(nv04->draw, (struct draw_vertex_shader *) shader); +} + +static void * +nv04_fp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *cso) +{ + struct nv04_fragment_program *fp; + + fp = CALLOC(1, sizeof(struct nv04_fragment_program)); + fp->pipe.tokens = tgsi_dup_tokens(cso->tokens); + + return (void *)fp; +} + +static void +nv04_fp_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv04_context *nv04 = nv04_context(pipe); + struct nv04_fragment_program *fp = hwcso; + + nv04->fragprog.current = fp; + nv04->dirty |= NV04_NEW_FRAGPROG; +} + +static void +nv04_fp_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv04_context *nv04 = nv04_context(pipe); + struct nv04_fragment_program *fp = hwcso; + + nv04_fragprog_destroy(nv04, fp); + free((void*)fp->pipe.tokens); + free(fp); +} + +static void +nv04_set_blend_color(struct pipe_context *pipe, + const struct pipe_blend_color *bcol) +{ +} + +static void +nv04_set_clip_state(struct pipe_context *pipe, + const struct pipe_clip_state *clip) +{ +} + +static void +nv04_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, + const struct pipe_constant_buffer *buf ) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + if (shader == PIPE_SHADER_VERTEX) { + nv04->vertprog.constant_buf = buf->buffer; + nv04->dirty |= NV04_NEW_VERTPROG; + } else + if (shader == PIPE_SHADER_FRAGMENT) { + nv04->fragprog.constant_buf = buf->buffer; + nv04->dirty |= NV04_NEW_FRAGPROG; + } +} + +static void +nv04_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct nv04_context *nv04 = nv04_context(pipe); + struct pipe_surface *rt, *zeta; + uint32_t rt_format, w, h; + int colour_format = 0, zeta_format = 0; + + w = fb->cbufs[0]->width; + h = fb->cbufs[0]->height; + colour_format = fb->cbufs[0]->format; + rt = fb->cbufs[0]; + + if (fb->zsbuf) { + if (colour_format) { + assert(w == fb->zsbuf->width); + assert(h == fb->zsbuf->height); + } else { + w = fb->zsbuf->width; + h = fb->zsbuf->height; + } + + zeta_format = fb->zsbuf->format; + zeta = fb->zsbuf; + } + + switch (colour_format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case 0: + rt_format = 0x108; + break; + case PIPE_FORMAT_R5G6B5_UNORM: + rt_format = 0x103; + break; + default: + assert(0); + } + + BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_FORMAT, 1); + OUT_RING(rt_format); + + /* FIXME pitches have to be aligned ! */ + BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_PITCH, 2); + OUT_RING(rt->stride|(zeta->stride<<16)); + OUT_RELOCl(rt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + if (fb->zsbuf) { + BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_OFFSET_ZETA, 1); + OUT_RELOCl(zeta->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + } +} + +static void +nv04_set_polygon_stipple(struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple) +{ + NOUVEAU_ERR("line stipple hahaha\n"); +} + +static void +nv04_set_scissor_state(struct pipe_context *pipe, + const struct pipe_scissor_state *s) +{ +/* struct nv04_context *nv04 = nv04_context(pipe); + + // XXX + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_SCISSOR_HORIZ, 2); + OUT_RING (((s->maxx - s->minx) << 16) | s->minx); + OUT_RING (((s->maxy - s->miny) << 16) | s->miny);*/ +} + +static void +nv04_set_viewport_state(struct pipe_context *pipe, + const struct pipe_viewport_state *viewport) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + nv04->viewport = *viewport; + + draw_set_viewport_state(nv04->draw, &nv04->viewport); +} + +static void +nv04_set_vertex_buffers(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_buffer *buffers) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + draw_flush(nv04->draw); + + memcpy(nv04->vertex_buffer, buffers, count * sizeof(buffers[0])); + nv04->num_vertex_buffers = count; + + draw_set_vertex_buffers(nv04->draw, count, buffers); +} + +static void +nv04_set_vertex_elements(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_element *elements) +{ + struct nv04_context *nv04 = nv04_context(pipe); + + draw_flush(nv04->draw); + + nv04->num_vertex_elements = count; + draw_set_vertex_elements(nv04->draw, count, elements); +} + +void +nv04_init_state_functions(struct nv04_context *nv04) +{ + nv04->pipe.create_blend_state = nv04_blend_state_create; + nv04->pipe.bind_blend_state = nv04_blend_state_bind; + nv04->pipe.delete_blend_state = nv04_blend_state_delete; + + nv04->pipe.create_sampler_state = nv04_sampler_state_create; + nv04->pipe.bind_sampler_states = nv04_sampler_state_bind; + nv04->pipe.delete_sampler_state = nv04_sampler_state_delete; + nv04->pipe.set_sampler_textures = nv04_set_sampler_texture; + + nv04->pipe.create_rasterizer_state = nv04_rasterizer_state_create; + nv04->pipe.bind_rasterizer_state = nv04_rasterizer_state_bind; + nv04->pipe.delete_rasterizer_state = nv04_rasterizer_state_delete; + + nv04->pipe.create_depth_stencil_alpha_state = nv04_depth_stencil_alpha_state_create; + nv04->pipe.bind_depth_stencil_alpha_state = nv04_depth_stencil_alpha_state_bind; + nv04->pipe.delete_depth_stencil_alpha_state = nv04_depth_stencil_alpha_state_delete; + + nv04->pipe.create_vs_state = nv04_vp_state_create; + nv04->pipe.bind_vs_state = nv04_vp_state_bind; + nv04->pipe.delete_vs_state = nv04_vp_state_delete; + + nv04->pipe.create_fs_state = nv04_fp_state_create; + nv04->pipe.bind_fs_state = nv04_fp_state_bind; + nv04->pipe.delete_fs_state = nv04_fp_state_delete; + + nv04->pipe.set_blend_color = nv04_set_blend_color; + nv04->pipe.set_clip_state = nv04_set_clip_state; + nv04->pipe.set_constant_buffer = nv04_set_constant_buffer; + nv04->pipe.set_framebuffer_state = nv04_set_framebuffer_state; + nv04->pipe.set_polygon_stipple = nv04_set_polygon_stipple; + nv04->pipe.set_scissor_state = nv04_set_scissor_state; + nv04->pipe.set_viewport_state = nv04_set_viewport_state; + + nv04->pipe.set_vertex_buffers = nv04_set_vertex_buffers; + nv04->pipe.set_vertex_elements = nv04_set_vertex_elements; +} + diff --git a/src/gallium/drivers/nv04/nv04_state.h b/src/gallium/drivers/nv04/nv04_state.h new file mode 100644 index 0000000000..399f750dbe --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_state.h @@ -0,0 +1,71 @@ +#ifndef __NV04_STATE_H__ +#define __NV04_STATE_H__ + +#include "pipe/p_state.h" +#include "tgsi/tgsi_scan.h" + +struct nv04_blend_state { + uint32_t b_enable; + uint32_t b_src; + uint32_t b_dst; +}; + +struct nv04_fragtex_state { + uint32_t format; +}; + +struct nv04_sampler_state { + uint32_t filter; + uint32_t format; +}; + +struct nv04_depth_stencil_alpha_state { + uint32_t control; +}; + +struct nv04_rasterizer_state { + uint32_t blend; + + const struct pipe_rasterizer_state *templ; +}; + +struct nv04_miptree { + struct pipe_texture base; + + struct pipe_buffer *buffer; + uint total_size; + + struct { + uint pitch; + uint *image_offset; + } level[PIPE_MAX_TEXTURE_LEVELS]; +}; + +struct nv04_fragment_program_data { + unsigned offset; + unsigned index; +}; + +struct nv04_fragment_program { + struct pipe_shader_state pipe; + struct tgsi_shader_info info; + + boolean translated; + boolean on_hw; + unsigned samplers; + + uint32_t *insn; + int insn_len; + + struct nv04_fragment_program_data *consts; + unsigned nr_consts; + + struct pipe_buffer *buffer; + + uint32_t fp_control; + uint32_t fp_reg_control; +}; + + + +#endif diff --git a/src/gallium/drivers/nv04/nv04_state_emit.c b/src/gallium/drivers/nv04/nv04_state_emit.c new file mode 100644 index 0000000000..0ad40a092e --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_state_emit.c @@ -0,0 +1,156 @@ +#include "nv04_context.h" +#include "nv04_state.h" + +static void nv04_vertex_layout(struct pipe_context* pipe) +{ + struct nv04_context *nv04 = nv04_context(pipe); + struct nv04_fragment_program *fp = nv04->fragprog.current; + uint32_t src = 0; + int i; + struct vertex_info vinfo; + + memset(&vinfo, 0, sizeof(vinfo)); + + for (i = 0; i < fp->info.num_inputs; i++) { + switch (i) { + case TGSI_SEMANTIC_POSITION: + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src++); + break; + case TGSI_SEMANTIC_COLOR: + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src++); + break; + default: + case TGSI_SEMANTIC_GENERIC: + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++); + break; + case TGSI_SEMANTIC_FOG: + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++); + break; + } + } + draw_compute_vertex_size(&vinfo); +} + +static uint32_t nv04_blend_func(uint32_t f) +{ + switch ( f ) { + case PIPE_BLENDFACTOR_ZERO: return 0x1; + case PIPE_BLENDFACTOR_ONE: return 0x2; + case PIPE_BLENDFACTOR_SRC_COLOR: return 0x3; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: return 0x4; + case PIPE_BLENDFACTOR_SRC_ALPHA: return 0x5; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: return 0x6; + case PIPE_BLENDFACTOR_DST_ALPHA: return 0x7; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: return 0x8; + case PIPE_BLENDFACTOR_DST_COLOR: return 0x9; + case PIPE_BLENDFACTOR_INV_DST_COLOR: return 0xA; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: return 0xB; + } + NOUVEAU_MSG("Unable to find the blend function 0x%x\n",f); + return 0; +} + +static void nv04_emit_control(struct nv04_context* nv04) +{ + uint32_t control = nv04->dsa->control; + + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_CONTROL, 1); + OUT_RING(control); +} + +static void nv04_emit_blend(struct nv04_context* nv04) +{ + uint32_t blend; + + blend=0x4; // texture MODULATE_ALPHA + blend|=0x20; // alpha is MSB + blend|=(2<<6); // flat shading + blend|=(1<<8); // persp correct + blend|=(0<<16); // no fog + blend|=(nv04->blend->b_enable<<20); + blend|=(nv04_blend_func(nv04->blend->b_src)<<24); + blend|=(nv04_blend_func(nv04->blend->b_dst)<<28); + + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_BLEND, 1); + OUT_RING(blend); +} + +static void nv04_emit_sampler(struct nv04_context *nv04, int unit) +{ + struct nv04_miptree *nv04mt = nv04->tex_miptree[unit]; + struct pipe_texture *pt = &nv04mt->base; + + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_OFFSET, 3); + OUT_RELOCl(nv04mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD); + OUT_RELOCd(nv04mt->buffer, (nv04->fragtex.format | nv04->sampler[unit]->format), NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_OR | NOUVEAU_BO_RD, 1/*VRAM*/,2/*TT*/); + OUT_RING(nv04->sampler[unit]->filter); +} + +void +nv04_emit_hw_state(struct nv04_context *nv04) +{ + int i; + + if (nv04->dirty & NV04_NEW_VERTPROG) { + //nv04_vertprog_bind(nv04, nv04->vertprog.current); + nv04->dirty &= ~NV04_NEW_VERTPROG; + } + + if (nv04->dirty & NV04_NEW_FRAGPROG) { + nv04_fragprog_bind(nv04, nv04->fragprog.current); + /*XXX: clear NV04_NEW_FRAGPROG if no new program uploaded */ + nv04->dirty_samplers |= (1<<10); + nv04->dirty_samplers = 0; + } + + if (nv04->dirty & NV04_NEW_CONTROL) { + nv04->dirty &= ~NV04_NEW_CONTROL; + + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_CONTROL, 1); + OUT_RING(nv04->dsa->control); + } + + if (nv04->dirty & NV04_NEW_BLEND) { + nv04->dirty &= ~NV04_NEW_BLEND; + + nv04_emit_blend(nv04); + } + + if (nv04->dirty & NV04_NEW_SAMPLER) { + nv04->dirty &= ~NV04_NEW_SAMPLER; + + nv04_emit_sampler(nv04, 0); + } + + if (nv04->dirty & NV04_NEW_VIEWPORT) { + nv04->dirty &= ~NV04_NEW_VIEWPORT; +// nv04_state_emit_viewport(nv04); + } + + /* Emit relocs for every referenced buffer. + * This is to ensure the bufmgr has an accurate idea of how + * the buffer is used. This isn't very efficient, but we don't + * seem to take a significant performance hit. Will be improved + * at some point. Vertex arrays are emitted by nv04_vbo.c + */ + + /* Render target */ +/* BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_PITCH, 2); + OUT_RING(rt->stride|(zeta->stride<<16)); + OUT_RELOCl(rt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + if (fb->zsbuf) { + BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_OFFSET_ZETA, 1); + OUT_RELOCl(zeta->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + }*/ + + /* Texture images */ + for (i = 0; i < 1; i++) { + if (!(nv04->fp_samplers & (1 << i))) + continue; + struct nv04_miptree *nv04mt = nv04->tex_miptree[i]; + BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_OFFSET, 2); + OUT_RELOCl(nv04mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD); + OUT_RELOCd(nv04mt->buffer, (nv04->fragtex.format | nv04->sampler[i]->format), NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_OR | NOUVEAU_BO_RD, 1/*VRAM*/,2/*TT*/); + } +} + diff --git a/src/gallium/drivers/nv04/nv04_surface.c b/src/gallium/drivers/nv04/nv04_surface.c new file mode 100644 index 0000000000..9d9943ed4e --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_surface.c @@ -0,0 +1,64 @@ + +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "nv04_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/p_inlines.h" +#include "util/u_tile.h" + +static void +nv04_surface_copy(struct pipe_context *pipe, boolean do_flip, + struct pipe_surface *dest, unsigned destx, unsigned desty, + struct pipe_surface *src, unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + struct nv04_context *nv04 = nv04_context(pipe); + struct nouveau_winsys *nvws = nv04->nvws; + + nvws->surface_copy(nvws, dest, destx, desty, src, srcx, srcy, + width, height); +} + +static void +nv04_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest, + unsigned destx, unsigned desty, unsigned width, + unsigned height, unsigned value) +{ + struct nv04_context *nv04 = nv04_context(pipe); + struct nouveau_winsys *nvws = nv04->nvws; + + nvws->surface_fill(nvws, dest, destx, desty, width, height, value); +} + +void +nv04_init_surface_functions(struct nv04_context *nv04) +{ + nv04->pipe.surface_copy = nv04_surface_copy; + nv04->pipe.surface_fill = nv04_surface_fill; +} diff --git a/src/gallium/drivers/nv04/nv04_vbo.c b/src/gallium/drivers/nv04/nv04_vbo.c new file mode 100644 index 0000000000..91f919d48e --- /dev/null +++ b/src/gallium/drivers/nv04/nv04_vbo.c @@ -0,0 +1,71 @@ +#include "draw/draw_context.h" +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +#include "nv04_context.h" +#include "nv04_state.h" + +#include "nouveau/nouveau_channel.h" +#include "nouveau/nouveau_pushbuf.h" + +boolean nv04_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) +{ + struct nv04_context *nv04 = nv04_context( pipe ); + struct draw_context *draw = nv04->draw; + unsigned i; + + /* + * Map vertex buffers + */ + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { + if (nv04->vertex_buffer[i].buffer) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + nv04->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); + } + } + /* Map index buffer, if present */ + if (indexBuffer) { + void *mapped_indexes + = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); + } + else { + /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + /* draw! */ + draw_arrays(nv04->draw, prim, start, count); + + /* + * unmap vertex/index buffers + */ + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { + if (nv04->vertex_buffer[i].buffer) { + pipe->winsys->buffer_unmap(pipe->winsys, nv04->vertex_buffer[i].buffer); + draw_set_mapped_vertex_buffer(draw, i, NULL); + } + } + if (indexBuffer) { + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + return TRUE; +} + +boolean nv04_draw_arrays( struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) +{ + return nv04_draw_elements(pipe, NULL, 0, prim, start, count); +} + + + diff --git a/src/gallium/drivers/nv10/Makefile b/src/gallium/drivers/nv10/Makefile new file mode 100644 index 0000000000..4ba7ce586d --- /dev/null +++ b/src/gallium/drivers/nv10/Makefile @@ -0,0 +1,28 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = nv10 + +DRIVER_SOURCES = \ + nv10_clear.c \ + nv10_context.c \ + nv10_fragprog.c \ + nv10_fragtex.c \ + nv10_miptree.c \ + nv10_prim_vbuf.c \ + nv10_screen.c \ + nv10_state.c \ + nv10_state_emit.c \ + nv10_surface.c \ + nv10_vbo.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/drivers/nv10/nv10_clear.c b/src/gallium/drivers/nv10/nv10_clear.c new file mode 100644 index 0000000000..be7e09cf4b --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_clear.c @@ -0,0 +1,12 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "nv10_context.h" + +void +nv10_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); +} diff --git a/src/gallium/drivers/nv10/nv10_context.c b/src/gallium/drivers/nv10/nv10_context.c new file mode 100644 index 0000000000..4eb4ed9185 --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_context.c @@ -0,0 +1,296 @@ +#include "draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" + +#include "nv10_context.h" +#include "nv10_screen.h" + +static void +nv10_flush(struct pipe_context *pipe, unsigned flags, + struct pipe_fence_handle **fence) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + draw_flush(nv10->draw); + + FIRE_RING(fence); +} + +static void +nv10_destroy(struct pipe_context *pipe) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + if (nv10->draw) + draw_destroy(nv10->draw); + + FREE(nv10); +} + +static void nv10_init_hwctx(struct nv10_context *nv10) +{ + struct nv10_screen *screen = nv10->screen; + struct nouveau_winsys *nvws = screen->nvws; + int i; + float projectionmatrix[16]; + + BEGIN_RING(celsius, NV10TCL_DMA_NOTIFY, 1); + OUT_RING (screen->sync->handle); + BEGIN_RING(celsius, NV10TCL_DMA_IN_MEMORY0, 2); + OUT_RING (nvws->channel->vram->handle); + OUT_RING (nvws->channel->gart->handle); + BEGIN_RING(celsius, NV10TCL_DMA_IN_MEMORY2, 2); + OUT_RING (nvws->channel->vram->handle); + OUT_RING (nvws->channel->vram->handle); + + BEGIN_RING(celsius, NV10TCL_NOP, 1); + OUT_RING (0); + + BEGIN_RING(celsius, NV10TCL_RT_HORIZ, 2); + OUT_RING (0); + OUT_RING (0); + + BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_HORIZ(0), 1); + OUT_RING ((0x7ff<<16)|0x800); + BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_VERT(0), 1); + OUT_RING ((0x7ff<<16)|0x800); + + for (i=1;i<8;i++) { + BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_HORIZ(i), 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_VERT(i), 1); + OUT_RING (0); + } + + BEGIN_RING(celsius, 0x290, 1); + OUT_RING ((0x10<<16)|1); + BEGIN_RING(celsius, 0x3f4, 1); + OUT_RING (0); + + BEGIN_RING(celsius, NV10TCL_NOP, 1); + OUT_RING (0); + + if (nv10->screen->celsius->grclass != NV10TCL) { + /* For nv11, nv17 */ + BEGIN_RING(celsius, 0x120, 3); + OUT_RING (0); + OUT_RING (1); + OUT_RING (2); + + BEGIN_RING(celsius, NV10TCL_NOP, 1); + OUT_RING (0); + } + + BEGIN_RING(celsius, NV10TCL_NOP, 1); + OUT_RING (0); + + /* Set state */ + BEGIN_RING(celsius, NV10TCL_FOG_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_FUNC, 2); + OUT_RING (0x207); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_TX_ENABLE(0), 2); + OUT_RING (0); + OUT_RING (0); + + BEGIN_RING(celsius, NV10TCL_RC_IN_ALPHA(0), 12); + OUT_RING (0x30141010); + OUT_RING (0); + OUT_RING (0x20040000); + OUT_RING (0); + OUT_RING (0); + OUT_RING (0); + OUT_RING (0x00000c00); + OUT_RING (0); + OUT_RING (0x00000c00); + OUT_RING (0x18000000); + OUT_RING (0x300e0300); + OUT_RING (0x0c091c80); + + BEGIN_RING(celsius, NV10TCL_BLEND_FUNC_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_DITHER_ENABLE, 2); + OUT_RING (1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_LINE_SMOOTH_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_VERTEX_WEIGHT_ENABLE, 2); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_BLEND_FUNC_SRC, 4); + OUT_RING (1); + OUT_RING (0); + OUT_RING (0); + OUT_RING (0x8006); + BEGIN_RING(celsius, NV10TCL_STENCIL_MASK, 8); + OUT_RING (0xff); + OUT_RING (0x207); + OUT_RING (0); + OUT_RING (0xff); + OUT_RING (0x1e00); + OUT_RING (0x1e00); + OUT_RING (0x1e00); + OUT_RING (0x1d01); + BEGIN_RING(celsius, NV10TCL_NORMALIZE_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_FOG_ENABLE, 2); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_LIGHT_MODEL, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_COLOR_CONTROL, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_ENABLED_LIGHTS, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_POLYGON_OFFSET_POINT_ENABLE, 3); + OUT_RING (0); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_DEPTH_FUNC, 1); + OUT_RING (0x201); + BEGIN_RING(celsius, NV10TCL_DEPTH_WRITE_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_DEPTH_TEST_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_POLYGON_OFFSET_FACTOR, 2); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_POINT_SIZE, 1); + OUT_RING (8); + BEGIN_RING(celsius, NV10TCL_POINT_PARAMETERS_ENABLE, 2); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_LINE_WIDTH, 1); + OUT_RING (8); + BEGIN_RING(celsius, NV10TCL_LINE_SMOOTH_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_POLYGON_MODE_FRONT, 2); + OUT_RING (0x1b02); + OUT_RING (0x1b02); + BEGIN_RING(celsius, NV10TCL_CULL_FACE, 2); + OUT_RING (0x405); + OUT_RING (0x901); + BEGIN_RING(celsius, NV10TCL_POLYGON_SMOOTH_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_CULL_FACE_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_TX_GEN_S(0), 8); + for (i=0;i<8;i++) { + OUT_RING (0); + } + BEGIN_RING(celsius, NV10TCL_FOG_EQUATION_CONSTANT, 3); + OUT_RING (0x3fc00000); /* -1.50 */ + OUT_RING (0xbdb8aa0a); /* -0.09 */ + OUT_RING (0); /* 0.00 */ + + BEGIN_RING(celsius, NV10TCL_NOP, 1); + OUT_RING (0); + + BEGIN_RING(celsius, NV10TCL_FOG_MODE, 2); + OUT_RING (0x802); + OUT_RING (2); + /* for some reason VIEW_MATRIX_ENABLE need to be 6 instead of 4 when + * using texturing, except when using the texture matrix + */ + BEGIN_RING(celsius, NV10TCL_VIEW_MATRIX_ENABLE, 1); + OUT_RING (6); + BEGIN_RING(celsius, NV10TCL_COLOR_MASK, 1); + OUT_RING (0x01010101); + + /* Set vertex component */ + BEGIN_RING(celsius, NV10TCL_VERTEX_COL_4F_R, 4); + OUT_RINGf (1.0); + OUT_RINGf (1.0); + OUT_RINGf (1.0); + OUT_RINGf (1.0); + BEGIN_RING(celsius, NV10TCL_VERTEX_COL2_3F_R, 3); + OUT_RING (0); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(celsius, NV10TCL_VERTEX_NOR_3F_X, 3); + OUT_RING (0); + OUT_RING (0); + OUT_RINGf (1.0); + BEGIN_RING(celsius, NV10TCL_VERTEX_TX0_4F_S, 4); + OUT_RINGf (0.0); + OUT_RINGf (0.0); + OUT_RINGf (0.0); + OUT_RINGf (1.0); + BEGIN_RING(celsius, NV10TCL_VERTEX_TX1_4F_S, 4); + OUT_RINGf (0.0); + OUT_RINGf (0.0); + OUT_RINGf (0.0); + OUT_RINGf (1.0); + BEGIN_RING(celsius, NV10TCL_VERTEX_FOG_1F, 1); + OUT_RINGf (0.0); + BEGIN_RING(celsius, NV10TCL_EDGEFLAG_ENABLE, 1); + OUT_RING (1); + + memset(projectionmatrix, 0, sizeof(projectionmatrix)); + BEGIN_RING(celsius, NV10TCL_PROJECTION_MATRIX(0), 16); + projectionmatrix[0*4+0] = 1.0; + projectionmatrix[1*4+1] = 1.0; + projectionmatrix[2*4+2] = 1.0; + projectionmatrix[3*4+3] = 1.0; + for (i=0;i<16;i++) { + OUT_RINGf (projectionmatrix[i]); + } + + BEGIN_RING(celsius, NV10TCL_DEPTH_RANGE_NEAR, 2); + OUT_RING (0.0); + OUT_RINGf (16777216.0); + + BEGIN_RING(celsius, NV10TCL_VIEWPORT_SCALE_X, 4); + OUT_RINGf (-2048.0); + OUT_RINGf (-2048.0); + OUT_RINGf (16777215.0 * 0.5); + OUT_RING (0); + + FIRE_RING (NULL); +} + +static void +nv10_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield) +{ +} + +struct pipe_context * +nv10_create(struct pipe_screen *pscreen, unsigned pctx_id) +{ + struct nv10_screen *screen = nv10_screen(pscreen); + struct pipe_winsys *ws = pscreen->winsys; + struct nv10_context *nv10; + struct nouveau_winsys *nvws = screen->nvws; + + nv10 = CALLOC(1, sizeof(struct nv10_context)); + if (!nv10) + return NULL; + nv10->screen = screen; + nv10->pctx_id = pctx_id; + + nv10->nvws = nvws; + + nv10->pipe.winsys = ws; + nv10->pipe.screen = pscreen; + nv10->pipe.destroy = nv10_destroy; + nv10->pipe.set_edgeflags = nv10_set_edgeflags; + nv10->pipe.draw_arrays = nv10_draw_arrays; + nv10->pipe.draw_elements = nv10_draw_elements; + nv10->pipe.clear = nv10_clear; + nv10->pipe.flush = nv10_flush; + + nv10_init_surface_functions(nv10); + nv10_init_state_functions(nv10); + + nv10->draw = draw_create(); + assert(nv10->draw); + draw_set_rasterize_stage(nv10->draw, nv10_draw_vbuf_stage(nv10)); + + nv10_init_hwctx(nv10); + + return &nv10->pipe; +} + diff --git a/src/gallium/drivers/nv10/nv10_context.h b/src/gallium/drivers/nv10/nv10_context.h new file mode 100644 index 0000000000..f3b56de25a --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_context.h @@ -0,0 +1,153 @@ +#ifndef __NV10_CONTEXT_H__ +#define __NV10_CONTEXT_H__ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_compiler.h" + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "draw/draw_vertex.h" + +#include "nouveau/nouveau_winsys.h" +#include "nouveau/nouveau_gldefs.h" + +#define NOUVEAU_PUSH_CONTEXT(ctx) \ + struct nv10_screen *ctx = nv10->screen +#include "nouveau/nouveau_push.h" + +#include "nv10_state.h" + +#define NOUVEAU_ERR(fmt, args...) \ + fprintf(stderr, "%s:%d - "fmt, __func__, __LINE__, ##args); +#define NOUVEAU_MSG(fmt, args...) \ + fprintf(stderr, "nouveau: "fmt, ##args); + +#define NV10_NEW_VERTPROG (1 << 0) +#define NV10_NEW_FRAGPROG (1 << 1) +#define NV10_NEW_VTXARRAYS (1 << 2) +#define NV10_NEW_BLEND (1 << 3) +#define NV10_NEW_BLENDCOL (1 << 4) +#define NV10_NEW_RAST (1 << 5) +#define NV10_NEW_DSA (1 << 6) +#define NV10_NEW_VIEWPORT (1 << 7) +#define NV10_NEW_SCISSOR (1 << 8) +#define NV10_NEW_FRAMEBUFFER (1 << 9) + +#include "nv10_screen.h" + +struct nv10_context { + struct pipe_context pipe; + + struct nouveau_winsys *nvws; + struct nv10_screen *screen; + unsigned pctx_id; + + struct draw_context *draw; + + uint32_t dirty; + + struct nv10_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS]; + struct nv10_miptree *tex_miptree[PIPE_MAX_SAMPLERS]; + unsigned dirty_samplers; + unsigned fp_samplers; + unsigned vp_samplers; + + uint32_t rt_enable; + struct pipe_buffer *rt[4]; + struct pipe_buffer *zeta; + uint32_t lma_offset; + + struct nv10_blend_state *blend; + struct pipe_blend_color *blend_color; + struct nv10_rasterizer_state *rast; + struct nv10_depth_stencil_alpha_state *dsa; + struct pipe_viewport_state *viewport; + struct pipe_scissor_state *scissor; + struct pipe_framebuffer_state *framebuffer; + + //struct pipe_buffer *constbuf[PIPE_SHADER_TYPES]; + float *constbuf[PIPE_SHADER_TYPES][32][4]; + unsigned constbuf_nr[PIPE_SHADER_TYPES]; + + struct vertex_info vertex_info; + + struct { + struct pipe_buffer *buffer; + uint32_t format; + } tex[2]; + + unsigned vb_enable; + struct { + struct pipe_buffer *buffer; + unsigned delta; + } vb[16]; + +/* struct { + + struct nouveau_resource *exec_heap; + struct nouveau_resource *data_heap; + + struct nv10_vertex_program *active; + + struct nv10_vertex_program *current; + } vertprog; +*/ + struct { + struct nv10_fragment_program *active; + + struct nv10_fragment_program *current; + struct pipe_buffer *constant_buf; + } fragprog; + + struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS]; + struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS]; +}; + +static INLINE struct nv10_context * +nv10_context(struct pipe_context *pipe) +{ + return (struct nv10_context *)pipe; +} + +extern void nv10_init_state_functions(struct nv10_context *nv10); +extern void nv10_init_surface_functions(struct nv10_context *nv10); + +extern void nv10_screen_init_miptree_functions(struct pipe_screen *pscreen); + +/* nv10_clear.c */ +extern void nv10_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue); + +/* nv10_draw.c */ +extern struct draw_stage *nv10_draw_render_stage(struct nv10_context *nv10); + +/* nv10_fragprog.c */ +extern void nv10_fragprog_bind(struct nv10_context *, + struct nv10_fragment_program *); +extern void nv10_fragprog_destroy(struct nv10_context *, + struct nv10_fragment_program *); + +/* nv10_fragtex.c */ +extern void nv10_fragtex_bind(struct nv10_context *); + +/* nv10_prim_vbuf.c */ +struct draw_stage *nv10_draw_vbuf_stage( struct nv10_context *nv10 ); +extern void nv10_vtxbuf_bind(struct nv10_context* nv10); + +/* nv10_state.c and friends */ +extern void nv10_emit_hw_state(struct nv10_context *nv10); +extern void nv10_state_tex_update(struct nv10_context *nv10); + +/* nv10_vbo.c */ +extern boolean nv10_draw_arrays(struct pipe_context *, unsigned mode, + unsigned start, unsigned count); +extern boolean nv10_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count); + + +#endif diff --git a/src/gallium/drivers/nv10/nv10_fragprog.c b/src/gallium/drivers/nv10/nv10_fragprog.c new file mode 100644 index 0000000000..698db5a16a --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_fragprog.c @@ -0,0 +1,21 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" + +#include "nv10_context.h" + +void +nv10_fragprog_bind(struct nv10_context *nv10, struct nv10_fragment_program *fp) +{ +} + +void +nv10_fragprog_destroy(struct nv10_context *nv10, + struct nv10_fragment_program *fp) +{ +} + diff --git a/src/gallium/drivers/nv10/nv10_fragtex.c b/src/gallium/drivers/nv10/nv10_fragtex.c new file mode 100644 index 0000000000..27f2f87584 --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_fragtex.c @@ -0,0 +1,124 @@ +#include "nv10_context.h" +#include "nouveau/nouveau_util.h" + +#define _(m,tf) \ +{ \ + TRUE, \ + PIPE_FORMAT_##m, \ + NV10TCL_TX_FORMAT_FORMAT_##tf, \ +} + +struct nv10_texture_format { + boolean defined; + uint pipe; + int format; +}; + +static struct nv10_texture_format +nv10_texture_formats[] = { + _(A8R8G8B8_UNORM, A8R8G8B8), + _(A1R5G5B5_UNORM, A1R5G5B5), + _(A4R4G4B4_UNORM, A4R4G4B4), + _(L8_UNORM , L8 ), + _(A8_UNORM , A8 ), + _(A8L8_UNORM , A8L8 ), +// _(RGB_DXT1 , DXT1, ), +// _(RGBA_DXT1 , DXT1, ), +// _(RGBA_DXT3 , DXT3, ), +// _(RGBA_DXT5 , DXT5, ), + {}, +}; + +static struct nv10_texture_format * +nv10_fragtex_format(uint pipe_format) +{ + struct nv10_texture_format *tf = nv10_texture_formats; + + while (tf->defined) { + if (tf->pipe == pipe_format) + return tf; + tf++; + } + + return NULL; +} + + +static void +nv10_fragtex_build(struct nv10_context *nv10, int unit) +{ +#if 0 + struct nv10_sampler_state *ps = nv10->tex_sampler[unit]; + struct nv10_miptree *nv10mt = nv10->tex_miptree[unit]; + struct pipe_texture *pt = &nv10mt->base; + struct nv10_texture_format *tf; + uint32_t txf, txs, txp; + + tf = nv10_fragtex_format(pt->format); + if (!tf || !tf->defined) { + NOUVEAU_ERR("Unsupported texture format: 0x%x\n", pt->format); + return; + } + + txf = tf->format << 8; + txf |= (pt->last_level + 1) << 16; + txf |= log2i(pt->width[0]) << 20; + txf |= log2i(pt->height[0]) << 24; + txf |= log2i(pt->depth[0]) << 28; + txf |= 8; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE: + txf |= NV10TCL_TX_FORMAT_CUBE_MAP; + /* fall-through */ + case PIPE_TEXTURE_2D: + txf |= (2<<4); + break; + case PIPE_TEXTURE_1D: + txf |= (1<<4); + break; + default: + NOUVEAU_ERR("Unknown target %d\n", pt->target); + return; + } + + BEGIN_RING(celsius, NV10TCL_TX_OFFSET(unit), 8); + OUT_RELOCl(nv10mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD); + OUT_RELOCd(nv10mt->buffer,txf,NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_OR | NOUVEAU_BO_RD, 1/*VRAM*/,2/*TT*/); + OUT_RING (ps->wrap); + OUT_RING (0x40000000); /* enable */ + OUT_RING (txs); + OUT_RING (ps->filt | 0x2000 /* magic */); + OUT_RING ((pt->width[0] << 16) | pt->height[0]); + OUT_RING (ps->bcol); +#endif +} + +void +nv10_fragtex_bind(struct nv10_context *nv10) +{ +#if 0 + struct nv10_fragment_program *fp = nv10->fragprog.active; + unsigned samplers, unit; + + samplers = nv10->fp_samplers & ~fp->samplers; + while (samplers) { + unit = ffs(samplers) - 1; + samplers &= ~(1 << unit); + + BEGIN_RING(celsius, NV10TCL_TX_ENABLE(unit), 1); + OUT_RING (0); + } + + samplers = nv10->dirty_samplers & fp->samplers; + while (samplers) { + unit = ffs(samplers) - 1; + samplers &= ~(1 << unit); + + nv10_fragtex_build(nv10, unit); + } + + nv10->fp_samplers = fp->samplers; +#endif +} + diff --git a/src/gallium/drivers/nv10/nv10_miptree.c b/src/gallium/drivers/nv10/nv10_miptree.c new file mode 100644 index 0000000000..943f9e21e9 --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_miptree.c @@ -0,0 +1,149 @@ +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "nv10_context.h" +#include "nv10_screen.h" + +static void +nv10_miptree_layout(struct nv10_miptree *nv10mt) +{ + struct pipe_texture *pt = &nv10mt->base; + boolean swizzled = FALSE; + uint width = pt->width[0], height = pt->height[0]; + uint offset = 0; + int nr_faces, l, f; + + if (pt->target == PIPE_TEXTURE_CUBE) { + nr_faces = 6; + } else { + nr_faces = 1; + } + + for (l = 0; l <= pt->last_level; l++) { + pt->width[l] = width; + pt->height[l] = height; + pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); + pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); + + if (swizzled) + nv10mt->level[l].pitch = pt->nblocksx[l] * pt->block.size; + else + nv10mt->level[l].pitch = pt->nblocksx[0] * pt->block.size; + nv10mt->level[l].pitch = (nv10mt->level[l].pitch + 63) & ~63; + + nv10mt->level[l].image_offset = + CALLOC(nr_faces, sizeof(unsigned)); + + width = MAX2(1, width >> 1); + height = MAX2(1, height >> 1); + + } + + for (f = 0; f < nr_faces; f++) { + for (l = 0; l <= pt->last_level; l++) { + nv10mt->level[l].image_offset[f] = offset; + offset += nv10mt->level[l].pitch * pt->height[l]; + } + } + + nv10mt->total_size = offset; +} + +static struct pipe_texture * +nv10_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) +{ + struct pipe_winsys *ws = screen->winsys; + struct nv10_miptree *mt; + + mt = MALLOC(sizeof(struct nv10_miptree)); + if (!mt) + return NULL; + mt->base = *pt; + mt->base.refcount = 1; + mt->base.screen = screen; + + nv10_miptree_layout(mt); + + mt->buffer = ws->buffer_create(ws, 256, PIPE_BUFFER_USAGE_PIXEL, + mt->total_size); + if (!mt->buffer) { + FREE(mt); + return NULL; + } + + return &mt->base; +} + +static void +nv10_miptree_release(struct pipe_screen *screen, struct pipe_texture **pt) +{ + struct pipe_texture *mt = *pt; + + *pt = NULL; + if (--mt->refcount <= 0) { + struct nv10_miptree *nv10mt = (struct nv10_miptree *)mt; + int l; + + pipe_buffer_reference(screen, &nv10mt->buffer, NULL); + for (l = 0; l <= mt->last_level; l++) { + if (nv10mt->level[l].image_offset) + FREE(nv10mt->level[l].image_offset); + } + FREE(nv10mt); + } +} + +static void +nv10_miptree_update(struct pipe_context *pipe, struct pipe_texture *mt, + uint face, uint levels) +{ +} + + +static struct pipe_surface * +nv10_miptree_surface_get(struct pipe_screen *screen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + unsigned flags) +{ + struct pipe_winsys *ws = screen->winsys; + struct nv10_miptree *nv10mt = (struct nv10_miptree *)pt; + struct pipe_surface *ps; + + ps = ws->surface_alloc(ws); + if (!ps) + return NULL; + pipe_buffer_reference(screen, &ps->buffer, nv10mt->buffer); + ps->format = pt->format; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->block = pt->block; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = nv10mt->level[level].pitch; + ps->refcount = 1; + ps->winsys = screen->winsys; + + if (pt->target == PIPE_TEXTURE_CUBE) { + ps->offset = nv10mt->level[level].image_offset[face]; + } else { + ps->offset = nv10mt->level[level].image_offset[0]; + } + + return ps; +} + +static void +nv10_miptree_surface_release(struct pipe_screen *screen, + struct pipe_surface **surface) +{ +} + +void nv10_screen_init_miptree_functions(struct pipe_screen *pscreen) +{ + pscreen->texture_create = nv10_miptree_create; + pscreen->texture_release = nv10_miptree_release; + pscreen->get_tex_surface = nv10_miptree_surface_get; + pscreen->tex_surface_release = nv10_miptree_surface_release; +} + diff --git a/src/gallium/drivers/nv10/nv10_prim_vbuf.c b/src/gallium/drivers/nv10/nv10_prim_vbuf.c new file mode 100644 index 0000000000..e7e81d3dff --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_prim_vbuf.c @@ -0,0 +1,245 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Build post-transformation, post-clipping vertex buffers and element + * lists by hooking into the end of the primitive pipeline and + * manipulating the vertex_id field in the vertex headers. + * + * XXX: work in progress + * + * \author José Fonseca <jrfonseca@tungstengraphics.com> + * \author Keith Whitwell <keith@tungstengraphics.com> + */ + + +#include "pipe/p_debug.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" + +#include "nv10_context.h" +#include "nv10_state.h" + +#include "draw/draw_vbuf.h" + +/** + * Primitive renderer for nv10. + */ +struct nv10_vbuf_render { + struct vbuf_render base; + + struct nv10_context *nv10; + + /** Vertex buffer */ + struct pipe_buffer* buffer; + + /** Vertex size in bytes */ + unsigned vertex_size; + + /** Hardware primitive */ + unsigned hwprim; +}; + + +void nv10_vtxbuf_bind( struct nv10_context* nv10 ) +{ + int i; + for(i = 0; i < 8; i++) { + BEGIN_RING(celsius, NV10TCL_VERTEX_ARRAY_ATTRIB_OFFSET(i), 1); + OUT_RING(0/*nv10->vtxbuf*/); + BEGIN_RING(celsius, NV10TCL_VERTEX_ARRAY_ATTRIB_FORMAT(i) ,1); + OUT_RING(0/*XXX*/); + } +} + +/** + * Basically a cast wrapper. + */ +static INLINE struct nv10_vbuf_render * +nv10_vbuf_render( struct vbuf_render *render ) +{ + assert(render); + return (struct nv10_vbuf_render *)render; +} + + +static const struct vertex_info * +nv10_vbuf_render_get_vertex_info( struct vbuf_render *render ) +{ + struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); + struct nv10_context *nv10 = nv10_render->nv10; + + nv10_emit_hw_state(nv10); + + return &nv10->vertex_info; +} + + +static void * +nv10_vbuf_render_allocate_vertices( struct vbuf_render *render, + ushort vertex_size, + ushort nr_vertices ) +{ + struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); + struct nv10_context *nv10 = nv10_render->nv10; + struct pipe_winsys *winsys = nv10->pipe.winsys; + size_t size = (size_t)vertex_size * (size_t)nr_vertices; + + assert(!nv10_render->buffer); + nv10_render->buffer = winsys->buffer_create(winsys, 64, PIPE_BUFFER_USAGE_VERTEX, size); + + nv10->dirty |= NV10_NEW_VTXARRAYS; + + return winsys->buffer_map(winsys, + nv10_render->buffer, + PIPE_BUFFER_USAGE_CPU_WRITE); +} + + +static boolean +nv10_vbuf_render_set_primitive( struct vbuf_render *render, + unsigned prim ) +{ + struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); + unsigned hwp = nvgl_primitive(prim); + if (hwp == 0) + return FALSE; + + nv10_render->hwprim = hwp; + return TRUE; +} + + +static void +nv10_vbuf_render_draw( struct vbuf_render *render, + const ushort *indices, + uint nr_indices) +{ + struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); + struct nv10_context *nv10 = nv10_render->nv10; + int push, i; + + nv10_emit_hw_state(nv10); + + BEGIN_RING(celsius, NV10TCL_VERTEX_ARRAY_OFFSET_POS, 1); + OUT_RELOCl(nv10_render->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD); + + BEGIN_RING(celsius, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1); + OUT_RING(nv10_render->hwprim); + + if (nr_indices & 1) { + BEGIN_RING(celsius, NV10TCL_VB_ELEMENT_U32, 1); + OUT_RING (indices[0]); + indices++; nr_indices--; + } + + while (nr_indices) { + // XXX too big/small ? check the size + push = MIN2(nr_indices, 1200 * 2); + + BEGIN_RING_NI(celsius, NV10TCL_VB_ELEMENT_U16, push >> 1); + for (i = 0; i < push; i+=2) + OUT_RING((indices[i+1] << 16) | indices[i]); + + nr_indices -= push; + indices += push; + } + + BEGIN_RING(celsius, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1); + OUT_RING (0); +} + + +static void +nv10_vbuf_render_release_vertices( struct vbuf_render *render, + void *vertices, + unsigned vertex_size, + unsigned vertices_used ) +{ + struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); + struct nv10_context *nv10 = nv10_render->nv10; + struct pipe_winsys *winsys = nv10->pipe.winsys; + struct pipe_screen *pscreen = &nv10->screen->pipe; + + assert(nv10_render->buffer); + winsys->buffer_unmap(winsys, nv10_render->buffer); + pipe_buffer_reference(pscreen, &nv10_render->buffer, NULL); +} + + +static void +nv10_vbuf_render_destroy( struct vbuf_render *render ) +{ + struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); + FREE(nv10_render); +} + + +/** + * Create a new primitive render. + */ +static struct vbuf_render * +nv10_vbuf_render_create( struct nv10_context *nv10 ) +{ + struct nv10_vbuf_render *nv10_render = CALLOC_STRUCT(nv10_vbuf_render); + + nv10_render->nv10 = nv10; + + nv10_render->base.max_vertex_buffer_bytes = 16*1024; + nv10_render->base.max_indices = 1024; + nv10_render->base.get_vertex_info = nv10_vbuf_render_get_vertex_info; + nv10_render->base.allocate_vertices = nv10_vbuf_render_allocate_vertices; + nv10_render->base.set_primitive = nv10_vbuf_render_set_primitive; + nv10_render->base.draw = nv10_vbuf_render_draw; + nv10_render->base.release_vertices = nv10_vbuf_render_release_vertices; + nv10_render->base.destroy = nv10_vbuf_render_destroy; + + return &nv10_render->base; +} + + +/** + * Create a new primitive vbuf/render stage. + */ +struct draw_stage *nv10_draw_vbuf_stage( struct nv10_context *nv10 ) +{ + struct vbuf_render *render; + struct draw_stage *stage; + + render = nv10_vbuf_render_create(nv10); + if(!render) + return NULL; + + stage = draw_vbuf_stage( nv10->draw, render ); + if(!stage) { + render->destroy(render); + return NULL; + } + + return stage; +} diff --git a/src/gallium/drivers/nv10/nv10_screen.c b/src/gallium/drivers/nv10/nv10_screen.c new file mode 100644 index 0000000000..4d9fbd4b5f --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_screen.c @@ -0,0 +1,210 @@ +#include "pipe/p_screen.h" + +#include "nv10_context.h" +#include "nv10_screen.h" + +static const char * +nv10_screen_get_name(struct pipe_screen *screen) +{ + struct nv10_screen *nv10screen = nv10_screen(screen); + struct nouveau_device *dev = nv10screen->nvws->channel->device; + static char buffer[128]; + + snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); + return buffer; +} + +static const char * +nv10_screen_get_vendor(struct pipe_screen *screen) +{ + return "nouveau"; +} + +static int +nv10_screen_get_param(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 2; + case PIPE_CAP_NPOT_TEXTURES: + return 0; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 0; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 1; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 0; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 12; + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 0; + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 12; + case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: + return 0; + case NOUVEAU_CAP_HW_VTXBUF: + case NOUVEAU_CAP_HW_IDXBUF: + return 0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0; + } +} + +static float +nv10_screen_get_paramf(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 10.0; + case PIPE_CAP_MAX_POINT_WIDTH: + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 64.0; + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 2.0; + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 4.0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0.0; + } +} + +static boolean +nv10_screen_is_format_supported(struct pipe_screen *screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, unsigned geom_flags) +{ + if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z16_UNORM: + return TRUE; + default: + break; + } + } else { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_A1R5G5B5_UNORM: + case PIPE_FORMAT_A4R4G4B4_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_A8_UNORM: + case PIPE_FORMAT_I8_UNORM: + return TRUE; + default: + break; + } + } + + return FALSE; +} + +static void * +nv10_surface_map(struct pipe_screen *screen, struct pipe_surface *surface, + unsigned flags ) +{ + struct pipe_winsys *ws = screen->winsys; + void *map; + + map = ws->buffer_map(ws, surface->buffer, flags); + if (!map) + return NULL; + + return map + surface->offset; +} + +static void +nv10_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) +{ + struct pipe_winsys *ws = screen->winsys; + + ws->buffer_unmap(ws, surface->buffer); +} + +static void +nv10_screen_destroy(struct pipe_screen *pscreen) +{ + struct nv10_screen *screen = nv10_screen(pscreen); + struct nouveau_winsys *nvws = screen->nvws; + + nvws->notifier_free(&screen->sync); + nvws->grobj_free(&screen->celsius); + + FREE(pscreen); +} + +struct pipe_screen * +nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +{ + struct nv10_screen *screen = CALLOC_STRUCT(nv10_screen); + unsigned celsius_class; + unsigned chipset = nvws->channel->device->chipset; + int ret; + + if (!screen) + return NULL; + screen->nvws = nvws; + + /* 3D object */ + if (chipset>=0x20) + celsius_class=NV11TCL; + else if (chipset>=0x17) + celsius_class=NV17TCL; + else if (chipset>=0x11) + celsius_class=NV11TCL; + else + celsius_class=NV10TCL; + + if (!celsius_class) { + NOUVEAU_ERR("Unknown nv1x chipset: nv%02x\n", chipset); + return NULL; + } + + ret = nvws->grobj_alloc(nvws, celsius_class, &screen->celsius); + if (ret) { + NOUVEAU_ERR("Error creating 3D object: %d\n", ret); + return FALSE; + } + + /* Notifier for sync purposes */ + ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + if (ret) { + NOUVEAU_ERR("Error creating notifier object: %d\n", ret); + nv10_screen_destroy(&screen->pipe); + return NULL; + } + + screen->pipe.winsys = ws; + screen->pipe.destroy = nv10_screen_destroy; + + screen->pipe.get_name = nv10_screen_get_name; + screen->pipe.get_vendor = nv10_screen_get_vendor; + screen->pipe.get_param = nv10_screen_get_param; + screen->pipe.get_paramf = nv10_screen_get_paramf; + + screen->pipe.is_format_supported = nv10_screen_is_format_supported; + + screen->pipe.surface_map = nv10_surface_map; + screen->pipe.surface_unmap = nv10_surface_unmap; + + nv10_screen_init_miptree_functions(&screen->pipe); + + return &screen->pipe; +} + diff --git a/src/gallium/drivers/nv10/nv10_screen.h b/src/gallium/drivers/nv10/nv10_screen.h new file mode 100644 index 0000000000..3f8750a13f --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_screen.h @@ -0,0 +1,22 @@ +#ifndef __NV10_SCREEN_H__ +#define __NV10_SCREEN_H__ + +#include "pipe/p_screen.h" + +struct nv10_screen { + struct pipe_screen pipe; + + struct nouveau_winsys *nvws; + + /* HW graphics objects */ + struct nouveau_grobj *celsius; + struct nouveau_notifier *sync; +}; + +static INLINE struct nv10_screen * +nv10_screen(struct pipe_screen *screen) +{ + return (struct nv10_screen *)screen; +} + +#endif diff --git a/src/gallium/drivers/nv10/nv10_state.c b/src/gallium/drivers/nv10/nv10_state.c new file mode 100644 index 0000000000..d2375aa2f6 --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_state.c @@ -0,0 +1,588 @@ +#include "draw/draw_context.h" +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/tgsi_parse.h" + +#include "nv10_context.h" +#include "nv10_state.h" + +static void * +nv10_blend_state_create(struct pipe_context *pipe, + const struct pipe_blend_state *cso) +{ + struct nv10_blend_state *cb; + + cb = MALLOC(sizeof(struct nv10_blend_state)); + + cb->b_enable = cso->blend_enable ? 1 : 0; + cb->b_srcfunc = ((nvgl_blend_func(cso->alpha_src_factor)<<16) | + (nvgl_blend_func(cso->rgb_src_factor))); + cb->b_dstfunc = ((nvgl_blend_func(cso->alpha_dst_factor)<<16) | + (nvgl_blend_func(cso->rgb_dst_factor))); + + cb->c_mask = (((cso->colormask & PIPE_MASK_A) ? (0x01<<24) : 0) | + ((cso->colormask & PIPE_MASK_R) ? (0x01<<16) : 0) | + ((cso->colormask & PIPE_MASK_G) ? (0x01<< 8) : 0) | + ((cso->colormask & PIPE_MASK_B) ? (0x01<< 0) : 0)); + + cb->d_enable = cso->dither ? 1 : 0; + + return (void *)cb; +} + +static void +nv10_blend_state_bind(struct pipe_context *pipe, void *blend) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + nv10->blend = (struct nv10_blend_state*)blend; + + nv10->dirty |= NV10_NEW_BLEND; +} + +static void +nv10_blend_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + + +static INLINE unsigned +wrap_mode(unsigned wrap) { + unsigned ret; + + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + ret = NV10TCL_TX_FORMAT_WRAP_S_REPEAT; + break; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + ret = NV10TCL_TX_FORMAT_WRAP_S_MIRRORED_REPEAT; + break; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + ret = NV10TCL_TX_FORMAT_WRAP_S_CLAMP_TO_EDGE; + break; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + ret = NV10TCL_TX_FORMAT_WRAP_S_CLAMP_TO_BORDER; + break; + case PIPE_TEX_WRAP_CLAMP: + ret = NV10TCL_TX_FORMAT_WRAP_S_CLAMP; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + case PIPE_TEX_WRAP_MIRROR_CLAMP: + default: + NOUVEAU_ERR("unknown wrap mode: %d\n", wrap); + ret = NV10TCL_TX_FORMAT_WRAP_S_REPEAT; + break; + } + + return ret >> NV10TCL_TX_FORMAT_WRAP_S_SHIFT; +} + +static void * +nv10_sampler_state_create(struct pipe_context *pipe, + const struct pipe_sampler_state *cso) +{ + struct nv10_sampler_state *ps; + uint32_t filter = 0; + + ps = MALLOC(sizeof(struct nv10_sampler_state)); + + ps->wrap = ((wrap_mode(cso->wrap_s) << NV10TCL_TX_FORMAT_WRAP_S_SHIFT) | + (wrap_mode(cso->wrap_t) << NV10TCL_TX_FORMAT_WRAP_T_SHIFT)); + + ps->en = 0; + if (cso->max_anisotropy > 1.0) { + /* no idea, binary driver sets it, works without it.. meh.. */ + ps->wrap |= (1 << 5); + +/* if (cso->max_anisotropy >= 16.0) { + ps->en |= NV10TCL_TX_ENABLE_ANISO_16X; + } else + if (cso->max_anisotropy >= 12.0) { + ps->en |= NV10TCL_TX_ENABLE_ANISO_12X; + } else + if (cso->max_anisotropy >= 10.0) { + ps->en |= NV10TCL_TX_ENABLE_ANISO_10X; + } else + if (cso->max_anisotropy >= 8.0) { + ps->en |= NV10TCL_TX_ENABLE_ANISO_8X; + } else + if (cso->max_anisotropy >= 6.0) { + ps->en |= NV10TCL_TX_ENABLE_ANISO_6X; + } else + if (cso->max_anisotropy >= 4.0) { + ps->en |= NV10TCL_TX_ENABLE_ANISO_4X; + } else { + ps->en |= NV10TCL_TX_ENABLE_ANISO_2X; + }*/ + } + + switch (cso->mag_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + filter |= NV10TCL_TX_FILTER_MAGNIFY_LINEAR; + break; + case PIPE_TEX_FILTER_NEAREST: + default: + filter |= NV10TCL_TX_FILTER_MAGNIFY_NEAREST; + break; + } + + switch (cso->min_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + filter |= NV10TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + filter |= NV10TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + filter |= NV10TCL_TX_FILTER_MINIFY_LINEAR; + break; + } + break; + case PIPE_TEX_FILTER_NEAREST: + default: + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + filter |= NV10TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + filter |= NV10TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + filter |= NV10TCL_TX_FILTER_MINIFY_NEAREST; + break; + } + break; + } + + ps->filt = filter; + +/* if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + switch (cso->compare_func) { + case PIPE_FUNC_NEVER: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_NEVER; + break; + case PIPE_FUNC_GREATER: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_GREATER; + break; + case PIPE_FUNC_EQUAL: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_EQUAL; + break; + case PIPE_FUNC_GEQUAL: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_GEQUAL; + break; + case PIPE_FUNC_LESS: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_LESS; + break; + case PIPE_FUNC_NOTEQUAL: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_NOTEQUAL; + break; + case PIPE_FUNC_LEQUAL: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_LEQUAL; + break; + case PIPE_FUNC_ALWAYS: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_ALWAYS; + break; + default: + break; + } + }*/ + + ps->bcol = ((float_to_ubyte(cso->border_color[3]) << 24) | + (float_to_ubyte(cso->border_color[0]) << 16) | + (float_to_ubyte(cso->border_color[1]) << 8) | + (float_to_ubyte(cso->border_color[2]) << 0)); + + return (void *)ps; +} + +static void +nv10_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler) +{ + struct nv10_context *nv10 = nv10_context(pipe); + unsigned unit; + + for (unit = 0; unit < nr; unit++) { + nv10->tex_sampler[unit] = sampler[unit]; + nv10->dirty_samplers |= (1 << unit); + } +} + +static void +nv10_sampler_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + +static void +nv10_set_sampler_texture(struct pipe_context *pipe, unsigned nr, + struct pipe_texture **miptree) +{ + struct nv10_context *nv10 = nv10_context(pipe); + unsigned unit; + + for (unit = 0; unit < nr; unit++) { + nv10->tex_miptree[unit] = (struct nv10_miptree *)miptree[unit]; + nv10->dirty_samplers |= (1 << unit); + } +} + +static void * +nv10_rasterizer_state_create(struct pipe_context *pipe, + const struct pipe_rasterizer_state *cso) +{ + struct nv10_rasterizer_state *rs; + int i; + + /*XXX: ignored: + * light_twoside + * offset_cw/ccw -nohw + * scissor + * point_smooth -nohw + * multisample + * offset_units / offset_scale + */ + rs = MALLOC(sizeof(struct nv10_rasterizer_state)); + + rs->templ = cso; + + rs->shade_model = cso->flatshade ? 0x1d00 : 0x1d01; + + rs->line_width = (unsigned char)(cso->line_width * 8.0) & 0xff; + rs->line_smooth_en = cso->line_smooth ? 1 : 0; + + rs->point_size = *(uint32_t*)&cso->point_size; + + rs->poly_smooth_en = cso->poly_smooth ? 1 : 0; + + if (cso->front_winding == PIPE_WINDING_CCW) { + rs->front_face = NV10TCL_FRONT_FACE_CCW; + rs->poly_mode_front = nvgl_polygon_mode(cso->fill_ccw); + rs->poly_mode_back = nvgl_polygon_mode(cso->fill_cw); + } else { + rs->front_face = NV10TCL_FRONT_FACE_CW; + rs->poly_mode_front = nvgl_polygon_mode(cso->fill_cw); + rs->poly_mode_back = nvgl_polygon_mode(cso->fill_ccw); + } + + switch (cso->cull_mode) { + case PIPE_WINDING_CCW: + rs->cull_face_en = 1; + if (cso->front_winding == PIPE_WINDING_CCW) + rs->cull_face = NV10TCL_CULL_FACE_FRONT; + else + rs->cull_face = NV10TCL_CULL_FACE_BACK; + break; + case PIPE_WINDING_CW: + rs->cull_face_en = 1; + if (cso->front_winding == PIPE_WINDING_CW) + rs->cull_face = NV10TCL_CULL_FACE_FRONT; + else + rs->cull_face = NV10TCL_CULL_FACE_BACK; + break; + case PIPE_WINDING_BOTH: + rs->cull_face_en = 1; + rs->cull_face = NV10TCL_CULL_FACE_FRONT_AND_BACK; + break; + case PIPE_WINDING_NONE: + default: + rs->cull_face_en = 0; + rs->cull_face = 0; + break; + } + + if (cso->point_sprite) { + rs->point_sprite = (1 << 0); + for (i = 0; i < 8; i++) { + if (cso->sprite_coord_mode[i] != PIPE_SPRITE_COORD_NONE) + rs->point_sprite |= (1 << (8 + i)); + } + } else { + rs->point_sprite = 0; + } + + return (void *)rs; +} + +static void +nv10_rasterizer_state_bind(struct pipe_context *pipe, void *rast) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + nv10->rast = (struct nv10_rasterizer_state*)rast; + + draw_set_rasterizer_state(nv10->draw, (nv10->rast ? nv10->rast->templ : NULL)); + + nv10->dirty |= NV10_NEW_RAST; +} + +static void +nv10_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + +static void * +nv10_depth_stencil_alpha_state_create(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *cso) +{ + struct nv10_depth_stencil_alpha_state *hw; + + hw = MALLOC(sizeof(struct nv10_depth_stencil_alpha_state)); + + hw->depth.func = nvgl_comparison_op(cso->depth.func); + hw->depth.write_enable = cso->depth.writemask ? 1 : 0; + hw->depth.test_enable = cso->depth.enabled ? 1 : 0; + + hw->stencil.enable = cso->stencil[0].enabled ? 1 : 0; + hw->stencil.wmask = cso->stencil[0].write_mask; + hw->stencil.func = nvgl_comparison_op(cso->stencil[0].func); + hw->stencil.ref = cso->stencil[0].ref_value; + hw->stencil.vmask = cso->stencil[0].value_mask; + hw->stencil.fail = nvgl_stencil_op(cso->stencil[0].fail_op); + hw->stencil.zfail = nvgl_stencil_op(cso->stencil[0].zfail_op); + hw->stencil.zpass = nvgl_stencil_op(cso->stencil[0].zpass_op); + + hw->alpha.enabled = cso->alpha.enabled ? 1 : 0; + hw->alpha.func = nvgl_comparison_op(cso->alpha.func); + hw->alpha.ref = float_to_ubyte(cso->alpha.ref); + + return (void *)hw; +} + +static void +nv10_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *dsa) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + nv10->dsa = (struct nv10_depth_stencil_alpha_state*)dsa; + + nv10->dirty |= NV10_NEW_DSA; +} + +static void +nv10_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + +static void * +nv10_vp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + return draw_create_vertex_shader(nv10->draw, templ); +} + +static void +nv10_vp_state_bind(struct pipe_context *pipe, void *shader) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + draw_bind_vertex_shader(nv10->draw, (struct draw_vertex_shader *) shader); + + nv10->dirty |= NV10_NEW_VERTPROG; +} + +static void +nv10_vp_state_delete(struct pipe_context *pipe, void *shader) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + draw_delete_vertex_shader(nv10->draw, (struct draw_vertex_shader *) shader); +} + +static void * +nv10_fp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *cso) +{ + struct nv10_fragment_program *fp; + + fp = CALLOC(1, sizeof(struct nv10_fragment_program)); + fp->pipe.tokens = tgsi_dup_tokens(cso->tokens); + + tgsi_scan_shader(cso->tokens, &fp->info); + + return (void *)fp; +} + +static void +nv10_fp_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv10_context *nv10 = nv10_context(pipe); + struct nv10_fragment_program *fp = hwcso; + + nv10->fragprog.current = fp; + nv10->dirty |= NV10_NEW_FRAGPROG; +} + +static void +nv10_fp_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv10_context *nv10 = nv10_context(pipe); + struct nv10_fragment_program *fp = hwcso; + + nv10_fragprog_destroy(nv10, fp); + FREE((void*)fp->pipe.tokens); + FREE(fp); +} + +static void +nv10_set_blend_color(struct pipe_context *pipe, + const struct pipe_blend_color *bcol) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + nv10->blend_color = (struct pipe_blend_color*)bcol; + + nv10->dirty |= NV10_NEW_BLENDCOL; +} + +static void +nv10_set_clip_state(struct pipe_context *pipe, + const struct pipe_clip_state *clip) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + draw_set_clip_state(nv10->draw, clip); +} + +static void +nv10_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, + const struct pipe_constant_buffer *buf ) +{ + struct nv10_context *nv10 = nv10_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + if (buf) { + void *mapped; + if (buf->size && (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ))) + { + memcpy(nv10->constbuf[shader], mapped, buf->size); + nv10->constbuf_nr[shader] = + buf->size / (4 * sizeof(float)); + ws->buffer_unmap(ws, buf->buffer); + } + } +} + +static void +nv10_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + nv10->framebuffer = (struct pipe_framebuffer_state*)fb; + + nv10->dirty |= NV10_NEW_FRAMEBUFFER; +} + +static void +nv10_set_polygon_stipple(struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple) +{ + NOUVEAU_ERR("line stipple hahaha\n"); +} + +static void +nv10_set_scissor_state(struct pipe_context *pipe, + const struct pipe_scissor_state *s) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + nv10->scissor = (struct pipe_scissor_state*)s; + + nv10->dirty |= NV10_NEW_SCISSOR; +} + +static void +nv10_set_viewport_state(struct pipe_context *pipe, + const struct pipe_viewport_state *vpt) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + nv10->viewport = (struct pipe_viewport_state*)vpt; + + draw_set_viewport_state(nv10->draw, nv10->viewport); + + nv10->dirty |= NV10_NEW_VIEWPORT; +} + +static void +nv10_set_vertex_buffers(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_buffer *vb) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + memcpy(nv10->vtxbuf, vb, sizeof(*vb) * count); + nv10->dirty |= NV10_NEW_VTXARRAYS; + + draw_set_vertex_buffers(nv10->draw, count, vb); +} + +static void +nv10_set_vertex_elements(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_element *ve) +{ + struct nv10_context *nv10 = nv10_context(pipe); + + memcpy(nv10->vtxelt, ve, sizeof(*ve) * count); + nv10->dirty |= NV10_NEW_VTXARRAYS; + + draw_set_vertex_elements(nv10->draw, count, ve); +} + +void +nv10_init_state_functions(struct nv10_context *nv10) +{ + nv10->pipe.create_blend_state = nv10_blend_state_create; + nv10->pipe.bind_blend_state = nv10_blend_state_bind; + nv10->pipe.delete_blend_state = nv10_blend_state_delete; + + nv10->pipe.create_sampler_state = nv10_sampler_state_create; + nv10->pipe.bind_sampler_states = nv10_sampler_state_bind; + nv10->pipe.delete_sampler_state = nv10_sampler_state_delete; + nv10->pipe.set_sampler_textures = nv10_set_sampler_texture; + + nv10->pipe.create_rasterizer_state = nv10_rasterizer_state_create; + nv10->pipe.bind_rasterizer_state = nv10_rasterizer_state_bind; + nv10->pipe.delete_rasterizer_state = nv10_rasterizer_state_delete; + + nv10->pipe.create_depth_stencil_alpha_state = + nv10_depth_stencil_alpha_state_create; + nv10->pipe.bind_depth_stencil_alpha_state = + nv10_depth_stencil_alpha_state_bind; + nv10->pipe.delete_depth_stencil_alpha_state = + nv10_depth_stencil_alpha_state_delete; + + nv10->pipe.create_vs_state = nv10_vp_state_create; + nv10->pipe.bind_vs_state = nv10_vp_state_bind; + nv10->pipe.delete_vs_state = nv10_vp_state_delete; + + nv10->pipe.create_fs_state = nv10_fp_state_create; + nv10->pipe.bind_fs_state = nv10_fp_state_bind; + nv10->pipe.delete_fs_state = nv10_fp_state_delete; + + nv10->pipe.set_blend_color = nv10_set_blend_color; + nv10->pipe.set_clip_state = nv10_set_clip_state; + nv10->pipe.set_constant_buffer = nv10_set_constant_buffer; + nv10->pipe.set_framebuffer_state = nv10_set_framebuffer_state; + nv10->pipe.set_polygon_stipple = nv10_set_polygon_stipple; + nv10->pipe.set_scissor_state = nv10_set_scissor_state; + nv10->pipe.set_viewport_state = nv10_set_viewport_state; + + nv10->pipe.set_vertex_buffers = nv10_set_vertex_buffers; + nv10->pipe.set_vertex_elements = nv10_set_vertex_elements; +} + diff --git a/src/gallium/drivers/nv10/nv10_state.h b/src/gallium/drivers/nv10/nv10_state.h new file mode 100644 index 0000000000..3a3fd0d4f4 --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_state.h @@ -0,0 +1,139 @@ +#ifndef __NV10_STATE_H__ +#define __NV10_STATE_H__ + +#include "pipe/p_state.h" +#include "tgsi/tgsi_scan.h" + +struct nv10_blend_state { + uint32_t b_enable; + uint32_t b_srcfunc; + uint32_t b_dstfunc; + + uint32_t c_mask; + + uint32_t d_enable; +}; + +struct nv10_sampler_state { + uint32_t wrap; + uint32_t en; + uint32_t filt; + uint32_t bcol; +}; + +struct nv10_rasterizer_state { + uint32_t shade_model; + + uint32_t line_width; + uint32_t line_smooth_en; + + uint32_t point_size; + + uint32_t poly_smooth_en; + + uint32_t poly_mode_front; + uint32_t poly_mode_back; + + uint32_t front_face; + uint32_t cull_face; + uint32_t cull_face_en; + + uint32_t point_sprite; + + const struct pipe_rasterizer_state *templ; +}; + +struct nv10_vertex_program_exec { + uint32_t data[4]; + boolean has_branch_offset; + int const_index; +}; + +struct nv10_vertex_program_data { + int index; /* immediates == -1 */ + float value[4]; +}; + +struct nv10_vertex_program { + const struct pipe_shader_state *pipe; + + boolean translated; + struct nv10_vertex_program_exec *insns; + unsigned nr_insns; + struct nv10_vertex_program_data *consts; + unsigned nr_consts; + + struct nouveau_resource *exec; + unsigned exec_start; + struct nouveau_resource *data; + unsigned data_start; + unsigned data_start_min; + + uint32_t ir; + uint32_t or; +}; + +struct nv10_fragment_program_data { + unsigned offset; + unsigned index; +}; + +struct nv10_fragment_program { + struct pipe_shader_state pipe; + struct tgsi_shader_info info; + + boolean translated; + boolean on_hw; + unsigned samplers; + + uint32_t *insn; + int insn_len; + + struct nv10_fragment_program_data *consts; + unsigned nr_consts; + + struct pipe_buffer *buffer; + + uint32_t fp_control; + uint32_t fp_reg_control; +}; + + +struct nv10_depth_stencil_alpha_state { + struct { + uint32_t func; + uint32_t write_enable; + uint32_t test_enable; + } depth; + + struct { + uint32_t enable; + uint32_t wmask; + uint32_t func; + uint32_t ref; + uint32_t vmask; + uint32_t fail; + uint32_t zfail; + uint32_t zpass; + } stencil; + + struct { + uint32_t enabled; + uint32_t func; + uint32_t ref; + } alpha; +}; + +struct nv10_miptree { + struct pipe_texture base; + + struct pipe_buffer *buffer; + uint total_size; + + struct { + uint pitch; + uint *image_offset; + } level[PIPE_MAX_TEXTURE_LEVELS]; +}; + +#endif diff --git a/src/gallium/drivers/nv10/nv10_state_emit.c b/src/gallium/drivers/nv10/nv10_state_emit.c new file mode 100644 index 0000000000..46c7e1d753 --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_state_emit.c @@ -0,0 +1,303 @@ +#include "nv10_context.h" +#include "nv10_state.h" + +static void nv10_state_emit_blend(struct nv10_context* nv10) +{ + struct nv10_blend_state *b = nv10->blend; + + BEGIN_RING(celsius, NV10TCL_DITHER_ENABLE, 1); + OUT_RING (b->d_enable); + + BEGIN_RING(celsius, NV10TCL_BLEND_FUNC_ENABLE, 3); + OUT_RING (b->b_enable); + OUT_RING (b->b_srcfunc); + OUT_RING (b->b_dstfunc); + + BEGIN_RING(celsius, NV10TCL_COLOR_MASK, 1); + OUT_RING (b->c_mask); +} + +static void nv10_state_emit_blend_color(struct nv10_context* nv10) +{ + struct pipe_blend_color *c = nv10->blend_color; + + BEGIN_RING(celsius, NV10TCL_BLEND_COLOR, 1); + OUT_RING ((float_to_ubyte(c->color[3]) << 24)| + (float_to_ubyte(c->color[0]) << 16)| + (float_to_ubyte(c->color[1]) << 8) | + (float_to_ubyte(c->color[2]) << 0)); +} + +static void nv10_state_emit_rast(struct nv10_context* nv10) +{ + struct nv10_rasterizer_state *r = nv10->rast; + + BEGIN_RING(celsius, NV10TCL_SHADE_MODEL, 2); + OUT_RING (r->shade_model); + OUT_RING (r->line_width); + + + BEGIN_RING(celsius, NV10TCL_POINT_SIZE, 1); + OUT_RING (r->point_size); + + BEGIN_RING(celsius, NV10TCL_POLYGON_MODE_FRONT, 2); + OUT_RING (r->poly_mode_front); + OUT_RING (r->poly_mode_back); + + + BEGIN_RING(celsius, NV10TCL_CULL_FACE, 2); + OUT_RING (r->cull_face); + OUT_RING (r->front_face); + + BEGIN_RING(celsius, NV10TCL_LINE_SMOOTH_ENABLE, 2); + OUT_RING (r->line_smooth_en); + OUT_RING (r->poly_smooth_en); + + BEGIN_RING(celsius, NV10TCL_CULL_FACE_ENABLE, 1); + OUT_RING (r->cull_face_en); +} + +static void nv10_state_emit_dsa(struct nv10_context* nv10) +{ + struct nv10_depth_stencil_alpha_state *d = nv10->dsa; + + BEGIN_RING(celsius, NV10TCL_DEPTH_FUNC, 1); + OUT_RING (d->depth.func); + + BEGIN_RING(celsius, NV10TCL_DEPTH_WRITE_ENABLE, 1); + OUT_RING (d->depth.write_enable); + + BEGIN_RING(celsius, NV10TCL_DEPTH_TEST_ENABLE, 1); + OUT_RING (d->depth.test_enable); + +#if 0 + BEGIN_RING(celsius, NV10TCL_STENCIL_ENABLE, 1); + OUT_RING (d->stencil.enable); + BEGIN_RING(celsius, NV10TCL_STENCIL_MASK, 7); + OUT_RINGp ((uint32_t *)&(d->stencil.wmask), 7); +#endif + + BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_ENABLE, 1); + OUT_RING (d->alpha.enabled); + + BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_FUNC, 1); + OUT_RING (d->alpha.func); + + BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_REF, 1); + OUT_RING (d->alpha.ref); +} + +static void nv10_state_emit_viewport(struct nv10_context* nv10) +{ +} + +static void nv10_state_emit_scissor(struct nv10_context* nv10) +{ + // XXX this is so not working +/* struct pipe_scissor_state *s = nv10->scissor; + BEGIN_RING(celsius, NV10TCL_SCISSOR_HORIZ, 2); + OUT_RING (((s->maxx - s->minx) << 16) | s->minx); + OUT_RING (((s->maxy - s->miny) << 16) | s->miny);*/ +} + +static void nv10_state_emit_framebuffer(struct nv10_context* nv10) +{ + struct pipe_framebuffer_state* fb = nv10->framebuffer; + struct pipe_surface *rt, *zeta = NULL; + uint32_t rt_format, w, h; + int colour_format = 0, zeta_format = 0; + + w = fb->cbufs[0]->width; + h = fb->cbufs[0]->height; + colour_format = fb->cbufs[0]->format; + rt = fb->cbufs[0]; + + if (fb->zsbuf) { + if (colour_format) { + assert(w == fb->zsbuf->width); + assert(h == fb->zsbuf->height); + } else { + w = fb->zsbuf->width; + h = fb->zsbuf->height; + } + + zeta_format = fb->zsbuf->format; + zeta = fb->zsbuf; + } + + rt_format = NV10TCL_RT_FORMAT_TYPE_LINEAR; + + switch (colour_format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case 0: + rt_format |= NV10TCL_RT_FORMAT_COLOR_A8R8G8B8; + break; + case PIPE_FORMAT_R5G6B5_UNORM: + rt_format |= NV10TCL_RT_FORMAT_COLOR_R5G6B5; + break; + default: + assert(0); + } + + if (zeta) { + BEGIN_RING(celsius, NV10TCL_RT_PITCH, 1); + OUT_RING (rt->stride | (zeta->stride << 16)); + } else { + BEGIN_RING(celsius, NV10TCL_RT_PITCH, 1); + OUT_RING (rt->stride | (rt->stride << 16)); + } + + nv10->rt[0] = rt->buffer; + + if (zeta_format) + { + nv10->zeta = zeta->buffer; + } + + BEGIN_RING(celsius, NV10TCL_RT_HORIZ, 3); + OUT_RING ((w << 16) | 0); + OUT_RING ((h << 16) | 0); + OUT_RING (rt_format); + BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_HORIZ(0), 2); + OUT_RING (((w - 1) << 16) | 0 | 0x08000800); + OUT_RING (((h - 1) << 16) | 0 | 0x08000800); +} + +static void nv10_vertex_layout(struct nv10_context *nv10) +{ + struct nv10_fragment_program *fp = nv10->fragprog.current; + uint32_t src = 0; + int i; + struct vertex_info vinfo; + + memset(&vinfo, 0, sizeof(vinfo)); + + for (i = 0; i < fp->info.num_inputs; i++) { + switch (fp->info.input_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src++); + break; + case TGSI_SEMANTIC_COLOR: + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src++); + break; + default: + case TGSI_SEMANTIC_GENERIC: + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++); + break; + case TGSI_SEMANTIC_FOG: + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++); + break; + } + } + draw_compute_vertex_size(&vinfo); +} + +void +nv10_emit_hw_state(struct nv10_context *nv10) +{ + int i; + + if (nv10->dirty & NV10_NEW_VERTPROG) { + //nv10_vertprog_bind(nv10, nv10->vertprog.current); + nv10->dirty &= ~NV10_NEW_VERTPROG; + } + + if (nv10->dirty & NV10_NEW_FRAGPROG) { + nv10_fragprog_bind(nv10, nv10->fragprog.current); + /*XXX: clear NV10_NEW_FRAGPROG if no new program uploaded */ + nv10->dirty_samplers |= (1<<10); + nv10->dirty_samplers = 0; + } + + if (nv10->dirty_samplers || (nv10->dirty & NV10_NEW_FRAGPROG)) { + nv10_fragtex_bind(nv10); + nv10->dirty &= ~NV10_NEW_FRAGPROG; + } + + if (nv10->dirty & NV10_NEW_VTXARRAYS) { + nv10->dirty &= ~NV10_NEW_VTXARRAYS; + nv10_vertex_layout(nv10); + nv10_vtxbuf_bind(nv10); + } + + if (nv10->dirty & NV10_NEW_BLEND) { + nv10->dirty &= ~NV10_NEW_BLEND; + nv10_state_emit_blend(nv10); + } + + if (nv10->dirty & NV10_NEW_BLENDCOL) { + nv10->dirty &= ~NV10_NEW_BLENDCOL; + nv10_state_emit_blend_color(nv10); + } + + if (nv10->dirty & NV10_NEW_RAST) { + nv10->dirty &= ~NV10_NEW_RAST; + nv10_state_emit_rast(nv10); + } + + if (nv10->dirty & NV10_NEW_DSA) { + nv10->dirty &= ~NV10_NEW_DSA; + nv10_state_emit_dsa(nv10); + } + + if (nv10->dirty & NV10_NEW_VIEWPORT) { + nv10->dirty &= ~NV10_NEW_VIEWPORT; + nv10_state_emit_viewport(nv10); + } + + if (nv10->dirty & NV10_NEW_SCISSOR) { + nv10->dirty &= ~NV10_NEW_SCISSOR; + nv10_state_emit_scissor(nv10); + } + + if (nv10->dirty & NV10_NEW_FRAMEBUFFER) { + nv10->dirty &= ~NV10_NEW_FRAMEBUFFER; + nv10_state_emit_framebuffer(nv10); + } + + /* Emit relocs for every referenced buffer. + * This is to ensure the bufmgr has an accurate idea of how + * the buffer is used. This isn't very efficient, but we don't + * seem to take a significant performance hit. Will be improved + * at some point. Vertex arrays are emitted by nv10_vbo.c + */ + + /* Render target */ +// XXX figre out who's who for NV10TCL_DMA_* and fill accordingly +// BEGIN_RING(celsius, NV10TCL_DMA_COLOR0, 1); +// OUT_RELOCo(nv10->rt[0], NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + BEGIN_RING(celsius, NV10TCL_COLOR_OFFSET, 1); + OUT_RELOCl(nv10->rt[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + + if (nv10->zeta) { +// XXX +// BEGIN_RING(celsius, NV10TCL_DMA_ZETA, 1); +// OUT_RELOCo(nv10->zeta, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + BEGIN_RING(celsius, NV10TCL_ZETA_OFFSET, 1); + OUT_RELOCl(nv10->zeta, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + /* XXX for when we allocate LMA on nv17 */ +/* BEGIN_RING(celsius, NV10TCL_LMA_DEPTH_BUFFER_OFFSET, 1); + OUT_RELOCl(nv10->zeta + lma_offset);*/ + } + + /* Vertex buffer */ + BEGIN_RING(celsius, NV10TCL_DMA_VTXBUF0, 1); + OUT_RELOCo(nv10->rt[0], NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + BEGIN_RING(celsius, NV10TCL_COLOR_OFFSET, 1); + OUT_RELOCl(nv10->rt[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + + /* Texture images */ + for (i = 0; i < 2; i++) { + if (!(nv10->fp_samplers & (1 << i))) + continue; + BEGIN_RING(celsius, NV10TCL_TX_OFFSET(i), 1); + OUT_RELOCl(nv10->tex[i].buffer, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_GART | NOUVEAU_BO_RD); + BEGIN_RING(celsius, NV10TCL_TX_FORMAT(i), 1); + OUT_RELOCd(nv10->tex[i].buffer, nv10->tex[i].format, + NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | + NOUVEAU_BO_OR, NV10TCL_TX_FORMAT_DMA0, + NV10TCL_TX_FORMAT_DMA1); + } +} + diff --git a/src/gallium/drivers/nv10/nv10_surface.c b/src/gallium/drivers/nv10/nv10_surface.c new file mode 100644 index 0000000000..be44c7bed5 --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_surface.c @@ -0,0 +1,64 @@ + +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "nv10_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/p_inlines.h" +#include "util/u_tile.h" + +static void +nv10_surface_copy(struct pipe_context *pipe, boolean do_flip, + struct pipe_surface *dest, unsigned destx, unsigned desty, + struct pipe_surface *src, unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + struct nv10_context *nv10 = nv10_context(pipe); + struct nouveau_winsys *nvws = nv10->nvws; + + nvws->surface_copy(nvws, dest, destx, desty, src, srcx, srcy, + width, height); +} + +static void +nv10_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest, + unsigned destx, unsigned desty, unsigned width, + unsigned height, unsigned value) +{ + struct nv10_context *nv10 = nv10_context(pipe); + struct nouveau_winsys *nvws = nv10->nvws; + + nvws->surface_fill(nvws, dest, destx, desty, width, height, value); +} + +void +nv10_init_surface_functions(struct nv10_context *nv10) +{ + nv10->pipe.surface_copy = nv10_surface_copy; + nv10->pipe.surface_fill = nv10_surface_fill; +} diff --git a/src/gallium/drivers/nv10/nv10_vbo.c b/src/gallium/drivers/nv10/nv10_vbo.c new file mode 100644 index 0000000000..d0e788ac03 --- /dev/null +++ b/src/gallium/drivers/nv10/nv10_vbo.c @@ -0,0 +1,77 @@ +#include "draw/draw_context.h" +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +#include "nv10_context.h" +#include "nv10_state.h" + +#include "nouveau/nouveau_channel.h" +#include "nouveau/nouveau_pushbuf.h" + +boolean nv10_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) +{ + struct nv10_context *nv10 = nv10_context( pipe ); + struct draw_context *draw = nv10->draw; + unsigned i; + + nv10_emit_hw_state(nv10); + + /* + * Map vertex buffers + */ + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { + if (nv10->vtxbuf[i].buffer) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + nv10->vtxbuf[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); + } + } + /* Map index buffer, if present */ + if (indexBuffer) { + void *mapped_indexes + = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); + } + else { + /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + draw_set_mapped_constant_buffer(draw, + nv10->constbuf[PIPE_SHADER_VERTEX], + nv10->constbuf_nr[PIPE_SHADER_VERTEX]); + + /* draw! */ + draw_arrays(nv10->draw, prim, start, count); + + /* + * unmap vertex/index buffers + */ + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { + if (nv10->vtxbuf[i].buffer) { + pipe->winsys->buffer_unmap(pipe->winsys, nv10->vtxbuf[i].buffer); + draw_set_mapped_vertex_buffer(draw, i, NULL); + } + } + if (indexBuffer) { + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + return TRUE; +} + +boolean nv10_draw_arrays( struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) +{ + return nv10_draw_elements(pipe, NULL, 0, prim, start, count); +} + + + diff --git a/src/gallium/drivers/nv20/Makefile b/src/gallium/drivers/nv20/Makefile new file mode 100644 index 0000000000..d777fd3d8b --- /dev/null +++ b/src/gallium/drivers/nv20/Makefile @@ -0,0 +1,29 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = nv20 + +DRIVER_SOURCES = \ + nv20_clear.c \ + nv20_context.c \ + nv20_fragprog.c \ + nv20_fragtex.c \ + nv20_miptree.c \ + nv20_prim_vbuf.c \ + nv20_screen.c \ + nv20_state.c \ + nv20_state_emit.c \ + nv20_surface.c \ + nv20_vbo.c +# nv20_vertprog.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/drivers/nv20/nv20_clear.c b/src/gallium/drivers/nv20/nv20_clear.c new file mode 100644 index 0000000000..81b6f3e78a --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_clear.c @@ -0,0 +1,12 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "nv20_context.h" + +void +nv20_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); +} diff --git a/src/gallium/drivers/nv20/nv20_context.c b/src/gallium/drivers/nv20/nv20_context.c new file mode 100644 index 0000000000..9a17f4af57 --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_context.c @@ -0,0 +1,419 @@ +#include "draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" + +#include "nv20_context.h" +#include "nv20_screen.h" + +static void +nv20_flush(struct pipe_context *pipe, unsigned flags, + struct pipe_fence_handle **fence) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + draw_flush(nv20->draw); + + FIRE_RING(fence); +} + +static void +nv20_destroy(struct pipe_context *pipe) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + if (nv20->draw) + draw_destroy(nv20->draw); + + FREE(nv20); +} + +static void nv20_init_hwctx(struct nv20_context *nv20) +{ + struct nv20_screen *screen = nv20->screen; + struct nouveau_winsys *nvws = screen->nvws; + int i; + float projectionmatrix[16]; + const boolean is_nv25tcl = (nv20->screen->kelvin->grclass == NV25TCL); + + BEGIN_RING(kelvin, NV20TCL_DMA_NOTIFY, 1); + OUT_RING (screen->sync->handle); + BEGIN_RING(kelvin, NV20TCL_DMA_TEXTURE0, 2); + OUT_RING (nvws->channel->vram->handle); + OUT_RING (nvws->channel->gart->handle); /* TEXTURE1 */ + BEGIN_RING(kelvin, NV20TCL_DMA_COLOR, 2); + OUT_RING (nvws->channel->vram->handle); + OUT_RING (nvws->channel->vram->handle); /* ZETA */ + + BEGIN_RING(kelvin, NV20TCL_DMA_QUERY, 1); + OUT_RING (0); /* renouveau: beef0351, unique */ + + BEGIN_RING(kelvin, NV20TCL_RT_HORIZ, 2); + OUT_RING (0); + OUT_RING (0); + + BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_HORIZ(0), 1); + OUT_RING ((0xfff << 16) | 0x0); + BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_VERT(0), 1); + OUT_RING ((0xfff << 16) | 0x0); + + for (i = 1; i < NV20TCL_VIEWPORT_CLIP_HORIZ__SIZE; i++) { + BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_HORIZ(i), 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_VERT(i), 1); + OUT_RING (0); + } + + BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_MODE, 1); + OUT_RING (0); + + BEGIN_RING(kelvin, 0x17e0, 3); + OUT_RINGf (0.0); + OUT_RINGf (0.0); + OUT_RINGf (1.0); + + if (is_nv25tcl) { + BEGIN_RING(kelvin, NV20TCL_TX_RCOMP, 1); + OUT_RING (NV20TCL_TX_RCOMP_LEQUAL | 0xdb0); + } else { + BEGIN_RING(kelvin, 0x1e68, 1); + OUT_RING (0x4b800000); /* 16777216.000000 */ + BEGIN_RING(kelvin, NV20TCL_TX_RCOMP, 1); + OUT_RING (NV20TCL_TX_RCOMP_LEQUAL); + } + + BEGIN_RING(kelvin, 0x290, 1); + OUT_RING ((0x10 << 16) | 1); + BEGIN_RING(kelvin, 0x9fc, 1); + OUT_RING (0); + BEGIN_RING(kelvin, 0x1d80, 1); + OUT_RING (1); + BEGIN_RING(kelvin, 0x9f8, 1); + OUT_RING (4); + BEGIN_RING(kelvin, 0x17ec, 3); + OUT_RINGf (0.0); + OUT_RINGf (1.0); + OUT_RINGf (0.0); + + if (is_nv25tcl) { + BEGIN_RING(kelvin, 0x1d88, 1); + OUT_RING (3); + + BEGIN_RING(kelvin, NV25TCL_DMA_IN_MEMORY9, 1); + OUT_RING (nvws->channel->vram->handle); + BEGIN_RING(kelvin, NV25TCL_DMA_IN_MEMORY8, 1); + OUT_RING (nvws->channel->vram->handle); + } + BEGIN_RING(kelvin, NV20TCL_DMA_FENCE, 1); + OUT_RING (0); /* renouveau: beef1e10 */ + + BEGIN_RING(kelvin, 0x1e98, 1); + OUT_RING (0); +#if 0 + if (is_nv25tcl) { + BEGIN_RING(NvSub3D, NV25TCL_DMA_IN_MEMORY4, 2); + OUT_RING (NvDmaTT); /* renouveau: beef0202 */ + OUT_RING (NvDmaFB); /* renouveau: beef0201 */ + + BEGIN_RING(NvSub3D, NV20TCL_DMA_TEXTURE1, 1); + OUT_RING (NvDmaTT); /* renouveau: beef0202 */ + } +#endif + BEGIN_RING(kelvin, NV20TCL_NOTIFY, 1); + OUT_RING (0); + + BEGIN_RING(kelvin, 0x120, 3); + OUT_RING (0); + OUT_RING (1); + OUT_RING (2); + +/* error: ILLEGAL_MTHD, PROTECTION_FAULT + BEGIN_RING(kelvin, NV20TCL_VIEWPORT_TRANSLATE_X, 4); + OUT_RINGf (0.0); + OUT_RINGf (512.0); + OUT_RINGf (0.0); + OUT_RINGf (0.0); +*/ + + if (is_nv25tcl) { + BEGIN_RING(kelvin, 0x022c, 2); + OUT_RING (0x280); + OUT_RING (0x07d28000); + } + +/* * illegal method, protection fault + BEGIN_RING(NvSub3D, 0x1c2c, 1); + OUT_RING (0); */ + + if (is_nv25tcl) { + BEGIN_RING(kelvin, 0x1da4, 1); + OUT_RING (0); + } + +/* * crashes with illegal method, protection fault + BEGIN_RING(NvSub3D, 0x1c18, 1); + OUT_RING (0x200); */ + + BEGIN_RING(kelvin, NV20TCL_RT_HORIZ, 2); + OUT_RING ((0 << 16) | 0); + OUT_RING ((0 << 16) | 0); + + /* *** Set state *** */ + + BEGIN_RING(kelvin, NV20TCL_ALPHA_FUNC_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_ALPHA_FUNC_FUNC, 2); + OUT_RING (NV20TCL_ALPHA_FUNC_FUNC_ALWAYS); + OUT_RING (0); /* NV20TCL_ALPHA_FUNC_REF */ + + for (i = 0; i < NV20TCL_TX_ENABLE__SIZE; ++i) { + BEGIN_RING(kelvin, NV20TCL_TX_ENABLE(i), 1); + OUT_RING (0); + } + BEGIN_RING(kelvin, NV20TCL_TX_SHADER_OP, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_TX_SHADER_CULL_MODE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_RC_IN_ALPHA(0), 4); + OUT_RING (0x30d410d0); + OUT_RING (0); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_RC_OUT_RGB(0), 4); + OUT_RING (0x00000c00); + OUT_RING (0); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_RC_ENABLE, 1); + OUT_RING (0x00011101); + BEGIN_RING(kelvin, NV20TCL_RC_FINAL0, 2); + OUT_RING (0x130e0300); + OUT_RING (0x0c091c80); + BEGIN_RING(kelvin, NV20TCL_RC_OUT_ALPHA(0), 4); + OUT_RING (0x00000c00); + OUT_RING (0); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_RC_IN_RGB(0), 4); + OUT_RING (0x20c400c0); + OUT_RING (0); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_RC_COLOR0, 2); + OUT_RING (0); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_RC_CONSTANT_COLOR0(0), 4); + OUT_RING (0x035125a0); + OUT_RING (0); + OUT_RING (0x40002000); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_MULTISAMPLE_CONTROL, 1); + OUT_RING (0xffff0000); + + BEGIN_RING(kelvin, NV20TCL_BLEND_FUNC_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_DITHER_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_STENCIL_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_BLEND_FUNC_SRC, 4); + OUT_RING (NV20TCL_BLEND_FUNC_SRC_ONE); + OUT_RING (NV20TCL_BLEND_FUNC_DST_ZERO); + OUT_RING (0); /* NV20TCL_BLEND_COLOR */ + OUT_RING (NV20TCL_BLEND_EQUATION_FUNC_ADD); + BEGIN_RING(kelvin, NV20TCL_STENCIL_MASK, 7); + OUT_RING (0xff); + OUT_RING (NV20TCL_STENCIL_FUNC_FUNC_ALWAYS); + OUT_RING (0); /* NV20TCL_STENCIL_FUNC_REF */ + OUT_RING (0xff); /* NV20TCL_STENCIL_FUNC_MASK */ + OUT_RING (NV20TCL_STENCIL_OP_FAIL_KEEP); + OUT_RING (NV20TCL_STENCIL_OP_ZFAIL_KEEP); + OUT_RING (NV20TCL_STENCIL_OP_ZPASS_KEEP); + + BEGIN_RING(kelvin, NV20TCL_COLOR_LOGIC_OP_ENABLE, 2); + OUT_RING (0); + OUT_RING (NV20TCL_COLOR_LOGIC_OP_OP_COPY); + BEGIN_RING(kelvin, 0x17cc, 1); + OUT_RING (0); + if (is_nv25tcl) { + BEGIN_RING(kelvin, 0x1d84, 1); + OUT_RING (1); + } + BEGIN_RING(kelvin, NV20TCL_LIGHTING_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_LIGHT_CONTROL, 1); + OUT_RING (0x00020000); + BEGIN_RING(kelvin, NV20TCL_SEPARATE_SPECULAR_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_LIGHT_MODEL_TWO_SIDE_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_ENABLED_LIGHTS, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_NORMALIZE_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_POLYGON_STIPPLE_PATTERN(0), + NV20TCL_POLYGON_STIPPLE_PATTERN__SIZE); + for (i = 0; i < NV20TCL_POLYGON_STIPPLE_PATTERN__SIZE; ++i) { + OUT_RING(0xffffffff); + } + + BEGIN_RING(kelvin, NV20TCL_POLYGON_OFFSET_POINT_ENABLE, 3); + OUT_RING (0); + OUT_RING (0); /* NV20TCL.POLYGON_OFFSET_LINE_ENABLE */ + OUT_RING (0); /* NV20TCL.POLYGON_OFFSET_FILL_ENABLE */ + BEGIN_RING(kelvin, NV20TCL_DEPTH_FUNC, 1); + OUT_RING (NV20TCL_DEPTH_FUNC_LESS); + BEGIN_RING(kelvin, NV20TCL_DEPTH_WRITE_ENABLE, 1); + OUT_RING (1); + BEGIN_RING(kelvin, NV20TCL_DEPTH_TEST_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_POLYGON_OFFSET_FACTOR, 2); + OUT_RINGf (0.0); + OUT_RINGf (0.0); /* NV20TCL.POLYGON_OFFSET_UNITS */ + BEGIN_RING(kelvin, NV20TCL_DEPTH_UNK17D8, 1); + OUT_RING (1); + if (!is_nv25tcl) { + BEGIN_RING(kelvin, 0x1d84, 1); + OUT_RING (3); + } + BEGIN_RING(kelvin, NV20TCL_POINT_SIZE, 1); + if (!is_nv25tcl) { + OUT_RING (8); + } else { + OUT_RINGf (1.0); + } + if (!is_nv25tcl) { + BEGIN_RING(kelvin, NV20TCL_POINT_PARAMETERS_ENABLE, 2); + OUT_RING (0); + OUT_RING (0); /* NV20TCL.POINT_SMOOTH_ENABLE */ + } else { + BEGIN_RING(kelvin, NV20TCL_POINT_PARAMETERS_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, 0x0a1c, 1); + OUT_RING (0x800); + } + BEGIN_RING(kelvin, NV20TCL_LINE_WIDTH, 1); + OUT_RING (8); + BEGIN_RING(kelvin, NV20TCL_LINE_SMOOTH_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_POLYGON_MODE_FRONT, 2); + OUT_RING (NV20TCL_POLYGON_MODE_FRONT_FILL); + OUT_RING (NV20TCL_POLYGON_MODE_BACK_FILL); + BEGIN_RING(kelvin, NV20TCL_CULL_FACE, 2); + OUT_RING (NV20TCL_CULL_FACE_BACK); + OUT_RING (NV20TCL_FRONT_FACE_CCW); + BEGIN_RING(kelvin, NV20TCL_POLYGON_SMOOTH_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_CULL_FACE_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_SHADE_MODEL, 1); + OUT_RING (NV20TCL_SHADE_MODEL_SMOOTH); + BEGIN_RING(kelvin, NV20TCL_POLYGON_STIPPLE_ENABLE, 1); + OUT_RING (0); + BEGIN_RING(kelvin, NV20TCL_TX_GEN_S(0), 4 * NV20TCL_TX_GEN_S__SIZE); + for (i=0; i < 4 * NV20TCL_TX_GEN_S__SIZE; ++i) { + OUT_RING(0); + } + BEGIN_RING(kelvin, NV20TCL_FOG_EQUATION_CONSTANT, 3); + OUT_RINGf (1.5); + OUT_RINGf (-0.090168); /* NV20TCL.FOG_EQUATION_LINEAR */ + OUT_RINGf (0.0); /* NV20TCL.FOG_EQUATION_QUADRATIC */ + BEGIN_RING(kelvin, NV20TCL_FOG_MODE, 2); + OUT_RING (NV20TCL_FOG_MODE_EXP_2); + OUT_RING (NV20TCL_FOG_COORD_DIST_COORD_FOG); + BEGIN_RING(kelvin, NV20TCL_FOG_ENABLE, 2); + OUT_RING (0); + OUT_RING (0); /* NV20TCL.FOG_COLOR */ + BEGIN_RING(kelvin, NV20TCL_ENGINE, 1); + OUT_RING (NV20TCL_ENGINE_FIXED); + + for (i = 0; i < NV20TCL_TX_MATRIX_ENABLE__SIZE; ++i) { + BEGIN_RING(kelvin, NV20TCL_TX_MATRIX_ENABLE(i), 1); + OUT_RING (0); + } + + BEGIN_RING(kelvin, NV20TCL_VTX_ATTR_4F_X(1), 4 * 15); + OUT_RINGf(1.0); OUT_RINGf(0.0); OUT_RINGf(0.0); OUT_RINGf(1.0); + OUT_RINGf(0.0); OUT_RINGf(0.0); OUT_RINGf(1.0); OUT_RINGf(1.0); + OUT_RINGf(1.0); OUT_RINGf(1.0); OUT_RINGf(1.0); OUT_RINGf(1.0); + for (i = 4; i < 16; ++i) { + OUT_RINGf(0.0); OUT_RINGf(0.0); OUT_RINGf(0.0); OUT_RINGf(1.0); + } + + BEGIN_RING(kelvin, NV20TCL_EDGEFLAG_ENABLE, 1); + OUT_RING (1); + BEGIN_RING(kelvin, NV20TCL_COLOR_MASK, 1); + OUT_RING (0x00010101); + BEGIN_RING(kelvin, NV20TCL_CLEAR_VALUE, 1); + OUT_RING (0); + + memset(projectionmatrix, 0, sizeof(projectionmatrix)); + projectionmatrix[0*4+0] = 1.0; + projectionmatrix[1*4+1] = 1.0; + projectionmatrix[2*4+2] = 1.0; + projectionmatrix[3*4+3] = 1.0; + BEGIN_RING(kelvin, NV20TCL_PROJECTION_MATRIX(0), 16); + for (i = 0; i < 16; i++) { + OUT_RINGf (projectionmatrix[i]); + } + + BEGIN_RING(kelvin, NV20TCL_DEPTH_RANGE_NEAR, 2); + OUT_RINGf (0.0); + OUT_RINGf (16777216.0); /* bpp dependant? */ + + BEGIN_RING(kelvin, NV20TCL_VIEWPORT_SCALE0_X, 4); + OUT_RINGf (-2048.0); + OUT_RINGf (-2048.0); + OUT_RINGf (16777215.0 * 0.5); + OUT_RING (0); + + BEGIN_RING(kelvin, NV20TCL_VIEWPORT_SCALE1_X, 4); + OUT_RINGf (-2048.0); + OUT_RINGf (-2048.0); + OUT_RINGf (16777215.0 * 0.5); + OUT_RING (0); + + FIRE_RING (NULL); +} + +static void +nv20_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield) +{ +} + +struct pipe_context * +nv20_create(struct pipe_screen *pscreen, unsigned pctx_id) +{ + struct nv20_screen *screen = nv20_screen(pscreen); + struct pipe_winsys *ws = pscreen->winsys; + struct nv20_context *nv20; + struct nouveau_winsys *nvws = screen->nvws; + + nv20 = CALLOC(1, sizeof(struct nv20_context)); + if (!nv20) + return NULL; + nv20->screen = screen; + nv20->pctx_id = pctx_id; + + nv20->nvws = nvws; + + nv20->pipe.winsys = ws; + nv20->pipe.screen = pscreen; + nv20->pipe.destroy = nv20_destroy; + nv20->pipe.set_edgeflags = nv20_set_edgeflags; + nv20->pipe.draw_arrays = nv20_draw_arrays; + nv20->pipe.draw_elements = nv20_draw_elements; + nv20->pipe.clear = nv20_clear; + nv20->pipe.flush = nv20_flush; + + nv20_init_surface_functions(nv20); + nv20_init_state_functions(nv20); + + nv20->draw = draw_create(); + assert(nv20->draw); + draw_set_rasterize_stage(nv20->draw, nv20_draw_vbuf_stage(nv20)); + + nv20_init_hwctx(nv20); + + return &nv20->pipe; +} + diff --git a/src/gallium/drivers/nv20/nv20_context.h b/src/gallium/drivers/nv20/nv20_context.h new file mode 100644 index 0000000000..8ad926db20 --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_context.h @@ -0,0 +1,153 @@ +#ifndef __NV20_CONTEXT_H__ +#define __NV20_CONTEXT_H__ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_compiler.h" + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "draw/draw_vertex.h" + +#include "nouveau/nouveau_winsys.h" +#include "nouveau/nouveau_gldefs.h" + +#define NOUVEAU_PUSH_CONTEXT(ctx) \ + struct nv20_screen *ctx = nv20->screen +#include "nouveau/nouveau_push.h" + +#include "nv20_state.h" + +#define NOUVEAU_ERR(fmt, args...) \ + fprintf(stderr, "%s:%d - "fmt, __func__, __LINE__, ##args); +#define NOUVEAU_MSG(fmt, args...) \ + fprintf(stderr, "nouveau: "fmt, ##args); + +#define NV20_NEW_VERTPROG (1 << 0) +#define NV20_NEW_FRAGPROG (1 << 1) +#define NV20_NEW_VTXARRAYS (1 << 2) +#define NV20_NEW_BLEND (1 << 3) +#define NV20_NEW_BLENDCOL (1 << 4) +#define NV20_NEW_RAST (1 << 5) +#define NV20_NEW_DSA (1 << 6) +#define NV20_NEW_VIEWPORT (1 << 7) +#define NV20_NEW_SCISSOR (1 << 8) +#define NV20_NEW_FRAMEBUFFER (1 << 9) + +#include "nv20_screen.h" + +struct nv20_context { + struct pipe_context pipe; + + struct nouveau_winsys *nvws; + struct nv20_screen *screen; + unsigned pctx_id; + + struct draw_context *draw; + + uint32_t dirty; + + struct nv20_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS]; + struct nv20_miptree *tex_miptree[PIPE_MAX_SAMPLERS]; + unsigned dirty_samplers; + unsigned fp_samplers; + unsigned vp_samplers; + + uint32_t rt_enable; + struct pipe_buffer *rt[4]; + struct pipe_buffer *zeta; + uint32_t lma_offset; + + struct nv20_blend_state *blend; + struct pipe_blend_color *blend_color; + struct nv20_rasterizer_state *rast; + struct nv20_depth_stencil_alpha_state *dsa; + struct pipe_viewport_state *viewport; + struct pipe_scissor_state *scissor; + struct pipe_framebuffer_state *framebuffer; + + //struct pipe_buffer *constbuf[PIPE_SHADER_TYPES]; + float *constbuf[PIPE_SHADER_TYPES][32][4]; + unsigned constbuf_nr[PIPE_SHADER_TYPES]; + + struct vertex_info vertex_info; + + struct { + struct pipe_buffer *buffer; + uint32_t format; + } tex[2]; + + unsigned vb_enable; + struct { + struct pipe_buffer *buffer; + unsigned delta; + } vb[16]; + +/* struct { + + struct nouveau_resource *exec_heap; + struct nouveau_resource *data_heap; + + struct nv20_vertex_program *active; + + struct nv20_vertex_program *current; + } vertprog; +*/ + struct { + struct nv20_fragment_program *active; + + struct nv20_fragment_program *current; + struct pipe_buffer *constant_buf; + } fragprog; + + struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS]; + struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS]; +}; + +static INLINE struct nv20_context * +nv20_context(struct pipe_context *pipe) +{ + return (struct nv20_context *)pipe; +} + +extern void nv20_init_state_functions(struct nv20_context *nv20); +extern void nv20_init_surface_functions(struct nv20_context *nv20); + +extern void nv20_screen_init_miptree_functions(struct pipe_screen *pscreen); + +/* nv20_clear.c */ +extern void nv20_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue); + +/* nv20_draw.c */ +extern struct draw_stage *nv20_draw_render_stage(struct nv20_context *nv20); + +/* nv20_fragprog.c */ +extern void nv20_fragprog_bind(struct nv20_context *, + struct nv20_fragment_program *); +extern void nv20_fragprog_destroy(struct nv20_context *, + struct nv20_fragment_program *); + +/* nv20_fragtex.c */ +extern void nv20_fragtex_bind(struct nv20_context *); + +/* nv20_prim_vbuf.c */ +struct draw_stage *nv20_draw_vbuf_stage( struct nv20_context *nv20 ); +extern void nv20_vtxbuf_bind(struct nv20_context* nv20); + +/* nv20_state.c and friends */ +extern void nv20_emit_hw_state(struct nv20_context *nv20); +extern void nv20_state_tex_update(struct nv20_context *nv20); + +/* nv20_vbo.c */ +extern boolean nv20_draw_arrays(struct pipe_context *, unsigned mode, + unsigned start, unsigned count); +extern boolean nv20_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count); + + +#endif diff --git a/src/gallium/drivers/nv20/nv20_fragprog.c b/src/gallium/drivers/nv20/nv20_fragprog.c new file mode 100644 index 0000000000..4f496369dd --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_fragprog.c @@ -0,0 +1,21 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" + +#include "nv20_context.h" + +void +nv20_fragprog_bind(struct nv20_context *nv20, struct nv20_fragment_program *fp) +{ +} + +void +nv20_fragprog_destroy(struct nv20_context *nv20, + struct nv20_fragment_program *fp) +{ +} + diff --git a/src/gallium/drivers/nv20/nv20_fragtex.c b/src/gallium/drivers/nv20/nv20_fragtex.c new file mode 100644 index 0000000000..495a7be912 --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_fragtex.c @@ -0,0 +1,124 @@ +#include "nv20_context.h" +#include "nouveau/nouveau_util.h" + +#define _(m,tf) \ +{ \ + TRUE, \ + PIPE_FORMAT_##m, \ + NV20TCL_TX_FORMAT_FORMAT_##tf, \ +} + +struct nv20_texture_format { + boolean defined; + uint pipe; + int format; +}; + +static struct nv20_texture_format +nv20_texture_formats[] = { + _(A8R8G8B8_UNORM, A8R8G8B8), + _(A1R5G5B5_UNORM, A1R5G5B5), + _(A4R4G4B4_UNORM, A4R4G4B4), + _(L8_UNORM , L8 ), + _(A8_UNORM , A8 ), + _(A8L8_UNORM , A8L8 ), +/* _(RGB_DXT1 , DXT1, ), */ +/* _(RGBA_DXT1 , DXT1, ), */ +/* _(RGBA_DXT3 , DXT3, ), */ +/* _(RGBA_DXT5 , DXT5, ), */ + {}, +}; + +static struct nv20_texture_format * +nv20_fragtex_format(uint pipe_format) +{ + struct nv20_texture_format *tf = nv20_texture_formats; + + while (tf->defined) { + if (tf->pipe == pipe_format) + return tf; + tf++; + } + + return NULL; +} + + +static void +nv20_fragtex_build(struct nv20_context *nv20, int unit) +{ +#if 0 + struct nv20_sampler_state *ps = nv20->tex_sampler[unit]; + struct nv20_miptree *nv20mt = nv20->tex_miptree[unit]; + struct pipe_texture *pt = &nv20mt->base; + struct nv20_texture_format *tf; + uint32_t txf, txs, txp; + + tf = nv20_fragtex_format(pt->format); + if (!tf || !tf->defined) { + NOUVEAU_ERR("Unsupported texture format: 0x%x\n", pt->format); + return; + } + + txf = tf->format << 8; + txf |= (pt->last_level + 1) << 16; + txf |= log2i(pt->width[0]) << 20; + txf |= log2i(pt->height[0]) << 24; + txf |= log2i(pt->depth[0]) << 28; + txf |= 8; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE: + txf |= NV10TCL_TX_FORMAT_CUBE_MAP; + /* fall-through */ + case PIPE_TEXTURE_2D: + txf |= (2<<4); + break; + case PIPE_TEXTURE_1D: + txf |= (1<<4); + break; + default: + NOUVEAU_ERR("Unknown target %d\n", pt->target); + return; + } + + BEGIN_RING(kelvin, NV10TCL_TX_OFFSET(unit), 8); + OUT_RELOCl(nv20mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD); + OUT_RELOCd(nv20mt->buffer,txf,NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_OR | NOUVEAU_BO_RD, 1/*VRAM*/,2/*TT*/); + OUT_RING (ps->wrap); + OUT_RING (0x40000000); /* enable */ + OUT_RING (txs); + OUT_RING (ps->filt | 0x2000 /* magic */); + OUT_RING ((pt->width[0] << 16) | pt->height[0]); + OUT_RING (ps->bcol); +#endif +} + +void +nv20_fragtex_bind(struct nv20_context *nv20) +{ +#if 0 + struct nv20_fragment_program *fp = nv20->fragprog.active; + unsigned samplers, unit; + + samplers = nv20->fp_samplers & ~fp->samplers; + while (samplers) { + unit = ffs(samplers) - 1; + samplers &= ~(1 << unit); + + BEGIN_RING(kelvin, NV10TCL_TX_ENABLE(unit), 1); + OUT_RING (0); + } + + samplers = nv20->dirty_samplers & fp->samplers; + while (samplers) { + unit = ffs(samplers) - 1; + samplers &= ~(1 << unit); + + nv20_fragtex_build(nv20, unit); + } + + nv20->fp_samplers = fp->samplers; +#endif +} + diff --git a/src/gallium/drivers/nv20/nv20_miptree.c b/src/gallium/drivers/nv20/nv20_miptree.c new file mode 100644 index 0000000000..c6106d58c4 --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_miptree.c @@ -0,0 +1,156 @@ +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "nv20_context.h" +#include "nv20_screen.h" + +static void +nv20_miptree_layout(struct nv20_miptree *nv20mt) +{ + struct pipe_texture *pt = &nv20mt->base; + boolean swizzled = FALSE; + uint width = pt->width[0], height = pt->height[0]; + uint offset = 0; + int nr_faces, l, f; + + if (pt->target == PIPE_TEXTURE_CUBE) { + nr_faces = 6; + } else { + nr_faces = 1; + } + + for (l = 0; l <= pt->last_level; l++) { + pt->width[l] = width; + pt->height[l] = height; + pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); + pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); + + if (swizzled) + nv20mt->level[l].pitch = pt->nblocksx[l] * pt->block.size; + else + nv20mt->level[l].pitch = pt->nblocksx[0] * pt->block.size; + nv20mt->level[l].pitch = (nv20mt->level[l].pitch + 63) & ~63; + + nv20mt->level[l].image_offset = + CALLOC(nr_faces, sizeof(unsigned)); + + width = MAX2(1, width >> 1); + height = MAX2(1, height >> 1); + + } + + for (f = 0; f < nr_faces; f++) { + for (l = 0; l <= pt->last_level; l++) { + nv20mt->level[l].image_offset[f] = offset; + offset += nv20mt->level[l].pitch * pt->height[l]; + } + } + + nv20mt->total_size = offset; +} + +static struct pipe_texture * +nv20_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) +{ + struct pipe_winsys *ws = screen->winsys; + struct nv20_miptree *mt; + + mt = MALLOC(sizeof(struct nv20_miptree)); + if (!mt) + return NULL; + mt->base = *pt; + mt->base.refcount = 1; + mt->base.screen = screen; + + nv20_miptree_layout(mt); + + mt->buffer = ws->buffer_create(ws, 256, PIPE_BUFFER_USAGE_PIXEL, + mt->total_size); + if (!mt->buffer) { + FREE(mt); + return NULL; + } + + return &mt->base; +} + +static void +nv20_miptree_release(struct pipe_screen *screen, struct pipe_texture **pt) +{ + struct pipe_texture *mt = *pt; + + *pt = NULL; + if (--mt->refcount <= 0) { + struct nv20_miptree *nv20mt = (struct nv20_miptree *)mt; + int l; + + pipe_buffer_reference(screen, &nv20mt->buffer, NULL); + for (l = 0; l <= mt->last_level; l++) { + if (nv20mt->level[l].image_offset) + FREE(nv20mt->level[l].image_offset); + } + FREE(nv20mt); + } +} + +static struct pipe_surface * +nv20_miptree_surface_get(struct pipe_screen *screen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + unsigned flags) +{ + struct nv20_miptree *nv20mt = (struct nv20_miptree *)pt; + struct pipe_surface *ps; + + ps = CALLOC_STRUCT(pipe_surface); + if (!ps) + return NULL; + pipe_texture_reference(&ps->texture, pt); + pipe_buffer_reference(screen, &ps->buffer, nv20mt->buffer); + ps->format = pt->format; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->block = pt->block; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = nv20mt->level[level].pitch; + ps->usage = flags; + ps->status = PIPE_SURFACE_STATUS_DEFINED; + ps->refcount = 1; + ps->winsys = screen->winsys; + + if (pt->target == PIPE_TEXTURE_CUBE) { + ps->offset = nv20mt->level[level].image_offset[face]; + } else + if (pt->target == PIPE_TEXTURE_3D) { + ps->offset = nv20mt->level[level].image_offset[zslice]; + } else { + ps->offset = nv20mt->level[level].image_offset[0]; + } + + return ps; +} + +static void +nv20_miptree_surface_release(struct pipe_screen *pscreen, + struct pipe_surface **psurface) +{ + struct pipe_surface *ps = *psurface; + + *psurface = NULL; + if (--ps->refcount > 0) + return; + + pipe_texture_reference(&ps->texture, NULL); + pipe_buffer_reference(pscreen, &ps->buffer, NULL); + FREE(ps); +} + +void nv20_screen_init_miptree_functions(struct pipe_screen *pscreen) +{ + pscreen->texture_create = nv20_miptree_create; + pscreen->texture_release = nv20_miptree_release; + pscreen->get_tex_surface = nv20_miptree_surface_get; + pscreen->tex_surface_release = nv20_miptree_surface_release; +} + diff --git a/src/gallium/drivers/nv20/nv20_prim_vbuf.c b/src/gallium/drivers/nv20/nv20_prim_vbuf.c new file mode 100644 index 0000000000..74540845a8 --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_prim_vbuf.c @@ -0,0 +1,402 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Build post-transformation, post-clipping vertex buffers and element + * lists by hooking into the end of the primitive pipeline and + * manipulating the vertex_id field in the vertex headers. + * + * XXX: work in progress + * + * \author José Fonseca <jrfonseca@tungstengraphics.com> + * \author Keith Whitwell <keith@tungstengraphics.com> + */ + + +#include "pipe/p_debug.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" + +#include "nv20_context.h" +#include "nv20_state.h" + +#include "draw/draw_vbuf.h" + +/** + * Primitive renderer for nv20. + */ +struct nv20_vbuf_render { + struct vbuf_render base; + + struct nv20_context *nv20; + + /** Vertex buffer in VRAM */ + struct pipe_buffer *pbuffer; + + /** Vertex buffer in normal memory */ + void *mbuffer; + + /** Vertex size in bytes */ + /*unsigned vertex_size;*/ + + /** Hardware primitive */ + unsigned hwprim; +}; + +/** + * Basically a cast wrapper. + */ +static INLINE struct nv20_vbuf_render * +nv20_vbuf_render(struct vbuf_render *render) +{ + assert(render); + return (struct nv20_vbuf_render *)render; +} + +void nv20_vtxbuf_bind( struct nv20_context* nv20 ) +{ +#if 0 + int i; + for(i = 0; i < NV20TCL_VTXBUF_ADDRESS__SIZE; i++) { + BEGIN_RING(kelvin, NV20TCL_VTXBUF_ADDRESS(i), 1); + OUT_RING(0/*nv20->vtxbuf*/); + BEGIN_RING(kelvin, NV20TCL_VTXFMT(i) ,1); + OUT_RING(0/*XXX*/); + } +#endif +} + +static const struct vertex_info * +nv20_vbuf_render_get_vertex_info( struct vbuf_render *render ) +{ + struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); + struct nv20_context *nv20 = nv20_render->nv20; + + nv20_emit_hw_state(nv20); + + return &nv20->vertex_info; +} + +static void * +nv20__allocate_mbuffer(struct nv20_vbuf_render *nv20_render, size_t size) +{ + nv20_render->mbuffer = MALLOC(size); + return nv20_render->mbuffer; +} + +static void * +nv20__allocate_pbuffer(struct nv20_vbuf_render *nv20_render, size_t size) +{ + struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys; + nv20_render->pbuffer = winsys->buffer_create(winsys, 64, + PIPE_BUFFER_USAGE_VERTEX, size); + return winsys->buffer_map(winsys, + nv20_render->pbuffer, + PIPE_BUFFER_USAGE_CPU_WRITE); +} + +static void * +nv20_vbuf_render_allocate_vertices( struct vbuf_render *render, + ushort vertex_size, + ushort nr_vertices ) +{ + struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); + size_t size = (size_t)vertex_size * (size_t)nr_vertices; + void *buf; + + assert(!nv20_render->pbuffer); + assert(!nv20_render->mbuffer); + + /* + * For small amount of vertices, don't bother with pipe vertex + * buffer, the data will be passed directly via the fifo. + */ + /* XXX: Pipe vertex buffers don't work. */ + if (0 && size > 16 * 1024) + buf = nv20__allocate_pbuffer(nv20_render, size); + else + buf = nv20__allocate_mbuffer(nv20_render, size); + + if (buf) + nv20_render->nv20->dirty |= NV20_NEW_VTXARRAYS; + + return buf; +} + +static boolean +nv20_vbuf_render_set_primitive( struct vbuf_render *render, + unsigned prim ) +{ + struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); + unsigned hwp = nvgl_primitive(prim); + if (hwp == 0) + return FALSE; + + nv20_render->hwprim = hwp; + return TRUE; +} + +static uint32_t +nv20__vtxhwformat(unsigned stride, unsigned fields, unsigned type) +{ + return (stride << NV20TCL_VTXFMT_STRIDE_SHIFT) | + (fields << NV20TCL_VTXFMT_SIZE_SHIFT) | + (type << NV20TCL_VTXFMT_TYPE_SHIFT); +} + +static unsigned +nv20__emit_format(struct nv20_context *nv20, enum attrib_emit type, int hwattr) +{ + uint32_t hwfmt = 0; + unsigned fields; + + switch (type) { + case EMIT_OMIT: + hwfmt = nv20__vtxhwformat(0, 0, 2); + fields = 0; + break; + case EMIT_1F: + hwfmt = nv20__vtxhwformat(4, 1, 2); + fields = 1; + break; + case EMIT_2F: + hwfmt = nv20__vtxhwformat(8, 2, 2); + fields = 2; + break; + case EMIT_3F: + hwfmt = nv20__vtxhwformat(12, 3, 2); + fields = 3; + break; + case EMIT_4F: + hwfmt = nv20__vtxhwformat(16, 4, 2); + fields = 4; + break; + default: + NOUVEAU_ERR("unhandled attrib_emit %d\n", type); + return 0; + } + + BEGIN_RING(kelvin, NV20TCL_VTXFMT(hwattr), 1); + OUT_RING(hwfmt); + return fields; +} + +static unsigned +nv20__emit_vertex_array_format(struct nv20_context *nv20) +{ + struct vertex_info *vinfo = &nv20->vertex_info; + int hwattr = NV20TCL_VTXFMT__SIZE; + int attr = 0; + unsigned nr_fields = 0; + + while (hwattr-- > 0) { + if (vinfo->hwfmt[0] & (1 << hwattr)) { + nr_fields += nv20__emit_format(nv20, + vinfo->attrib[attr].emit, hwattr); + attr++; + } else + nv20__emit_format(nv20, EMIT_OMIT, hwattr); + } + + return nr_fields; +} + +static void +nv20__draw_mbuffer(struct nv20_vbuf_render *nv20_render, + const ushort *indices, + uint nr_indices) +{ + struct nv20_context *nv20 = nv20_render->nv20; + struct vertex_info *vinfo = &nv20->vertex_info; + unsigned nr_fields; + int max_push; + ubyte *data = nv20_render->mbuffer; + int vsz = 4 * vinfo->size; + + nr_fields = nv20__emit_vertex_array_format(nv20); + + BEGIN_RING(kelvin, NV20TCL_VERTEX_BEGIN_END, 1); + OUT_RING(nv20_render->hwprim); + + max_push = 1200 / nr_fields; + while (nr_indices) { + int i; + int push = MIN2(nr_indices, max_push); + + BEGIN_RING_NI(kelvin, NV20TCL_VERTEX_DATA, push * nr_fields); + for (i = 0; i < push; i++) { + /* XXX: fixme to handle other than floats? */ + int f = nr_fields; + float *attrv = (float*)&data[indices[i] * vsz]; + while (f-- > 0) + OUT_RINGf(*attrv++); + } + + nr_indices -= push; + indices += push; + } + + BEGIN_RING(kelvin, NV20TCL_VERTEX_BEGIN_END, 1); + OUT_RING(NV20TCL_VERTEX_BEGIN_END_STOP); +} + +static void +nv20__draw_pbuffer(struct nv20_vbuf_render *nv20_render, + const ushort *indices, + uint nr_indices) +{ + struct nv20_context *nv20 = nv20_render->nv20; + int push, i; + + NOUVEAU_ERR("nv20__draw_pbuffer: this path is broken.\n"); + + BEGIN_RING(kelvin, NV10TCL_VERTEX_ARRAY_OFFSET_POS, 1); + OUT_RELOCl(nv20_render->pbuffer, 0, + NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD); + + BEGIN_RING(kelvin, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1); + OUT_RING(nv20_render->hwprim); + + if (nr_indices & 1) { + BEGIN_RING(kelvin, NV10TCL_VB_ELEMENT_U32, 1); + OUT_RING (indices[0]); + indices++; nr_indices--; + } + + while (nr_indices) { + // XXX too big/small ? check the size + push = MIN2(nr_indices, 1200 * 2); + + BEGIN_RING_NI(kelvin, NV10TCL_VB_ELEMENT_U16, push >> 1); + for (i = 0; i < push; i+=2) + OUT_RING((indices[i+1] << 16) | indices[i]); + + nr_indices -= push; + indices += push; + } + + BEGIN_RING(kelvin, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1); + OUT_RING (0); +} + +static void +nv20_vbuf_render_draw( struct vbuf_render *render, + const ushort *indices, + uint nr_indices) +{ + struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); + + nv20_emit_hw_state(nv20_render->nv20); + + if (nv20_render->pbuffer) + nv20__draw_pbuffer(nv20_render, indices, nr_indices); + else if (nv20_render->mbuffer) + nv20__draw_mbuffer(nv20_render, indices, nr_indices); + else + assert(0); +} + + +static void +nv20_vbuf_render_release_vertices( struct vbuf_render *render, + void *vertices, + unsigned vertex_size, + unsigned vertices_used ) +{ + struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); + struct nv20_context *nv20 = nv20_render->nv20; + struct pipe_winsys *winsys = nv20->pipe.winsys; + struct pipe_screen *pscreen = &nv20->screen->pipe; + + if (nv20_render->pbuffer) { + winsys->buffer_unmap(winsys, nv20_render->pbuffer); + pipe_buffer_reference(pscreen, &nv20_render->pbuffer, NULL); + } else if (nv20_render->mbuffer) { + FREE(nv20_render->mbuffer); + nv20_render->mbuffer = NULL; + } else + assert(0); +} + + +static void +nv20_vbuf_render_destroy( struct vbuf_render *render ) +{ + struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); + + assert(!nv20_render->pbuffer); + assert(!nv20_render->mbuffer); + + FREE(nv20_render); +} + + +/** + * Create a new primitive render. + */ +static struct vbuf_render * +nv20_vbuf_render_create( struct nv20_context *nv20 ) +{ + struct nv20_vbuf_render *nv20_render = CALLOC_STRUCT(nv20_vbuf_render); + + nv20_render->nv20 = nv20; + + nv20_render->base.max_vertex_buffer_bytes = 16*1024; + nv20_render->base.max_indices = 1024; + nv20_render->base.get_vertex_info = nv20_vbuf_render_get_vertex_info; + nv20_render->base.allocate_vertices = + nv20_vbuf_render_allocate_vertices; + nv20_render->base.set_primitive = nv20_vbuf_render_set_primitive; + nv20_render->base.draw = nv20_vbuf_render_draw; + nv20_render->base.release_vertices = nv20_vbuf_render_release_vertices; + nv20_render->base.destroy = nv20_vbuf_render_destroy; + + return &nv20_render->base; +} + + +/** + * Create a new primitive vbuf/render stage. + */ +struct draw_stage *nv20_draw_vbuf_stage( struct nv20_context *nv20 ) +{ + struct vbuf_render *render; + struct draw_stage *stage; + + render = nv20_vbuf_render_create(nv20); + if(!render) + return NULL; + + stage = draw_vbuf_stage( nv20->draw, render ); + if(!stage) { + render->destroy(render); + return NULL; + } + + return stage; +} diff --git a/src/gallium/drivers/nv20/nv20_screen.c b/src/gallium/drivers/nv20/nv20_screen.c new file mode 100644 index 0000000000..2ca6e6b149 --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_screen.c @@ -0,0 +1,206 @@ +#include "pipe/p_screen.h" + +#include "nv20_context.h" +#include "nv20_screen.h" + +static const char * +nv20_screen_get_name(struct pipe_screen *screen) +{ + struct nv20_screen *nv20screen = nv20_screen(screen); + struct nouveau_device *dev = nv20screen->nvws->channel->device; + static char buffer[128]; + + snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); + return buffer; +} + +static const char * +nv20_screen_get_vendor(struct pipe_screen *screen) +{ + return "nouveau"; +} + +static int +nv20_screen_get_param(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 2; + case PIPE_CAP_NPOT_TEXTURES: + return 0; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 0; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 1; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 0; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 12; + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 0; + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 12; + case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: + return 0; + case NOUVEAU_CAP_HW_VTXBUF: + case NOUVEAU_CAP_HW_IDXBUF: + return 0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0; + } +} + +static float +nv20_screen_get_paramf(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 10.0; + case PIPE_CAP_MAX_POINT_WIDTH: + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 64.0; + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 2.0; + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 4.0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0.0; + } +} + +static boolean +nv20_screen_is_format_supported(struct pipe_screen *screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, unsigned geom_flags) +{ + if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z16_UNORM: + return TRUE; + default: + break; + } + } else { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_A1R5G5B5_UNORM: + case PIPE_FORMAT_A4R4G4B4_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_A8_UNORM: + case PIPE_FORMAT_I8_UNORM: + return TRUE; + default: + break; + } + } + + return FALSE; +} + +static void * +nv20_surface_map(struct pipe_screen *screen, struct pipe_surface *surface, + unsigned flags ) +{ + struct pipe_winsys *ws = screen->winsys; + void *map; + + map = ws->buffer_map(ws, surface->buffer, flags); + if (!map) + return NULL; + + return map + surface->offset; +} + +static void +nv20_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) +{ + struct pipe_winsys *ws = screen->winsys; + + ws->buffer_unmap(ws, surface->buffer); +} + +static void +nv20_screen_destroy(struct pipe_screen *pscreen) +{ + struct nv20_screen *screen = nv20_screen(pscreen); + struct nouveau_winsys *nvws = screen->nvws; + + nvws->notifier_free(&screen->sync); + nvws->grobj_free(&screen->kelvin); + + FREE(pscreen); +} + +struct pipe_screen * +nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +{ + struct nv20_screen *screen = CALLOC_STRUCT(nv20_screen); + unsigned kelvin_class = 0; + unsigned chipset = nvws->channel->device->chipset; + int ret; + + if (!screen) + return NULL; + screen->nvws = nvws; + + /* 3D object */ + if (chipset >= 0x25) + kelvin_class = NV25TCL; + else if (chipset >= 0x20) + kelvin_class = NV20TCL; + + if (!kelvin_class || chipset >= 0x30) { + NOUVEAU_ERR("Unknown nv2x chipset: nv%02x\n", chipset); + return NULL; + } + + ret = nvws->grobj_alloc(nvws, kelvin_class, &screen->kelvin); + if (ret) { + NOUVEAU_ERR("Error creating 3D object: %d\n", ret); + return FALSE; + } + + /* Notifier for sync purposes */ + ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + if (ret) { + NOUVEAU_ERR("Error creating notifier object: %d\n", ret); + nv20_screen_destroy(&screen->pipe); + return NULL; + } + + screen->pipe.winsys = ws; + screen->pipe.destroy = nv20_screen_destroy; + + screen->pipe.get_name = nv20_screen_get_name; + screen->pipe.get_vendor = nv20_screen_get_vendor; + screen->pipe.get_param = nv20_screen_get_param; + screen->pipe.get_paramf = nv20_screen_get_paramf; + + screen->pipe.is_format_supported = nv20_screen_is_format_supported; + + screen->pipe.surface_map = nv20_surface_map; + screen->pipe.surface_unmap = nv20_surface_unmap; + + nv20_screen_init_miptree_functions(&screen->pipe); + + return &screen->pipe; +} + diff --git a/src/gallium/drivers/nv20/nv20_screen.h b/src/gallium/drivers/nv20/nv20_screen.h new file mode 100644 index 0000000000..8f2f2e341d --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_screen.h @@ -0,0 +1,22 @@ +#ifndef __NV20_SCREEN_H__ +#define __NV20_SCREEN_H__ + +#include "pipe/p_screen.h" + +struct nv20_screen { + struct pipe_screen pipe; + + struct nouveau_winsys *nvws; + + /* HW graphics objects */ + struct nouveau_grobj *kelvin; + struct nouveau_notifier *sync; +}; + +static INLINE struct nv20_screen * +nv20_screen(struct pipe_screen *screen) +{ + return (struct nv20_screen *)screen; +} + +#endif diff --git a/src/gallium/drivers/nv20/nv20_state.c b/src/gallium/drivers/nv20/nv20_state.c new file mode 100644 index 0000000000..21bde5b81f --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_state.c @@ -0,0 +1,581 @@ +#include "draw/draw_context.h" +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/tgsi_parse.h" + +#include "nv20_context.h" +#include "nv20_state.h" + +static void * +nv20_blend_state_create(struct pipe_context *pipe, + const struct pipe_blend_state *cso) +{ + struct nv20_blend_state *cb; + + cb = MALLOC(sizeof(struct nv20_blend_state)); + + cb->b_enable = cso->blend_enable ? 1 : 0; + cb->b_srcfunc = ((nvgl_blend_func(cso->alpha_src_factor)<<16) | + (nvgl_blend_func(cso->rgb_src_factor))); + cb->b_dstfunc = ((nvgl_blend_func(cso->alpha_dst_factor)<<16) | + (nvgl_blend_func(cso->rgb_dst_factor))); + + cb->c_mask = (((cso->colormask & PIPE_MASK_A) ? (0x01<<24) : 0) | + ((cso->colormask & PIPE_MASK_R) ? (0x01<<16) : 0) | + ((cso->colormask & PIPE_MASK_G) ? (0x01<< 8) : 0) | + ((cso->colormask & PIPE_MASK_B) ? (0x01<< 0) : 0)); + + cb->d_enable = cso->dither ? 1 : 0; + + return (void *)cb; +} + +static void +nv20_blend_state_bind(struct pipe_context *pipe, void *blend) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + nv20->blend = (struct nv20_blend_state*)blend; + + nv20->dirty |= NV20_NEW_BLEND; +} + +static void +nv20_blend_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + + +static INLINE unsigned +wrap_mode(unsigned wrap) { + unsigned ret; + + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + ret = NV20TCL_TX_WRAP_S_REPEAT; + break; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + ret = NV20TCL_TX_WRAP_S_MIRRORED_REPEAT; + break; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + ret = NV20TCL_TX_WRAP_S_CLAMP_TO_EDGE; + break; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + ret = NV20TCL_TX_WRAP_S_CLAMP_TO_BORDER; + break; + case PIPE_TEX_WRAP_CLAMP: + ret = NV20TCL_TX_WRAP_S_CLAMP; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + case PIPE_TEX_WRAP_MIRROR_CLAMP: + default: + NOUVEAU_ERR("unknown wrap mode: %d\n", wrap); + ret = NV20TCL_TX_WRAP_S_REPEAT; + break; + } + + return (ret >> NV20TCL_TX_WRAP_S_SHIFT); +} + +static void * +nv20_sampler_state_create(struct pipe_context *pipe, + const struct pipe_sampler_state *cso) +{ + struct nv20_sampler_state *ps; + uint32_t filter = 0; + + ps = MALLOC(sizeof(struct nv20_sampler_state)); + + ps->wrap = ((wrap_mode(cso->wrap_s) << NV20TCL_TX_WRAP_S_SHIFT) | + (wrap_mode(cso->wrap_t) << NV20TCL_TX_WRAP_T_SHIFT)); + + ps->en = 0; + if (cso->max_anisotropy > 1.0) { + /* no idea, binary driver sets it, works without it.. meh.. */ + ps->wrap |= (1 << 5); + +/* if (cso->max_anisotropy >= 8.0) { + ps->en |= NV20TCL_TX_ENABLE_ANISO_8X; + } else + if (cso->max_anisotropy >= 4.0) { + ps->en |= NV20TCL_TX_ENABLE_ANISO_4X; + } else { + ps->en |= NV20TCL_TX_ENABLE_ANISO_2X; + }*/ + } + + switch (cso->mag_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + filter |= NV20TCL_TX_FILTER_MAGNIFY_LINEAR; + break; + case PIPE_TEX_FILTER_NEAREST: + default: + filter |= NV20TCL_TX_FILTER_MAGNIFY_NEAREST; + break; + } + + switch (cso->min_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + filter |= + NV20TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + filter |= NV20TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + filter |= NV20TCL_TX_FILTER_MINIFY_LINEAR; + break; + } + break; + case PIPE_TEX_FILTER_NEAREST: + default: + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + filter |= + NV20TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + filter |= + NV20TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + filter |= NV20TCL_TX_FILTER_MINIFY_NEAREST; + break; + } + break; + } + + ps->filt = filter; + +/* if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + switch (cso->compare_func) { + case PIPE_FUNC_NEVER: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_NEVER; + break; + case PIPE_FUNC_GREATER: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_GREATER; + break; + case PIPE_FUNC_EQUAL: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_EQUAL; + break; + case PIPE_FUNC_GEQUAL: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_GEQUAL; + break; + case PIPE_FUNC_LESS: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_LESS; + break; + case PIPE_FUNC_NOTEQUAL: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_NOTEQUAL; + break; + case PIPE_FUNC_LEQUAL: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_LEQUAL; + break; + case PIPE_FUNC_ALWAYS: + ps->wrap |= NV10TCL_TX_WRAP_RCOMP_ALWAYS; + break; + default: + break; + } + }*/ + + ps->bcol = ((float_to_ubyte(cso->border_color[3]) << 24) | + (float_to_ubyte(cso->border_color[0]) << 16) | + (float_to_ubyte(cso->border_color[1]) << 8) | + (float_to_ubyte(cso->border_color[2]) << 0)); + + return (void *)ps; +} + +static void +nv20_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler) +{ + struct nv20_context *nv20 = nv20_context(pipe); + unsigned unit; + + for (unit = 0; unit < nr; unit++) { + nv20->tex_sampler[unit] = sampler[unit]; + nv20->dirty_samplers |= (1 << unit); + } +} + +static void +nv20_sampler_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + +static void +nv20_set_sampler_texture(struct pipe_context *pipe, unsigned nr, + struct pipe_texture **miptree) +{ + struct nv20_context *nv20 = nv20_context(pipe); + unsigned unit; + + for (unit = 0; unit < nr; unit++) { + nv20->tex_miptree[unit] = (struct nv20_miptree *)miptree[unit]; + nv20->dirty_samplers |= (1 << unit); + } +} + +static void * +nv20_rasterizer_state_create(struct pipe_context *pipe, + const struct pipe_rasterizer_state *cso) +{ + struct nv20_rasterizer_state *rs; + int i; + + /*XXX: ignored: + * light_twoside + * offset_cw/ccw -nohw + * scissor + * point_smooth -nohw + * multisample + * offset_units / offset_scale + */ + rs = MALLOC(sizeof(struct nv20_rasterizer_state)); + + rs->templ = cso; + + rs->shade_model = cso->flatshade ? NV20TCL_SHADE_MODEL_FLAT : + NV20TCL_SHADE_MODEL_SMOOTH; + + rs->line_width = (unsigned char)(cso->line_width * 8.0) & 0xff; + rs->line_smooth_en = cso->line_smooth ? 1 : 0; + + /* XXX: nv20 and nv25 different! */ + rs->point_size = *(uint32_t*)&cso->point_size; + + rs->poly_smooth_en = cso->poly_smooth ? 1 : 0; + + if (cso->front_winding == PIPE_WINDING_CCW) { + rs->front_face = NV20TCL_FRONT_FACE_CCW; + rs->poly_mode_front = nvgl_polygon_mode(cso->fill_ccw); + rs->poly_mode_back = nvgl_polygon_mode(cso->fill_cw); + } else { + rs->front_face = NV20TCL_FRONT_FACE_CW; + rs->poly_mode_front = nvgl_polygon_mode(cso->fill_cw); + rs->poly_mode_back = nvgl_polygon_mode(cso->fill_ccw); + } + + switch (cso->cull_mode) { + case PIPE_WINDING_CCW: + rs->cull_face_en = 1; + if (cso->front_winding == PIPE_WINDING_CCW) + rs->cull_face = NV20TCL_CULL_FACE_FRONT; + else + rs->cull_face = NV20TCL_CULL_FACE_BACK; + break; + case PIPE_WINDING_CW: + rs->cull_face_en = 1; + if (cso->front_winding == PIPE_WINDING_CW) + rs->cull_face = NV20TCL_CULL_FACE_FRONT; + else + rs->cull_face = NV20TCL_CULL_FACE_BACK; + break; + case PIPE_WINDING_BOTH: + rs->cull_face_en = 1; + rs->cull_face = NV20TCL_CULL_FACE_FRONT_AND_BACK; + break; + case PIPE_WINDING_NONE: + default: + rs->cull_face_en = 0; + rs->cull_face = 0; + break; + } + + if (cso->point_sprite) { + rs->point_sprite = (1 << 0); + for (i = 0; i < 8; i++) { + if (cso->sprite_coord_mode[i] != PIPE_SPRITE_COORD_NONE) + rs->point_sprite |= (1 << (8 + i)); + } + } else { + rs->point_sprite = 0; + } + + return (void *)rs; +} + +static void +nv20_rasterizer_state_bind(struct pipe_context *pipe, void *rast) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + nv20->rast = (struct nv20_rasterizer_state*)rast; + + draw_set_rasterizer_state(nv20->draw, (nv20->rast ? nv20->rast->templ : NULL)); + + nv20->dirty |= NV20_NEW_RAST; +} + +static void +nv20_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + +static void * +nv20_depth_stencil_alpha_state_create(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *cso) +{ + struct nv20_depth_stencil_alpha_state *hw; + + hw = MALLOC(sizeof(struct nv20_depth_stencil_alpha_state)); + + hw->depth.func = nvgl_comparison_op(cso->depth.func); + hw->depth.write_enable = cso->depth.writemask ? 1 : 0; + hw->depth.test_enable = cso->depth.enabled ? 1 : 0; + + hw->stencil.enable = cso->stencil[0].enabled ? 1 : 0; + hw->stencil.wmask = cso->stencil[0].write_mask; + hw->stencil.func = nvgl_comparison_op(cso->stencil[0].func); + hw->stencil.ref = cso->stencil[0].ref_value; + hw->stencil.vmask = cso->stencil[0].value_mask; + hw->stencil.fail = nvgl_stencil_op(cso->stencil[0].fail_op); + hw->stencil.zfail = nvgl_stencil_op(cso->stencil[0].zfail_op); + hw->stencil.zpass = nvgl_stencil_op(cso->stencil[0].zpass_op); + + hw->alpha.enabled = cso->alpha.enabled ? 1 : 0; + hw->alpha.func = nvgl_comparison_op(cso->alpha.func); + hw->alpha.ref = float_to_ubyte(cso->alpha.ref); + + return (void *)hw; +} + +static void +nv20_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *dsa) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + nv20->dsa = (struct nv20_depth_stencil_alpha_state*)dsa; + + nv20->dirty |= NV20_NEW_DSA; +} + +static void +nv20_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + +static void * +nv20_vp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + return draw_create_vertex_shader(nv20->draw, templ); +} + +static void +nv20_vp_state_bind(struct pipe_context *pipe, void *shader) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + draw_bind_vertex_shader(nv20->draw, (struct draw_vertex_shader *) shader); + + nv20->dirty |= NV20_NEW_VERTPROG; +} + +static void +nv20_vp_state_delete(struct pipe_context *pipe, void *shader) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + draw_delete_vertex_shader(nv20->draw, (struct draw_vertex_shader *) shader); +} + +static void * +nv20_fp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *cso) +{ + struct nv20_fragment_program *fp; + + fp = CALLOC(1, sizeof(struct nv20_fragment_program)); + fp->pipe.tokens = tgsi_dup_tokens(cso->tokens); + + tgsi_scan_shader(cso->tokens, &fp->info); + + return (void *)fp; +} + +static void +nv20_fp_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv20_context *nv20 = nv20_context(pipe); + struct nv20_fragment_program *fp = hwcso; + + nv20->fragprog.current = fp; + nv20->dirty |= NV20_NEW_FRAGPROG; +} + +static void +nv20_fp_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv20_context *nv20 = nv20_context(pipe); + struct nv20_fragment_program *fp = hwcso; + + nv20_fragprog_destroy(nv20, fp); + FREE((void*)fp->pipe.tokens); + FREE(fp); +} + +static void +nv20_set_blend_color(struct pipe_context *pipe, + const struct pipe_blend_color *bcol) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + nv20->blend_color = (struct pipe_blend_color*)bcol; + + nv20->dirty |= NV20_NEW_BLENDCOL; +} + +static void +nv20_set_clip_state(struct pipe_context *pipe, + const struct pipe_clip_state *clip) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + draw_set_clip_state(nv20->draw, clip); +} + +static void +nv20_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, + const struct pipe_constant_buffer *buf ) +{ + struct nv20_context *nv20 = nv20_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + if (buf) { + void *mapped; + if (buf->size && (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ))) + { + memcpy(nv20->constbuf[shader], mapped, buf->size); + nv20->constbuf_nr[shader] = + buf->size / (4 * sizeof(float)); + ws->buffer_unmap(ws, buf->buffer); + } + } +} + +static void +nv20_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + nv20->framebuffer = (struct pipe_framebuffer_state*)fb; + + nv20->dirty |= NV20_NEW_FRAMEBUFFER; +} + +static void +nv20_set_polygon_stipple(struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple) +{ + NOUVEAU_ERR("line stipple hahaha\n"); +} + +static void +nv20_set_scissor_state(struct pipe_context *pipe, + const struct pipe_scissor_state *s) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + nv20->scissor = (struct pipe_scissor_state*)s; + + nv20->dirty |= NV20_NEW_SCISSOR; +} + +static void +nv20_set_viewport_state(struct pipe_context *pipe, + const struct pipe_viewport_state *vpt) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + nv20->viewport = (struct pipe_viewport_state*)vpt; + + draw_set_viewport_state(nv20->draw, nv20->viewport); + + nv20->dirty |= NV20_NEW_VIEWPORT; +} + +static void +nv20_set_vertex_buffers(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_buffer *vb) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + memcpy(nv20->vtxbuf, vb, sizeof(*vb) * count); + nv20->dirty |= NV20_NEW_VTXARRAYS; + + draw_set_vertex_buffers(nv20->draw, count, vb); +} + +static void +nv20_set_vertex_elements(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_element *ve) +{ + struct nv20_context *nv20 = nv20_context(pipe); + + memcpy(nv20->vtxelt, ve, sizeof(*ve) * count); + nv20->dirty |= NV20_NEW_VTXARRAYS; + + draw_set_vertex_elements(nv20->draw, count, ve); +} + +void +nv20_init_state_functions(struct nv20_context *nv20) +{ + nv20->pipe.create_blend_state = nv20_blend_state_create; + nv20->pipe.bind_blend_state = nv20_blend_state_bind; + nv20->pipe.delete_blend_state = nv20_blend_state_delete; + + nv20->pipe.create_sampler_state = nv20_sampler_state_create; + nv20->pipe.bind_sampler_states = nv20_sampler_state_bind; + nv20->pipe.delete_sampler_state = nv20_sampler_state_delete; + nv20->pipe.set_sampler_textures = nv20_set_sampler_texture; + + nv20->pipe.create_rasterizer_state = nv20_rasterizer_state_create; + nv20->pipe.bind_rasterizer_state = nv20_rasterizer_state_bind; + nv20->pipe.delete_rasterizer_state = nv20_rasterizer_state_delete; + + nv20->pipe.create_depth_stencil_alpha_state = + nv20_depth_stencil_alpha_state_create; + nv20->pipe.bind_depth_stencil_alpha_state = + nv20_depth_stencil_alpha_state_bind; + nv20->pipe.delete_depth_stencil_alpha_state = + nv20_depth_stencil_alpha_state_delete; + + nv20->pipe.create_vs_state = nv20_vp_state_create; + nv20->pipe.bind_vs_state = nv20_vp_state_bind; + nv20->pipe.delete_vs_state = nv20_vp_state_delete; + + nv20->pipe.create_fs_state = nv20_fp_state_create; + nv20->pipe.bind_fs_state = nv20_fp_state_bind; + nv20->pipe.delete_fs_state = nv20_fp_state_delete; + + nv20->pipe.set_blend_color = nv20_set_blend_color; + nv20->pipe.set_clip_state = nv20_set_clip_state; + nv20->pipe.set_constant_buffer = nv20_set_constant_buffer; + nv20->pipe.set_framebuffer_state = nv20_set_framebuffer_state; + nv20->pipe.set_polygon_stipple = nv20_set_polygon_stipple; + nv20->pipe.set_scissor_state = nv20_set_scissor_state; + nv20->pipe.set_viewport_state = nv20_set_viewport_state; + + nv20->pipe.set_vertex_buffers = nv20_set_vertex_buffers; + nv20->pipe.set_vertex_elements = nv20_set_vertex_elements; +} + diff --git a/src/gallium/drivers/nv20/nv20_state.h b/src/gallium/drivers/nv20/nv20_state.h new file mode 100644 index 0000000000..34f402fdcb --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_state.h @@ -0,0 +1,139 @@ +#ifndef __NV20_STATE_H__ +#define __NV20_STATE_H__ + +#include "pipe/p_state.h" +#include "tgsi/tgsi_scan.h" + +struct nv20_blend_state { + uint32_t b_enable; + uint32_t b_srcfunc; + uint32_t b_dstfunc; + + uint32_t c_mask; + + uint32_t d_enable; +}; + +struct nv20_sampler_state { + uint32_t wrap; + uint32_t en; + uint32_t filt; + uint32_t bcol; +}; + +struct nv20_rasterizer_state { + uint32_t shade_model; + + uint32_t line_width; + uint32_t line_smooth_en; + + uint32_t point_size; + + uint32_t poly_smooth_en; + + uint32_t poly_mode_front; + uint32_t poly_mode_back; + + uint32_t front_face; + uint32_t cull_face; + uint32_t cull_face_en; + + uint32_t point_sprite; + + const struct pipe_rasterizer_state *templ; +}; + +struct nv20_vertex_program_exec { + uint32_t data[4]; + boolean has_branch_offset; + int const_index; +}; + +struct nv20_vertex_program_data { + int index; /* immediates == -1 */ + float value[4]; +}; + +struct nv20_vertex_program { + const struct pipe_shader_state *pipe; + + boolean translated; + struct nv20_vertex_program_exec *insns; + unsigned nr_insns; + struct nv20_vertex_program_data *consts; + unsigned nr_consts; + + struct nouveau_resource *exec; + unsigned exec_start; + struct nouveau_resource *data; + unsigned data_start; + unsigned data_start_min; + + uint32_t ir; + uint32_t or; +}; + +struct nv20_fragment_program_data { + unsigned offset; + unsigned index; +}; + +struct nv20_fragment_program { + struct pipe_shader_state pipe; + struct tgsi_shader_info info; + + boolean translated; + boolean on_hw; + unsigned samplers; + + uint32_t *insn; + int insn_len; + + struct nv20_fragment_program_data *consts; + unsigned nr_consts; + + struct pipe_buffer *buffer; + + uint32_t fp_control; + uint32_t fp_reg_control; +}; + + +struct nv20_depth_stencil_alpha_state { + struct { + uint32_t func; + uint32_t write_enable; + uint32_t test_enable; + } depth; + + struct { + uint32_t enable; + uint32_t wmask; + uint32_t func; + uint32_t ref; + uint32_t vmask; + uint32_t fail; + uint32_t zfail; + uint32_t zpass; + } stencil; + + struct { + uint32_t enabled; + uint32_t func; + uint32_t ref; + } alpha; +}; + +struct nv20_miptree { + struct pipe_texture base; + + struct pipe_buffer *buffer; + uint total_size; + + struct { + uint pitch; + uint *image_offset; + } level[PIPE_MAX_TEXTURE_LEVELS]; +}; + +#endif diff --git a/src/gallium/drivers/nv20/nv20_state_emit.c b/src/gallium/drivers/nv20/nv20_state_emit.c new file mode 100644 index 0000000000..5265bf3a31 --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_state_emit.c @@ -0,0 +1,359 @@ +#include "nv20_context.h" +#include "nv20_state.h" +#include "draw/draw_context.h" + +static void nv20_state_emit_blend(struct nv20_context* nv20) +{ + struct nv20_blend_state *b = nv20->blend; + + BEGIN_RING(kelvin, NV20TCL_DITHER_ENABLE, 1); + OUT_RING (b->d_enable); + + BEGIN_RING(kelvin, NV20TCL_BLEND_FUNC_ENABLE, 1); + OUT_RING (b->b_enable); + + BEGIN_RING(kelvin, NV20TCL_BLEND_FUNC_SRC, 2); + OUT_RING (b->b_srcfunc); + OUT_RING (b->b_dstfunc); + + BEGIN_RING(kelvin, NV20TCL_COLOR_MASK, 1); + OUT_RING (b->c_mask); +} + +static void nv20_state_emit_blend_color(struct nv20_context* nv20) +{ + struct pipe_blend_color *c = nv20->blend_color; + + BEGIN_RING(kelvin, NV20TCL_BLEND_COLOR, 1); + OUT_RING ((float_to_ubyte(c->color[3]) << 24)| + (float_to_ubyte(c->color[0]) << 16)| + (float_to_ubyte(c->color[1]) << 8) | + (float_to_ubyte(c->color[2]) << 0)); +} + +static void nv20_state_emit_rast(struct nv20_context* nv20) +{ + struct nv20_rasterizer_state *r = nv20->rast; + + BEGIN_RING(kelvin, NV20TCL_SHADE_MODEL, 2); + OUT_RING (r->shade_model); + OUT_RING (r->line_width); + + + BEGIN_RING(kelvin, NV20TCL_POINT_SIZE, 1); + OUT_RING (r->point_size); + + BEGIN_RING(kelvin, NV20TCL_POLYGON_MODE_FRONT, 2); + OUT_RING (r->poly_mode_front); + OUT_RING (r->poly_mode_back); + + + BEGIN_RING(kelvin, NV20TCL_CULL_FACE, 2); + OUT_RING (r->cull_face); + OUT_RING (r->front_face); + + BEGIN_RING(kelvin, NV20TCL_LINE_SMOOTH_ENABLE, 2); + OUT_RING (r->line_smooth_en); + OUT_RING (r->poly_smooth_en); + + BEGIN_RING(kelvin, NV20TCL_CULL_FACE_ENABLE, 1); + OUT_RING (r->cull_face_en); +} + +static void nv20_state_emit_dsa(struct nv20_context* nv20) +{ + struct nv20_depth_stencil_alpha_state *d = nv20->dsa; + + BEGIN_RING(kelvin, NV20TCL_DEPTH_FUNC, 1); + OUT_RING (d->depth.func); + + BEGIN_RING(kelvin, NV20TCL_DEPTH_WRITE_ENABLE, 1); + OUT_RING (d->depth.write_enable); + + BEGIN_RING(kelvin, NV20TCL_DEPTH_TEST_ENABLE, 1); + OUT_RING (d->depth.test_enable); + +#if 0 + BEGIN_RING(kelvin, NV20TCL_STENCIL_ENABLE, 1); + OUT_RING (d->stencil.enable); + BEGIN_RING(kelvin, NV20TCL_STENCIL_MASK, 7); + OUT_RINGp ((uint32_t *)&(d->stencil.wmask), 7); +#endif + + BEGIN_RING(kelvin, NV20TCL_ALPHA_FUNC_ENABLE, 1); + OUT_RING (d->alpha.enabled); + + BEGIN_RING(kelvin, NV20TCL_ALPHA_FUNC_FUNC, 1); + OUT_RING (d->alpha.func); + + BEGIN_RING(kelvin, NV20TCL_ALPHA_FUNC_REF, 1); + OUT_RING (d->alpha.ref); +} + +static void nv20_state_emit_viewport(struct nv20_context* nv20) +{ +} + +static void nv20_state_emit_scissor(struct nv20_context* nv20) +{ + /* NV20TCL_SCISSOR_* is probably a software method */ +/* struct pipe_scissor_state *s = nv20->scissor; + BEGIN_RING(kelvin, NV20TCL_SCISSOR_HORIZ, 2); + OUT_RING (((s->maxx - s->minx) << 16) | s->minx); + OUT_RING (((s->maxy - s->miny) << 16) | s->miny);*/ +} + +static void nv20_state_emit_framebuffer(struct nv20_context* nv20) +{ + struct pipe_framebuffer_state* fb = nv20->framebuffer; + struct pipe_surface *rt, *zeta = NULL; + uint32_t rt_format, w, h; + int colour_format = 0, zeta_format = 0; + + w = fb->cbufs[0]->width; + h = fb->cbufs[0]->height; + colour_format = fb->cbufs[0]->format; + rt = fb->cbufs[0]; + + if (fb->zsbuf) { + if (colour_format) { + assert(w == fb->zsbuf->width); + assert(h == fb->zsbuf->height); + } else { + w = fb->zsbuf->width; + h = fb->zsbuf->height; + } + + zeta_format = fb->zsbuf->format; + zeta = fb->zsbuf; + } + + rt_format = NV20TCL_RT_FORMAT_TYPE_LINEAR | 0x20; + + switch (colour_format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case 0: + rt_format |= NV20TCL_RT_FORMAT_COLOR_A8R8G8B8; + break; + case PIPE_FORMAT_R5G6B5_UNORM: + rt_format |= NV20TCL_RT_FORMAT_COLOR_R5G6B5; + break; + default: + assert(0); + } + + if (zeta) { + BEGIN_RING(kelvin, NV20TCL_RT_PITCH, 1); + OUT_RING (rt->stride | (zeta->stride << 16)); + } else { + BEGIN_RING(kelvin, NV20TCL_RT_PITCH, 1); + OUT_RING (rt->stride | (rt->stride << 16)); + } + + nv20->rt[0] = rt->buffer; + + if (zeta_format) + { + nv20->zeta = zeta->buffer; + } + + BEGIN_RING(kelvin, NV20TCL_RT_HORIZ, 3); + OUT_RING ((w << 16) | 0); + OUT_RING ((h << 16) | 0); + OUT_RING (rt_format); + BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_HORIZ(0), 2); + OUT_RING (((w - 1) << 16) | 0); + OUT_RING (((h - 1) << 16) | 0); +} + +static void nv20_vertex_layout(struct nv20_context *nv20) +{ + struct nv20_fragment_program *fp = nv20->fragprog.current; + struct draw_context *dc = nv20->draw; + uint32_t src; + int i; + struct vertex_info *vinfo = &nv20->vertex_info; + const enum interp_mode colorInterp = INTERP_LINEAR; + boolean colors[2] = { FALSE }; + boolean generics[4] = { FALSE }; + boolean fog = FALSE; + + memset(vinfo, 0, sizeof(*vinfo)); + + /* + * NV10 hardware vertex attribute order: + * fog, weight, normal, tex1, tex0, 2nd color, color, position + * vinfo->hwfmt[0] has a used-bit corresponding to each of these. + * relation to TGSI_SEMANTIC_*: + * - POSITION: position (always used) + * - COLOR: 2nd color, color + * - GENERIC: weight, normal, tex1, tex0 + * - FOG: fog + */ + + for (i = 0; i < fp->info.num_inputs; i++) { + int isn = fp->info.input_semantic_name[i]; + int isi = fp->info.input_semantic_index[i]; + switch (isn) { + case TGSI_SEMANTIC_POSITION: + break; + case TGSI_SEMANTIC_COLOR: + assert(isi < 2); + colors[isi] = TRUE; + break; + case TGSI_SEMANTIC_GENERIC: + assert(isi < 4); + generics[isi] = TRUE; + break; + case TGSI_SEMANTIC_FOG: + fog = TRUE; + break; + default: + assert(0 && "unknown input_semantic_name"); + } + } + + if (fog) { + int src = draw_find_vs_output(dc, TGSI_SEMANTIC_FOG, 0); + draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_PERSPECTIVE, src); + vinfo->hwfmt[0] |= (1 << 7); + } + + for (i = 3; i >= 0; i--) { + int src; + if (!generics[i]) + continue; + src = draw_find_vs_output(dc, TGSI_SEMANTIC_GENERIC, i); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); + vinfo->hwfmt[0] |= (1 << (i + 3)); + } + + if (colors[1]) { + int src = draw_find_vs_output(dc, TGSI_SEMANTIC_COLOR, 1); + draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); + vinfo->hwfmt[0] |= (1 << 2); + } + + if (colors[0]) { + int src = draw_find_vs_output(dc, TGSI_SEMANTIC_COLOR, 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); + vinfo->hwfmt[0] |= (1 << 1); + } + + /* always do position */ + src = draw_find_vs_output(dc, TGSI_SEMANTIC_POSITION, 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_LINEAR, src); + vinfo->hwfmt[0] |= (1 << 0); + + draw_compute_vertex_size(vinfo); +} + +void +nv20_emit_hw_state(struct nv20_context *nv20) +{ + int i; + + if (nv20->dirty & NV20_NEW_VERTPROG) { + //nv20_vertprog_bind(nv20, nv20->vertprog.current); + nv20->dirty &= ~NV20_NEW_VERTPROG; + } + + if (nv20->dirty & NV20_NEW_FRAGPROG) { + nv20_fragprog_bind(nv20, nv20->fragprog.current); + /*XXX: clear NV20_NEW_FRAGPROG if no new program uploaded */ + nv20->dirty_samplers |= (1<<10); + nv20->dirty_samplers = 0; + } + + if (nv20->dirty_samplers || (nv20->dirty & NV20_NEW_FRAGPROG)) { + nv20_fragtex_bind(nv20); + nv20->dirty &= ~NV20_NEW_FRAGPROG; + } + + if (nv20->dirty & NV20_NEW_VTXARRAYS) { + nv20->dirty &= ~NV20_NEW_VTXARRAYS; + nv20_vertex_layout(nv20); + nv20_vtxbuf_bind(nv20); + } + + if (nv20->dirty & NV20_NEW_BLEND) { + nv20->dirty &= ~NV20_NEW_BLEND; + nv20_state_emit_blend(nv20); + } + + if (nv20->dirty & NV20_NEW_BLENDCOL) { + nv20->dirty &= ~NV20_NEW_BLENDCOL; + nv20_state_emit_blend_color(nv20); + } + + if (nv20->dirty & NV20_NEW_RAST) { + nv20->dirty &= ~NV20_NEW_RAST; + nv20_state_emit_rast(nv20); + } + + if (nv20->dirty & NV20_NEW_DSA) { + nv20->dirty &= ~NV20_NEW_DSA; + nv20_state_emit_dsa(nv20); + } + + if (nv20->dirty & NV20_NEW_VIEWPORT) { + nv20->dirty &= ~NV20_NEW_VIEWPORT; + nv20_state_emit_viewport(nv20); + } + + if (nv20->dirty & NV20_NEW_SCISSOR) { + nv20->dirty &= ~NV20_NEW_SCISSOR; + nv20_state_emit_scissor(nv20); + } + + if (nv20->dirty & NV20_NEW_FRAMEBUFFER) { + nv20->dirty &= ~NV20_NEW_FRAMEBUFFER; + nv20_state_emit_framebuffer(nv20); + } + + /* Emit relocs for every referenced buffer. + * This is to ensure the bufmgr has an accurate idea of how + * the buffer is used. This isn't very efficient, but we don't + * seem to take a significant performance hit. Will be improved + * at some point. Vertex arrays are emitted by nv20_vbo.c + */ + + /* Render target */ +/* XXX figre out who's who for NV10TCL_DMA_* and fill accordingly + * BEGIN_RING(kelvin, NV20TCL_DMA_COLOR0, 1); + * OUT_RELOCo(nv20->rt[0], NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); */ + BEGIN_RING(kelvin, NV20TCL_COLOR_OFFSET, 1); + OUT_RELOCl(nv20->rt[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + + if (nv20->zeta) { +/* XXX + * BEGIN_RING(kelvin, NV20TCL_DMA_ZETA, 1); + * OUT_RELOCo(nv20->zeta, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); */ + BEGIN_RING(kelvin, NV20TCL_ZETA_OFFSET, 1); + OUT_RELOCl(nv20->zeta, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + /* XXX for when we allocate LMA on nv17 */ +/* BEGIN_RING(kelvin, NV10TCL_LMA_DEPTH_BUFFER_OFFSET, 1); + OUT_RELOCl(nv20->zeta + lma_offset);*/ + } + + /* Vertex buffer */ + BEGIN_RING(kelvin, NV20TCL_DMA_VTXBUF0, 1); + OUT_RELOCo(nv20->rt[0], NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + BEGIN_RING(kelvin, NV20TCL_COLOR_OFFSET, 1); + OUT_RELOCl(nv20->rt[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + + /* Texture images */ + for (i = 0; i < 2; i++) { + if (!(nv20->fp_samplers & (1 << i))) + continue; + BEGIN_RING(kelvin, NV20TCL_TX_OFFSET(i), 1); + OUT_RELOCl(nv20->tex[i].buffer, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_GART | NOUVEAU_BO_RD); + BEGIN_RING(kelvin, NV20TCL_TX_FORMAT(i), 1); + OUT_RELOCd(nv20->tex[i].buffer, nv20->tex[i].format, + NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | + NOUVEAU_BO_OR, NV20TCL_TX_FORMAT_DMA0, + NV20TCL_TX_FORMAT_DMA1); + } +} + diff --git a/src/gallium/drivers/nv20/nv20_surface.c b/src/gallium/drivers/nv20/nv20_surface.c new file mode 100644 index 0000000000..7bc68d0ca2 --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_surface.c @@ -0,0 +1,64 @@ + +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "nv20_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/p_inlines.h" +#include "util/u_tile.h" + +static void +nv20_surface_copy(struct pipe_context *pipe, boolean do_flip, + struct pipe_surface *dest, unsigned destx, unsigned desty, + struct pipe_surface *src, unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + struct nv20_context *nv20 = nv20_context(pipe); + struct nouveau_winsys *nvws = nv20->nvws; + + nvws->surface_copy(nvws, dest, destx, desty, src, srcx, srcy, + width, height); +} + +static void +nv20_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest, + unsigned destx, unsigned desty, unsigned width, + unsigned height, unsigned value) +{ + struct nv20_context *nv20 = nv20_context(pipe); + struct nouveau_winsys *nvws = nv20->nvws; + + nvws->surface_fill(nvws, dest, destx, desty, width, height, value); +} + +void +nv20_init_surface_functions(struct nv20_context *nv20) +{ + nv20->pipe.surface_copy = nv20_surface_copy; + nv20->pipe.surface_fill = nv20_surface_fill; +} diff --git a/src/gallium/drivers/nv20/nv20_vbo.c b/src/gallium/drivers/nv20/nv20_vbo.c new file mode 100644 index 0000000000..4edc4efebd --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_vbo.c @@ -0,0 +1,77 @@ +#include "draw/draw_context.h" +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +#include "nv20_context.h" +#include "nv20_state.h" + +#include "nouveau/nouveau_channel.h" +#include "nouveau/nouveau_pushbuf.h" + +boolean nv20_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) +{ + struct nv20_context *nv20 = nv20_context( pipe ); + struct draw_context *draw = nv20->draw; + unsigned i; + + nv20_emit_hw_state(nv20); + + /* + * Map vertex buffers + */ + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { + if (nv20->vtxbuf[i].buffer) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + nv20->vtxbuf[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); + } + } + /* Map index buffer, if present */ + if (indexBuffer) { + void *mapped_indexes + = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); + } + else { + /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + draw_set_mapped_constant_buffer(draw, + nv20->constbuf[PIPE_SHADER_VERTEX], + nv20->constbuf_nr[PIPE_SHADER_VERTEX]); + + /* draw! */ + draw_arrays(nv20->draw, prim, start, count); + + /* + * unmap vertex/index buffers + */ + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { + if (nv20->vtxbuf[i].buffer) { + pipe->winsys->buffer_unmap(pipe->winsys, nv20->vtxbuf[i].buffer); + draw_set_mapped_vertex_buffer(draw, i, NULL); + } + } + if (indexBuffer) { + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + return TRUE; +} + +boolean nv20_draw_arrays( struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) +{ + return nv20_draw_elements(pipe, NULL, 0, prim, start, count); +} + + + diff --git a/src/gallium/drivers/nv20/nv20_vertprog.c b/src/gallium/drivers/nv20/nv20_vertprog.c new file mode 100644 index 0000000000..a885fcd7a5 --- /dev/null +++ b/src/gallium/drivers/nv20/nv20_vertprog.c @@ -0,0 +1,838 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_dump.h" + +#include "nv20_context.h" +#include "nv20_state.h" + +/* TODO (at least...): + * 1. Indexed consts + ARL + * 2. Arb. swz/negation + * 3. NV_vp11, NV_vp2, NV_vp3 features + * - extra arith opcodes + * - branching + * - texture sampling + * - indexed attribs + * - indexed results + * 4. bugs + */ + +#define SWZ_X 0 +#define SWZ_Y 1 +#define SWZ_Z 2 +#define SWZ_W 3 +#define MASK_X 8 +#define MASK_Y 4 +#define MASK_Z 2 +#define MASK_W 1 +#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W) +#define DEF_SCALE 0 +#define DEF_CTEST 0 +#include "nv20_shader.h" + +#define swz(s,x,y,z,w) nv20_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w) +#define neg(s) nv20_sr_neg((s)) +#define abs(s) nv20_sr_abs((s)) + +struct nv20_vpc { + struct nv20_vertex_program *vp; + + struct nv20_vertex_program_exec *vpi; + + unsigned output_map[PIPE_MAX_SHADER_OUTPUTS]; + + int high_temp; + int temp_temp_count; + + struct nv20_sreg *imm; + unsigned nr_imm; +}; + +static struct nv20_sreg +temp(struct nv20_vpc *vpc) +{ + int idx; + + idx = vpc->temp_temp_count++; + idx += vpc->high_temp + 1; + return nv20_sr(NV30SR_TEMP, idx); +} + +static struct nv20_sreg +constant(struct nv20_vpc *vpc, int pipe, float x, float y, float z, float w) +{ + struct nv20_vertex_program *vp = vpc->vp; + struct nv20_vertex_program_data *vpd; + int idx; + + if (pipe >= 0) { + for (idx = 0; idx < vp->nr_consts; idx++) { + if (vp->consts[idx].index == pipe) + return nv20_sr(NV30SR_CONST, idx); + } + } + + idx = vp->nr_consts++; + vp->consts = realloc(vp->consts, sizeof(*vpd) * vp->nr_consts); + vpd = &vp->consts[idx]; + + vpd->index = pipe; + vpd->value[0] = x; + vpd->value[1] = y; + vpd->value[2] = z; + vpd->value[3] = w; + return nv20_sr(NV30SR_CONST, idx); +} + +#define arith(cc,s,o,d,m,s0,s1,s2) \ + nv20_vp_arith((cc), (s), NV30_VP_INST_##o, (d), (m), (s0), (s1), (s2)) + +static void +emit_src(struct nv20_vpc *vpc, uint32_t *hw, int pos, struct nv20_sreg src) +{ + struct nv20_vertex_program *vp = vpc->vp; + uint32_t sr = 0; + + switch (src.type) { + case NV30SR_TEMP: + sr |= (NV30_VP_SRC_REG_TYPE_TEMP << NV30_VP_SRC_REG_TYPE_SHIFT); + sr |= (src.index << NV30_VP_SRC_TEMP_SRC_SHIFT); + break; + case NV30SR_INPUT: + sr |= (NV30_VP_SRC_REG_TYPE_INPUT << + NV30_VP_SRC_REG_TYPE_SHIFT); + vp->ir |= (1 << src.index); + hw[1] |= (src.index << NV30_VP_INST_INPUT_SRC_SHIFT); + break; + case NV30SR_CONST: + sr |= (NV30_VP_SRC_REG_TYPE_CONST << + NV30_VP_SRC_REG_TYPE_SHIFT); + assert(vpc->vpi->const_index == -1 || + vpc->vpi->const_index == src.index); + vpc->vpi->const_index = src.index; + break; + case NV30SR_NONE: + sr |= (NV30_VP_SRC_REG_TYPE_INPUT << + NV30_VP_SRC_REG_TYPE_SHIFT); + break; + default: + assert(0); + } + + if (src.negate) + sr |= NV30_VP_SRC_NEGATE; + + if (src.abs) + hw[0] |= (1 << (21 + pos)); + + sr |= ((src.swz[0] << NV30_VP_SRC_SWZ_X_SHIFT) | + (src.swz[1] << NV30_VP_SRC_SWZ_Y_SHIFT) | + (src.swz[2] << NV30_VP_SRC_SWZ_Z_SHIFT) | + (src.swz[3] << NV30_VP_SRC_SWZ_W_SHIFT)); + +/* + * |VVV| + * d�.�b + * \u/ + * + */ + + switch (pos) { + case 0: + hw[1] |= ((sr & NV30_VP_SRC0_HIGH_MASK) >> + NV30_VP_SRC0_HIGH_SHIFT) << NV30_VP_INST_SRC0H_SHIFT; + hw[2] |= (sr & NV30_VP_SRC0_LOW_MASK) << + NV30_VP_INST_SRC0L_SHIFT; + break; + case 1: + hw[2] |= sr << NV30_VP_INST_SRC1_SHIFT; + break; + case 2: + hw[2] |= ((sr & NV30_VP_SRC2_HIGH_MASK) >> + NV30_VP_SRC2_HIGH_SHIFT) << NV30_VP_INST_SRC2H_SHIFT; + hw[3] |= (sr & NV30_VP_SRC2_LOW_MASK) << + NV30_VP_INST_SRC2L_SHIFT; + break; + default: + assert(0); + } +} + +static void +emit_dst(struct nv20_vpc *vpc, uint32_t *hw, int slot, struct nv20_sreg dst) +{ + struct nv20_vertex_program *vp = vpc->vp; + + switch (dst.type) { + case NV30SR_TEMP: + hw[0] |= (dst.index << NV30_VP_INST_DEST_TEMP_ID_SHIFT); + break; + case NV30SR_OUTPUT: + switch (dst.index) { + case NV30_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break; + case NV30_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break; + case NV30_VP_INST_DEST_BFC0 : vp->or |= (1 << 2); break; + case NV30_VP_INST_DEST_BFC1 : vp->or |= (1 << 3); break; + case NV30_VP_INST_DEST_FOGC : vp->or |= (1 << 4); break; + case NV30_VP_INST_DEST_PSZ : vp->or |= (1 << 5); break; + case NV30_VP_INST_DEST_TC(0): vp->or |= (1 << 14); break; + case NV30_VP_INST_DEST_TC(1): vp->or |= (1 << 15); break; + case NV30_VP_INST_DEST_TC(2): vp->or |= (1 << 16); break; + case NV30_VP_INST_DEST_TC(3): vp->or |= (1 << 17); break; + case NV30_VP_INST_DEST_TC(4): vp->or |= (1 << 18); break; + case NV30_VP_INST_DEST_TC(5): vp->or |= (1 << 19); break; + case NV30_VP_INST_DEST_TC(6): vp->or |= (1 << 20); break; + case NV30_VP_INST_DEST_TC(7): vp->or |= (1 << 21); break; + default: + break; + } + + hw[3] |= (dst.index << NV30_VP_INST_DEST_SHIFT); + hw[0] |= NV30_VP_INST_VEC_DEST_TEMP_MASK | (1<<20); + + /*XXX: no way this is entirely correct, someone needs to + * figure out what exactly it is. + */ + hw[3] |= 0x800; + break; + default: + assert(0); + } +} + +static void +nv20_vp_arith(struct nv20_vpc *vpc, int slot, int op, + struct nv20_sreg dst, int mask, + struct nv20_sreg s0, struct nv20_sreg s1, + struct nv20_sreg s2) +{ + struct nv20_vertex_program *vp = vpc->vp; + uint32_t *hw; + + vp->insns = realloc(vp->insns, ++vp->nr_insns * sizeof(*vpc->vpi)); + vpc->vpi = &vp->insns[vp->nr_insns - 1]; + memset(vpc->vpi, 0, sizeof(*vpc->vpi)); + vpc->vpi->const_index = -1; + + hw = vpc->vpi->data; + + hw[0] |= (NV30_VP_INST_COND_TR << NV30_VP_INST_COND_SHIFT); + hw[0] |= ((0 << NV30_VP_INST_COND_SWZ_X_SHIFT) | + (1 << NV30_VP_INST_COND_SWZ_Y_SHIFT) | + (2 << NV30_VP_INST_COND_SWZ_Z_SHIFT) | + (3 << NV30_VP_INST_COND_SWZ_W_SHIFT)); + + hw[1] |= (op << NV30_VP_INST_VEC_OPCODE_SHIFT); +// hw[3] |= NV30_VP_INST_SCA_DEST_TEMP_MASK; +// hw[3] |= (mask << NV30_VP_INST_VEC_WRITEMASK_SHIFT); + + if (dst.type == NV30SR_OUTPUT) { + if (slot) + hw[3] |= (mask << NV30_VP_INST_SDEST_WRITEMASK_SHIFT); + else + hw[3] |= (mask << NV30_VP_INST_VDEST_WRITEMASK_SHIFT); + } else { + if (slot) + hw[3] |= (mask << NV30_VP_INST_STEMP_WRITEMASK_SHIFT); + else + hw[3] |= (mask << NV30_VP_INST_VTEMP_WRITEMASK_SHIFT); + } + + emit_dst(vpc, hw, slot, dst); + emit_src(vpc, hw, 0, s0); + emit_src(vpc, hw, 1, s1); + emit_src(vpc, hw, 2, s2); +} + +static INLINE struct nv20_sreg +tgsi_src(struct nv20_vpc *vpc, const struct tgsi_full_src_register *fsrc) { + struct nv20_sreg src; + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + src = nv20_sr(NV30SR_INPUT, fsrc->SrcRegister.Index); + break; + case TGSI_FILE_CONSTANT: + src = constant(vpc, fsrc->SrcRegister.Index, 0, 0, 0, 0); + break; + case TGSI_FILE_IMMEDIATE: + src = vpc->imm[fsrc->SrcRegister.Index]; + break; + case TGSI_FILE_TEMPORARY: + if (vpc->high_temp < fsrc->SrcRegister.Index) + vpc->high_temp = fsrc->SrcRegister.Index; + src = nv20_sr(NV30SR_TEMP, fsrc->SrcRegister.Index); + break; + default: + NOUVEAU_ERR("bad src file\n"); + break; + } + + src.abs = fsrc->SrcRegisterExtMod.Absolute; + src.negate = fsrc->SrcRegister.Negate; + src.swz[0] = fsrc->SrcRegister.SwizzleX; + src.swz[1] = fsrc->SrcRegister.SwizzleY; + src.swz[2] = fsrc->SrcRegister.SwizzleZ; + src.swz[3] = fsrc->SrcRegister.SwizzleW; + return src; +} + +static INLINE struct nv20_sreg +tgsi_dst(struct nv20_vpc *vpc, const struct tgsi_full_dst_register *fdst) { + struct nv20_sreg dst; + + switch (fdst->DstRegister.File) { + case TGSI_FILE_OUTPUT: + dst = nv20_sr(NV30SR_OUTPUT, + vpc->output_map[fdst->DstRegister.Index]); + + break; + case TGSI_FILE_TEMPORARY: + dst = nv20_sr(NV30SR_TEMP, fdst->DstRegister.Index); + if (vpc->high_temp < dst.index) + vpc->high_temp = dst.index; + break; + default: + NOUVEAU_ERR("bad dst file\n"); + break; + } + + return dst; +} + +static INLINE int +tgsi_mask(uint tgsi) +{ + int mask = 0; + + if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X; + if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y; + if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z; + if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W; + return mask; +} + +static boolean +nv20_vertprog_parse_instruction(struct nv20_vpc *vpc, + const struct tgsi_full_instruction *finst) +{ + struct nv20_sreg src[3], dst, tmp; + struct nv20_sreg none = nv20_sr(NV30SR_NONE, 0); + int mask; + int ai = -1, ci = -1; + int i; + + if (finst->Instruction.Opcode == TGSI_OPCODE_END) + return TRUE; + + vpc->temp_temp_count = 0; + for (i = 0; i < finst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *fsrc; + + fsrc = &finst->FullSrcRegisters[i]; + if (fsrc->SrcRegister.File == TGSI_FILE_TEMPORARY) { + src[i] = tgsi_src(vpc, fsrc); + } + } + + for (i = 0; i < finst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *fsrc; + + fsrc = &finst->FullSrcRegisters[i]; + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + if (ai == -1 || ai == fsrc->SrcRegister.Index) { + ai = fsrc->SrcRegister.Index; + src[i] = tgsi_src(vpc, fsrc); + } else { + src[i] = temp(vpc); + arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + tgsi_src(vpc, fsrc), none, none); + } + break; + /*XXX: index comparison is broken now that consts come from + * two different register files. + */ + case TGSI_FILE_CONSTANT: + case TGSI_FILE_IMMEDIATE: + if (ci == -1 || ci == fsrc->SrcRegister.Index) { + ci = fsrc->SrcRegister.Index; + src[i] = tgsi_src(vpc, fsrc); + } else { + src[i] = temp(vpc); + arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + tgsi_src(vpc, fsrc), none, none); + } + break; + case TGSI_FILE_TEMPORARY: + /* handled above */ + break; + default: + NOUVEAU_ERR("bad src file\n"); + return FALSE; + } + } + + dst = tgsi_dst(vpc, &finst->FullDstRegisters[0]); + mask = tgsi_mask(finst->FullDstRegisters[0].DstRegister.WriteMask); + + switch (finst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + arith(vpc, 0, OP_MOV, dst, mask, abs(src[0]), none, none); + break; + case TGSI_OPCODE_ADD: + arith(vpc, 0, OP_ADD, dst, mask, src[0], none, src[1]); + break; + case TGSI_OPCODE_ARL: + arith(vpc, 0, OP_ARL, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_DP3: + arith(vpc, 0, OP_DP3, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DP4: + arith(vpc, 0, OP_DP4, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DPH: + arith(vpc, 0, OP_DPH, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DST: + arith(vpc, 0, OP_DST, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_EX2: + arith(vpc, 1, OP_EX2, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_EXP: + arith(vpc, 1, OP_EXP, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_FLR: + arith(vpc, 0, OP_FLR, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_FRC: + arith(vpc, 0, OP_FRC, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_LG2: + arith(vpc, 1, OP_LG2, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_LIT: + arith(vpc, 1, OP_LIT, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_LOG: + arith(vpc, 1, OP_LOG, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_MAD: + arith(vpc, 0, OP_MAD, dst, mask, src[0], src[1], src[2]); + break; + case TGSI_OPCODE_MAX: + arith(vpc, 0, OP_MAX, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_MIN: + arith(vpc, 0, OP_MIN, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_MOV: + arith(vpc, 0, OP_MOV, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_MUL: + arith(vpc, 0, OP_MUL, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_POW: + tmp = temp(vpc); + arith(vpc, 1, OP_LG2, tmp, MASK_X, none, none, + swz(src[0], X, X, X, X)); + arith(vpc, 0, OP_MUL, tmp, MASK_X, swz(tmp, X, X, X, X), + swz(src[1], X, X, X, X), none); + arith(vpc, 1, OP_EX2, dst, mask, none, none, + swz(tmp, X, X, X, X)); + break; + case TGSI_OPCODE_RCP: + arith(vpc, 1, OP_RCP, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_RET: + break; + case TGSI_OPCODE_RSQ: + arith(vpc, 1, OP_RSQ, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_SGE: + arith(vpc, 0, OP_SGE, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SGT: + arith(vpc, 0, OP_SGT, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SLT: + arith(vpc, 0, OP_SLT, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SUB: + arith(vpc, 0, OP_ADD, dst, mask, src[0], none, neg(src[1])); + break; + case TGSI_OPCODE_XPD: + tmp = temp(vpc); + arith(vpc, 0, OP_MUL, tmp, mask, + swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none); + arith(vpc, 0, OP_MAD, dst, (mask & ~MASK_W), + swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y), + neg(tmp)); + break; + default: + NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode); + return FALSE; + } + + return TRUE; +} + +static boolean +nv20_vertprog_parse_decl_output(struct nv20_vpc *vpc, + const struct tgsi_full_declaration *fdec) +{ + int hw; + + switch (fdec->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: + hw = NV30_VP_INST_DEST_POS; + break; + case TGSI_SEMANTIC_COLOR: + if (fdec->Semantic.SemanticIndex == 0) { + hw = NV30_VP_INST_DEST_COL0; + } else + if (fdec->Semantic.SemanticIndex == 1) { + hw = NV30_VP_INST_DEST_COL1; + } else { + NOUVEAU_ERR("bad colour semantic index\n"); + return FALSE; + } + break; + case TGSI_SEMANTIC_BCOLOR: + if (fdec->Semantic.SemanticIndex == 0) { + hw = NV30_VP_INST_DEST_BFC0; + } else + if (fdec->Semantic.SemanticIndex == 1) { + hw = NV30_VP_INST_DEST_BFC1; + } else { + NOUVEAU_ERR("bad bcolour semantic index\n"); + return FALSE; + } + break; + case TGSI_SEMANTIC_FOG: + hw = NV30_VP_INST_DEST_FOGC; + break; + case TGSI_SEMANTIC_PSIZE: + hw = NV30_VP_INST_DEST_PSZ; + break; + case TGSI_SEMANTIC_GENERIC: + if (fdec->Semantic.SemanticIndex <= 7) { + hw = NV30_VP_INST_DEST_TC(fdec->Semantic.SemanticIndex); + } else { + NOUVEAU_ERR("bad generic semantic index\n"); + return FALSE; + } + break; + default: + NOUVEAU_ERR("bad output semantic\n"); + return FALSE; + } + + vpc->output_map[fdec->DeclarationRange.First] = hw; + return TRUE; +} + +static boolean +nv20_vertprog_prepare(struct nv20_vpc *vpc) +{ + struct tgsi_parse_context p; + int nr_imm = 0; + + tgsi_parse_init(&p, vpc->vp->pipe.tokens); + while (!tgsi_parse_end_of_tokens(&p)) { + const union tgsi_full_token *tok = &p.FullToken; + + tgsi_parse_token(&p); + switch(tok->Token.Type) { + case TGSI_TOKEN_TYPE_IMMEDIATE: + nr_imm++; + break; + default: + break; + } + } + tgsi_parse_free(&p); + + if (nr_imm) { + vpc->imm = CALLOC(nr_imm, sizeof(struct nv20_sreg)); + assert(vpc->imm); + } + + return TRUE; +} + +static void +nv20_vertprog_translate(struct nv20_context *nv20, + struct nv20_vertex_program *vp) +{ + struct tgsi_parse_context parse; + struct nv20_vpc *vpc = NULL; + + tgsi_dump(vp->pipe.tokens,0); + + vpc = CALLOC(1, sizeof(struct nv20_vpc)); + if (!vpc) + return; + vpc->vp = vp; + vpc->high_temp = -1; + + if (!nv20_vertprog_prepare(vpc)) { + FREE(vpc); + return; + } + + tgsi_parse_init(&parse, vp->pipe.tokens); + + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + const struct tgsi_full_declaration *fdec; + fdec = &parse.FullToken.FullDeclaration; + switch (fdec->Declaration.File) { + case TGSI_FILE_OUTPUT: + if (!nv20_vertprog_parse_decl_output(vpc, fdec)) + goto out_err; + break; + default: + break; + } + } + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + const struct tgsi_full_immediate *imm; + + imm = &parse.FullToken.FullImmediate; + assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32); +// assert(imm->Immediate.Size == 4); + vpc->imm[vpc->nr_imm++] = + constant(vpc, -1, + imm->u.ImmediateFloat32[0].Float, + imm->u.ImmediateFloat32[1].Float, + imm->u.ImmediateFloat32[2].Float, + imm->u.ImmediateFloat32[3].Float); + } + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + const struct tgsi_full_instruction *finst; + finst = &parse.FullToken.FullInstruction; + if (!nv20_vertprog_parse_instruction(vpc, finst)) + goto out_err; + } + break; + default: + break; + } + } + + vp->insns[vp->nr_insns - 1].data[3] |= NV30_VP_INST_LAST; + vp->translated = TRUE; +out_err: + tgsi_parse_free(&parse); + FREE(vpc); +} + +static boolean +nv20_vertprog_validate(struct nv20_context *nv20) +{ + struct nouveau_winsys *nvws = nv20->nvws; + struct pipe_winsys *ws = nv20->pipe.winsys; + struct nouveau_grobj *rankine = nv20->screen->rankine; + struct nv20_vertex_program *vp; + struct pipe_buffer *constbuf; + boolean upload_code = FALSE, upload_data = FALSE; + int i; + + vp = nv20->vertprog; + constbuf = nv20->constbuf[PIPE_SHADER_VERTEX]; + + /* Translate TGSI shader into hw bytecode */ + if (!vp->translated) { + nv20_vertprog_translate(nv20, vp); + if (!vp->translated) + return FALSE; + } + + /* Allocate hw vtxprog exec slots */ + if (!vp->exec) { + struct nouveau_resource *heap = nv20->screen->vp_exec_heap; + struct nouveau_stateobj *so; + uint vplen = vp->nr_insns; + + if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) { + while (heap->next && heap->size < vplen) { + struct nv20_vertex_program *evict; + + evict = heap->next->priv; + nvws->res_free(&evict->exec); + } + + if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) + assert(0); + } + + so = so_new(2, 0); + so_method(so, rankine, NV34TCL_VP_START_FROM_ID, 1); + so_data (so, vp->exec->start); + so_ref(so, &vp->so); + + upload_code = TRUE; + } + + /* Allocate hw vtxprog const slots */ + if (vp->nr_consts && !vp->data) { + struct nouveau_resource *heap = nv20->screen->vp_data_heap; + + if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) { + while (heap->next && heap->size < vp->nr_consts) { + struct nv20_vertex_program *evict; + + evict = heap->next->priv; + nvws->res_free(&evict->data); + } + + if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) + assert(0); + } + + /*XXX: handle this some day */ + assert(vp->data->start >= vp->data_start_min); + + upload_data = TRUE; + if (vp->data_start != vp->data->start) + upload_code = TRUE; + } + + /* If exec or data segments moved we need to patch the program to + * fixup offsets and register IDs. + */ + if (vp->exec_start != vp->exec->start) { + for (i = 0; i < vp->nr_insns; i++) { + struct nv20_vertex_program_exec *vpi = &vp->insns[i]; + + if (vpi->has_branch_offset) { + assert(0); + } + } + + vp->exec_start = vp->exec->start; + } + + if (vp->nr_consts && vp->data_start != vp->data->start) { + for (i = 0; i < vp->nr_insns; i++) { + struct nv20_vertex_program_exec *vpi = &vp->insns[i]; + + if (vpi->const_index >= 0) { + vpi->data[1] &= ~NV30_VP_INST_CONST_SRC_MASK; + vpi->data[1] |= + (vpi->const_index + vp->data->start) << + NV30_VP_INST_CONST_SRC_SHIFT; + + } + } + + vp->data_start = vp->data->start; + } + + /* Update + Upload constant values */ + if (vp->nr_consts) { + float *map = NULL; + + if (constbuf) { + map = ws->buffer_map(ws, constbuf, + PIPE_BUFFER_USAGE_CPU_READ); + } + + for (i = 0; i < vp->nr_consts; i++) { + struct nv20_vertex_program_data *vpd = &vp->consts[i]; + + if (vpd->index >= 0) { + if (!upload_data && + !memcmp(vpd->value, &map[vpd->index * 4], + 4 * sizeof(float))) + continue; + memcpy(vpd->value, &map[vpd->index * 4], + 4 * sizeof(float)); + } + + BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_CONST_ID, 5); + OUT_RING (i + vp->data->start); + OUT_RINGp ((uint32_t *)vpd->value, 4); + } + + if (constbuf) { + ws->buffer_unmap(ws, constbuf); + } + } + + /* Upload vtxprog */ + if (upload_code) { +#if 0 + for (i = 0; i < vp->nr_insns; i++) { + NOUVEAU_MSG("VP inst %d: 0x%08x 0x%08x 0x%08x 0x%08x\n", + i, vp->insns[i].data[0], vp->insns[i].data[1], + vp->insns[i].data[2], vp->insns[i].data[3]); + } +#endif + BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_FROM_ID, 1); + OUT_RING (vp->exec->start); + for (i = 0; i < vp->nr_insns; i++) { + BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_INST(0), 4); + OUT_RINGp (vp->insns[i].data, 4); + } + } + + if (vp->so != nv20->state.hw[NV30_STATE_VERTPROG]) { + so_ref(vp->so, &nv20->state.hw[NV30_STATE_VERTPROG]); + return TRUE; + } + + return FALSE; +} + +void +nv20_vertprog_destroy(struct nv20_context *nv20, struct nv20_vertex_program *vp) +{ + struct nouveau_winsys *nvws = nv20->screen->nvws; + + vp->translated = FALSE; + + if (vp->nr_insns) { + FREE(vp->insns); + vp->insns = NULL; + vp->nr_insns = 0; + } + + if (vp->nr_consts) { + FREE(vp->consts); + vp->consts = NULL; + vp->nr_consts = 0; + } + + nvws->res_free(&vp->exec); + vp->exec_start = 0; + nvws->res_free(&vp->data); + vp->data_start = 0; + vp->data_start_min = 0; + + vp->ir = vp->or = 0; + so_ref(NULL, &vp->so); +} + +struct nv20_state_entry nv20_state_vertprog = { + .validate = nv20_vertprog_validate, + .dirty = { + .pipe = NV30_NEW_VERTPROG /*| NV30_NEW_UCP*/, + .hw = NV30_STATE_VERTPROG, + } +}; diff --git a/src/gallium/drivers/nv30/Makefile b/src/gallium/drivers/nv30/Makefile new file mode 100644 index 0000000000..69f2790dfe --- /dev/null +++ b/src/gallium/drivers/nv30/Makefile @@ -0,0 +1,37 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = nv30 + +DRIVER_SOURCES = \ + nv30_clear.c \ + nv30_context.c \ + nv30_draw.c \ + nv30_fragprog.c \ + nv30_fragtex.c \ + nv30_miptree.c \ + nv30_query.c \ + nv30_screen.c \ + nv30_state.c \ + nv30_state_blend.c \ + nv30_state_emit.c \ + nv30_state_fb.c \ + nv30_state_rasterizer.c \ + nv30_state_scissor.c \ + nv30_state_stipple.c \ + nv30_state_viewport.c \ + nv30_state_zsa.c \ + nv30_surface.c \ + nv30_vbo.c \ + nv30_vertprog.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/drivers/nv30/nv30_clear.c b/src/gallium/drivers/nv30/nv30_clear.c new file mode 100644 index 0000000000..8c3ca204d5 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_clear.c @@ -0,0 +1,13 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "nv30_context.h" + +void +nv30_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); + ps->status = PIPE_SURFACE_STATUS_CLEAR; +} diff --git a/src/gallium/drivers/nv30/nv30_context.c b/src/gallium/drivers/nv30/nv30_context.c new file mode 100644 index 0000000000..2bff28aca9 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_context.c @@ -0,0 +1,72 @@ +#include "draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" + +#include "nv30_context.h" +#include "nv30_screen.h" + +static void +nv30_flush(struct pipe_context *pipe, unsigned flags, + struct pipe_fence_handle **fence) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + if (flags & PIPE_FLUSH_TEXTURE_CACHE) { + BEGIN_RING(rankine, 0x1fd8, 1); + OUT_RING (2); + BEGIN_RING(rankine, 0x1fd8, 1); + OUT_RING (1); + } + + FIRE_RING(fence); +} + +static void +nv30_destroy(struct pipe_context *pipe) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + if (nv30->draw) + draw_destroy(nv30->draw); + FREE(nv30); +} + +struct pipe_context * +nv30_create(struct pipe_screen *pscreen, unsigned pctx_id) +{ + struct nv30_screen *screen = nv30_screen(pscreen); + struct pipe_winsys *ws = pscreen->winsys; + struct nv30_context *nv30; + struct nouveau_winsys *nvws = screen->nvws; + + nv30 = CALLOC(1, sizeof(struct nv30_context)); + if (!nv30) + return NULL; + nv30->screen = screen; + nv30->pctx_id = pctx_id; + + nv30->nvws = nvws; + + nv30->pipe.winsys = ws; + nv30->pipe.screen = pscreen; + nv30->pipe.destroy = nv30_destroy; + nv30->pipe.draw_arrays = nv30_draw_arrays; + nv30->pipe.draw_elements = nv30_draw_elements; + nv30->pipe.clear = nv30_clear; + nv30->pipe.flush = nv30_flush; + + nv30_init_query_functions(nv30); + nv30_init_surface_functions(nv30); + nv30_init_state_functions(nv30); + + /* Create, configure, and install fallback swtnl path */ + nv30->draw = draw_create(); + draw_wide_point_threshold(nv30->draw, 9999999.0); + draw_wide_line_threshold(nv30->draw, 9999999.0); + draw_enable_line_stipple(nv30->draw, FALSE); + draw_enable_point_sprites(nv30->draw, FALSE); + draw_set_rasterize_stage(nv30->draw, nv30_draw_render_stage(nv30)); + + return &nv30->pipe; +} + diff --git a/src/gallium/drivers/nv30/nv30_context.h b/src/gallium/drivers/nv30/nv30_context.h new file mode 100644 index 0000000000..b933769700 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_context.h @@ -0,0 +1,212 @@ +#ifndef __NV30_CONTEXT_H__ +#define __NV30_CONTEXT_H__ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_compiler.h" + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "draw/draw_vertex.h" + +#include "nouveau/nouveau_winsys.h" +#include "nouveau/nouveau_gldefs.h" + +#define NOUVEAU_PUSH_CONTEXT(ctx) \ + struct nv30_screen *ctx = nv30->screen +#include "nouveau/nouveau_push.h" +#include "nouveau/nouveau_stateobj.h" + +#include "nv30_state.h" + +#define NOUVEAU_ERR(fmt, args...) \ + fprintf(stderr, "%s:%d - "fmt, __func__, __LINE__, ##args); +#define NOUVEAU_MSG(fmt, args...) \ + fprintf(stderr, "nouveau: "fmt, ##args); + +enum nv30_state_index { + NV30_STATE_FB = 0, + NV30_STATE_VIEWPORT = 1, + NV30_STATE_BLEND = 2, + NV30_STATE_RAST = 3, + NV30_STATE_ZSA = 4, + NV30_STATE_BCOL = 5, + NV30_STATE_CLIP = 6, + NV30_STATE_SCISSOR = 7, + NV30_STATE_STIPPLE = 8, + NV30_STATE_FRAGPROG = 9, + NV30_STATE_VERTPROG = 10, + NV30_STATE_FRAGTEX0 = 11, + NV30_STATE_FRAGTEX1 = 12, + NV30_STATE_FRAGTEX2 = 13, + NV30_STATE_FRAGTEX3 = 14, + NV30_STATE_FRAGTEX4 = 15, + NV30_STATE_FRAGTEX5 = 16, + NV30_STATE_FRAGTEX6 = 17, + NV30_STATE_FRAGTEX7 = 18, + NV30_STATE_FRAGTEX8 = 19, + NV30_STATE_FRAGTEX9 = 20, + NV30_STATE_FRAGTEX10 = 21, + NV30_STATE_FRAGTEX11 = 22, + NV30_STATE_FRAGTEX12 = 23, + NV30_STATE_FRAGTEX13 = 24, + NV30_STATE_FRAGTEX14 = 25, + NV30_STATE_FRAGTEX15 = 26, + NV30_STATE_VERTTEX0 = 27, + NV30_STATE_VERTTEX1 = 28, + NV30_STATE_VERTTEX2 = 29, + NV30_STATE_VERTTEX3 = 30, + NV30_STATE_VTXBUF = 31, + NV30_STATE_VTXFMT = 32, + NV30_STATE_VTXATTR = 33, + NV30_STATE_MAX = 34 +}; + +#include "nv30_screen.h" + +#define NV30_NEW_BLEND (1 << 0) +#define NV30_NEW_RAST (1 << 1) +#define NV30_NEW_ZSA (1 << 2) +#define NV30_NEW_SAMPLER (1 << 3) +#define NV30_NEW_FB (1 << 4) +#define NV30_NEW_STIPPLE (1 << 5) +#define NV30_NEW_SCISSOR (1 << 6) +#define NV30_NEW_VIEWPORT (1 << 7) +#define NV30_NEW_BCOL (1 << 8) +#define NV30_NEW_VERTPROG (1 << 9) +#define NV30_NEW_FRAGPROG (1 << 10) +#define NV30_NEW_ARRAYS (1 << 11) +#define NV30_NEW_UCP (1 << 12) + +struct nv30_rasterizer_state { + struct pipe_rasterizer_state pipe; + struct nouveau_stateobj *so; +}; + +struct nv30_zsa_state { + struct pipe_depth_stencil_alpha_state pipe; + struct nouveau_stateobj *so; +}; + +struct nv30_blend_state { + struct pipe_blend_state pipe; + struct nouveau_stateobj *so; +}; + + +struct nv30_state { + unsigned scissor_enabled; + unsigned stipple_enabled; + unsigned viewport_bypass; + unsigned fp_samplers; + + uint64_t dirty; + struct nouveau_stateobj *hw[NV30_STATE_MAX]; +}; + +struct nv30_context { + struct pipe_context pipe; + + struct nouveau_winsys *nvws; + struct nv30_screen *screen; + unsigned pctx_id; + + struct draw_context *draw; + + /* HW state derived from pipe states */ + struct nv30_state state; + + /* Context state */ + unsigned dirty; + struct pipe_scissor_state scissor; + unsigned stipple[32]; + struct nv30_vertex_program *vertprog; + struct nv30_fragment_program *fragprog; + struct pipe_buffer *constbuf[PIPE_SHADER_TYPES]; + unsigned constbuf_nr[PIPE_SHADER_TYPES]; + struct nv30_rasterizer_state *rasterizer; + struct nv30_zsa_state *zsa; + struct nv30_blend_state *blend; + struct pipe_blend_color blend_colour; + struct pipe_viewport_state viewport; + struct pipe_framebuffer_state framebuffer; + struct pipe_buffer *idxbuf; + unsigned idxbuf_format; + struct nv30_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS]; + struct nv30_miptree *tex_miptree[PIPE_MAX_SAMPLERS]; + unsigned nr_samplers; + unsigned nr_textures; + unsigned dirty_samplers; + struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS]; + unsigned vtxbuf_nr; + struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS]; + unsigned vtxelt_nr; + const unsigned *edgeflags; +}; + +static INLINE struct nv30_context * +nv30_context(struct pipe_context *pipe) +{ + return (struct nv30_context *)pipe; +} + +struct nv30_state_entry { + boolean (*validate)(struct nv30_context *nv30); + struct { + unsigned pipe; + unsigned hw; + } dirty; +}; + +extern void nv30_init_state_functions(struct nv30_context *nv30); +extern void nv30_init_surface_functions(struct nv30_context *nv30); +extern void nv30_init_query_functions(struct nv30_context *nv30); + +extern void nv30_screen_init_miptree_functions(struct pipe_screen *pscreen); + +/* nv30_draw.c */ +extern struct draw_stage *nv30_draw_render_stage(struct nv30_context *nv30); + +/* nv30_vertprog.c */ +extern void nv30_vertprog_destroy(struct nv30_context *, + struct nv30_vertex_program *); + +/* nv30_fragprog.c */ +extern void nv30_fragprog_destroy(struct nv30_context *, + struct nv30_fragment_program *); + +/* nv30_fragtex.c */ +extern void nv30_fragtex_bind(struct nv30_context *); + +/* nv30_state.c and friends */ +extern boolean nv30_state_validate(struct nv30_context *nv30); +extern void nv30_state_emit(struct nv30_context *nv30); +extern struct nv30_state_entry nv30_state_rasterizer; +extern struct nv30_state_entry nv30_state_scissor; +extern struct nv30_state_entry nv30_state_stipple; +extern struct nv30_state_entry nv30_state_fragprog; +extern struct nv30_state_entry nv30_state_vertprog; +extern struct nv30_state_entry nv30_state_blend; +extern struct nv30_state_entry nv30_state_blend_colour; +extern struct nv30_state_entry nv30_state_zsa; +extern struct nv30_state_entry nv30_state_viewport; +extern struct nv30_state_entry nv30_state_framebuffer; +extern struct nv30_state_entry nv30_state_fragtex; +extern struct nv30_state_entry nv30_state_vbo; + +/* nv30_vbo.c */ +extern boolean nv30_draw_arrays(struct pipe_context *, unsigned mode, + unsigned start, unsigned count); +extern boolean nv30_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, + unsigned count); + +/* nv30_clear.c */ +extern void nv30_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue); + +#endif diff --git a/src/gallium/drivers/nv30/nv30_draw.c b/src/gallium/drivers/nv30/nv30_draw.c new file mode 100644 index 0000000000..74fc138c05 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_draw.c @@ -0,0 +1,61 @@ +#include "draw/draw_pipe.h" + +#include "nv30_context.h" + +struct nv30_draw_stage { + struct draw_stage draw; + struct nv30_context *nv30; +}; + +static void +nv30_draw_point(struct draw_stage *draw, struct prim_header *prim) +{ + NOUVEAU_ERR("\n"); +} + +static void +nv30_draw_line(struct draw_stage *draw, struct prim_header *prim) +{ + NOUVEAU_ERR("\n"); +} + +static void +nv30_draw_tri(struct draw_stage *draw, struct prim_header *prim) +{ + NOUVEAU_ERR("\n"); +} + +static void +nv30_draw_flush(struct draw_stage *draw, unsigned flags) +{ +} + +static void +nv30_draw_reset_stipple_counter(struct draw_stage *draw) +{ + NOUVEAU_ERR("\n"); +} + +static void +nv30_draw_destroy(struct draw_stage *draw) +{ + FREE(draw); +} + +struct draw_stage * +nv30_draw_render_stage(struct nv30_context *nv30) +{ + struct nv30_draw_stage *nv30draw = CALLOC_STRUCT(nv30_draw_stage); + + nv30draw->nv30 = nv30; + nv30draw->draw.draw = nv30->draw; + nv30draw->draw.point = nv30_draw_point; + nv30draw->draw.line = nv30_draw_line; + nv30draw->draw.tri = nv30_draw_tri; + nv30draw->draw.flush = nv30_draw_flush; + nv30draw->draw.reset_stipple_counter = nv30_draw_reset_stipple_counter; + nv30draw->draw.destroy = nv30_draw_destroy; + + return &nv30draw->draw; +} + diff --git a/src/gallium/drivers/nv30/nv30_fragprog.c b/src/gallium/drivers/nv30/nv30_fragprog.c new file mode 100644 index 0000000000..320ba3f4bf --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_fragprog.c @@ -0,0 +1,911 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" + +#include "nv30_context.h" + +#define SWZ_X 0 +#define SWZ_Y 1 +#define SWZ_Z 2 +#define SWZ_W 3 +#define MASK_X 1 +#define MASK_Y 2 +#define MASK_Z 4 +#define MASK_W 8 +#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W) +#define DEF_SCALE NV30_FP_OP_DST_SCALE_1X +#define DEF_CTEST NV30_FP_OP_COND_TR +#include "nv30_shader.h" + +#define swz(s,x,y,z,w) nv30_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w) +#define neg(s) nv30_sr_neg((s)) +#define abs(s) nv30_sr_abs((s)) +#define scale(s,v) nv30_sr_scale((s), NV30_FP_OP_DST_SCALE_##v) + +#define MAX_CONSTS 128 +#define MAX_IMM 32 +struct nv30_fpc { + struct nv30_fragment_program *fp; + + uint attrib_map[PIPE_MAX_SHADER_INPUTS]; + + int high_temp; + int temp_temp_count; + int num_regs; + + uint depth_id; + uint colour_id; + + unsigned inst_offset; + + struct { + int pipe; + float vals[4]; + } consts[MAX_CONSTS]; + int nr_consts; + + struct nv30_sreg imm[MAX_IMM]; + unsigned nr_imm; +}; + +static INLINE struct nv30_sreg +temp(struct nv30_fpc *fpc) +{ + int idx; + + idx = fpc->temp_temp_count++; + idx += fpc->high_temp + 1; + return nv30_sr(NV30SR_TEMP, idx); +} + +static INLINE struct nv30_sreg +constant(struct nv30_fpc *fpc, int pipe, float vals[4]) +{ + int idx; + + if (fpc->nr_consts == MAX_CONSTS) + assert(0); + idx = fpc->nr_consts++; + + fpc->consts[idx].pipe = pipe; + if (pipe == -1) + memcpy(fpc->consts[idx].vals, vals, 4 * sizeof(float)); + return nv30_sr(NV30SR_CONST, idx); +} + +#define arith(cc,s,o,d,m,s0,s1,s2) \ + nv30_fp_arith((cc), (s), NV30_FP_OP_OPCODE_##o, \ + (d), (m), (s0), (s1), (s2)) +#define tex(cc,s,o,u,d,m,s0,s1,s2) \ + nv30_fp_tex((cc), (s), NV30_FP_OP_OPCODE_##o, (u), \ + (d), (m), (s0), none, none) + +static void +grow_insns(struct nv30_fpc *fpc, int size) +{ + struct nv30_fragment_program *fp = fpc->fp; + + fp->insn_len += size; + fp->insn = realloc(fp->insn, sizeof(uint32_t) * fp->insn_len); +} + +static void +emit_src(struct nv30_fpc *fpc, int pos, struct nv30_sreg src) +{ + struct nv30_fragment_program *fp = fpc->fp; + uint32_t *hw = &fp->insn[fpc->inst_offset]; + uint32_t sr = 0; + + switch (src.type) { + case NV30SR_INPUT: + sr |= (NV30_FP_REG_TYPE_INPUT << NV30_FP_REG_TYPE_SHIFT); + hw[0] |= (src.index << NV30_FP_OP_INPUT_SRC_SHIFT); + break; + case NV30SR_OUTPUT: + sr |= NV30_FP_REG_SRC_HALF; + /* fall-through */ + case NV30SR_TEMP: + sr |= (NV30_FP_REG_TYPE_TEMP << NV30_FP_REG_TYPE_SHIFT); + sr |= (src.index << NV30_FP_REG_SRC_SHIFT); + break; + case NV30SR_CONST: + grow_insns(fpc, 4); + hw = &fp->insn[fpc->inst_offset]; + if (fpc->consts[src.index].pipe >= 0) { + struct nv30_fragment_program_data *fpd; + + fp->consts = realloc(fp->consts, ++fp->nr_consts * + sizeof(*fpd)); + fpd = &fp->consts[fp->nr_consts - 1]; + fpd->offset = fpc->inst_offset + 4; + fpd->index = fpc->consts[src.index].pipe; + memset(&fp->insn[fpd->offset], 0, sizeof(uint32_t) * 4); + } else { + memcpy(&fp->insn[fpc->inst_offset + 4], + fpc->consts[src.index].vals, + sizeof(uint32_t) * 4); + } + + sr |= (NV30_FP_REG_TYPE_CONST << NV30_FP_REG_TYPE_SHIFT); + break; + case NV30SR_NONE: + sr |= (NV30_FP_REG_TYPE_INPUT << NV30_FP_REG_TYPE_SHIFT); + break; + default: + assert(0); + } + + if (src.negate) + sr |= NV30_FP_REG_NEGATE; + + if (src.abs) + hw[1] |= (1 << (29 + pos)); + + sr |= ((src.swz[0] << NV30_FP_REG_SWZ_X_SHIFT) | + (src.swz[1] << NV30_FP_REG_SWZ_Y_SHIFT) | + (src.swz[2] << NV30_FP_REG_SWZ_Z_SHIFT) | + (src.swz[3] << NV30_FP_REG_SWZ_W_SHIFT)); + + hw[pos + 1] |= sr; +} + +static void +emit_dst(struct nv30_fpc *fpc, struct nv30_sreg dst) +{ + struct nv30_fragment_program *fp = fpc->fp; + uint32_t *hw = &fp->insn[fpc->inst_offset]; + + switch (dst.type) { + case NV30SR_TEMP: + if (fpc->num_regs < (dst.index + 1)) + fpc->num_regs = dst.index + 1; + break; + case NV30SR_OUTPUT: + if (dst.index == 1) { + fp->fp_control |= 0xe; + } else { + hw[0] |= NV30_FP_OP_OUT_REG_HALF; + } + break; + case NV30SR_NONE: + hw[0] |= (1 << 30); + break; + default: + assert(0); + } + + hw[0] |= (dst.index << NV30_FP_OP_OUT_REG_SHIFT); +} + +static void +nv30_fp_arith(struct nv30_fpc *fpc, int sat, int op, + struct nv30_sreg dst, int mask, + struct nv30_sreg s0, struct nv30_sreg s1, struct nv30_sreg s2) +{ + struct nv30_fragment_program *fp = fpc->fp; + uint32_t *hw; + + fpc->inst_offset = fp->insn_len; + grow_insns(fpc, 4); + hw = &fp->insn[fpc->inst_offset]; + memset(hw, 0, sizeof(uint32_t) * 4); + + if (op == NV30_FP_OP_OPCODE_KIL) + fp->fp_control |= NV34TCL_FP_CONTROL_USES_KIL; + hw[0] |= (op << NV30_FP_OP_OPCODE_SHIFT); + hw[0] |= (mask << NV30_FP_OP_OUTMASK_SHIFT); + hw[2] |= (dst.dst_scale << NV30_FP_OP_DST_SCALE_SHIFT); + + if (sat) + hw[0] |= NV30_FP_OP_OUT_SAT; + + if (dst.cc_update) + hw[0] |= NV30_FP_OP_COND_WRITE_ENABLE; + hw[1] |= (dst.cc_test << NV30_FP_OP_COND_SHIFT); + hw[1] |= ((dst.cc_swz[0] << NV30_FP_OP_COND_SWZ_X_SHIFT) | + (dst.cc_swz[1] << NV30_FP_OP_COND_SWZ_Y_SHIFT) | + (dst.cc_swz[2] << NV30_FP_OP_COND_SWZ_Z_SHIFT) | + (dst.cc_swz[3] << NV30_FP_OP_COND_SWZ_W_SHIFT)); + + emit_dst(fpc, dst); + emit_src(fpc, 0, s0); + emit_src(fpc, 1, s1); + emit_src(fpc, 2, s2); +} + +static void +nv30_fp_tex(struct nv30_fpc *fpc, int sat, int op, int unit, + struct nv30_sreg dst, int mask, + struct nv30_sreg s0, struct nv30_sreg s1, struct nv30_sreg s2) +{ + struct nv30_fragment_program *fp = fpc->fp; + + nv30_fp_arith(fpc, sat, op, dst, mask, s0, s1, s2); + + fp->insn[fpc->inst_offset] |= (unit << NV30_FP_OP_TEX_UNIT_SHIFT); + fp->samplers |= (1 << unit); +} + +static INLINE struct nv30_sreg +tgsi_src(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc) +{ + struct nv30_sreg src; + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + src = nv30_sr(NV30SR_INPUT, + fpc->attrib_map[fsrc->SrcRegister.Index]); + break; + case TGSI_FILE_CONSTANT: + src = constant(fpc, fsrc->SrcRegister.Index, NULL); + break; + case TGSI_FILE_IMMEDIATE: + assert(fsrc->SrcRegister.Index < fpc->nr_imm); + src = fpc->imm[fsrc->SrcRegister.Index]; + break; + case TGSI_FILE_TEMPORARY: + src = nv30_sr(NV30SR_TEMP, fsrc->SrcRegister.Index + 1); + if (fpc->high_temp < src.index) + fpc->high_temp = src.index; + break; + /* This is clearly insane, but gallium hands us shaders like this. + * Luckily fragprog results are just temp regs.. + */ + case TGSI_FILE_OUTPUT: + if (fsrc->SrcRegister.Index == fpc->colour_id) + return nv30_sr(NV30SR_OUTPUT, 0); + else + return nv30_sr(NV30SR_OUTPUT, 1); + break; + default: + NOUVEAU_ERR("bad src file\n"); + break; + } + + src.abs = fsrc->SrcRegisterExtMod.Absolute; + src.negate = fsrc->SrcRegister.Negate; + src.swz[0] = fsrc->SrcRegister.SwizzleX; + src.swz[1] = fsrc->SrcRegister.SwizzleY; + src.swz[2] = fsrc->SrcRegister.SwizzleZ; + src.swz[3] = fsrc->SrcRegister.SwizzleW; + return src; +} + +static INLINE struct nv30_sreg +tgsi_dst(struct nv30_fpc *fpc, const struct tgsi_full_dst_register *fdst) { + int idx; + + switch (fdst->DstRegister.File) { + case TGSI_FILE_OUTPUT: + if (fdst->DstRegister.Index == fpc->colour_id) + return nv30_sr(NV30SR_OUTPUT, 0); + else + return nv30_sr(NV30SR_OUTPUT, 1); + break; + case TGSI_FILE_TEMPORARY: + idx = fdst->DstRegister.Index + 1; + if (fpc->high_temp < idx) + fpc->high_temp = idx; + return nv30_sr(NV30SR_TEMP, idx); + case TGSI_FILE_NULL: + return nv30_sr(NV30SR_NONE, 0); + default: + NOUVEAU_ERR("bad dst file %d\n", fdst->DstRegister.File); + return nv30_sr(NV30SR_NONE, 0); + } +} + +static INLINE int +tgsi_mask(uint tgsi) +{ + int mask = 0; + + if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X; + if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y; + if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z; + if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W; + return mask; +} + +static boolean +src_native_swz(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc, + struct nv30_sreg *src) +{ + const struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0); + struct nv30_sreg tgsi = tgsi_src(fpc, fsrc); + uint mask = 0, zero_mask = 0, one_mask = 0, neg_mask = 0; + uint neg[4] = { fsrc->SrcRegisterExtSwz.NegateX, + fsrc->SrcRegisterExtSwz.NegateY, + fsrc->SrcRegisterExtSwz.NegateZ, + fsrc->SrcRegisterExtSwz.NegateW }; + uint c; + + for (c = 0; c < 4; c++) { + switch (tgsi_util_get_full_src_register_extswizzle(fsrc, c)) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + mask |= (1 << c); + break; + case TGSI_EXTSWIZZLE_ZERO: + zero_mask |= (1 << c); + tgsi.swz[c] = SWZ_X; + break; + case TGSI_EXTSWIZZLE_ONE: + one_mask |= (1 << c); + tgsi.swz[c] = SWZ_X; + break; + default: + assert(0); + } + + if (!tgsi.negate && neg[c]) + neg_mask |= (1 << c); + } + + if (mask == MASK_ALL && !neg_mask) + return TRUE; + + *src = temp(fpc); + + if (mask) + arith(fpc, 0, MOV, *src, mask, tgsi, none, none); + + if (zero_mask) + arith(fpc, 0, SFL, *src, zero_mask, *src, none, none); + + if (one_mask) + arith(fpc, 0, STR, *src, one_mask, *src, none, none); + + if (neg_mask) { + struct nv30_sreg one = temp(fpc); + arith(fpc, 0, STR, one, neg_mask, one, none, none); + arith(fpc, 0, MUL, *src, neg_mask, *src, neg(one), none); + } + + return FALSE; +} + +static boolean +nv30_fragprog_parse_instruction(struct nv30_fpc *fpc, + const struct tgsi_full_instruction *finst) +{ + const struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0); + struct nv30_sreg src[3], dst, tmp; + int mask, sat, unit = 0; + int ai = -1, ci = -1; + int i; + + if (finst->Instruction.Opcode == TGSI_OPCODE_END) + return TRUE; + + fpc->temp_temp_count = 0; + for (i = 0; i < finst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *fsrc; + + fsrc = &finst->FullSrcRegisters[i]; + if (fsrc->SrcRegister.File == TGSI_FILE_TEMPORARY) { + src[i] = tgsi_src(fpc, fsrc); + } + } + + for (i = 0; i < finst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *fsrc; + + fsrc = &finst->FullSrcRegisters[i]; + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + case TGSI_FILE_CONSTANT: + case TGSI_FILE_TEMPORARY: + if (!src_native_swz(fpc, fsrc, &src[i])) + continue; + break; + default: + break; + } + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + if (ai == -1 || ai == fsrc->SrcRegister.Index) { + ai = fsrc->SrcRegister.Index; + src[i] = tgsi_src(fpc, fsrc); + } else { + NOUVEAU_MSG("extra src attr %d\n", + fsrc->SrcRegister.Index); + src[i] = temp(fpc); + arith(fpc, 0, MOV, src[i], MASK_ALL, + tgsi_src(fpc, fsrc), none, none); + } + break; + case TGSI_FILE_CONSTANT: + case TGSI_FILE_IMMEDIATE: + if (ci == -1 || ci == fsrc->SrcRegister.Index) { + ci = fsrc->SrcRegister.Index; + src[i] = tgsi_src(fpc, fsrc); + } else { + src[i] = temp(fpc); + arith(fpc, 0, MOV, src[i], MASK_ALL, + tgsi_src(fpc, fsrc), none, none); + } + break; + case TGSI_FILE_TEMPORARY: + /* handled above */ + break; + case TGSI_FILE_SAMPLER: + unit = fsrc->SrcRegister.Index; + break; + case TGSI_FILE_OUTPUT: + break; + default: + NOUVEAU_ERR("bad src file\n"); + return FALSE; + } + } + + dst = tgsi_dst(fpc, &finst->FullDstRegisters[0]); + mask = tgsi_mask(finst->FullDstRegisters[0].DstRegister.WriteMask); + sat = (finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE); + + switch (finst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + arith(fpc, sat, MOV, dst, mask, abs(src[0]), none, none); + break; + case TGSI_OPCODE_ADD: + arith(fpc, sat, ADD, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_CMP: + tmp = temp(fpc); + arith(fpc, sat, MOV, dst, mask, src[2], none, none); + tmp.cc_update = 1; + arith(fpc, 0, MOV, tmp, 0xf, src[0], none, none); + dst.cc_test = NV30_VP_INST_COND_LT; + arith(fpc, sat, MOV, dst, mask, src[1], none, none); + break; + case TGSI_OPCODE_COS: + arith(fpc, sat, COS, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_DP3: + arith(fpc, sat, DP3, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DP4: + arith(fpc, sat, DP4, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DPH: + tmp = temp(fpc); + arith(fpc, 0, DP3, tmp, MASK_X, src[0], src[1], none); + arith(fpc, sat, ADD, dst, mask, swz(tmp, X, X, X, X), + swz(src[1], W, W, W, W), none); + break; + case TGSI_OPCODE_DST: + arith(fpc, sat, DST, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_EX2: + arith(fpc, sat, EX2, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_FLR: + arith(fpc, sat, FLR, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_FRC: + arith(fpc, sat, FRC, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_KILP: + arith(fpc, 0, KIL, none, 0, none, none, none); + break; + case TGSI_OPCODE_KIL: + dst = nv30_sr(NV30SR_NONE, 0); + dst.cc_update = 1; + arith(fpc, 0, MOV, dst, MASK_ALL, src[0], none, none); + dst.cc_update = 0; dst.cc_test = NV30_FP_OP_COND_LT; + arith(fpc, 0, KIL, dst, 0, none, none, none); + break; + case TGSI_OPCODE_LG2: + arith(fpc, sat, LG2, dst, mask, src[0], none, none); + break; +// case TGSI_OPCODE_LIT: + case TGSI_OPCODE_LRP: + arith(fpc, sat, LRP, dst, mask, src[0], src[1], src[2]); + break; + case TGSI_OPCODE_MAD: + arith(fpc, sat, MAD, dst, mask, src[0], src[1], src[2]); + break; + case TGSI_OPCODE_MAX: + arith(fpc, sat, MAX, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_MIN: + arith(fpc, sat, MIN, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_MOV: + arith(fpc, sat, MOV, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_MUL: + arith(fpc, sat, MUL, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_NOISE1: + case TGSI_OPCODE_NOISE2: + case TGSI_OPCODE_NOISE3: + case TGSI_OPCODE_NOISE4: + arith(fpc, sat, SFL, dst, mask, none, none, none); + break; + case TGSI_OPCODE_POW: + arith(fpc, sat, POW, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_RCP: + arith(fpc, sat, RCP, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_RET: + assert(0); + break; + case TGSI_OPCODE_RFL: + arith(fpc, 0, RFL, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_RSQ: + arith(fpc, sat, RSQ, dst, mask, abs(swz(src[0], X, X, X, X)), none, none); + break; + case TGSI_OPCODE_SCS: + if (mask & MASK_X) { + arith(fpc, sat, COS, dst, MASK_X, + swz(src[0], X, X, X, X), none, none); + } + if (mask & MASK_Y) { + arith(fpc, sat, SIN, dst, MASK_Y, + swz(src[0], X, X, X, X), none, none); + } + break; + case TGSI_OPCODE_SIN: + arith(fpc, sat, SIN, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_SGE: + arith(fpc, sat, SGE, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SGT: + arith(fpc, sat, SGT, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SLT: + arith(fpc, sat, SLT, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SUB: + arith(fpc, sat, ADD, dst, mask, src[0], neg(src[1]), none); + break; + case TGSI_OPCODE_TEX: + tex(fpc, sat, TEX, unit, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_TXB: + tex(fpc, sat, TXB, unit, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_TXP: + tex(fpc, sat, TXP, unit, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_XPD: + tmp = temp(fpc); + arith(fpc, 0, MUL, tmp, mask, + swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none); + arith(fpc, sat, MAD, dst, (mask & ~MASK_W), + swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y), + neg(tmp)); + break; + default: + NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode); + return FALSE; + } + + return TRUE; +} + +static boolean +nv30_fragprog_parse_decl_attrib(struct nv30_fpc *fpc, + const struct tgsi_full_declaration *fdec) +{ + int hw; + + switch (fdec->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: + hw = NV30_FP_OP_INPUT_SRC_POSITION; + break; + case TGSI_SEMANTIC_COLOR: + if (fdec->Semantic.SemanticIndex == 0) { + hw = NV30_FP_OP_INPUT_SRC_COL0; + } else + if (fdec->Semantic.SemanticIndex == 1) { + hw = NV30_FP_OP_INPUT_SRC_COL1; + } else { + NOUVEAU_ERR("bad colour semantic index\n"); + return FALSE; + } + break; + case TGSI_SEMANTIC_FOG: + hw = NV30_FP_OP_INPUT_SRC_FOGC; + break; + case TGSI_SEMANTIC_GENERIC: + if (fdec->Semantic.SemanticIndex <= 7) { + hw = NV30_FP_OP_INPUT_SRC_TC(fdec->Semantic. + SemanticIndex); + } else { + NOUVEAU_ERR("bad generic semantic index\n"); + return FALSE; + } + break; + default: + NOUVEAU_ERR("bad input semantic\n"); + return FALSE; + } + + fpc->attrib_map[fdec->DeclarationRange.First] = hw; + return TRUE; +} + +static boolean +nv30_fragprog_parse_decl_output(struct nv30_fpc *fpc, + const struct tgsi_full_declaration *fdec) +{ + switch (fdec->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: + fpc->depth_id = fdec->DeclarationRange.First; + break; + case TGSI_SEMANTIC_COLOR: + fpc->colour_id = fdec->DeclarationRange.First; + break; + default: + NOUVEAU_ERR("bad output semantic\n"); + return FALSE; + } + + return TRUE; +} + +static boolean +nv30_fragprog_prepare(struct nv30_fpc *fpc) +{ + struct tgsi_parse_context p; + /*int high_temp = -1, i;*/ + + tgsi_parse_init(&p, fpc->fp->pipe.tokens); + while (!tgsi_parse_end_of_tokens(&p)) { + const union tgsi_full_token *tok = &p.FullToken; + + tgsi_parse_token(&p); + switch(tok->Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + const struct tgsi_full_declaration *fdec; + fdec = &p.FullToken.FullDeclaration; + switch (fdec->Declaration.File) { + case TGSI_FILE_INPUT: + if (!nv30_fragprog_parse_decl_attrib(fpc, fdec)) + goto out_err; + break; + case TGSI_FILE_OUTPUT: + if (!nv30_fragprog_parse_decl_output(fpc, fdec)) + goto out_err; + break; + /*case TGSI_FILE_TEMPORARY: + if (fdec->DeclarationRange.Last > high_temp) { + high_temp = + fdec->DeclarationRange.Last; + } + break;*/ + default: + break; + } + } + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *imm; + float vals[4]; + + imm = &p.FullToken.FullImmediate; + assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32); + assert(fpc->nr_imm < MAX_IMM); + + vals[0] = imm->u.ImmediateFloat32[0].Float; + vals[1] = imm->u.ImmediateFloat32[1].Float; + vals[2] = imm->u.ImmediateFloat32[2].Float; + vals[3] = imm->u.ImmediateFloat32[3].Float; + fpc->imm[fpc->nr_imm++] = constant(fpc, -1, vals); + } + break; + default: + break; + } + } + tgsi_parse_free(&p); + + /*if (++high_temp) { + fpc->r_temp = CALLOC(high_temp, sizeof(struct nv30_sreg)); + for (i = 0; i < high_temp; i++) + fpc->r_temp[i] = temp(fpc); + fpc->r_temps_discard = 0; + }*/ + + return TRUE; + +out_err: + /*if (fpc->r_temp) + FREE(fpc->r_temp);*/ + tgsi_parse_free(&p); + return FALSE; +} + +static void +nv30_fragprog_translate(struct nv30_context *nv30, + struct nv30_fragment_program *fp) +{ + struct tgsi_parse_context parse; + struct nv30_fpc *fpc = NULL; + + tgsi_dump(fp->pipe.tokens,0); + + fpc = CALLOC(1, sizeof(struct nv30_fpc)); + if (!fpc) + return; + fpc->fp = fp; + fpc->high_temp = -1; + fpc->num_regs = 2; + + if (!nv30_fragprog_prepare(fpc)) { + FREE(fpc); + return; + } + + tgsi_parse_init(&parse, fp->pipe.tokens); + + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + const struct tgsi_full_instruction *finst; + + finst = &parse.FullToken.FullInstruction; + if (!nv30_fragprog_parse_instruction(fpc, finst)) + goto out_err; + } + break; + default: + break; + } + } + + fp->fp_control |= (fpc->num_regs-1)/2; + fp->fp_reg_control = (1<<16)|0x4; + + /* Terminate final instruction */ + fp->insn[fpc->inst_offset] |= 0x00000001; + + /* Append NOP + END instruction, may or may not be necessary. */ + fpc->inst_offset = fp->insn_len; + grow_insns(fpc, 4); + fp->insn[fpc->inst_offset + 0] = 0x00000001; + fp->insn[fpc->inst_offset + 1] = 0x00000000; + fp->insn[fpc->inst_offset + 2] = 0x00000000; + fp->insn[fpc->inst_offset + 3] = 0x00000000; + + fp->translated = TRUE; + fp->on_hw = FALSE; +out_err: + tgsi_parse_free(&parse); + FREE(fpc); +} + +static void +nv30_fragprog_upload(struct nv30_context *nv30, + struct nv30_fragment_program *fp) +{ + struct pipe_winsys *ws = nv30->pipe.winsys; + const uint32_t le = 1; + uint32_t *map; + int i; + + map = ws->buffer_map(ws, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); + +#if 0 + for (i = 0; i < fp->insn_len; i++) { + fflush(stdout); fflush(stderr); + NOUVEAU_ERR("%d 0x%08x\n", i, fp->insn[i]); + fflush(stdout); fflush(stderr); + } +#endif + + if ((*(const uint8_t *)&le)) { + for (i = 0; i < fp->insn_len; i++) { + map[i] = fp->insn[i]; + } + } else { + /* Weird swapping for big-endian chips */ + for (i = 0; i < fp->insn_len; i++) { + map[i] = ((fp->insn[i] & 0xffff) << 16) | + ((fp->insn[i] >> 16) & 0xffff); + } + } + + ws->buffer_unmap(ws, fp->buffer); +} + +static boolean +nv30_fragprog_validate(struct nv30_context *nv30) +{ + struct nv30_fragment_program *fp = nv30->fragprog; + struct pipe_buffer *constbuf = + nv30->constbuf[PIPE_SHADER_FRAGMENT]; + struct pipe_winsys *ws = nv30->pipe.winsys; + struct nouveau_stateobj *so; + boolean new_consts = FALSE; + int i; + + if (fp->translated) + goto update_constants; + + /*nv30->fallback_swrast &= ~NV30_NEW_FRAGPROG;*/ + nv30_fragprog_translate(nv30, fp); + if (!fp->translated) { + /*nv30->fallback_swrast |= NV30_NEW_FRAGPROG;*/ + return FALSE; + } + + fp->buffer = ws->buffer_create(ws, 0x100, 0, fp->insn_len * 4); + nv30_fragprog_upload(nv30, fp); + + so = so_new(8, 1); + so_method(so, nv30->screen->rankine, NV34TCL_FP_ACTIVE_PROGRAM, 1); + so_reloc (so, fp->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | + NOUVEAU_BO_RD | NOUVEAU_BO_LOW | NOUVEAU_BO_OR, + NV34TCL_FP_ACTIVE_PROGRAM_DMA0, NV34TCL_FP_ACTIVE_PROGRAM_DMA1); + so_method(so, nv30->screen->rankine, NV34TCL_FP_CONTROL, 1); + so_data (so, fp->fp_control); + so_method(so, nv30->screen->rankine, NV34TCL_FP_REG_CONTROL, 1); + so_data (so, fp->fp_reg_control); + so_method(so, nv30->screen->rankine, NV34TCL_TX_UNITS_ENABLE, 1); + so_data (so, fp->samplers); + so_ref(so, &fp->so); + +update_constants: + if (fp->nr_consts) { + float *map; + + map = ws->buffer_map(ws, constbuf, PIPE_BUFFER_USAGE_CPU_READ); + for (i = 0; i < fp->nr_consts; i++) { + struct nv30_fragment_program_data *fpd = &fp->consts[i]; + uint32_t *p = &fp->insn[fpd->offset]; + uint32_t *cb = (uint32_t *)&map[fpd->index * 4]; + + if (!memcmp(p, cb, 4 * sizeof(float))) + continue; + memcpy(p, cb, 4 * sizeof(float)); + new_consts = TRUE; + } + ws->buffer_unmap(ws, constbuf); + + if (new_consts) + nv30_fragprog_upload(nv30, fp); + } + + if (new_consts || fp->so != nv30->state.hw[NV30_STATE_FRAGPROG]) { + so_ref(fp->so, &nv30->state.hw[NV30_STATE_FRAGPROG]); + return TRUE; + } + + return FALSE; +} + +void +nv30_fragprog_destroy(struct nv30_context *nv30, + struct nv30_fragment_program *fp) +{ + if (fp->insn_len) + FREE(fp->insn); +} + +struct nv30_state_entry nv30_state_fragprog = { + .validate = nv30_fragprog_validate, + .dirty = { + .pipe = NV30_NEW_FRAGPROG, + .hw = NV30_STATE_FRAGPROG + } +}; diff --git a/src/gallium/drivers/nv30/nv30_fragtex.c b/src/gallium/drivers/nv30/nv30_fragtex.c new file mode 100644 index 0000000000..b1d2663af3 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_fragtex.c @@ -0,0 +1,163 @@ +#include "nv30_context.h" +#include "nouveau/nouveau_util.h" + +#define _(m,tf,ts0x,ts0y,ts0z,ts0w,ts1x,ts1y,ts1z,ts1w) \ +{ \ + TRUE, \ + PIPE_FORMAT_##m, \ + NV34TCL_TX_FORMAT_FORMAT_##tf, \ + (NV34TCL_TX_SWIZZLE_S0_X_##ts0x | NV34TCL_TX_SWIZZLE_S0_Y_##ts0y | \ + NV34TCL_TX_SWIZZLE_S0_Z_##ts0z | NV34TCL_TX_SWIZZLE_S0_W_##ts0w | \ + NV34TCL_TX_SWIZZLE_S1_X_##ts1x | NV34TCL_TX_SWIZZLE_S1_Y_##ts1y | \ + NV34TCL_TX_SWIZZLE_S1_Z_##ts1z | NV34TCL_TX_SWIZZLE_S1_W_##ts1w) \ +} + +struct nv30_texture_format { + boolean defined; + uint pipe; + int format; + int swizzle; +}; + +static struct nv30_texture_format +nv30_texture_formats[] = { + _(A8R8G8B8_UNORM, A8R8G8B8, S1, S1, S1, S1, X, Y, Z, W), + _(A1R5G5B5_UNORM, A1R5G5B5, S1, S1, S1, S1, X, Y, Z, W), + _(A4R4G4B4_UNORM, A4R4G4B4, S1, S1, S1, S1, X, Y, Z, W), + _(R5G6B5_UNORM , R5G6B5 , S1, S1, S1, ONE, X, Y, Z, W), + _(L8_UNORM , L8 , S1, S1, S1, ONE, X, X, X, X), + _(A8_UNORM , L8 , ZERO, ZERO, ZERO, S1, X, X, X, X), + _(I8_UNORM , L8 , S1, S1, S1, S1, X, X, X, X), + _(A8L8_UNORM , A8L8 , S1, S1, S1, S1, X, X, X, Y), +// _(Z16_UNORM , Z16 , S1, S1, S1, ONE, X, X, X, X), +// _(Z24S8_UNORM , Z24 , S1, S1, S1, ONE, X, X, X, X), + _(DXT1_RGB , DXT1 , S1, S1, S1, ONE, X, Y, Z, W), + _(DXT1_RGBA , DXT1 , S1, S1, S1, S1, X, Y, Z, W), + _(DXT3_RGBA , DXT3 , S1, S1, S1, S1, X, Y, Z, W), + _(DXT5_RGBA , DXT5 , S1, S1, S1, S1, X, Y, Z, W), + {}, +}; + +static struct nv30_texture_format * +nv30_fragtex_format(uint pipe_format) +{ + struct nv30_texture_format *tf = nv30_texture_formats; + char fs[128]; + + while (tf->defined) { + if (tf->pipe == pipe_format) + return tf; + tf++; + } + + NOUVEAU_ERR("unknown texture format %s\n", pf_name(pipe_format)); + return NULL; +} + + +static struct nouveau_stateobj * +nv30_fragtex_build(struct nv30_context *nv30, int unit) +{ + struct nv30_sampler_state *ps = nv30->tex_sampler[unit]; + struct nv30_miptree *nv30mt = nv30->tex_miptree[unit]; + struct pipe_texture *pt = &nv30mt->base; + struct nv30_texture_format *tf; + struct nouveau_stateobj *so; + uint32_t txf, txs , txp; + unsigned tex_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD; + + tf = nv30_fragtex_format(pt->format); + if (!tf) + assert(0); + + txf = tf->format; + txf |= ((pt->last_level>0) ? NV34TCL_TX_FORMAT_MIPMAP : 0); + txf |= log2i(pt->width[0]) << 20; + txf |= log2i(pt->height[0]) << 24; + txf |= log2i(pt->depth[0]) << 28; + txf |= NV34TCL_TX_FORMAT_NO_BORDER | 0x10000; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE: + txf |= NV34TCL_TX_FORMAT_CUBIC; + /* fall-through */ + case PIPE_TEXTURE_2D: + txf |= NV34TCL_TX_FORMAT_DIMS_2D; + break; + case PIPE_TEXTURE_3D: + txf |= NV34TCL_TX_FORMAT_DIMS_3D; + break; + case PIPE_TEXTURE_1D: + txf |= NV34TCL_TX_FORMAT_DIMS_1D; + break; + default: + NOUVEAU_ERR("Unknown target %d\n", pt->target); + return NULL; + } + + if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + txp = 0; + } else { + txp = nv30mt->level[0].pitch; + txf |= (1<<13) /*FIXME: NV34TCL_TX_FORMAT_LINEAR ? */; + } + + txs = tf->swizzle; + + so = so_new(16, 2); + so_method(so, nv30->screen->rankine, NV34TCL_TX_OFFSET(unit), 8); + so_reloc (so, nv30mt->buffer, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, nv30mt->buffer, txf, tex_flags | NOUVEAU_BO_OR, + NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1); + so_data (so, ps->wrap); + so_data (so, NV34TCL_TX_ENABLE_ENABLE | ps->en); + so_data (so, txs); + so_data (so, ps->filt | 0x2000 /*voodoo*/); + so_data (so, (pt->width[0] << NV34TCL_TX_NPOT_SIZE_W_SHIFT) | + pt->height[0]); + so_data (so, ps->bcol); + + return so; +} + +static boolean +nv30_fragtex_validate(struct nv30_context *nv30) +{ + struct nv30_fragment_program *fp = nv30->fragprog; + struct nv30_state *state = &nv30->state; + struct nouveau_stateobj *so; + unsigned samplers, unit; + + samplers = state->fp_samplers & ~fp->samplers; + while (samplers) { + unit = ffs(samplers) - 1; + samplers &= ~(1 << unit); + + so = so_new(2, 0); + so_method(so, nv30->screen->rankine, NV34TCL_TX_ENABLE(unit), 1); + so_data (so, 0); + so_ref(so, &nv30->state.hw[NV30_STATE_FRAGTEX0 + unit]); + state->dirty |= (1ULL << (NV30_STATE_FRAGTEX0 + unit)); + } + + samplers = nv30->dirty_samplers & fp->samplers; + while (samplers) { + unit = ffs(samplers) - 1; + samplers &= ~(1 << unit); + + so = nv30_fragtex_build(nv30, unit); + so_ref(so, &nv30->state.hw[NV30_STATE_FRAGTEX0 + unit]); + state->dirty |= (1ULL << (NV30_STATE_FRAGTEX0 + unit)); + } + + nv30->state.fp_samplers = fp->samplers; + return FALSE; +} + +struct nv30_state_entry nv30_state_fragtex = { + .validate = nv30_fragtex_validate, + .dirty = { + .pipe = NV30_NEW_SAMPLER | NV30_NEW_FRAGPROG, + .hw = 0 + } +}; diff --git a/src/gallium/drivers/nv30/nv30_miptree.c b/src/gallium/drivers/nv30/nv30_miptree.c new file mode 100644 index 0000000000..aa670b9a45 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_miptree.c @@ -0,0 +1,195 @@ +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "nv30_context.h" + +static void +nv30_miptree_layout(struct nv30_miptree *nv30mt) +{ + struct pipe_texture *pt = &nv30mt->base; + boolean swizzled = FALSE; + uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; + uint offset = 0; + int nr_faces, l, f, pitch; + + if (pt->target == PIPE_TEXTURE_CUBE) { + nr_faces = 6; + } else + if (pt->target == PIPE_TEXTURE_3D) { + nr_faces = pt->depth[0]; + } else { + nr_faces = 1; + } + + pitch = pt->width[0]; + for (l = 0; l <= pt->last_level; l++) { + pt->width[l] = width; + pt->height[l] = height; + pt->depth[l] = depth; + pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); + pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); + + if (swizzled) + pitch = pt->nblocksx[l]; + pitch = align(pitch, 64); + + nv30mt->level[l].pitch = pitch * pt->block.size; + nv30mt->level[l].image_offset = + CALLOC(nr_faces, sizeof(unsigned)); + + width = MAX2(1, width >> 1); + height = MAX2(1, height >> 1); + depth = MAX2(1, depth >> 1); + } + + for (f = 0; f < nr_faces; f++) { + for (l = 0; l <= pt->last_level; l++) { + nv30mt->level[l].image_offset[f] = offset; + offset += nv30mt->level[l].pitch * pt->height[l]; + } + } + + nv30mt->total_size = offset; +} + +static struct pipe_texture * +nv30_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) +{ + struct pipe_winsys *ws = pscreen->winsys; + struct nv30_miptree *mt; + + mt = MALLOC(sizeof(struct nv30_miptree)); + if (!mt) + return NULL; + mt->base = *pt; + mt->base.refcount = 1; + mt->base.screen = pscreen; + mt->shadow_tex = NULL; + mt->shadow_surface = NULL; + + /* Swizzled textures must be POT */ + if (pt->width[0] & (pt->width[0] - 1) || + pt->height[0] & (pt->height[0] - 1)) + mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; + else + if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY | + PIPE_TEXTURE_USAGE_DISPLAY_TARGET)) + mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; + else { + switch (pt->format) { + /* TODO: Figure out which formats can be swizzled */ + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_X8R8G8B8_UNORM: + /* XXX: Re-enable when SIFM size limits are fixed */ + /*case PIPE_FORMAT_R16_SNORM:*/ + break; + default: + mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; + } + } + + nv30_miptree_layout(mt); + + mt->buffer = ws->buffer_create(ws, 256, + PIPE_BUFFER_USAGE_PIXEL | + NOUVEAU_BUFFER_USAGE_TEXTURE, + mt->total_size); + if (!mt->buffer) { + FREE(mt); + return NULL; + } + + return &mt->base; +} + +static void +nv30_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt) +{ + struct pipe_texture *pt = *ppt; + struct nv30_miptree *mt = (struct nv30_miptree *)pt; + int l; + + *ppt = NULL; + if (--pt->refcount) + return; + + pipe_buffer_reference(pscreen, &mt->buffer, NULL); + for (l = 0; l <= pt->last_level; l++) { + if (mt->level[l].image_offset) + FREE(mt->level[l].image_offset); + } + + if (mt->shadow_tex) { + assert(mt->shadow_surface); + pscreen->tex_surface_release(pscreen, &mt->shadow_surface); + nv30_miptree_release(pscreen, &mt->shadow_tex); + } + + FREE(mt); +} + +static struct pipe_surface * +nv30_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + unsigned flags) +{ + struct nv30_miptree *nv30mt = (struct nv30_miptree *)pt; + struct pipe_surface *ps; + + ps = CALLOC_STRUCT(pipe_surface); + if (!ps) + return NULL; + pipe_texture_reference(&ps->texture, pt); + pipe_buffer_reference(pscreen, &ps->buffer, nv30mt->buffer); + ps->format = pt->format; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->block = pt->block; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = nv30mt->level[level].pitch; + ps->usage = flags; + ps->status = PIPE_SURFACE_STATUS_DEFINED; + ps->refcount = 1; + ps->winsys = pscreen->winsys; + ps->face = face; + ps->level = level; + ps->zslice = zslice; + + if (pt->target == PIPE_TEXTURE_CUBE) { + ps->offset = nv30mt->level[level].image_offset[face]; + } else + if (pt->target == PIPE_TEXTURE_3D) { + ps->offset = nv30mt->level[level].image_offset[zslice]; + } else { + ps->offset = nv30mt->level[level].image_offset[0]; + } + + return ps; +} + +static void +nv30_miptree_surface_del(struct pipe_screen *pscreen, + struct pipe_surface **psurface) +{ + struct pipe_surface *ps = *psurface; + + *psurface = NULL; + if (--ps->refcount > 0) + return; + + pipe_texture_reference(&ps->texture, NULL); + pipe_buffer_reference(pscreen->winsys, &ps->buffer, NULL); + FREE(ps); +} + +void +nv30_screen_init_miptree_functions(struct pipe_screen *pscreen) +{ + pscreen->texture_create = nv30_miptree_create; + pscreen->texture_release = nv30_miptree_release; + pscreen->get_tex_surface = nv30_miptree_surface_new; + pscreen->tex_surface_release = nv30_miptree_surface_del; +} + diff --git a/src/gallium/drivers/nv30/nv30_query.c b/src/gallium/drivers/nv30/nv30_query.c new file mode 100644 index 0000000000..d40d75f264 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_query.c @@ -0,0 +1,122 @@ +#include "pipe/p_context.h" + +#include "nv30_context.h" + +struct nv30_query { + struct nouveau_resource *object; + unsigned type; + boolean ready; + uint64_t result; +}; + +static INLINE struct nv30_query * +nv30_query(struct pipe_query *pipe) +{ + return (struct nv30_query *)pipe; +} + +static struct pipe_query * +nv30_query_create(struct pipe_context *pipe, unsigned query_type) +{ + struct nv30_query *q; + + q = CALLOC(1, sizeof(struct nv30_query)); + q->type = query_type; + + return (struct pipe_query *)q; +} + +static void +nv30_query_destroy(struct pipe_context *pipe, struct pipe_query *pq) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nv30_query *q = nv30_query(pq); + + if (q->object) + nv30->nvws->res_free(&q->object); + FREE(q); +} + +static void +nv30_query_begin(struct pipe_context *pipe, struct pipe_query *pq) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nv30_query *q = nv30_query(pq); + + assert(q->type == PIPE_QUERY_OCCLUSION_COUNTER); + + /* Happens when end_query() is called, then another begin_query() + * without querying the result in-between. For now we'll wait for + * the existing query to notify completion, but it could be better. + */ + if (q->object) { + uint64 tmp; + pipe->get_query_result(pipe, pq, 1, &tmp); + } + + if (nv30->nvws->res_alloc(nv30->screen->query_heap, 1, NULL, &q->object)) + assert(0); + nv30->nvws->notifier_reset(nv30->screen->query, q->object->start); + + BEGIN_RING(rankine, NV34TCL_QUERY_RESET, 1); + OUT_RING (1); + BEGIN_RING(rankine, NV34TCL_QUERY_UNK17CC, 1); + OUT_RING (1); + + q->ready = FALSE; +} + +static void +nv30_query_end(struct pipe_context *pipe, struct pipe_query *pq) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nv30_query *q = nv30_query(pq); + + BEGIN_RING(rankine, NV34TCL_QUERY_GET, 1); + OUT_RING ((0x01 << NV34TCL_QUERY_GET_UNK24_SHIFT) | + ((q->object->start * 32) << NV34TCL_QUERY_GET_OFFSET_SHIFT)); + FIRE_RING(NULL); +} + +static boolean +nv30_query_result(struct pipe_context *pipe, struct pipe_query *pq, + boolean wait, uint64 *result) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nv30_query *q = nv30_query(pq); + struct nouveau_winsys *nvws = nv30->nvws; + + assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER); + + if (!q->ready) { + unsigned status; + + status = nvws->notifier_status(nv30->screen->query, + q->object->start); + if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) { + if (wait == FALSE) + return FALSE; + nvws->notifier_wait(nv30->screen->query, q->object->start, + NV_NOTIFY_STATE_STATUS_COMPLETED, + 0); + } + + q->result = nvws->notifier_retval(nv30->screen->query, + q->object->start); + q->ready = TRUE; + nvws->res_free(&q->object); + } + + *result = q->result; + return TRUE; +} + +void +nv30_init_query_functions(struct nv30_context *nv30) +{ + nv30->pipe.create_query = nv30_query_create; + nv30->pipe.destroy_query = nv30_query_destroy; + nv30->pipe.begin_query = nv30_query_begin; + nv30->pipe.end_query = nv30_query_end; + nv30->pipe.get_query_result = nv30_query_result; +} diff --git a/src/gallium/drivers/nv30/nv30_screen.c b/src/gallium/drivers/nv30/nv30_screen.c new file mode 100644 index 0000000000..d754892299 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_screen.c @@ -0,0 +1,385 @@ +#include "pipe/p_screen.h" + +#include "nv30_context.h" +#include "nv30_screen.h" + +#define NV30TCL_CHIPSET_3X_MASK 0x00000003 +#define NV34TCL_CHIPSET_3X_MASK 0x00000010 +#define NV35TCL_CHIPSET_3X_MASK 0x000001e0 + +static const char * +nv30_screen_get_name(struct pipe_screen *pscreen) +{ + struct nv30_screen *screen = nv30_screen(pscreen); + struct nouveau_device *dev = screen->nvws->channel->device; + static char buffer[128]; + + snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); + return buffer; +} + +static const char * +nv30_screen_get_vendor(struct pipe_screen *pscreen) +{ + return "nouveau"; +} + +static int +nv30_screen_get_param(struct pipe_screen *pscreen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 16; + case PIPE_CAP_NPOT_TEXTURES: + return 0; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 1; + case PIPE_CAP_POINT_SPRITE: + return 1; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 2; + case PIPE_CAP_OCCLUSION_QUERY: + return 1; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 13; + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 10; + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 13; + case PIPE_CAP_TEXTURE_MIRROR_CLAMP: + return 0; + case PIPE_CAP_TEXTURE_MIRROR_REPEAT: + return 1; + case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: + return 0; + case NOUVEAU_CAP_HW_VTXBUF: + case NOUVEAU_CAP_HW_IDXBUF: + return 1; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0; + } +} + +static float +nv30_screen_get_paramf(struct pipe_screen *pscreen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 10.0; + case PIPE_CAP_MAX_POINT_WIDTH: + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 64.0; + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 8.0; + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 4.0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0.0; + } +} + +static boolean +nv30_screen_surface_format_supported(struct pipe_screen *pscreen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, unsigned geom_flags) +{ + if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z16_UNORM: + return TRUE; + default: + break; + } + } else { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_A1R5G5B5_UNORM: + case PIPE_FORMAT_A4R4G4B4_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_A8_UNORM: + case PIPE_FORMAT_I8_UNORM: + case PIPE_FORMAT_A8L8_UNORM: + case PIPE_FORMAT_Z16_UNORM: + case PIPE_FORMAT_Z24S8_UNORM: + return TRUE; + default: + break; + } + } + + return FALSE; +} + +static void * +nv30_surface_map(struct pipe_screen *screen, struct pipe_surface *surface, + unsigned flags ) +{ + struct pipe_winsys *ws = screen->winsys; + struct pipe_surface *surface_to_map; + void *map; + + if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + struct nv30_miptree *mt = (struct nv30_miptree *)surface->texture; + + if (!mt->shadow_tex) { + unsigned old_tex_usage = surface->texture->tex_usage; + surface->texture->tex_usage = NOUVEAU_TEXTURE_USAGE_LINEAR; + mt->shadow_tex = screen->texture_create(screen, surface->texture); + surface->texture->tex_usage = old_tex_usage; + + assert(mt->shadow_tex->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR); + mt->shadow_surface = screen->get_tex_surface + ( + screen, mt->shadow_tex, + surface->face, surface->level, surface->zslice, + surface->usage + ); + } + + surface_to_map = mt->shadow_surface; + } + else + surface_to_map = surface; + + assert(surface_to_map); + + map = ws->buffer_map(ws, surface_to_map->buffer, flags); + if (!map) + return NULL; + + return map + surface_to_map->offset; +} + +static void +nv30_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) +{ + struct pipe_winsys *ws = screen->winsys; + struct pipe_surface *surface_to_unmap; + + /* TODO: Copy from shadow just before push buffer is flushed instead. + There are probably some programs that map/unmap excessively + before rendering. */ + if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + struct nv30_miptree *mt = (struct nv30_miptree *)surface->texture; + + assert(mt->shadow_tex); + + surface_to_unmap = mt->shadow_surface; + } + else + surface_to_unmap = surface; + + assert(surface_to_unmap); + + ws->buffer_unmap(ws, surface_to_unmap->buffer); + + if (surface_to_unmap != surface) { + struct nv30_screen *nvscreen = nv30_screen(screen); + + nvscreen->nvws->surface_copy(nvscreen->nvws, + surface, 0, 0, + surface_to_unmap, 0, 0, + surface->width, surface->height); + } +} + +static void +nv30_screen_destroy(struct pipe_screen *pscreen) +{ + struct nv30_screen *screen = nv30_screen(pscreen); + struct nouveau_winsys *nvws = screen->nvws; + + nvws->res_free(&screen->vp_exec_heap); + nvws->res_free(&screen->vp_data_heap); + nvws->res_free(&screen->query_heap); + nvws->notifier_free(&screen->query); + nvws->notifier_free(&screen->sync); + nvws->grobj_free(&screen->rankine); + + FREE(pscreen); +} + +struct pipe_screen * +nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +{ + struct nv30_screen *screen = CALLOC_STRUCT(nv30_screen); + struct nouveau_stateobj *so; + unsigned rankine_class = 0; + unsigned chipset = nvws->channel->device->chipset; + int ret, i; + + if (!screen) + return NULL; + screen->nvws = nvws; + + /* 3D object */ + switch (chipset & 0xf0) { + case 0x30: + if (NV30TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f))) + rankine_class = 0x0397; + else + if (NV34TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f))) + rankine_class = 0x0697; + else + if (NV35TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f))) + rankine_class = 0x0497; + break; + default: + break; + } + + if (!rankine_class) { + NOUVEAU_ERR("Unknown nv3x chipset: nv%02x\n", chipset); + return NULL; + } + + ret = nvws->grobj_alloc(nvws, rankine_class, &screen->rankine); + if (ret) { + NOUVEAU_ERR("Error creating 3D object: %d\n", ret); + return FALSE; + } + + /* Notifier for sync purposes */ + ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + if (ret) { + NOUVEAU_ERR("Error creating notifier object: %d\n", ret); + nv30_screen_destroy(&screen->pipe); + return NULL; + } + + /* Query objects */ + ret = nvws->notifier_alloc(nvws, 32, &screen->query); + if (ret) { + NOUVEAU_ERR("Error initialising query objects: %d\n", ret); + nv30_screen_destroy(&screen->pipe); + return NULL; + } + + ret = nvws->res_init(&screen->query_heap, 0, 32); + if (ret) { + NOUVEAU_ERR("Error initialising query object heap: %d\n", ret); + nv30_screen_destroy(&screen->pipe); + return NULL; + } + + /* Vtxprog resources */ + if (nvws->res_init(&screen->vp_exec_heap, 0, 256) || + nvws->res_init(&screen->vp_data_heap, 0, 256)) { + nv30_screen_destroy(&screen->pipe); + return NULL; + } + + /* Static rankine initialisation */ + so = so_new(128, 0); + so_method(so, screen->rankine, NV34TCL_DMA_NOTIFY, 1); + so_data (so, screen->sync->handle); + so_method(so, screen->rankine, NV34TCL_DMA_TEXTURE0, 2); + so_data (so, nvws->channel->vram->handle); + so_data (so, nvws->channel->gart->handle); + so_method(so, screen->rankine, NV34TCL_DMA_COLOR1, 1); + so_data (so, nvws->channel->vram->handle); + so_method(so, screen->rankine, NV34TCL_DMA_COLOR0, 2); + so_data (so, nvws->channel->vram->handle); + so_data (so, nvws->channel->vram->handle); + so_method(so, screen->rankine, NV34TCL_DMA_VTXBUF0, 2); + so_data (so, nvws->channel->vram->handle); + so_data (so, nvws->channel->gart->handle); +/* so_method(so, screen->rankine, NV34TCL_DMA_FENCE, 2); + so_data (so, 0); + so_data (so, screen->query->handle);*/ + so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY7, 1); + so_data (so, nvws->channel->vram->handle); + so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY8, 1); + so_data (so, nvws->channel->vram->handle); + + for (i=1; i<8; i++) { + so_method(so, screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(i), 1); + so_data (so, 0); + so_method(so, screen->rankine, NV34TCL_VIEWPORT_CLIP_VERT(i), 1); + so_data (so, 0); + } + + so_method(so, screen->rankine, 0x220, 1); + so_data (so, 1); + + so_method(so, screen->rankine, 0x03b0, 1); + so_data (so, 0x00100000); + so_method(so, screen->rankine, 0x1454, 1); + so_data (so, 0); + so_method(so, screen->rankine, 0x1d80, 1); + so_data (so, 3); + so_method(so, screen->rankine, 0x1450, 1); + so_data (so, 0x00030004); + + /* NEW */ + so_method(so, screen->rankine, 0x1e98, 1); + so_data (so, 0); + so_method(so, screen->rankine, 0x17e0, 3); + so_data (so, fui(0.0)); + so_data (so, fui(0.0)); + so_data (so, fui(1.0)); + so_method(so, screen->rankine, 0x1f80, 16); + for (i=0; i<16; i++) { + so_data (so, (i==8) ? 0x0000ffff : 0); + } + + so_method(so, screen->rankine, 0x120, 3); + so_data (so, 0); + so_data (so, 1); + so_data (so, 2); + + so_method(so, screen->rankine, 0x1d88, 1); + so_data (so, 0x00001200); + + so_method(so, screen->rankine, NV34TCL_RC_ENABLE, 1); + so_data (so, 0); + + so_method(so, screen->rankine, NV34TCL_DEPTH_RANGE_NEAR, 2); + so_data (so, fui(0.0)); + so_data (so, fui(1.0)); + + so_method(so, screen->rankine, NV34TCL_MULTISAMPLE_CONTROL, 1); + so_data (so, 0xffff0000); + + /* enables use of vp rather than fixed-function somehow */ + so_method(so, screen->rankine, 0x1e94, 1); + so_data (so, 0x13); + + so_emit(nvws, so); + so_ref(NULL, &so); + nvws->push_flush(nvws, 0, NULL); + + screen->pipe.winsys = ws; + screen->pipe.destroy = nv30_screen_destroy; + + screen->pipe.get_name = nv30_screen_get_name; + screen->pipe.get_vendor = nv30_screen_get_vendor; + screen->pipe.get_param = nv30_screen_get_param; + screen->pipe.get_paramf = nv30_screen_get_paramf; + + screen->pipe.is_format_supported = nv30_screen_surface_format_supported; + + screen->pipe.surface_map = nv30_surface_map; + screen->pipe.surface_unmap = nv30_surface_unmap; + + nv30_screen_init_miptree_functions(&screen->pipe); + + return &screen->pipe; +} + diff --git a/src/gallium/drivers/nv30/nv30_screen.h b/src/gallium/drivers/nv30/nv30_screen.h new file mode 100644 index 0000000000..b7ddc2a959 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_screen.h @@ -0,0 +1,35 @@ +#ifndef __NV30_SCREEN_H__ +#define __NV30_SCREEN_H__ + +#include "pipe/p_screen.h" + +struct nv30_screen { + struct pipe_screen pipe; + + struct nouveau_winsys *nvws; + + unsigned cur_pctx; + + /* HW graphics objects */ + struct nouveau_grobj *rankine; + struct nouveau_notifier *sync; + + /* Query object resources */ + struct nouveau_notifier *query; + struct nouveau_resource *query_heap; + + /* Vtxprog resources */ + struct nouveau_resource *vp_exec_heap; + struct nouveau_resource *vp_data_heap; + + /* Current 3D state of channel */ + struct nouveau_stateobj *state[NV30_STATE_MAX]; +}; + +static INLINE struct nv30_screen * +nv30_screen(struct pipe_screen *screen) +{ + return (struct nv30_screen *)screen; +} + +#endif diff --git a/src/gallium/drivers/nv30/nv30_shader.h b/src/gallium/drivers/nv30/nv30_shader.h new file mode 100644 index 0000000000..dd3a36f78f --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_shader.h @@ -0,0 +1,490 @@ +#ifndef __NV30_SHADER_H__ +#define __NV30_SHADER_H__ + +/* Vertex programs instruction set + * + * 128bit opcodes, split into 4 32-bit ones for ease of use. + * + * Non-native instructions + * ABS - MOV + NV40_VP_INST0_DEST_ABS + * POW - EX2 + MUL + LG2 + * SUB - ADD, second source negated + * SWZ - MOV + * XPD - + * + * Register access + * - Only one INPUT can be accessed per-instruction (move extras into TEMPs) + * - Only one CONST can be accessed per-instruction (move extras into TEMPs) + * + * Relative Addressing + * According to the value returned for + * MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB + * + * there are only two address registers available. The destination in the + * ARL instruction is set to TEMP <n> (The temp isn't actually written). + * + * When using vanilla ARB_v_p, the proprietary driver will squish both the + * available ADDRESS regs into the first hardware reg in the X and Y + * components. + * + * To use an address reg as an index into consts, the CONST_SRC is set to + * (const_base + offset) and INDEX_CONST is set. + * + * To access the second address reg use ADDR_REG_SELECT_1. A particular + * component of the address regs is selected with ADDR_SWZ. + * + * Only one address register can be accessed per instruction. + * + * Conditional execution (see NV_vertex_program{2,3} for details) Conditional + * execution of an instruction is enabled by setting COND_TEST_ENABLE, and + * selecting the condition which will allow the test to pass with + * COND_{FL,LT,...}. It is possible to swizzle the values in the condition + * register, which allows for testing against an individual component. + * + * Branching: + * + * The BRA/CAL instructions seem to follow a slightly different opcode + * layout. The destination instruction ID (IADDR) overlaps a source field. + * Instruction ID's seem to be numbered based on the UPLOAD_FROM_ID FIFO + * command, and is incremented automatically on each UPLOAD_INST FIFO + * command. + * + * Conditional branching is achieved by using the condition tests described + * above. There doesn't appear to be dedicated looping instructions, but + * this can be done using a temp reg + conditional branching. + * + * Subroutines may be uploaded before the main program itself, but the first + * executed instruction is determined by the PROGRAM_START_ID FIFO command. + * + */ + +/* DWORD 0 */ + +#define NV30_VP_INST_ADDR_REG_SELECT_1 (1 << 24) +#define NV30_VP_INST_SRC2_ABS (1 << 23) /* guess */ +#define NV30_VP_INST_SRC1_ABS (1 << 22) /* guess */ +#define NV30_VP_INST_SRC0_ABS (1 << 21) /* guess */ +#define NV30_VP_INST_VEC_RESULT (1 << 20) +#define NV30_VP_INST_DEST_TEMP_ID_SHIFT 16 +#define NV30_VP_INST_DEST_TEMP_ID_MASK (0x0F << 16) +#define NV30_VP_INST_COND_UPDATE_ENABLE (1<<15) +#define NV30_VP_INST_VEC_DEST_TEMP_MASK (0xF << 16) +#define NV30_VP_INST_COND_TEST_ENABLE (1<<14) +#define NV30_VP_INST_COND_SHIFT 11 +#define NV30_VP_INST_COND_MASK (0x07 << 11) +# define NV30_VP_INST_COND_FL 0 /* guess */ +# define NV30_VP_INST_COND_LT 1 +# define NV30_VP_INST_COND_EQ 2 +# define NV30_VP_INST_COND_LE 3 +# define NV30_VP_INST_COND_GT 4 +# define NV30_VP_INST_COND_NE 5 +# define NV30_VP_INST_COND_GE 6 +# define NV30_VP_INST_COND_TR 7 /* guess */ +#define NV30_VP_INST_COND_SWZ_X_SHIFT 9 +#define NV30_VP_INST_COND_SWZ_X_MASK (0x03 << 9) +#define NV30_VP_INST_COND_SWZ_Y_SHIFT 7 +#define NV30_VP_INST_COND_SWZ_Y_MASK (0x03 << 7) +#define NV30_VP_INST_COND_SWZ_Z_SHIFT 5 +#define NV30_VP_INST_COND_SWZ_Z_MASK (0x03 << 5) +#define NV30_VP_INST_COND_SWZ_W_SHIFT 3 +#define NV30_VP_INST_COND_SWZ_W_MASK (0x03 << 3) +#define NV30_VP_INST_COND_SWZ_ALL_SHIFT 3 +#define NV30_VP_INST_COND_SWZ_ALL_MASK (0xFF << 3) +#define NV30_VP_INST_ADDR_SWZ_SHIFT 1 +#define NV30_VP_INST_ADDR_SWZ_MASK (0x03 << 1) +#define NV30_VP_INST_SCA_OPCODEH_SHIFT 0 +#define NV30_VP_INST_SCA_OPCODEH_MASK (0x01 << 0) + +/* DWORD 1 */ +#define NV30_VP_INST_SCA_OPCODEL_SHIFT 28 +#define NV30_VP_INST_SCA_OPCODEL_MASK (0x0F << 28) +# define NV30_VP_INST_OP_NOP 0x00 +# define NV30_VP_INST_OP_RCP 0x02 +# define NV30_VP_INST_OP_RCC 0x03 +# define NV30_VP_INST_OP_RSQ 0x04 +# define NV30_VP_INST_OP_EXP 0x05 +# define NV30_VP_INST_OP_LOG 0x06 +# define NV30_VP_INST_OP_LIT 0x07 +# define NV30_VP_INST_OP_BRA 0x09 +# define NV30_VP_INST_OP_CAL 0x0B +# define NV30_VP_INST_OP_RET 0x0C +# define NV30_VP_INST_OP_LG2 0x0D +# define NV30_VP_INST_OP_EX2 0x0E +# define NV30_VP_INST_OP_SIN 0x0F +# define NV30_VP_INST_OP_COS 0x10 +#define NV30_VP_INST_VEC_OPCODE_SHIFT 23 +#define NV30_VP_INST_VEC_OPCODE_MASK (0x1F << 23) +# define NV30_VP_INST_OP_NOPV 0x00 +# define NV30_VP_INST_OP_MOV 0x01 +# define NV30_VP_INST_OP_MUL 0x02 +# define NV30_VP_INST_OP_ADD 0x03 +# define NV30_VP_INST_OP_MAD 0x04 +# define NV30_VP_INST_OP_DP3 0x05 +# define NV30_VP_INST_OP_DP4 0x07 +# define NV30_VP_INST_OP_DPH 0x06 +# define NV30_VP_INST_OP_DST 0x08 +# define NV30_VP_INST_OP_MIN 0x09 +# define NV30_VP_INST_OP_MAX 0x0A +# define NV30_VP_INST_OP_SLT 0x0B +# define NV30_VP_INST_OP_SGE 0x0C +# define NV30_VP_INST_OP_ARL 0x0D +# define NV30_VP_INST_OP_FRC 0x0E +# define NV30_VP_INST_OP_FLR 0x0F +# define NV30_VP_INST_OP_SEQ 0x10 +# define NV30_VP_INST_OP_SFL 0x11 +# define NV30_VP_INST_OP_SGT 0x12 +# define NV30_VP_INST_OP_SLE 0x13 +# define NV30_VP_INST_OP_SNE 0x14 +# define NV30_VP_INST_OP_STR 0x15 +# define NV30_VP_INST_OP_SSG 0x16 +# define NV30_VP_INST_OP_ARR 0x17 +# define NV30_VP_INST_OP_ARA 0x18 +#define NV30_VP_INST_CONST_SRC_SHIFT 14 +#define NV30_VP_INST_CONST_SRC_MASK (0xFF << 14) +#define NV30_VP_INST_INPUT_SRC_SHIFT 9 /*NV20*/ +#define NV30_VP_INST_INPUT_SRC_MASK (0x0F << 9) /*NV20*/ +# define NV30_VP_INST_IN_POS 0 /* These seem to match the bindings specified in */ +# define NV30_VP_INST_IN_WEIGHT 1 /* the ARB_v_p spec (2.14.3.1) */ +# define NV30_VP_INST_IN_NORMAL 2 +# define NV30_VP_INST_IN_COL0 3 /* Should probably confirm them all though */ +# define NV30_VP_INST_IN_COL1 4 +# define NV30_VP_INST_IN_FOGC 5 +# define NV30_VP_INST_IN_TC0 8 +# define NV30_VP_INST_IN_TC(n) (8+n) +#define NV30_VP_INST_SRC0H_SHIFT 0 /*NV20*/ +#define NV30_VP_INST_SRC0H_MASK (0x1FF << 0) /*NV20*/ + +/* Please note: the IADDR fields overlap other fields because they are used + * only for branch instructions. See Branching: label above + * + * DWORD 2 + */ +#define NV30_VP_INST_SRC0L_SHIFT 26 /*NV20*/ +#define NV30_VP_INST_SRC0L_MASK (0x3F <<26) /* NV30_VP_SRC0_LOW_MASK << 26 */ +#define NV30_VP_INST_SRC1_SHIFT 11 /*NV20*/ +#define NV30_VP_INST_SRC1_MASK (0x7FFF<<11) /*NV20*/ +#define NV30_VP_INST_SRC2H_SHIFT 0 /*NV20*/ +#define NV30_VP_INST_SRC2H_MASK (0x7FF << 0) /* NV30_VP_SRC2_HIGH_MASK >> 4*/ +#define NV30_VP_INST_IADDR_SHIFT 2 +#define NV30_VP_INST_IADDR_MASK (0xF << 28) /* NV30_VP_SRC2_LOW_MASK << 28 */ + +/* DWORD 3 */ +#define NV30_VP_INST_SRC2L_SHIFT 28 /*NV20*/ +#define NV30_VP_INST_SRC2L_MASK (0x0F <<28) /*NV20*/ +#define NV30_VP_INST_STEMP_WRITEMASK_SHIFT 24 +#define NV30_VP_INST_STEMP_WRITEMASK_MASK (0x0F << 24) +#define NV30_VP_INST_VTEMP_WRITEMASK_SHIFT 20 +#define NV30_VP_INST_VTEMP_WRITEMASK_MASK (0x0F << 20) +#define NV30_VP_INST_SDEST_WRITEMASK_SHIFT 16 +#define NV30_VP_INST_SDEST_WRITEMASK_MASK (0x0F << 16) +#define NV30_VP_INST_VDEST_WRITEMASK_SHIFT 12 /*NV20*/ +#define NV30_VP_INST_VDEST_WRITEMASK_MASK (0x0F << 12) /*NV20*/ +#define NV30_VP_INST_DEST_SHIFT 2 +#define NV30_VP_INST_DEST_MASK (0x0F << 2) +# define NV30_VP_INST_DEST_POS 0 +# define NV30_VP_INST_DEST_BFC0 1 +# define NV30_VP_INST_DEST_BFC1 2 +# define NV30_VP_INST_DEST_COL0 3 +# define NV30_VP_INST_DEST_COL1 4 +# define NV30_VP_INST_DEST_FOGC 5 +# define NV30_VP_INST_DEST_PSZ 6 +# define NV30_VP_INST_DEST_TC(n) (8+n) + +#define NV30_VP_INST_LAST (1 << 0) + +/* Useful to split the source selection regs into their pieces */ +#define NV30_VP_SRC0_HIGH_SHIFT 6 +#define NV30_VP_SRC0_HIGH_MASK 0x00007FC0 +#define NV30_VP_SRC0_LOW_MASK 0x0000003F +#define NV30_VP_SRC2_HIGH_SHIFT 4 +#define NV30_VP_SRC2_HIGH_MASK 0x00007FF0 +#define NV30_VP_SRC2_LOW_MASK 0x0000000F + + +/* Source-register definition - matches NV20 exactly */ +#define NV30_VP_SRC_NEGATE (1<<14) +#define NV30_VP_SRC_SWZ_X_SHIFT 12 +#define NV30_VP_SRC_REG_SWZ_X_MASK (0x03 <<12) +#define NV30_VP_SRC_SWZ_Y_SHIFT 10 +#define NV30_VP_SRC_REG_SWZ_Y_MASK (0x03 <<10) +#define NV30_VP_SRC_SWZ_Z_SHIFT 8 +#define NV30_VP_SRC_REG_SWZ_Z_MASK (0x03 << 8) +#define NV30_VP_SRC_SWZ_W_SHIFT 6 +#define NV30_VP_SRC_REG_SWZ_W_MASK (0x03 << 6) +#define NV30_VP_SRC_REG_SWZ_ALL_SHIFT 6 +#define NV30_VP_SRC_REG_SWZ_ALL_MASK (0xFF << 6) +#define NV30_VP_SRC_TEMP_SRC_SHIFT 2 +#define NV30_VP_SRC_REG_TEMP_ID_MASK (0x0F << 0) +#define NV30_VP_SRC_REG_TYPE_SHIFT 0 +#define NV30_VP_SRC_REG_TYPE_MASK (0x03 << 0) +#define NV30_VP_SRC_REG_TYPE_TEMP 1 +#define NV30_VP_SRC_REG_TYPE_INPUT 2 +#define NV30_VP_SRC_REG_TYPE_CONST 3 /* guess */ + +/* + * Each fragment program opcode appears to be comprised of 4 32-bit values. + * + * 0 - Opcode, output reg/mask, ATTRIB source + * 1 - Source 0 + * 2 - Source 1 + * 3 - Source 2 + * + * There appears to be no special difference between result regs and temp regs. + * result.color == R0.xyzw + * result.depth == R1.z + * When the fragprog contains instructions to write depth, NV30_TCL_PRIMITIVE_3D_UNK1D78=0 + * otherwise it is set to 1. + * + * Constants are inserted directly after the instruction that uses them. + * + * It appears that it's not possible to use two input registers in one + * instruction as the input sourcing is done in the instruction dword + * and not the source selection dwords. As such instructions such as: + * + * ADD result.color, fragment.color, fragment.texcoord[0]; + * + * must be split into two MOV's and then an ADD (nvidia does this) but + * I'm not sure why it's not just one MOV and then source the second input + * in the ADD instruction.. + * + * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary + * negation requires multiplication with a const. + * + * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO/SWIZZLE_ONE + * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as SWIZZLE_ZERO + * is implemented simply by not writing to the relevant components of the destination. + * + * Conditional execution + * TODO + * + * Non-native instructions: + * LIT + * LRP - MAD+MAD + * SUB - ADD, negate second source + * RSQ - LG2 + EX2 + * POW - LG2 + MUL + EX2 + * SCS - COS + SIN + * XPD + */ + +//== Opcode / Destination selection == +#define NV30_FP_OP_PROGRAM_END (1 << 0) +#define NV30_FP_OP_OUT_REG_SHIFT 1 +#define NV30_FP_OP_OUT_REG_MASK (31 << 1) /* uncertain */ +/* Needs to be set when writing outputs to get expected result.. */ +#define NV30_FP_OP_OUT_REG_HALF (1 << 7) +#define NV30_FP_OP_COND_WRITE_ENABLE (1 << 8) +#define NV30_FP_OP_OUTMASK_SHIFT 9 +#define NV30_FP_OP_OUTMASK_MASK (0xF << 9) +# define NV30_FP_OP_OUT_X (1<<9) +# define NV30_FP_OP_OUT_Y (1<<10) +# define NV30_FP_OP_OUT_Z (1<<11) +# define NV30_FP_OP_OUT_W (1<<12) +/* Uncertain about these, especially the input_src values.. it's possible that + * they can be dynamically changed. + */ +#define NV30_FP_OP_INPUT_SRC_SHIFT 13 +#define NV30_FP_OP_INPUT_SRC_MASK (15 << 13) +# define NV30_FP_OP_INPUT_SRC_POSITION 0x0 +# define NV30_FP_OP_INPUT_SRC_COL0 0x1 +# define NV30_FP_OP_INPUT_SRC_COL1 0x2 +# define NV30_FP_OP_INPUT_SRC_FOGC 0x3 +# define NV30_FP_OP_INPUT_SRC_TC0 0x4 +# define NV30_FP_OP_INPUT_SRC_TC(n) (0x4 + n) +#define NV30_FP_OP_TEX_UNIT_SHIFT 17 +#define NV30_FP_OP_TEX_UNIT_MASK (0xF << 17) /* guess */ +#define NV30_FP_OP_PRECISION_SHIFT 22 +#define NV30_FP_OP_PRECISION_MASK (3 << 22) +# define NV30_FP_PRECISION_FP32 0 +# define NV30_FP_PRECISION_FP16 1 +# define NV30_FP_PRECISION_FX12 2 +#define NV30_FP_OP_OPCODE_SHIFT 24 +#define NV30_FP_OP_OPCODE_MASK (0x3F << 24) +# define NV30_FP_OP_OPCODE_NOP 0x00 +# define NV30_FP_OP_OPCODE_MOV 0x01 +# define NV30_FP_OP_OPCODE_MUL 0x02 +# define NV30_FP_OP_OPCODE_ADD 0x03 +# define NV30_FP_OP_OPCODE_MAD 0x04 +# define NV30_FP_OP_OPCODE_DP3 0x05 +# define NV30_FP_OP_OPCODE_DP4 0x06 +# define NV30_FP_OP_OPCODE_DST 0x07 +# define NV30_FP_OP_OPCODE_MIN 0x08 +# define NV30_FP_OP_OPCODE_MAX 0x09 +# define NV30_FP_OP_OPCODE_SLT 0x0A +# define NV30_FP_OP_OPCODE_SGE 0x0B +# define NV30_FP_OP_OPCODE_SLE 0x0C +# define NV30_FP_OP_OPCODE_SGT 0x0D +# define NV30_FP_OP_OPCODE_SNE 0x0E +# define NV30_FP_OP_OPCODE_SEQ 0x0F +# define NV30_FP_OP_OPCODE_FRC 0x10 +# define NV30_FP_OP_OPCODE_FLR 0x11 +# define NV30_FP_OP_OPCODE_KIL 0x12 +# define NV30_FP_OP_OPCODE_PK4B 0x13 +# define NV30_FP_OP_OPCODE_UP4B 0x14 +# define NV30_FP_OP_OPCODE_DDX 0x15 /* can only write XY */ +# define NV30_FP_OP_OPCODE_DDY 0x16 /* can only write XY */ +# define NV30_FP_OP_OPCODE_TEX 0x17 +# define NV30_FP_OP_OPCODE_TXP 0x18 +# define NV30_FP_OP_OPCODE_TXD 0x19 +# define NV30_FP_OP_OPCODE_RCP 0x1A +# define NV30_FP_OP_OPCODE_RSQ 0x1B +# define NV30_FP_OP_OPCODE_EX2 0x1C +# define NV30_FP_OP_OPCODE_LG2 0x1D +# define NV30_FP_OP_OPCODE_LIT 0x1E +# define NV30_FP_OP_OPCODE_LRP 0x1F +# define NV30_FP_OP_OPCODE_STR 0x20 +# define NV30_FP_OP_OPCODE_SFL 0x21 +# define NV30_FP_OP_OPCODE_COS 0x22 +# define NV30_FP_OP_OPCODE_SIN 0x23 +# define NV30_FP_OP_OPCODE_PK2H 0x24 +# define NV30_FP_OP_OPCODE_UP2H 0x25 +# define NV30_FP_OP_OPCODE_POW 0x26 +# define NV30_FP_OP_OPCODE_PK4UB 0x27 +# define NV30_FP_OP_OPCODE_UP4UB 0x28 +# define NV30_FP_OP_OPCODE_PK2US 0x29 +# define NV30_FP_OP_OPCODE_UP2US 0x2A +# define NV30_FP_OP_OPCODE_DP2A 0x2E +# define NV30_FP_OP_OPCODE_TXB 0x31 +# define NV30_FP_OP_OPCODE_RFL 0x36 +# define NV30_FP_OP_OPCODE_DIV 0x3A +#define NV30_FP_OP_OUT_SAT (1 << 31) + +/* high order bits of SRC0 */ +#define NV30_FP_OP_OUT_ABS (1 << 29) +#define NV30_FP_OP_COND_SWZ_W_SHIFT 27 +#define NV30_FP_OP_COND_SWZ_W_MASK (3 << 27) +#define NV30_FP_OP_COND_SWZ_Z_SHIFT 25 +#define NV30_FP_OP_COND_SWZ_Z_MASK (3 << 25) +#define NV30_FP_OP_COND_SWZ_Y_SHIFT 23 +#define NV30_FP_OP_COND_SWZ_Y_MASK (3 << 23) +#define NV30_FP_OP_COND_SWZ_X_SHIFT 21 +#define NV30_FP_OP_COND_SWZ_X_MASK (3 << 21) +#define NV30_FP_OP_COND_SWZ_ALL_SHIFT 21 +#define NV30_FP_OP_COND_SWZ_ALL_MASK (0xFF << 21) +#define NV30_FP_OP_COND_SHIFT 18 +#define NV30_FP_OP_COND_MASK (0x07 << 18) +# define NV30_FP_OP_COND_FL 0 +# define NV30_FP_OP_COND_LT 1 +# define NV30_FP_OP_COND_EQ 2 +# define NV30_FP_OP_COND_LE 3 +# define NV30_FP_OP_COND_GT 4 +# define NV30_FP_OP_COND_NE 5 +# define NV30_FP_OP_COND_GE 6 +# define NV30_FP_OP_COND_TR 7 + +/* high order bits of SRC1 */ +#define NV30_FP_OP_DST_SCALE_SHIFT 28 +#define NV30_FP_OP_DST_SCALE_MASK (3 << 28) +#define NV30_FP_OP_DST_SCALE_1X 0 +#define NV30_FP_OP_DST_SCALE_2X 1 +#define NV30_FP_OP_DST_SCALE_4X 2 +#define NV30_FP_OP_DST_SCALE_8X 3 +#define NV30_FP_OP_DST_SCALE_INV_2X 5 +#define NV30_FP_OP_DST_SCALE_INV_4X 6 +#define NV30_FP_OP_DST_SCALE_INV_8X 7 + + +/* high order bits of SRC2 */ +#define NV30_FP_OP_INDEX_INPUT (1 << 30) + +//== Register selection == +#define NV30_FP_REG_TYPE_SHIFT 0 +#define NV30_FP_REG_TYPE_MASK (3 << 0) +# define NV30_FP_REG_TYPE_TEMP 0 +# define NV30_FP_REG_TYPE_INPUT 1 +# define NV30_FP_REG_TYPE_CONST 2 +#define NV30_FP_REG_SRC_SHIFT 2 /* uncertain */ +#define NV30_FP_REG_SRC_MASK (31 << 2) +#define NV30_FP_REG_SRC_HALF (1 << 8) +#define NV30_FP_REG_SWZ_ALL_SHIFT 9 +#define NV30_FP_REG_SWZ_ALL_MASK (255 << 9) +#define NV30_FP_REG_SWZ_X_SHIFT 9 +#define NV30_FP_REG_SWZ_X_MASK (3 << 9) +#define NV30_FP_REG_SWZ_Y_SHIFT 11 +#define NV30_FP_REG_SWZ_Y_MASK (3 << 11) +#define NV30_FP_REG_SWZ_Z_SHIFT 13 +#define NV30_FP_REG_SWZ_Z_MASK (3 << 13) +#define NV30_FP_REG_SWZ_W_SHIFT 15 +#define NV30_FP_REG_SWZ_W_MASK (3 << 15) +# define NV30_FP_SWIZZLE_X 0 +# define NV30_FP_SWIZZLE_Y 1 +# define NV30_FP_SWIZZLE_Z 2 +# define NV30_FP_SWIZZLE_W 3 +#define NV30_FP_REG_NEGATE (1 << 17) + +#define NV30SR_NONE 0 +#define NV30SR_OUTPUT 1 +#define NV30SR_INPUT 2 +#define NV30SR_TEMP 3 +#define NV30SR_CONST 4 + +struct nv30_sreg { + int type; + int index; + + int dst_scale; + + int negate; + int abs; + int swz[4]; + + int cc_update; + int cc_update_reg; + int cc_test; + int cc_test_reg; + int cc_swz[4]; +}; + +static INLINE struct nv30_sreg +nv30_sr(int type, int index) +{ + struct nv30_sreg temp = { + .type = type, + .index = index, + .dst_scale = DEF_SCALE, + .abs = 0, + .negate = 0, + .swz = { 0, 1, 2, 3 }, + .cc_update = 0, + .cc_update_reg = 0, + .cc_test = DEF_CTEST, + .cc_test_reg = 0, + .cc_swz = { 0, 1, 2, 3 }, + }; + return temp; +} + +static INLINE struct nv30_sreg +nv30_sr_swz(struct nv30_sreg src, int x, int y, int z, int w) +{ + struct nv30_sreg dst = src; + + dst.swz[SWZ_X] = src.swz[x]; + dst.swz[SWZ_Y] = src.swz[y]; + dst.swz[SWZ_Z] = src.swz[z]; + dst.swz[SWZ_W] = src.swz[w]; + return dst; +} + +static INLINE struct nv30_sreg +nv30_sr_neg(struct nv30_sreg src) +{ + src.negate = !src.negate; + return src; +} + +static INLINE struct nv30_sreg +nv30_sr_abs(struct nv30_sreg src) +{ + src.abs = 1; + return src; +} + +static INLINE struct nv30_sreg +nv30_sr_scale(struct nv30_sreg src, int scale) +{ + src.dst_scale = scale; + return src; +} + +#endif diff --git a/src/gallium/drivers/nv30/nv30_state.c b/src/gallium/drivers/nv30/nv30_state.c new file mode 100644 index 0000000000..fc66075c83 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_state.c @@ -0,0 +1,724 @@ +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "tgsi/tgsi_parse.h" + +#include "nv30_context.h" +#include "nv30_state.h" + +static void * +nv30_blend_state_create(struct pipe_context *pipe, + const struct pipe_blend_state *cso) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nouveau_grobj *rankine = nv30->screen->rankine; + struct nv30_blend_state *bso = CALLOC(1, sizeof(*bso)); + struct nouveau_stateobj *so = so_new(16, 0); + + if (cso->blend_enable) { + so_method(so, rankine, NV34TCL_BLEND_FUNC_ENABLE, 3); + so_data (so, 1); + so_data (so, (nvgl_blend_func(cso->alpha_src_factor) << 16) | + nvgl_blend_func(cso->rgb_src_factor)); + so_data (so, nvgl_blend_func(cso->alpha_dst_factor) << 16 | + nvgl_blend_func(cso->rgb_dst_factor)); + so_method(so, rankine, NV34TCL_BLEND_EQUATION, 1); + so_data (so, nvgl_blend_eqn(cso->alpha_func) << 16 | + nvgl_blend_eqn(cso->rgb_func)); + } else { + so_method(so, rankine, NV34TCL_BLEND_FUNC_ENABLE, 1); + so_data (so, 0); + } + + so_method(so, rankine, NV34TCL_COLOR_MASK, 1); + so_data (so, (((cso->colormask & PIPE_MASK_A) ? (0x01 << 24) : 0) | + ((cso->colormask & PIPE_MASK_R) ? (0x01 << 16) : 0) | + ((cso->colormask & PIPE_MASK_G) ? (0x01 << 8) : 0) | + ((cso->colormask & PIPE_MASK_B) ? (0x01 << 0) : 0))); + + if (cso->logicop_enable) { + so_method(so, rankine, NV34TCL_COLOR_LOGIC_OP_ENABLE, 2); + so_data (so, 1); + so_data (so, nvgl_logicop_func(cso->logicop_func)); + } else { + so_method(so, rankine, NV34TCL_COLOR_LOGIC_OP_ENABLE, 1); + so_data (so, 0); + } + + so_method(so, rankine, NV34TCL_DITHER_ENABLE, 1); + so_data (so, cso->dither ? 1 : 0); + + so_ref(so, &bso->so); + bso->pipe = *cso; + return (void *)bso; +} + +static void +nv30_blend_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->blend = hwcso; + nv30->dirty |= NV30_NEW_BLEND; +} + +static void +nv30_blend_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv30_blend_state *bso = hwcso; + + so_ref(NULL, &bso->so); + FREE(bso); +} + + +static INLINE unsigned +wrap_mode(unsigned wrap) { + unsigned ret; + + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + ret = NV34TCL_TX_WRAP_S_REPEAT; + break; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + ret = NV34TCL_TX_WRAP_S_MIRRORED_REPEAT; + break; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + ret = NV34TCL_TX_WRAP_S_CLAMP_TO_EDGE; + break; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + ret = NV34TCL_TX_WRAP_S_CLAMP_TO_BORDER; + break; + case PIPE_TEX_WRAP_CLAMP: + ret = NV34TCL_TX_WRAP_S_CLAMP; + break; +/* case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + ret = NV34TCL_TX_WRAP_S_MIRROR_CLAMP_TO_EDGE; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + ret = NV34TCL_TX_WRAP_S_MIRROR_CLAMP_TO_BORDER; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + ret = NV34TCL_TX_WRAP_S_MIRROR_CLAMP; + break;*/ + default: + NOUVEAU_ERR("unknown wrap mode: %d\n", wrap); + ret = NV34TCL_TX_WRAP_S_REPEAT; + break; + } + + return ret >> NV34TCL_TX_WRAP_S_SHIFT; +} + +static void * +nv30_sampler_state_create(struct pipe_context *pipe, + const struct pipe_sampler_state *cso) +{ + struct nv30_sampler_state *ps; + uint32_t filter = 0; + + ps = MALLOC(sizeof(struct nv30_sampler_state)); + + ps->fmt = 0; + /* TODO: Not all RECTs formats have this bit set, bits 15-8 of format + are the tx format to use. We should store normalized coord flag + in sampler state structure, and set appropriate format in + nvxx_fragtex_build() + */ + /*NV34TCL_TX_FORMAT_RECT*/ + /*if (!cso->normalized_coords) { + ps->fmt |= (1<<14) ; + }*/ + + ps->wrap = ((wrap_mode(cso->wrap_s) << NV34TCL_TX_WRAP_S_SHIFT) | + (wrap_mode(cso->wrap_t) << NV34TCL_TX_WRAP_T_SHIFT) | + (wrap_mode(cso->wrap_r) << NV34TCL_TX_WRAP_R_SHIFT)); + + ps->en = 0; + + if (cso->max_anisotropy >= 8.0) { + ps->en |= NV34TCL_TX_ENABLE_ANISO_8X; + } else + if (cso->max_anisotropy >= 4.0) { + ps->en |= NV34TCL_TX_ENABLE_ANISO_4X; + } else + if (cso->max_anisotropy >= 2.0) { + ps->en |= NV34TCL_TX_ENABLE_ANISO_2X; + } + + switch (cso->mag_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + filter |= NV34TCL_TX_FILTER_MAGNIFY_LINEAR; + break; + case PIPE_TEX_FILTER_NEAREST: + default: + filter |= NV34TCL_TX_FILTER_MAGNIFY_NEAREST; + break; + } + + switch (cso->min_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR; + break; + } + break; + case PIPE_TEX_FILTER_NEAREST: + default: + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST; + break; + } + break; + } + + ps->filt = filter; + + { + float limit; + + limit = CLAMP(cso->lod_bias, -16.0, 15.0); + ps->filt |= (int)(cso->lod_bias * 256.0) & 0x1fff; + + limit = CLAMP(cso->max_lod, 0.0, 15.0); + ps->en |= (int)(limit) << 14 /*NV34TCL_TX_ENABLE_MIPMAP_MAX_LOD_SHIFT*/; + + limit = CLAMP(cso->min_lod, 0.0, 15.0); + ps->en |= (int)(limit) << 26 /*NV34TCL_TX_ENABLE_MIPMAP_MIN_LOD_SHIFT*/; + } + + if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + switch (cso->compare_func) { + case PIPE_FUNC_NEVER: + ps->wrap |= NV34TCL_TX_WRAP_RCOMP_NEVER; + break; + case PIPE_FUNC_GREATER: + ps->wrap |= NV34TCL_TX_WRAP_RCOMP_GREATER; + break; + case PIPE_FUNC_EQUAL: + ps->wrap |= NV34TCL_TX_WRAP_RCOMP_EQUAL; + break; + case PIPE_FUNC_GEQUAL: + ps->wrap |= NV34TCL_TX_WRAP_RCOMP_GEQUAL; + break; + case PIPE_FUNC_LESS: + ps->wrap |= NV34TCL_TX_WRAP_RCOMP_LESS; + break; + case PIPE_FUNC_NOTEQUAL: + ps->wrap |= NV34TCL_TX_WRAP_RCOMP_NOTEQUAL; + break; + case PIPE_FUNC_LEQUAL: + ps->wrap |= NV34TCL_TX_WRAP_RCOMP_LEQUAL; + break; + case PIPE_FUNC_ALWAYS: + ps->wrap |= NV34TCL_TX_WRAP_RCOMP_ALWAYS; + break; + default: + break; + } + } + + ps->bcol = ((float_to_ubyte(cso->border_color[3]) << 24) | + (float_to_ubyte(cso->border_color[0]) << 16) | + (float_to_ubyte(cso->border_color[1]) << 8) | + (float_to_ubyte(cso->border_color[2]) << 0)); + + return (void *)ps; +} + +static void +nv30_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler) +{ + struct nv30_context *nv30 = nv30_context(pipe); + unsigned unit; + + for (unit = 0; unit < nr; unit++) { + nv30->tex_sampler[unit] = sampler[unit]; + nv30->dirty_samplers |= (1 << unit); + } + + for (unit = nr; unit < nv30->nr_samplers; unit++) { + nv30->tex_sampler[unit] = NULL; + nv30->dirty_samplers |= (1 << unit); + } + + nv30->nr_samplers = nr; + nv30->dirty |= NV30_NEW_SAMPLER; +} + +static void +nv30_sampler_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + +static void +nv30_set_sampler_texture(struct pipe_context *pipe, unsigned nr, + struct pipe_texture **miptree) +{ + struct nv30_context *nv30 = nv30_context(pipe); + unsigned unit; + + for (unit = 0; unit < nr; unit++) { + pipe_texture_reference((struct pipe_texture **) + &nv30->tex_miptree[unit], miptree[unit]); + nv30->dirty_samplers |= (1 << unit); + } + + for (unit = nr; unit < nv30->nr_textures; unit++) { + pipe_texture_reference((struct pipe_texture **) + &nv30->tex_miptree[unit], NULL); + nv30->dirty_samplers |= (1 << unit); + } + + nv30->nr_textures = nr; + nv30->dirty |= NV30_NEW_SAMPLER; +} + +static void * +nv30_rasterizer_state_create(struct pipe_context *pipe, + const struct pipe_rasterizer_state *cso) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nv30_rasterizer_state *rsso = CALLOC(1, sizeof(*rsso)); + struct nouveau_stateobj *so = so_new(32, 0); + struct nouveau_grobj *rankine = nv30->screen->rankine; + + /*XXX: ignored: + * light_twoside + * point_smooth -nohw + * multisample + */ + + so_method(so, rankine, NV34TCL_SHADE_MODEL, 1); + so_data (so, cso->flatshade ? NV34TCL_SHADE_MODEL_FLAT : + NV34TCL_SHADE_MODEL_SMOOTH); + + so_method(so, rankine, NV34TCL_LINE_WIDTH, 2); + so_data (so, (unsigned char)(cso->line_width * 8.0) & 0xff); + so_data (so, cso->line_smooth ? 1 : 0); + so_method(so, rankine, NV34TCL_LINE_STIPPLE_ENABLE, 2); + so_data (so, cso->line_stipple_enable ? 1 : 0); + so_data (so, (cso->line_stipple_pattern << 16) | + cso->line_stipple_factor); + + so_method(so, rankine, NV34TCL_POINT_SIZE, 1); + so_data (so, fui(cso->point_size)); + + so_method(so, rankine, NV34TCL_POLYGON_MODE_FRONT, 6); + if (cso->front_winding == PIPE_WINDING_CCW) { + so_data(so, nvgl_polygon_mode(cso->fill_ccw)); + so_data(so, nvgl_polygon_mode(cso->fill_cw)); + switch (cso->cull_mode) { + case PIPE_WINDING_CCW: + so_data(so, NV34TCL_CULL_FACE_FRONT); + break; + case PIPE_WINDING_CW: + so_data(so, NV34TCL_CULL_FACE_BACK); + break; + case PIPE_WINDING_BOTH: + so_data(so, NV34TCL_CULL_FACE_FRONT_AND_BACK); + break; + default: + so_data(so, NV34TCL_CULL_FACE_BACK); + break; + } + so_data(so, NV34TCL_FRONT_FACE_CCW); + } else { + so_data(so, nvgl_polygon_mode(cso->fill_cw)); + so_data(so, nvgl_polygon_mode(cso->fill_ccw)); + switch (cso->cull_mode) { + case PIPE_WINDING_CCW: + so_data(so, NV34TCL_CULL_FACE_BACK); + break; + case PIPE_WINDING_CW: + so_data(so, NV34TCL_CULL_FACE_FRONT); + break; + case PIPE_WINDING_BOTH: + so_data(so, NV34TCL_CULL_FACE_FRONT_AND_BACK); + break; + default: + so_data(so, NV34TCL_CULL_FACE_BACK); + break; + } + so_data(so, NV34TCL_FRONT_FACE_CW); + } + so_data(so, cso->poly_smooth ? 1 : 0); + so_data(so, (cso->cull_mode != PIPE_WINDING_NONE) ? 1 : 0); + + so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1); + so_data (so, cso->poly_stipple_enable ? 1 : 0); + + so_method(so, rankine, NV34TCL_POLYGON_OFFSET_POINT_ENABLE, 3); + if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) || + (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT)) + so_data(so, 1); + else + so_data(so, 0); + if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) || + (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE)) + so_data(so, 1); + else + so_data(so, 0); + if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) || + (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL)) + so_data(so, 1); + else + so_data(so, 0); + if (cso->offset_cw || cso->offset_ccw) { + so_method(so, rankine, NV34TCL_POLYGON_OFFSET_FACTOR, 2); + so_data (so, fui(cso->offset_scale)); + so_data (so, fui(cso->offset_units * 2)); + } + + so_method(so, rankine, NV34TCL_POINT_SPRITE, 1); + if (cso->point_sprite) { + unsigned psctl = (1 << 0), i; + + for (i = 0; i < 8; i++) { + if (cso->sprite_coord_mode[i] != PIPE_SPRITE_COORD_NONE) + psctl |= (1 << (8 + i)); + } + + so_data(so, psctl); + } else { + so_data(so, 0); + } + + so_ref(so, &rsso->so); + rsso->pipe = *cso; + return (void *)rsso; +} + +static void +nv30_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->rasterizer = hwcso; + nv30->dirty |= NV30_NEW_RAST; + /*nv30->draw_dirty |= NV30_NEW_RAST;*/ +} + +static void +nv30_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv30_rasterizer_state *rsso = hwcso; + + so_ref(NULL, &rsso->so); + FREE(rsso); +} + +static void * +nv30_depth_stencil_alpha_state_create(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *cso) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nv30_zsa_state *zsaso = CALLOC(1, sizeof(*zsaso)); + struct nouveau_stateobj *so = so_new(32, 0); + struct nouveau_grobj *rankine = nv30->screen->rankine; + + so_method(so, rankine, NV34TCL_DEPTH_FUNC, 3); + so_data (so, nvgl_comparison_op(cso->depth.func)); + so_data (so, cso->depth.writemask ? 1 : 0); + so_data (so, cso->depth.enabled ? 1 : 0); + + so_method(so, rankine, NV34TCL_ALPHA_FUNC_ENABLE, 3); + so_data (so, cso->alpha.enabled ? 1 : 0); + so_data (so, nvgl_comparison_op(cso->alpha.func)); + so_data (so, float_to_ubyte(cso->alpha.ref)); + + if (cso->stencil[0].enabled) { + so_method(so, rankine, NV34TCL_STENCIL_FRONT_ENABLE, 8); + so_data (so, cso->stencil[0].enabled ? 1 : 0); + so_data (so, cso->stencil[0].write_mask); + so_data (so, nvgl_comparison_op(cso->stencil[0].func)); + so_data (so, cso->stencil[0].ref_value); + so_data (so, cso->stencil[0].value_mask); + so_data (so, nvgl_stencil_op(cso->stencil[0].fail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[0].zfail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[0].zpass_op)); + } else { + so_method(so, rankine, NV34TCL_STENCIL_FRONT_ENABLE, 1); + so_data (so, 0); + } + + if (cso->stencil[1].enabled) { + so_method(so, rankine, NV34TCL_STENCIL_BACK_ENABLE, 8); + so_data (so, cso->stencil[1].enabled ? 1 : 0); + so_data (so, cso->stencil[1].write_mask); + so_data (so, nvgl_comparison_op(cso->stencil[1].func)); + so_data (so, cso->stencil[1].ref_value); + so_data (so, cso->stencil[1].value_mask); + so_data (so, nvgl_stencil_op(cso->stencil[1].fail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[1].zfail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[1].zpass_op)); + } else { + so_method(so, rankine, NV34TCL_STENCIL_BACK_ENABLE, 1); + so_data (so, 0); + } + + so_ref(so, &zsaso->so); + zsaso->pipe = *cso; + return (void *)zsaso; +} + +static void +nv30_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->zsa = hwcso; + nv30->dirty |= NV30_NEW_ZSA; +} + +static void +nv30_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv30_zsa_state *zsaso = hwcso; + + so_ref(NULL, &zsaso->so); + FREE(zsaso); +} + +static void * +nv30_vp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *cso) +{ + /*struct nv30_context *nv30 = nv30_context(pipe);*/ + struct nv30_vertex_program *vp; + + vp = CALLOC(1, sizeof(struct nv30_vertex_program)); + vp->pipe.tokens = tgsi_dup_tokens(cso->tokens); + /*vp->draw = draw_create_vertex_shader(nv30->draw, &vp->pipe);*/ + + return (void *)vp; +} + +static void +nv30_vp_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->vertprog = hwcso; + nv30->dirty |= NV30_NEW_VERTPROG; + /*nv30->draw_dirty |= NV30_NEW_VERTPROG;*/ +} + +static void +nv30_vp_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nv30_vertex_program *vp = hwcso; + + /*draw_delete_vertex_shader(nv30->draw, vp->draw);*/ + nv30_vertprog_destroy(nv30, vp); + FREE((void*)vp->pipe.tokens); + FREE(vp); +} + +static void * +nv30_fp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *cso) +{ + struct nv30_fragment_program *fp; + + fp = CALLOC(1, sizeof(struct nv30_fragment_program)); + fp->pipe.tokens = tgsi_dup_tokens(cso->tokens); + + tgsi_scan_shader(fp->pipe.tokens, &fp->info); + + return (void *)fp; +} + +static void +nv30_fp_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->fragprog = hwcso; + nv30->dirty |= NV30_NEW_FRAGPROG; +} + +static void +nv30_fp_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nv30_fragment_program *fp = hwcso; + + nv30_fragprog_destroy(nv30, fp); + FREE((void*)fp->pipe.tokens); + FREE(fp); +} + +static void +nv30_set_blend_color(struct pipe_context *pipe, + const struct pipe_blend_color *bcol) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->blend_colour = *bcol; + nv30->dirty |= NV30_NEW_BCOL; +} + +static void +nv30_set_clip_state(struct pipe_context *pipe, + const struct pipe_clip_state *clip) +{ +} + +static void +nv30_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, + const struct pipe_constant_buffer *buf ) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->constbuf[shader] = buf->buffer; + nv30->constbuf_nr[shader] = buf->size / (4 * sizeof(float)); + + if (shader == PIPE_SHADER_VERTEX) { + nv30->dirty |= NV30_NEW_VERTPROG; + } else + if (shader == PIPE_SHADER_FRAGMENT) { + nv30->dirty |= NV30_NEW_FRAGPROG; + } +} + +static void +nv30_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->framebuffer = *fb; + nv30->dirty |= NV30_NEW_FB; +} + +static void +nv30_set_polygon_stipple(struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + memcpy(nv30->stipple, stipple->stipple, 4 * 32); + nv30->dirty |= NV30_NEW_STIPPLE; +} + +static void +nv30_set_scissor_state(struct pipe_context *pipe, + const struct pipe_scissor_state *s) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->scissor = *s; + nv30->dirty |= NV30_NEW_SCISSOR; +} + +static void +nv30_set_viewport_state(struct pipe_context *pipe, + const struct pipe_viewport_state *vpt) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->viewport = *vpt; + nv30->dirty |= NV30_NEW_VIEWPORT; + /*nv30->draw_dirty |= NV30_NEW_VIEWPORT;*/ +} + +static void +nv30_set_vertex_buffers(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_buffer *vb) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + memcpy(nv30->vtxbuf, vb, sizeof(*vb) * count); + nv30->vtxbuf_nr = count; + + nv30->dirty |= NV30_NEW_ARRAYS; + /*nv30->draw_dirty |= NV30_NEW_ARRAYS;*/ +} + +static void +nv30_set_vertex_elements(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_element *ve) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + memcpy(nv30->vtxelt, ve, sizeof(*ve) * count); + nv30->vtxelt_nr = count; + + nv30->dirty |= NV30_NEW_ARRAYS; + /*nv30->draw_dirty |= NV30_NEW_ARRAYS;*/ +} + +static void +nv30_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield) +{ + struct nv30_context *nv30 = nv30_context(pipe); + + nv30->edgeflags = bitfield; + nv30->dirty |= NV30_NEW_ARRAYS; + /*nv30->draw_dirty |= NV30_NEW_ARRAYS;*/ +} + +void +nv30_init_state_functions(struct nv30_context *nv30) +{ + nv30->pipe.create_blend_state = nv30_blend_state_create; + nv30->pipe.bind_blend_state = nv30_blend_state_bind; + nv30->pipe.delete_blend_state = nv30_blend_state_delete; + + nv30->pipe.create_sampler_state = nv30_sampler_state_create; + nv30->pipe.bind_sampler_states = nv30_sampler_state_bind; + nv30->pipe.delete_sampler_state = nv30_sampler_state_delete; + nv30->pipe.set_sampler_textures = nv30_set_sampler_texture; + + nv30->pipe.create_rasterizer_state = nv30_rasterizer_state_create; + nv30->pipe.bind_rasterizer_state = nv30_rasterizer_state_bind; + nv30->pipe.delete_rasterizer_state = nv30_rasterizer_state_delete; + + nv30->pipe.create_depth_stencil_alpha_state = + nv30_depth_stencil_alpha_state_create; + nv30->pipe.bind_depth_stencil_alpha_state = + nv30_depth_stencil_alpha_state_bind; + nv30->pipe.delete_depth_stencil_alpha_state = + nv30_depth_stencil_alpha_state_delete; + + nv30->pipe.create_vs_state = nv30_vp_state_create; + nv30->pipe.bind_vs_state = nv30_vp_state_bind; + nv30->pipe.delete_vs_state = nv30_vp_state_delete; + + nv30->pipe.create_fs_state = nv30_fp_state_create; + nv30->pipe.bind_fs_state = nv30_fp_state_bind; + nv30->pipe.delete_fs_state = nv30_fp_state_delete; + + nv30->pipe.set_blend_color = nv30_set_blend_color; + nv30->pipe.set_clip_state = nv30_set_clip_state; + nv30->pipe.set_constant_buffer = nv30_set_constant_buffer; + nv30->pipe.set_framebuffer_state = nv30_set_framebuffer_state; + nv30->pipe.set_polygon_stipple = nv30_set_polygon_stipple; + nv30->pipe.set_scissor_state = nv30_set_scissor_state; + nv30->pipe.set_viewport_state = nv30_set_viewport_state; + + nv30->pipe.set_edgeflags = nv30_set_edgeflags; + nv30->pipe.set_vertex_buffers = nv30_set_vertex_buffers; + nv30->pipe.set_vertex_elements = nv30_set_vertex_elements; +} + diff --git a/src/gallium/drivers/nv30/nv30_state.h b/src/gallium/drivers/nv30/nv30_state.h new file mode 100644 index 0000000000..2023278e37 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_state.h @@ -0,0 +1,88 @@ +#ifndef __NV30_STATE_H__ +#define __NV30_STATE_H__ + +#include "pipe/p_state.h" +#include "tgsi/tgsi_scan.h" + +struct nv30_sampler_state { + uint32_t fmt; + uint32_t wrap; + uint32_t en; + uint32_t filt; + uint32_t bcol; +}; + +struct nv30_vertex_program_exec { + uint32_t data[4]; + boolean has_branch_offset; + int const_index; +}; + +struct nv30_vertex_program_data { + int index; /* immediates == -1 */ + float value[4]; +}; + +struct nv30_vertex_program { + struct pipe_shader_state pipe; + + boolean translated; + + struct nv30_vertex_program_exec *insns; + unsigned nr_insns; + struct nv30_vertex_program_data *consts; + unsigned nr_consts; + + struct nouveau_resource *exec; + unsigned exec_start; + struct nouveau_resource *data; + unsigned data_start; + unsigned data_start_min; + + uint32_t ir; + uint32_t or; + struct nouveau_stateobj *so; +}; + +struct nv30_fragment_program_data { + unsigned offset; + unsigned index; +}; + +struct nv30_fragment_program { + struct pipe_shader_state pipe; + struct tgsi_shader_info info; + + boolean translated; + boolean on_hw; + unsigned samplers; + + uint32_t *insn; + int insn_len; + + struct nv30_fragment_program_data *consts; + unsigned nr_consts; + + struct pipe_buffer *buffer; + + uint32_t fp_control; + uint32_t fp_reg_control; + struct nouveau_stateobj *so; +}; + +struct nv30_miptree { + struct pipe_texture base; + + struct pipe_buffer *buffer; + uint total_size; + + struct pipe_texture *shadow_tex; + struct pipe_surface *shadow_surface; + + struct { + uint pitch; + uint *image_offset; + } level[PIPE_MAX_TEXTURE_LEVELS]; +}; + +#endif diff --git a/src/gallium/drivers/nv30/nv30_state_blend.c b/src/gallium/drivers/nv30/nv30_state_blend.c new file mode 100644 index 0000000000..44d43e132a --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_state_blend.c @@ -0,0 +1,40 @@ +#include "nv30_context.h" + +static boolean +nv30_state_blend_validate(struct nv30_context *nv30) +{ + so_ref(nv30->blend->so, &nv30->state.hw[NV30_STATE_BLEND]); + return TRUE; +} + +struct nv30_state_entry nv30_state_blend = { + .validate = nv30_state_blend_validate, + .dirty = { + .pipe = NV30_NEW_BLEND, + .hw = NV30_STATE_BLEND + } +}; + +static boolean +nv30_state_blend_colour_validate(struct nv30_context *nv30) +{ + struct nouveau_stateobj *so = so_new(2, 0); + struct pipe_blend_color *bcol = &nv30->blend_colour; + + so_method(so, nv30->screen->rankine, NV34TCL_BLEND_COLOR, 1); + so_data (so, ((float_to_ubyte(bcol->color[3]) << 24) | + (float_to_ubyte(bcol->color[0]) << 16) | + (float_to_ubyte(bcol->color[1]) << 8) | + (float_to_ubyte(bcol->color[2]) << 0))); + + so_ref(so, &nv30->state.hw[NV30_STATE_BCOL]); + return TRUE; +} + +struct nv30_state_entry nv30_state_blend_colour = { + .validate = nv30_state_blend_colour_validate, + .dirty = { + .pipe = NV30_NEW_BCOL, + .hw = NV30_STATE_BCOL + } +}; diff --git a/src/gallium/drivers/nv30/nv30_state_emit.c b/src/gallium/drivers/nv30/nv30_state_emit.c new file mode 100644 index 0000000000..40fed621b2 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_state_emit.c @@ -0,0 +1,118 @@ +#include "nv30_context.h" +#include "nv30_state.h" + +static struct nv30_state_entry *render_states[] = { + &nv30_state_framebuffer, + &nv30_state_rasterizer, + &nv30_state_scissor, + &nv30_state_stipple, + &nv30_state_fragprog, + &nv30_state_fragtex, + &nv30_state_vertprog, + &nv30_state_blend, + &nv30_state_blend_colour, + &nv30_state_zsa, + &nv30_state_viewport, + &nv30_state_vbo, + NULL +}; + +static void +nv30_state_do_validate(struct nv30_context *nv30, + struct nv30_state_entry **states) +{ + const struct pipe_framebuffer_state *fb = &nv30->framebuffer; + unsigned i; + + for (i = 0; i < fb->num_cbufs; i++) + fb->cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED; + if (fb->zsbuf) + fb->zsbuf->status = PIPE_SURFACE_STATUS_DEFINED; + + while (*states) { + struct nv30_state_entry *e = *states; + + if (nv30->dirty & e->dirty.pipe) { + if (e->validate(nv30)) { + nv30->state.dirty |= (1ULL << e->dirty.hw); + } + } + + states++; + } + nv30->dirty = 0; +} + +void +nv30_state_emit(struct nv30_context *nv30) +{ + struct nv30_state *state = &nv30->state; + struct nv30_screen *screen = nv30->screen; + unsigned i, samplers; + uint64 states; + + if (nv30->pctx_id != screen->cur_pctx) { + for (i = 0; i < NV30_STATE_MAX; i++) { + if (state->hw[i] && screen->state[i] != state->hw[i]) + state->dirty |= (1ULL << i); + } + + screen->cur_pctx = nv30->pctx_id; + } + + for (i = 0, states = state->dirty; states; i++) { + if (!(states & (1ULL << i))) + continue; + so_ref (state->hw[i], &nv30->screen->state[i]); + if (state->hw[i]) + so_emit(nv30->nvws, nv30->screen->state[i]); + states &= ~(1ULL << i); + } + + state->dirty = 0; + + so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_FB]); + for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) { + if (!(samplers & (1 << i))) + continue; + so_emit_reloc_markers(nv30->nvws, + state->hw[NV30_STATE_FRAGTEX0+i]); + samplers &= ~(1ULL << i); + } + so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_FRAGPROG]); + if (state->hw[NV30_STATE_VTXBUF] /*&& nv30->render_mode == HW*/) + so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_VTXBUF]); +} + +boolean +nv30_state_validate(struct nv30_context *nv30) +{ +#if 0 + boolean was_sw = nv30->fallback_swtnl ? TRUE : FALSE; + + if (nv30->render_mode != HW) { + /* Don't even bother trying to go back to hw if none + * of the states that caused swtnl previously have changed. + */ + if ((nv30->fallback_swtnl & nv30->dirty) + != nv30->fallback_swtnl) + return FALSE; + + /* Attempt to go to hwtnl again */ + nv30->pipe.flush(&nv30->pipe, 0, NULL); + nv30->dirty |= (NV30_NEW_VIEWPORT | + NV30_NEW_VERTPROG | + NV30_NEW_ARRAYS); + nv30->render_mode = HW; + } +#endif + nv30_state_do_validate(nv30, render_states); +#if 0 + if (nv30->fallback_swtnl || nv30->fallback_swrast) + return FALSE; + + if (was_sw) + NOUVEAU_ERR("swtnl->hw\n"); +#endif + return TRUE; +} diff --git a/src/gallium/drivers/nv30/nv30_state_fb.c b/src/gallium/drivers/nv30/nv30_state_fb.c new file mode 100644 index 0000000000..73bdf7e56c --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_state_fb.c @@ -0,0 +1,140 @@ +#include "nv30_context.h" +#include "nouveau/nouveau_util.h" + +static boolean +nv30_state_framebuffer_validate(struct nv30_context *nv30) +{ + struct pipe_framebuffer_state *fb = &nv30->framebuffer; + struct pipe_surface *rt[2], *zeta = NULL; + uint32_t rt_enable, rt_format; + int i, colour_format = 0, zeta_format = 0; + struct nouveau_stateobj *so = so_new(64, 10); + unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM; + unsigned w = fb->width; + unsigned h = fb->height; + + rt_enable = 0; + for (i = 0; i < fb->num_cbufs; i++) { + if (colour_format) { + assert(colour_format == fb->cbufs[i]->format); + } else { + colour_format = fb->cbufs[i]->format; + rt_enable |= (NV34TCL_RT_ENABLE_COLOR0 << i); + rt[i] = fb->cbufs[i]; + } + } + + if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) + rt_enable |= NV34TCL_RT_ENABLE_MRT; + + if (fb->zsbuf) { + zeta_format = fb->zsbuf->format; + zeta = fb->zsbuf; + } + + if (!(rt[0]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1))); + for (i = 1; i < fb->num_cbufs; i++) + assert(!(rt[i]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)); + + /* FIXME: NV34TCL_RT_FORMAT_LOG2_[WIDTH/HEIGHT] */ + rt_format = NV34TCL_RT_FORMAT_TYPE_SWIZZLED | + log2i(fb->width) << 16 /*NV34TCL_RT_FORMAT_LOG2_WIDTH_SHIFT*/ | + log2i(fb->height) << 24 /*NV34TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT*/; + } + else + rt_format = NV34TCL_RT_FORMAT_TYPE_LINEAR; + + switch (colour_format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case 0: + rt_format |= NV34TCL_RT_FORMAT_COLOR_A8R8G8B8; + break; + case PIPE_FORMAT_R5G6B5_UNORM: + rt_format |= NV34TCL_RT_FORMAT_COLOR_R5G6B5; + break; + default: + assert(0); + } + + switch (zeta_format) { + case PIPE_FORMAT_Z16_UNORM: + rt_format |= NV34TCL_RT_FORMAT_ZETA_Z16; + break; + case PIPE_FORMAT_Z24S8_UNORM: + case 0: + rt_format |= NV34TCL_RT_FORMAT_ZETA_Z24S8; + break; + default: + assert(0); + } + + if (rt_enable & NV34TCL_RT_ENABLE_COLOR0) { + uint32_t pitch = rt[0]->stride; + if (zeta) { + pitch |= (zeta->stride << 16); + } else { + pitch |= (pitch << 16); + } + + so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR0, 1); + so_reloc (so, rt[0]->buffer, 0, rt_flags | NOUVEAU_BO_OR, + nv30->nvws->channel->vram->handle, + nv30->nvws->channel->gart->handle); + so_method(so, nv30->screen->rankine, NV34TCL_COLOR0_PITCH, 2); + so_data (so, pitch); + so_reloc (so, rt[0]->buffer, rt[0]->offset, rt_flags | + NOUVEAU_BO_LOW, 0, 0); + } + + if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) { + so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR1, 1); + so_reloc (so, rt[1]->buffer, 0, rt_flags | NOUVEAU_BO_OR, + nv30->nvws->channel->vram->handle, + nv30->nvws->channel->gart->handle); + so_method(so, nv30->screen->rankine, NV34TCL_COLOR1_OFFSET, 2); + so_reloc (so, rt[1]->buffer, rt[1]->offset, rt_flags | + NOUVEAU_BO_LOW, 0, 0); + so_data (so, rt[1]->stride); + } + + if (zeta_format) { + so_method(so, nv30->screen->rankine, NV34TCL_DMA_ZETA, 1); + so_reloc (so, zeta->buffer, 0, rt_flags | NOUVEAU_BO_OR, + nv30->nvws->channel->vram->handle, + nv30->nvws->channel->gart->handle); + so_method(so, nv30->screen->rankine, NV34TCL_ZETA_OFFSET, 1); + so_reloc (so, zeta->buffer, zeta->offset, rt_flags | + NOUVEAU_BO_LOW, 0, 0); + /* TODO: allocate LMA depth buffer */ + } + + so_method(so, nv30->screen->rankine, NV34TCL_RT_ENABLE, 1); + so_data (so, rt_enable); + so_method(so, nv30->screen->rankine, NV34TCL_RT_HORIZ, 3); + so_data (so, (w << 16) | 0); + so_data (so, (h << 16) | 0); + so_data (so, rt_format); + so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_HORIZ, 2); + so_data (so, (w << 16) | 0); + so_data (so, (h << 16) | 0); + so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2); + so_data (so, ((w - 1) << 16) | 0); + so_data (so, ((h - 1) << 16) | 0); + so_method(so, nv30->screen->rankine, 0x1d88, 1); + so_data (so, (1 << 12) | h); + /* Wonder why this is needed, context should all be set to zero on init */ + so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_TX_ORIGIN, 1); + so_data (so, 0); + + so_ref(so, &nv30->state.hw[NV30_STATE_FB]); + return TRUE; +} + +struct nv30_state_entry nv30_state_framebuffer = { + .validate = nv30_state_framebuffer_validate, + .dirty = { + .pipe = NV30_NEW_FB, + .hw = NV30_STATE_FB + } +}; diff --git a/src/gallium/drivers/nv30/nv30_state_rasterizer.c b/src/gallium/drivers/nv30/nv30_state_rasterizer.c new file mode 100644 index 0000000000..6d1b60e043 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_state_rasterizer.c @@ -0,0 +1,17 @@ +#include "nv30_context.h" + +static boolean +nv30_state_rasterizer_validate(struct nv30_context *nv30) +{ + so_ref(nv30->rasterizer->so, + &nv30->state.hw[NV30_STATE_RAST]); + return TRUE; +} + +struct nv30_state_entry nv30_state_rasterizer = { + .validate = nv30_state_rasterizer_validate, + .dirty = { + .pipe = NV30_NEW_RAST, + .hw = NV30_STATE_RAST + } +}; diff --git a/src/gallium/drivers/nv30/nv30_state_scissor.c b/src/gallium/drivers/nv30/nv30_state_scissor.c new file mode 100644 index 0000000000..1db9bc1795 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_state_scissor.c @@ -0,0 +1,35 @@ +#include "nv30_context.h" + +static boolean +nv30_state_scissor_validate(struct nv30_context *nv30) +{ + struct pipe_rasterizer_state *rast = &nv30->rasterizer->pipe; + struct pipe_scissor_state *s = &nv30->scissor; + struct nouveau_stateobj *so; + + if (nv30->state.hw[NV30_STATE_SCISSOR] && + (rast->scissor == 0 && nv30->state.scissor_enabled == 0)) + return FALSE; + nv30->state.scissor_enabled = rast->scissor; + + so = so_new(3, 0); + so_method(so, nv30->screen->rankine, NV34TCL_SCISSOR_HORIZ, 2); + if (nv30->state.scissor_enabled) { + so_data (so, ((s->maxx - s->minx) << 16) | s->minx); + so_data (so, ((s->maxy - s->miny) << 16) | s->miny); + } else { + so_data (so, 4096 << 16); + so_data (so, 4096 << 16); + } + + so_ref(so, &nv30->state.hw[NV30_STATE_SCISSOR]); + return TRUE; +} + +struct nv30_state_entry nv30_state_scissor = { + .validate = nv30_state_scissor_validate, + .dirty = { + .pipe = NV30_NEW_SCISSOR | NV30_NEW_RAST, + .hw = NV30_STATE_SCISSOR + } +}; diff --git a/src/gallium/drivers/nv30/nv30_state_stipple.c b/src/gallium/drivers/nv30/nv30_state_stipple.c new file mode 100644 index 0000000000..41b42813b4 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_state_stipple.c @@ -0,0 +1,39 @@ +#include "nv30_context.h" + +static boolean +nv30_state_stipple_validate(struct nv30_context *nv30) +{ + struct pipe_rasterizer_state *rast = &nv30->rasterizer->pipe; + struct nouveau_grobj *rankine = nv30->screen->rankine; + struct nouveau_stateobj *so; + + if (nv30->state.hw[NV30_STATE_STIPPLE] && + (rast->poly_stipple_enable == 0 && nv30->state.stipple_enabled == 0)) + return FALSE; + + if (rast->poly_stipple_enable) { + unsigned i; + + so = so_new(35, 0); + so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1); + so_data (so, 1); + so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_PATTERN(0), 32); + for (i = 0; i < 32; i++) + so_data(so, nv30->stipple[i]); + } else { + so = so_new(2, 0); + so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1); + so_data (so, 0); + } + + so_ref(so, &nv30->state.hw[NV30_STATE_STIPPLE]); + return TRUE; +} + +struct nv30_state_entry nv30_state_stipple = { + .validate = nv30_state_stipple_validate, + .dirty = { + .pipe = NV30_NEW_STIPPLE | NV30_NEW_RAST, + .hw = NV30_STATE_STIPPLE, + } +}; diff --git a/src/gallium/drivers/nv30/nv30_state_viewport.c b/src/gallium/drivers/nv30/nv30_state_viewport.c new file mode 100644 index 0000000000..951d40ebfd --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_state_viewport.c @@ -0,0 +1,70 @@ +#include "nv30_context.h" + +static boolean +nv30_state_viewport_validate(struct nv30_context *nv30) +{ + struct pipe_viewport_state *vpt = &nv30->viewport; + struct nouveau_stateobj *so; + unsigned bypass; + + if (/*nv30->render_mode == HW &&*/ !nv30->rasterizer->pipe.bypass_clipping) + bypass = 0; + else + bypass = 1; + + if (nv30->state.hw[NV30_STATE_VIEWPORT] && + (bypass || !(nv30->dirty & NV30_NEW_VIEWPORT)) && + nv30->state.viewport_bypass == bypass) + return FALSE; + nv30->state.viewport_bypass = bypass; + + so = so_new(11, 0); + if (!bypass) { + so_method(so, nv30->screen->rankine, + NV34TCL_VIEWPORT_TRANSLATE_X, 8); + so_data (so, fui(vpt->translate[0])); + so_data (so, fui(vpt->translate[1])); + so_data (so, fui(vpt->translate[2])); + so_data (so, fui(vpt->translate[3])); + so_data (so, fui(vpt->scale[0])); + so_data (so, fui(vpt->scale[1])); + so_data (so, fui(vpt->scale[2])); + so_data (so, fui(vpt->scale[3])); +/* so_method(so, nv30->screen->rankine, 0x1d78, 1); + so_data (so, 1); +*/ } else { + so_method(so, nv30->screen->rankine, + NV34TCL_VIEWPORT_TRANSLATE_X, 8); + so_data (so, fui(0.0)); + so_data (so, fui(0.0)); + so_data (so, fui(0.0)); + so_data (so, fui(0.0)); + so_data (so, fui(1.0)); + so_data (so, fui(1.0)); + so_data (so, fui(1.0)); + so_data (so, fui(0.0)); + /* Not entirely certain what this is yet. The DDX uses this + * value also as it fixes rendering when you pass + * pre-transformed vertices to the GPU. My best gusss is that + * this bypasses some culling/clipping stage. Might be worth + * noting that points/lines are uneffected by whatever this + * value fixes, only filled polygons are effected. + */ +/* so_method(so, nv30->screen->rankine, 0x1d78, 1); + so_data (so, 0x110); +*/ } + /* TODO/FIXME: never saw value 0x0110 in renouveau dumps, only 0x0001 */ + so_method(so, nv30->screen->rankine, 0x1d78, 1); + so_data (so, 1); + + so_ref(so, &nv30->state.hw[NV30_STATE_VIEWPORT]); + return TRUE; +} + +struct nv30_state_entry nv30_state_viewport = { + .validate = nv30_state_viewport_validate, + .dirty = { + .pipe = NV30_NEW_VIEWPORT | NV30_NEW_RAST, + .hw = NV30_STATE_VIEWPORT + } +}; diff --git a/src/gallium/drivers/nv30/nv30_state_zsa.c b/src/gallium/drivers/nv30/nv30_state_zsa.c new file mode 100644 index 0000000000..0940b7269b --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_state_zsa.c @@ -0,0 +1,17 @@ +#include "nv30_context.h" + +static boolean +nv30_state_zsa_validate(struct nv30_context *nv30) +{ + so_ref(nv30->zsa->so, + &nv30->state.hw[NV30_STATE_ZSA]); + return TRUE; +} + +struct nv30_state_entry nv30_state_zsa = { + .validate = nv30_state_zsa_validate, + .dirty = { + .pipe = NV30_NEW_ZSA, + .hw = NV30_STATE_ZSA + } +}; diff --git a/src/gallium/drivers/nv30/nv30_surface.c b/src/gallium/drivers/nv30/nv30_surface.c new file mode 100644 index 0000000000..d3376a73bf --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_surface.c @@ -0,0 +1,77 @@ + +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "nv30_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/p_inlines.h" + +#include "util/u_tile.h" + +static void +nv30_surface_copy(struct pipe_context *pipe, boolean do_flip, + struct pipe_surface *dest, unsigned destx, unsigned desty, + struct pipe_surface *src, unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nouveau_winsys *nvws = nv30->nvws; + + if (do_flip) { + /*XXX: This dodgyness will do for now for correctness. But, + * need to investigate whether the 2D engine is able to + * manage a flip (perhaps SIFM?), if not, use the 3D engine + */ + desty += height; + while (height--) { + nvws->surface_copy(nvws, dest, destx, desty--, src, + srcx, srcy++, width, 1); + } + } else { + nvws->surface_copy(nvws, dest, destx, desty, src, srcx, srcy, + width, height); + } +} + +static void +nv30_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest, + unsigned destx, unsigned desty, unsigned width, + unsigned height, unsigned value) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nouveau_winsys *nvws = nv30->nvws; + + nvws->surface_fill(nvws, dest, destx, desty, width, height, value); +} + +void +nv30_init_surface_functions(struct nv30_context *nv30) +{ + nv30->pipe.surface_copy = nv30_surface_copy; + nv30->pipe.surface_fill = nv30_surface_fill; +} diff --git a/src/gallium/drivers/nv30/nv30_vbo.c b/src/gallium/drivers/nv30/nv30_vbo.c new file mode 100644 index 0000000000..556f981d4a --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_vbo.c @@ -0,0 +1,556 @@ +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +#include "nv30_context.h" +#include "nv30_state.h" + +#include "nouveau/nouveau_channel.h" +#include "nouveau/nouveau_pushbuf.h" +#include "nouveau/nouveau_util.h" + +#define FORCE_SWTNL 0 + +static INLINE int +nv30_vbo_format_to_hw(enum pipe_format pipe, unsigned *fmt, unsigned *ncomp) +{ + switch (pipe) { + case PIPE_FORMAT_R32_FLOAT: + case PIPE_FORMAT_R32G32_FLOAT: + case PIPE_FORMAT_R32G32B32_FLOAT: + case PIPE_FORMAT_R32G32B32A32_FLOAT: + *fmt = NV34TCL_VTXFMT_TYPE_FLOAT; + break; + case PIPE_FORMAT_R8_UNORM: + case PIPE_FORMAT_R8G8_UNORM: + case PIPE_FORMAT_R8G8B8_UNORM: + case PIPE_FORMAT_R8G8B8A8_UNORM: + *fmt = NV34TCL_VTXFMT_TYPE_UBYTE; + break; + case PIPE_FORMAT_R16_SSCALED: + case PIPE_FORMAT_R16G16_SSCALED: + case PIPE_FORMAT_R16G16B16_SSCALED: + case PIPE_FORMAT_R16G16B16A16_SSCALED: + *fmt = NV34TCL_VTXFMT_TYPE_USHORT; + break; + default: + NOUVEAU_ERR("Unknown format %s\n", pf_name(pipe)); + return 1; + } + + switch (pipe) { + case PIPE_FORMAT_R8_UNORM: + case PIPE_FORMAT_R32_FLOAT: + case PIPE_FORMAT_R16_SSCALED: + *ncomp = 1; + break; + case PIPE_FORMAT_R8G8_UNORM: + case PIPE_FORMAT_R32G32_FLOAT: + case PIPE_FORMAT_R16G16_SSCALED: + *ncomp = 2; + break; + case PIPE_FORMAT_R8G8B8_UNORM: + case PIPE_FORMAT_R32G32B32_FLOAT: + case PIPE_FORMAT_R16G16B16_SSCALED: + *ncomp = 3; + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + case PIPE_FORMAT_R32G32B32A32_FLOAT: + case PIPE_FORMAT_R16G16B16A16_SSCALED: + *ncomp = 4; + break; + default: + NOUVEAU_ERR("Unknown format %s\n", pf_name(pipe)); + return 1; + } + + return 0; +} + +static boolean +nv30_vbo_set_idxbuf(struct nv30_context *nv30, struct pipe_buffer *ib, + unsigned ib_size) +{ + struct pipe_screen *pscreen = &nv30->screen->pipe; + unsigned type; + + if (!ib) { + nv30->idxbuf = NULL; + nv30->idxbuf_format = 0xdeadbeef; + return FALSE; + } + + if (!pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF) || ib_size == 1) + return FALSE; + + switch (ib_size) { + case 2: + type = NV34TCL_IDXBUF_FORMAT_TYPE_U16; + break; + case 4: + type = NV34TCL_IDXBUF_FORMAT_TYPE_U32; + break; + default: + return FALSE; + } + + if (ib != nv30->idxbuf || + type != nv30->idxbuf_format) { + nv30->dirty |= NV30_NEW_ARRAYS; + nv30->idxbuf = ib; + nv30->idxbuf_format = type; + } + + return TRUE; +} + +static boolean +nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so, + int attrib, struct pipe_vertex_element *ve, + struct pipe_vertex_buffer *vb) +{ + struct pipe_winsys *ws = nv30->pipe.winsys; + struct nouveau_grobj *rankine = nv30->screen->rankine; + unsigned type, ncomp; + void *map; + + if (nv30_vbo_format_to_hw(ve->src_format, &type, &ncomp)) + return FALSE; + + map = ws->buffer_map(ws, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ); + map += vb->buffer_offset + ve->src_offset; + + switch (type) { + case NV34TCL_VTXFMT_TYPE_FLOAT: + { + float *v = map; + + switch (ncomp) { + case 4: + so_method(so, rankine, NV34TCL_VTX_ATTR_4F_X(attrib), 4); + so_data (so, fui(v[0])); + so_data (so, fui(v[1])); + so_data (so, fui(v[2])); + so_data (so, fui(v[3])); + break; + case 3: + so_method(so, rankine, NV34TCL_VTX_ATTR_3F_X(attrib), 3); + so_data (so, fui(v[0])); + so_data (so, fui(v[1])); + so_data (so, fui(v[2])); + break; + case 2: + so_method(so, rankine, NV34TCL_VTX_ATTR_2F_X(attrib), 2); + so_data (so, fui(v[0])); + so_data (so, fui(v[1])); + break; + case 1: + so_method(so, rankine, NV34TCL_VTX_ATTR_1F(attrib), 1); + so_data (so, fui(v[0])); + break; + default: + ws->buffer_unmap(ws, vb->buffer); + return FALSE; + } + } + break; + default: + ws->buffer_unmap(ws, vb->buffer); + return FALSE; + } + + ws->buffer_unmap(ws, vb->buffer); + + return TRUE; +} + +boolean +nv30_draw_arrays(struct pipe_context *pipe, + unsigned mode, unsigned start, unsigned count) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nouveau_channel *chan = nv30->nvws->channel; + unsigned restart = 0; + + nv30_vbo_set_idxbuf(nv30, NULL, 0); + if (FORCE_SWTNL || !nv30_state_validate(nv30)) { + /*return nv30_draw_elements_swtnl(pipe, NULL, 0, + mode, start, count);*/ + return FALSE; + } + + while (count) { + unsigned vc, nr; + + nv30_state_emit(nv30); + + vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 256, + mode, start, count, &restart); + if (!vc) { + FIRE_RING(NULL); + continue; + } + + BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1); + OUT_RING (nvgl_primitive(mode)); + + nr = (vc & 0xff); + if (nr) { + BEGIN_RING(rankine, NV34TCL_VB_VERTEX_BATCH, 1); + OUT_RING (((nr - 1) << 24) | start); + start += nr; + } + + nr = vc >> 8; + while (nr) { + unsigned push = nr > 2047 ? 2047 : nr; + + nr -= push; + + BEGIN_RING_NI(rankine, NV34TCL_VB_VERTEX_BATCH, push); + while (push--) { + OUT_RING(((0x100 - 1) << 24) | start); + start += 0x100; + } + } + + BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1); + OUT_RING (0); + + count -= vc; + start = restart; + } + + pipe->flush(pipe, 0, NULL); + return TRUE; +} + +static INLINE void +nv30_draw_elements_u08(struct nv30_context *nv30, void *ib, + unsigned mode, unsigned start, unsigned count) +{ + struct nouveau_channel *chan = nv30->nvws->channel; + + while (count) { + uint8_t *elts = (uint8_t *)ib + start; + unsigned vc, push, restart = 0; + + nv30_state_emit(nv30); + + vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 2, + mode, start, count, &restart); + if (vc == 0) { + FIRE_RING(NULL); + continue; + } + count -= vc; + + BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1); + OUT_RING (nvgl_primitive(mode)); + + if (vc & 1) { + BEGIN_RING(rankine, NV34TCL_VB_ELEMENT_U32, 1); + OUT_RING (elts[0]); + elts++; vc--; + } + + while (vc) { + unsigned i; + + push = MIN2(vc, 2047 * 2); + + BEGIN_RING_NI(rankine, NV34TCL_VB_ELEMENT_U16, push >> 1); + for (i = 0; i < push; i+=2) + OUT_RING((elts[i+1] << 16) | elts[i]); + + vc -= push; + elts += push; + } + + BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1); + OUT_RING (0); + + start = restart; + } +} + +static INLINE void +nv30_draw_elements_u16(struct nv30_context *nv30, void *ib, + unsigned mode, unsigned start, unsigned count) +{ + struct nouveau_channel *chan = nv30->nvws->channel; + + while (count) { + uint16_t *elts = (uint16_t *)ib + start; + unsigned vc, push, restart = 0; + + nv30_state_emit(nv30); + + vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 2, + mode, start, count, &restart); + if (vc == 0) { + FIRE_RING(NULL); + continue; + } + count -= vc; + + BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1); + OUT_RING (nvgl_primitive(mode)); + + if (vc & 1) { + BEGIN_RING(rankine, NV34TCL_VB_ELEMENT_U32, 1); + OUT_RING (elts[0]); + elts++; vc--; + } + + while (vc) { + unsigned i; + + push = MIN2(vc, 2047 * 2); + + BEGIN_RING_NI(rankine, NV34TCL_VB_ELEMENT_U16, push >> 1); + for (i = 0; i < push; i+=2) + OUT_RING((elts[i+1] << 16) | elts[i]); + + vc -= push; + elts += push; + } + + BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1); + OUT_RING (0); + + start = restart; + } +} + +static INLINE void +nv30_draw_elements_u32(struct nv30_context *nv30, void *ib, + unsigned mode, unsigned start, unsigned count) +{ + struct nouveau_channel *chan = nv30->nvws->channel; + + while (count) { + uint32_t *elts = (uint32_t *)ib + start; + unsigned vc, push, restart = 0; + + nv30_state_emit(nv30); + + vc = nouveau_vbuf_split(chan->pushbuf->remaining, 5, 1, + mode, start, count, &restart); + if (vc == 0) { + FIRE_RING(NULL); + continue; + } + count -= vc; + + BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1); + OUT_RING (nvgl_primitive(mode)); + + while (vc) { + push = MIN2(vc, 2047); + + BEGIN_RING_NI(rankine, NV34TCL_VB_ELEMENT_U32, push); + OUT_RINGp (elts, push); + + vc -= push; + elts += push; + } + + BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1); + OUT_RING (0); + + start = restart; + } +} + +static boolean +nv30_draw_elements_inline(struct pipe_context *pipe, + struct pipe_buffer *ib, unsigned ib_size, + unsigned mode, unsigned start, unsigned count) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + void *map; + + map = ws->buffer_map(ws, ib, PIPE_BUFFER_USAGE_CPU_READ); + if (!ib) { + NOUVEAU_ERR("failed mapping ib\n"); + return FALSE; + } + + switch (ib_size) { + case 1: + nv30_draw_elements_u08(nv30, map, mode, start, count); + break; + case 2: + nv30_draw_elements_u16(nv30, map, mode, start, count); + break; + case 4: + nv30_draw_elements_u32(nv30, map, mode, start, count); + break; + default: + NOUVEAU_ERR("invalid idxbuf fmt %d\n", ib_size); + break; + } + + ws->buffer_unmap(ws, ib); + return TRUE; +} + +static boolean +nv30_draw_elements_vbo(struct pipe_context *pipe, + unsigned mode, unsigned start, unsigned count) +{ + struct nv30_context *nv30 = nv30_context(pipe); + struct nouveau_channel *chan = nv30->nvws->channel; + unsigned restart = 0; + + while (count) { + unsigned nr, vc; + + nv30_state_emit(nv30); + + vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 256, + mode, start, count, &restart); + if (!vc) { + FIRE_RING(NULL); + continue; + } + + BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1); + OUT_RING (nvgl_primitive(mode)); + + nr = (vc & 0xff); + if (nr) { + BEGIN_RING(rankine, NV34TCL_VB_INDEX_BATCH, 1); + OUT_RING (((nr - 1) << 24) | start); + start += nr; + } + + nr = vc >> 8; + while (nr) { + unsigned push = nr > 2047 ? 2047 : nr; + + nr -= push; + + BEGIN_RING_NI(rankine, NV34TCL_VB_INDEX_BATCH, push); + while (push--) { + OUT_RING(((0x100 - 1) << 24) | start); + start += 0x100; + } + } + + BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1); + OUT_RING (0); + + count -= vc; + start = restart; + } + + return TRUE; +} + +boolean +nv30_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, unsigned indexSize, + unsigned mode, unsigned start, unsigned count) +{ + struct nv30_context *nv30 = nv30_context(pipe); + boolean idxbuf; + + idxbuf = nv30_vbo_set_idxbuf(nv30, indexBuffer, indexSize); + if (FORCE_SWTNL || !nv30_state_validate(nv30)) { + /*return nv30_draw_elements_swtnl(pipe, NULL, 0, + mode, start, count);*/ + return FALSE; + } + + if (idxbuf) { + nv30_draw_elements_vbo(pipe, mode, start, count); + } else { + nv30_draw_elements_inline(pipe, indexBuffer, indexSize, + mode, start, count); + } + + pipe->flush(pipe, 0, NULL); + return TRUE; +} + +static boolean +nv30_vbo_validate(struct nv30_context *nv30) +{ + struct nouveau_stateobj *vtxbuf, *vtxfmt, *sattr = NULL; + struct nouveau_grobj *rankine = nv30->screen->rankine; + struct pipe_buffer *ib = nv30->idxbuf; + unsigned ib_format = nv30->idxbuf_format; + unsigned vb_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD; + int hw; + + if (nv30->edgeflags) { + /*nv30->fallback_swtnl |= NV30_NEW_ARRAYS;*/ + return FALSE; + } + + vtxbuf = so_new(20, 18); + so_method(vtxbuf, rankine, NV34TCL_VTXBUF_ADDRESS(0), nv30->vtxelt_nr); + vtxfmt = so_new(17, 0); + so_method(vtxfmt, rankine, NV34TCL_VTXFMT(0), nv30->vtxelt_nr); + + for (hw = 0; hw < nv30->vtxelt_nr; hw++) { + struct pipe_vertex_element *ve; + struct pipe_vertex_buffer *vb; + unsigned type, ncomp; + + ve = &nv30->vtxelt[hw]; + vb = &nv30->vtxbuf[ve->vertex_buffer_index]; + + if (!vb->pitch) { + if (!sattr) + sattr = so_new(16 * 5, 0); + + if (nv30_vbo_static_attrib(nv30, sattr, hw, ve, vb)) { + so_data(vtxbuf, 0); + so_data(vtxfmt, NV34TCL_VTXFMT_TYPE_FLOAT); + continue; + } + } + + if (nv30_vbo_format_to_hw(ve->src_format, &type, &ncomp)) { + /*nv30->fallback_swtnl |= NV30_NEW_ARRAYS;*/ + so_ref(NULL, &vtxbuf); + so_ref(NULL, &vtxfmt); + return FALSE; + } + + so_reloc(vtxbuf, vb->buffer, vb->buffer_offset + ve->src_offset, + vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR, + 0, NV34TCL_VTXBUF_ADDRESS_DMA1); + so_data (vtxfmt, ((vb->pitch << NV34TCL_VTXFMT_STRIDE_SHIFT) | + (ncomp << NV34TCL_VTXFMT_SIZE_SHIFT) | type)); + } + + if (ib) { + so_method(vtxbuf, rankine, NV34TCL_IDXBUF_ADDRESS, 2); + so_reloc (vtxbuf, ib, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (vtxbuf, ib, ib_format, vb_flags | NOUVEAU_BO_OR, + 0, NV34TCL_IDXBUF_FORMAT_DMA1); + } + + so_method(vtxbuf, rankine, 0x1710, 1); + so_data (vtxbuf, 0); + + so_ref(vtxbuf, &nv30->state.hw[NV30_STATE_VTXBUF]); + nv30->state.dirty |= (1ULL << NV30_STATE_VTXBUF); + so_ref(vtxfmt, &nv30->state.hw[NV30_STATE_VTXFMT]); + nv30->state.dirty |= (1ULL << NV30_STATE_VTXFMT); + so_ref(sattr, &nv30->state.hw[NV30_STATE_VTXATTR]); + nv30->state.dirty |= (1ULL << NV30_STATE_VTXATTR); + return FALSE; +} + +struct nv30_state_entry nv30_state_vbo = { + .validate = nv30_vbo_validate, + .dirty = { + .pipe = NV30_NEW_ARRAYS, + .hw = 0, + } +}; diff --git a/src/gallium/drivers/nv30/nv30_vertprog.c b/src/gallium/drivers/nv30/nv30_vertprog.c new file mode 100644 index 0000000000..72824559e8 --- /dev/null +++ b/src/gallium/drivers/nv30/nv30_vertprog.c @@ -0,0 +1,838 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_dump.h" + +#include "nv30_context.h" +#include "nv30_state.h" + +/* TODO (at least...): + * 1. Indexed consts + ARL + * 2. Arb. swz/negation + * 3. NV_vp11, NV_vp2, NV_vp3 features + * - extra arith opcodes + * - branching + * - texture sampling + * - indexed attribs + * - indexed results + * 4. bugs + */ + +#define SWZ_X 0 +#define SWZ_Y 1 +#define SWZ_Z 2 +#define SWZ_W 3 +#define MASK_X 8 +#define MASK_Y 4 +#define MASK_Z 2 +#define MASK_W 1 +#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W) +#define DEF_SCALE 0 +#define DEF_CTEST 0 +#include "nv30_shader.h" + +#define swz(s,x,y,z,w) nv30_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w) +#define neg(s) nv30_sr_neg((s)) +#define abs(s) nv30_sr_abs((s)) + +struct nv30_vpc { + struct nv30_vertex_program *vp; + + struct nv30_vertex_program_exec *vpi; + + unsigned output_map[PIPE_MAX_SHADER_OUTPUTS]; + + int high_temp; + int temp_temp_count; + + struct nv30_sreg *imm; + unsigned nr_imm; +}; + +static struct nv30_sreg +temp(struct nv30_vpc *vpc) +{ + int idx; + + idx = vpc->temp_temp_count++; + idx += vpc->high_temp + 1; + return nv30_sr(NV30SR_TEMP, idx); +} + +static struct nv30_sreg +constant(struct nv30_vpc *vpc, int pipe, float x, float y, float z, float w) +{ + struct nv30_vertex_program *vp = vpc->vp; + struct nv30_vertex_program_data *vpd; + int idx; + + if (pipe >= 0) { + for (idx = 0; idx < vp->nr_consts; idx++) { + if (vp->consts[idx].index == pipe) + return nv30_sr(NV30SR_CONST, idx); + } + } + + idx = vp->nr_consts++; + vp->consts = realloc(vp->consts, sizeof(*vpd) * vp->nr_consts); + vpd = &vp->consts[idx]; + + vpd->index = pipe; + vpd->value[0] = x; + vpd->value[1] = y; + vpd->value[2] = z; + vpd->value[3] = w; + return nv30_sr(NV30SR_CONST, idx); +} + +#define arith(cc,s,o,d,m,s0,s1,s2) \ + nv30_vp_arith((cc), (s), NV30_VP_INST_##o, (d), (m), (s0), (s1), (s2)) + +static void +emit_src(struct nv30_vpc *vpc, uint32_t *hw, int pos, struct nv30_sreg src) +{ + struct nv30_vertex_program *vp = vpc->vp; + uint32_t sr = 0; + + switch (src.type) { + case NV30SR_TEMP: + sr |= (NV30_VP_SRC_REG_TYPE_TEMP << NV30_VP_SRC_REG_TYPE_SHIFT); + sr |= (src.index << NV30_VP_SRC_TEMP_SRC_SHIFT); + break; + case NV30SR_INPUT: + sr |= (NV30_VP_SRC_REG_TYPE_INPUT << + NV30_VP_SRC_REG_TYPE_SHIFT); + vp->ir |= (1 << src.index); + hw[1] |= (src.index << NV30_VP_INST_INPUT_SRC_SHIFT); + break; + case NV30SR_CONST: + sr |= (NV30_VP_SRC_REG_TYPE_CONST << + NV30_VP_SRC_REG_TYPE_SHIFT); + assert(vpc->vpi->const_index == -1 || + vpc->vpi->const_index == src.index); + vpc->vpi->const_index = src.index; + break; + case NV30SR_NONE: + sr |= (NV30_VP_SRC_REG_TYPE_INPUT << + NV30_VP_SRC_REG_TYPE_SHIFT); + break; + default: + assert(0); + } + + if (src.negate) + sr |= NV30_VP_SRC_NEGATE; + + if (src.abs) + hw[0] |= (1 << (21 + pos)); + + sr |= ((src.swz[0] << NV30_VP_SRC_SWZ_X_SHIFT) | + (src.swz[1] << NV30_VP_SRC_SWZ_Y_SHIFT) | + (src.swz[2] << NV30_VP_SRC_SWZ_Z_SHIFT) | + (src.swz[3] << NV30_VP_SRC_SWZ_W_SHIFT)); + +/* + * |VVV| + * d�.�b + * \u/ + * + */ + + switch (pos) { + case 0: + hw[1] |= ((sr & NV30_VP_SRC0_HIGH_MASK) >> + NV30_VP_SRC0_HIGH_SHIFT) << NV30_VP_INST_SRC0H_SHIFT; + hw[2] |= (sr & NV30_VP_SRC0_LOW_MASK) << + NV30_VP_INST_SRC0L_SHIFT; + break; + case 1: + hw[2] |= sr << NV30_VP_INST_SRC1_SHIFT; + break; + case 2: + hw[2] |= ((sr & NV30_VP_SRC2_HIGH_MASK) >> + NV30_VP_SRC2_HIGH_SHIFT) << NV30_VP_INST_SRC2H_SHIFT; + hw[3] |= (sr & NV30_VP_SRC2_LOW_MASK) << + NV30_VP_INST_SRC2L_SHIFT; + break; + default: + assert(0); + } +} + +static void +emit_dst(struct nv30_vpc *vpc, uint32_t *hw, int slot, struct nv30_sreg dst) +{ + struct nv30_vertex_program *vp = vpc->vp; + + switch (dst.type) { + case NV30SR_TEMP: + hw[0] |= (dst.index << NV30_VP_INST_DEST_TEMP_ID_SHIFT); + break; + case NV30SR_OUTPUT: + switch (dst.index) { + case NV30_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break; + case NV30_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break; + case NV30_VP_INST_DEST_BFC0 : vp->or |= (1 << 2); break; + case NV30_VP_INST_DEST_BFC1 : vp->or |= (1 << 3); break; + case NV30_VP_INST_DEST_FOGC : vp->or |= (1 << 4); break; + case NV30_VP_INST_DEST_PSZ : vp->or |= (1 << 5); break; + case NV30_VP_INST_DEST_TC(0): vp->or |= (1 << 14); break; + case NV30_VP_INST_DEST_TC(1): vp->or |= (1 << 15); break; + case NV30_VP_INST_DEST_TC(2): vp->or |= (1 << 16); break; + case NV30_VP_INST_DEST_TC(3): vp->or |= (1 << 17); break; + case NV30_VP_INST_DEST_TC(4): vp->or |= (1 << 18); break; + case NV30_VP_INST_DEST_TC(5): vp->or |= (1 << 19); break; + case NV30_VP_INST_DEST_TC(6): vp->or |= (1 << 20); break; + case NV30_VP_INST_DEST_TC(7): vp->or |= (1 << 21); break; + default: + break; + } + + hw[3] |= (dst.index << NV30_VP_INST_DEST_SHIFT); + hw[0] |= NV30_VP_INST_VEC_DEST_TEMP_MASK | (1<<20); + + /*XXX: no way this is entirely correct, someone needs to + * figure out what exactly it is. + */ + hw[3] |= 0x800; + break; + default: + assert(0); + } +} + +static void +nv30_vp_arith(struct nv30_vpc *vpc, int slot, int op, + struct nv30_sreg dst, int mask, + struct nv30_sreg s0, struct nv30_sreg s1, + struct nv30_sreg s2) +{ + struct nv30_vertex_program *vp = vpc->vp; + uint32_t *hw; + + vp->insns = realloc(vp->insns, ++vp->nr_insns * sizeof(*vpc->vpi)); + vpc->vpi = &vp->insns[vp->nr_insns - 1]; + memset(vpc->vpi, 0, sizeof(*vpc->vpi)); + vpc->vpi->const_index = -1; + + hw = vpc->vpi->data; + + hw[0] |= (NV30_VP_INST_COND_TR << NV30_VP_INST_COND_SHIFT); + hw[0] |= ((0 << NV30_VP_INST_COND_SWZ_X_SHIFT) | + (1 << NV30_VP_INST_COND_SWZ_Y_SHIFT) | + (2 << NV30_VP_INST_COND_SWZ_Z_SHIFT) | + (3 << NV30_VP_INST_COND_SWZ_W_SHIFT)); + + hw[1] |= (op << NV30_VP_INST_VEC_OPCODE_SHIFT); +// hw[3] |= NV30_VP_INST_SCA_DEST_TEMP_MASK; +// hw[3] |= (mask << NV30_VP_INST_VEC_WRITEMASK_SHIFT); + + if (dst.type == NV30SR_OUTPUT) { + if (slot) + hw[3] |= (mask << NV30_VP_INST_SDEST_WRITEMASK_SHIFT); + else + hw[3] |= (mask << NV30_VP_INST_VDEST_WRITEMASK_SHIFT); + } else { + if (slot) + hw[3] |= (mask << NV30_VP_INST_STEMP_WRITEMASK_SHIFT); + else + hw[3] |= (mask << NV30_VP_INST_VTEMP_WRITEMASK_SHIFT); + } + + emit_dst(vpc, hw, slot, dst); + emit_src(vpc, hw, 0, s0); + emit_src(vpc, hw, 1, s1); + emit_src(vpc, hw, 2, s2); +} + +static INLINE struct nv30_sreg +tgsi_src(struct nv30_vpc *vpc, const struct tgsi_full_src_register *fsrc) { + struct nv30_sreg src; + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + src = nv30_sr(NV30SR_INPUT, fsrc->SrcRegister.Index); + break; + case TGSI_FILE_CONSTANT: + src = constant(vpc, fsrc->SrcRegister.Index, 0, 0, 0, 0); + break; + case TGSI_FILE_IMMEDIATE: + src = vpc->imm[fsrc->SrcRegister.Index]; + break; + case TGSI_FILE_TEMPORARY: + if (vpc->high_temp < fsrc->SrcRegister.Index) + vpc->high_temp = fsrc->SrcRegister.Index; + src = nv30_sr(NV30SR_TEMP, fsrc->SrcRegister.Index); + break; + default: + NOUVEAU_ERR("bad src file\n"); + break; + } + + src.abs = fsrc->SrcRegisterExtMod.Absolute; + src.negate = fsrc->SrcRegister.Negate; + src.swz[0] = fsrc->SrcRegister.SwizzleX; + src.swz[1] = fsrc->SrcRegister.SwizzleY; + src.swz[2] = fsrc->SrcRegister.SwizzleZ; + src.swz[3] = fsrc->SrcRegister.SwizzleW; + return src; +} + +static INLINE struct nv30_sreg +tgsi_dst(struct nv30_vpc *vpc, const struct tgsi_full_dst_register *fdst) { + struct nv30_sreg dst; + + switch (fdst->DstRegister.File) { + case TGSI_FILE_OUTPUT: + dst = nv30_sr(NV30SR_OUTPUT, + vpc->output_map[fdst->DstRegister.Index]); + + break; + case TGSI_FILE_TEMPORARY: + dst = nv30_sr(NV30SR_TEMP, fdst->DstRegister.Index); + if (vpc->high_temp < dst.index) + vpc->high_temp = dst.index; + break; + default: + NOUVEAU_ERR("bad dst file\n"); + break; + } + + return dst; +} + +static INLINE int +tgsi_mask(uint tgsi) +{ + int mask = 0; + + if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X; + if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y; + if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z; + if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W; + return mask; +} + +static boolean +nv30_vertprog_parse_instruction(struct nv30_vpc *vpc, + const struct tgsi_full_instruction *finst) +{ + struct nv30_sreg src[3], dst, tmp; + struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0); + int mask; + int ai = -1, ci = -1; + int i; + + if (finst->Instruction.Opcode == TGSI_OPCODE_END) + return TRUE; + + vpc->temp_temp_count = 0; + for (i = 0; i < finst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *fsrc; + + fsrc = &finst->FullSrcRegisters[i]; + if (fsrc->SrcRegister.File == TGSI_FILE_TEMPORARY) { + src[i] = tgsi_src(vpc, fsrc); + } + } + + for (i = 0; i < finst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *fsrc; + + fsrc = &finst->FullSrcRegisters[i]; + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + if (ai == -1 || ai == fsrc->SrcRegister.Index) { + ai = fsrc->SrcRegister.Index; + src[i] = tgsi_src(vpc, fsrc); + } else { + src[i] = temp(vpc); + arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + tgsi_src(vpc, fsrc), none, none); + } + break; + /*XXX: index comparison is broken now that consts come from + * two different register files. + */ + case TGSI_FILE_CONSTANT: + case TGSI_FILE_IMMEDIATE: + if (ci == -1 || ci == fsrc->SrcRegister.Index) { + ci = fsrc->SrcRegister.Index; + src[i] = tgsi_src(vpc, fsrc); + } else { + src[i] = temp(vpc); + arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + tgsi_src(vpc, fsrc), none, none); + } + break; + case TGSI_FILE_TEMPORARY: + /* handled above */ + break; + default: + NOUVEAU_ERR("bad src file\n"); + return FALSE; + } + } + + dst = tgsi_dst(vpc, &finst->FullDstRegisters[0]); + mask = tgsi_mask(finst->FullDstRegisters[0].DstRegister.WriteMask); + + switch (finst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + arith(vpc, 0, OP_MOV, dst, mask, abs(src[0]), none, none); + break; + case TGSI_OPCODE_ADD: + arith(vpc, 0, OP_ADD, dst, mask, src[0], none, src[1]); + break; + case TGSI_OPCODE_ARL: + arith(vpc, 0, OP_ARL, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_DP3: + arith(vpc, 0, OP_DP3, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DP4: + arith(vpc, 0, OP_DP4, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DPH: + arith(vpc, 0, OP_DPH, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DST: + arith(vpc, 0, OP_DST, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_EX2: + arith(vpc, 1, OP_EX2, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_EXP: + arith(vpc, 1, OP_EXP, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_FLR: + arith(vpc, 0, OP_FLR, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_FRC: + arith(vpc, 0, OP_FRC, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_LG2: + arith(vpc, 1, OP_LG2, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_LIT: + arith(vpc, 1, OP_LIT, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_LOG: + arith(vpc, 1, OP_LOG, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_MAD: + arith(vpc, 0, OP_MAD, dst, mask, src[0], src[1], src[2]); + break; + case TGSI_OPCODE_MAX: + arith(vpc, 0, OP_MAX, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_MIN: + arith(vpc, 0, OP_MIN, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_MOV: + arith(vpc, 0, OP_MOV, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_MUL: + arith(vpc, 0, OP_MUL, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_POW: + tmp = temp(vpc); + arith(vpc, 1, OP_LG2, tmp, MASK_X, none, none, + swz(src[0], X, X, X, X)); + arith(vpc, 0, OP_MUL, tmp, MASK_X, swz(tmp, X, X, X, X), + swz(src[1], X, X, X, X), none); + arith(vpc, 1, OP_EX2, dst, mask, none, none, + swz(tmp, X, X, X, X)); + break; + case TGSI_OPCODE_RCP: + arith(vpc, 1, OP_RCP, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_RET: + break; + case TGSI_OPCODE_RSQ: + arith(vpc, 1, OP_RSQ, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_SGE: + arith(vpc, 0, OP_SGE, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SGT: + arith(vpc, 0, OP_SGT, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SLT: + arith(vpc, 0, OP_SLT, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SUB: + arith(vpc, 0, OP_ADD, dst, mask, src[0], none, neg(src[1])); + break; + case TGSI_OPCODE_XPD: + tmp = temp(vpc); + arith(vpc, 0, OP_MUL, tmp, mask, + swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none); + arith(vpc, 0, OP_MAD, dst, (mask & ~MASK_W), + swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y), + neg(tmp)); + break; + default: + NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode); + return FALSE; + } + + return TRUE; +} + +static boolean +nv30_vertprog_parse_decl_output(struct nv30_vpc *vpc, + const struct tgsi_full_declaration *fdec) +{ + int hw; + + switch (fdec->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: + hw = NV30_VP_INST_DEST_POS; + break; + case TGSI_SEMANTIC_COLOR: + if (fdec->Semantic.SemanticIndex == 0) { + hw = NV30_VP_INST_DEST_COL0; + } else + if (fdec->Semantic.SemanticIndex == 1) { + hw = NV30_VP_INST_DEST_COL1; + } else { + NOUVEAU_ERR("bad colour semantic index\n"); + return FALSE; + } + break; + case TGSI_SEMANTIC_BCOLOR: + if (fdec->Semantic.SemanticIndex == 0) { + hw = NV30_VP_INST_DEST_BFC0; + } else + if (fdec->Semantic.SemanticIndex == 1) { + hw = NV30_VP_INST_DEST_BFC1; + } else { + NOUVEAU_ERR("bad bcolour semantic index\n"); + return FALSE; + } + break; + case TGSI_SEMANTIC_FOG: + hw = NV30_VP_INST_DEST_FOGC; + break; + case TGSI_SEMANTIC_PSIZE: + hw = NV30_VP_INST_DEST_PSZ; + break; + case TGSI_SEMANTIC_GENERIC: + if (fdec->Semantic.SemanticIndex <= 7) { + hw = NV30_VP_INST_DEST_TC(fdec->Semantic.SemanticIndex); + } else { + NOUVEAU_ERR("bad generic semantic index\n"); + return FALSE; + } + break; + default: + NOUVEAU_ERR("bad output semantic\n"); + return FALSE; + } + + vpc->output_map[fdec->DeclarationRange.First] = hw; + return TRUE; +} + +static boolean +nv30_vertprog_prepare(struct nv30_vpc *vpc) +{ + struct tgsi_parse_context p; + int nr_imm = 0; + + tgsi_parse_init(&p, vpc->vp->pipe.tokens); + while (!tgsi_parse_end_of_tokens(&p)) { + const union tgsi_full_token *tok = &p.FullToken; + + tgsi_parse_token(&p); + switch(tok->Token.Type) { + case TGSI_TOKEN_TYPE_IMMEDIATE: + nr_imm++; + break; + default: + break; + } + } + tgsi_parse_free(&p); + + if (nr_imm) { + vpc->imm = CALLOC(nr_imm, sizeof(struct nv30_sreg)); + assert(vpc->imm); + } + + return TRUE; +} + +static void +nv30_vertprog_translate(struct nv30_context *nv30, + struct nv30_vertex_program *vp) +{ + struct tgsi_parse_context parse; + struct nv30_vpc *vpc = NULL; + + tgsi_dump(vp->pipe.tokens,0); + + vpc = CALLOC(1, sizeof(struct nv30_vpc)); + if (!vpc) + return; + vpc->vp = vp; + vpc->high_temp = -1; + + if (!nv30_vertprog_prepare(vpc)) { + FREE(vpc); + return; + } + + tgsi_parse_init(&parse, vp->pipe.tokens); + + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + const struct tgsi_full_declaration *fdec; + fdec = &parse.FullToken.FullDeclaration; + switch (fdec->Declaration.File) { + case TGSI_FILE_OUTPUT: + if (!nv30_vertprog_parse_decl_output(vpc, fdec)) + goto out_err; + break; + default: + break; + } + } + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + const struct tgsi_full_immediate *imm; + + imm = &parse.FullToken.FullImmediate; + assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32); +// assert(imm->Immediate.Size == 4); + vpc->imm[vpc->nr_imm++] = + constant(vpc, -1, + imm->u.ImmediateFloat32[0].Float, + imm->u.ImmediateFloat32[1].Float, + imm->u.ImmediateFloat32[2].Float, + imm->u.ImmediateFloat32[3].Float); + } + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + const struct tgsi_full_instruction *finst; + finst = &parse.FullToken.FullInstruction; + if (!nv30_vertprog_parse_instruction(vpc, finst)) + goto out_err; + } + break; + default: + break; + } + } + + vp->insns[vp->nr_insns - 1].data[3] |= NV30_VP_INST_LAST; + vp->translated = TRUE; +out_err: + tgsi_parse_free(&parse); + FREE(vpc); +} + +static boolean +nv30_vertprog_validate(struct nv30_context *nv30) +{ + struct nouveau_winsys *nvws = nv30->nvws; + struct pipe_winsys *ws = nv30->pipe.winsys; + struct nouveau_grobj *rankine = nv30->screen->rankine; + struct nv30_vertex_program *vp; + struct pipe_buffer *constbuf; + boolean upload_code = FALSE, upload_data = FALSE; + int i; + + vp = nv30->vertprog; + constbuf = nv30->constbuf[PIPE_SHADER_VERTEX]; + + /* Translate TGSI shader into hw bytecode */ + if (!vp->translated) { + nv30_vertprog_translate(nv30, vp); + if (!vp->translated) + return FALSE; + } + + /* Allocate hw vtxprog exec slots */ + if (!vp->exec) { + struct nouveau_resource *heap = nv30->screen->vp_exec_heap; + struct nouveau_stateobj *so; + uint vplen = vp->nr_insns; + + if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) { + while (heap->next && heap->size < vplen) { + struct nv30_vertex_program *evict; + + evict = heap->next->priv; + nvws->res_free(&evict->exec); + } + + if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) + assert(0); + } + + so = so_new(2, 0); + so_method(so, rankine, NV34TCL_VP_START_FROM_ID, 1); + so_data (so, vp->exec->start); + so_ref(so, &vp->so); + + upload_code = TRUE; + } + + /* Allocate hw vtxprog const slots */ + if (vp->nr_consts && !vp->data) { + struct nouveau_resource *heap = nv30->screen->vp_data_heap; + + if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) { + while (heap->next && heap->size < vp->nr_consts) { + struct nv30_vertex_program *evict; + + evict = heap->next->priv; + nvws->res_free(&evict->data); + } + + if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) + assert(0); + } + + /*XXX: handle this some day */ + assert(vp->data->start >= vp->data_start_min); + + upload_data = TRUE; + if (vp->data_start != vp->data->start) + upload_code = TRUE; + } + + /* If exec or data segments moved we need to patch the program to + * fixup offsets and register IDs. + */ + if (vp->exec_start != vp->exec->start) { + for (i = 0; i < vp->nr_insns; i++) { + struct nv30_vertex_program_exec *vpi = &vp->insns[i]; + + if (vpi->has_branch_offset) { + assert(0); + } + } + + vp->exec_start = vp->exec->start; + } + + if (vp->nr_consts && vp->data_start != vp->data->start) { + for (i = 0; i < vp->nr_insns; i++) { + struct nv30_vertex_program_exec *vpi = &vp->insns[i]; + + if (vpi->const_index >= 0) { + vpi->data[1] &= ~NV30_VP_INST_CONST_SRC_MASK; + vpi->data[1] |= + (vpi->const_index + vp->data->start) << + NV30_VP_INST_CONST_SRC_SHIFT; + + } + } + + vp->data_start = vp->data->start; + } + + /* Update + Upload constant values */ + if (vp->nr_consts) { + float *map = NULL; + + if (constbuf) { + map = ws->buffer_map(ws, constbuf, + PIPE_BUFFER_USAGE_CPU_READ); + } + + for (i = 0; i < vp->nr_consts; i++) { + struct nv30_vertex_program_data *vpd = &vp->consts[i]; + + if (vpd->index >= 0) { + if (!upload_data && + !memcmp(vpd->value, &map[vpd->index * 4], + 4 * sizeof(float))) + continue; + memcpy(vpd->value, &map[vpd->index * 4], + 4 * sizeof(float)); + } + + BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_CONST_ID, 5); + OUT_RING (i + vp->data->start); + OUT_RINGp ((uint32_t *)vpd->value, 4); + } + + if (constbuf) { + ws->buffer_unmap(ws, constbuf); + } + } + + /* Upload vtxprog */ + if (upload_code) { +#if 0 + for (i = 0; i < vp->nr_insns; i++) { + NOUVEAU_MSG("VP inst %d: 0x%08x 0x%08x 0x%08x 0x%08x\n", + i, vp->insns[i].data[0], vp->insns[i].data[1], + vp->insns[i].data[2], vp->insns[i].data[3]); + } +#endif + BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_FROM_ID, 1); + OUT_RING (vp->exec->start); + for (i = 0; i < vp->nr_insns; i++) { + BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_INST(0), 4); + OUT_RINGp (vp->insns[i].data, 4); + } + } + + if (vp->so != nv30->state.hw[NV30_STATE_VERTPROG]) { + so_ref(vp->so, &nv30->state.hw[NV30_STATE_VERTPROG]); + return TRUE; + } + + return FALSE; +} + +void +nv30_vertprog_destroy(struct nv30_context *nv30, struct nv30_vertex_program *vp) +{ + struct nouveau_winsys *nvws = nv30->screen->nvws; + + vp->translated = FALSE; + + if (vp->nr_insns) { + FREE(vp->insns); + vp->insns = NULL; + vp->nr_insns = 0; + } + + if (vp->nr_consts) { + FREE(vp->consts); + vp->consts = NULL; + vp->nr_consts = 0; + } + + nvws->res_free(&vp->exec); + vp->exec_start = 0; + nvws->res_free(&vp->data); + vp->data_start = 0; + vp->data_start_min = 0; + + vp->ir = vp->or = 0; + so_ref(NULL, &vp->so); +} + +struct nv30_state_entry nv30_state_vertprog = { + .validate = nv30_vertprog_validate, + .dirty = { + .pipe = NV30_NEW_VERTPROG /*| NV30_NEW_UCP*/, + .hw = NV30_STATE_VERTPROG, + } +}; diff --git a/src/gallium/drivers/nv40/Makefile b/src/gallium/drivers/nv40/Makefile new file mode 100644 index 0000000000..9c8eadf7e4 --- /dev/null +++ b/src/gallium/drivers/nv40/Makefile @@ -0,0 +1,37 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = nv40 + +DRIVER_SOURCES = \ + nv40_clear.c \ + nv40_context.c \ + nv40_draw.c \ + nv40_fragprog.c \ + nv40_fragtex.c \ + nv40_miptree.c \ + nv40_query.c \ + nv40_screen.c \ + nv40_state.c \ + nv40_state_blend.c \ + nv40_state_emit.c \ + nv40_state_fb.c \ + nv40_state_rasterizer.c \ + nv40_state_scissor.c \ + nv40_state_stipple.c \ + nv40_state_viewport.c \ + nv40_state_zsa.c \ + nv40_surface.c \ + nv40_vbo.c \ + nv40_vertprog.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/drivers/nv40/nv40_clear.c b/src/gallium/drivers/nv40/nv40_clear.c new file mode 100644 index 0000000000..59efd620e3 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_clear.c @@ -0,0 +1,13 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "nv40_context.h" + +void +nv40_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); + ps->status = PIPE_SURFACE_STATUS_CLEAR; +} diff --git a/src/gallium/drivers/nv40/nv40_context.c b/src/gallium/drivers/nv40/nv40_context.c new file mode 100644 index 0000000000..cc63dd734b --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_context.c @@ -0,0 +1,72 @@ +#include "draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" + +#include "nv40_context.h" +#include "nv40_screen.h" + +static void +nv40_flush(struct pipe_context *pipe, unsigned flags, + struct pipe_fence_handle **fence) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + if (flags & PIPE_FLUSH_TEXTURE_CACHE) { + BEGIN_RING(curie, 0x1fd8, 1); + OUT_RING (2); + BEGIN_RING(curie, 0x1fd8, 1); + OUT_RING (1); + } + + FIRE_RING(fence); +} + +static void +nv40_destroy(struct pipe_context *pipe) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + if (nv40->draw) + draw_destroy(nv40->draw); + FREE(nv40); +} + +struct pipe_context * +nv40_create(struct pipe_screen *pscreen, unsigned pctx_id) +{ + struct nv40_screen *screen = nv40_screen(pscreen); + struct pipe_winsys *ws = pscreen->winsys; + struct nv40_context *nv40; + struct nouveau_winsys *nvws = screen->nvws; + + nv40 = CALLOC(1, sizeof(struct nv40_context)); + if (!nv40) + return NULL; + nv40->screen = screen; + nv40->pctx_id = pctx_id; + + nv40->nvws = nvws; + + nv40->pipe.winsys = ws; + nv40->pipe.screen = pscreen; + nv40->pipe.destroy = nv40_destroy; + nv40->pipe.draw_arrays = nv40_draw_arrays; + nv40->pipe.draw_elements = nv40_draw_elements; + nv40->pipe.clear = nv40_clear; + nv40->pipe.flush = nv40_flush; + + nv40_init_query_functions(nv40); + nv40_init_surface_functions(nv40); + nv40_init_state_functions(nv40); + + /* Create, configure, and install fallback swtnl path */ + nv40->draw = draw_create(); + draw_wide_point_threshold(nv40->draw, 9999999.0); + draw_wide_line_threshold(nv40->draw, 9999999.0); + draw_enable_line_stipple(nv40->draw, FALSE); + draw_enable_point_sprites(nv40->draw, FALSE); + draw_set_rasterize_stage(nv40->draw, nv40_draw_render_stage(nv40)); + + return &nv40->pipe; +} + diff --git a/src/gallium/drivers/nv40/nv40_context.h b/src/gallium/drivers/nv40/nv40_context.h new file mode 100644 index 0000000000..adcfbdd85a --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_context.h @@ -0,0 +1,233 @@ +#ifndef __NV40_CONTEXT_H__ +#define __NV40_CONTEXT_H__ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_compiler.h" + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "draw/draw_vertex.h" + +#include "nouveau/nouveau_winsys.h" +#include "nouveau/nouveau_gldefs.h" + +#define NOUVEAU_PUSH_CONTEXT(ctx) \ + struct nv40_screen *ctx = nv40->screen +#include "nouveau/nouveau_push.h" +#include "nouveau/nouveau_stateobj.h" + +#include "nv40_state.h" + +#define NOUVEAU_ERR(fmt, args...) \ + fprintf(stderr, "%s:%d - "fmt, __func__, __LINE__, ##args); +#define NOUVEAU_MSG(fmt, args...) \ + fprintf(stderr, "nouveau: "fmt, ##args); + +enum nv40_state_index { + NV40_STATE_FB = 0, + NV40_STATE_VIEWPORT = 1, + NV40_STATE_BLEND = 2, + NV40_STATE_RAST = 3, + NV40_STATE_ZSA = 4, + NV40_STATE_BCOL = 5, + NV40_STATE_CLIP = 6, + NV40_STATE_SCISSOR = 7, + NV40_STATE_STIPPLE = 8, + NV40_STATE_FRAGPROG = 9, + NV40_STATE_VERTPROG = 10, + NV40_STATE_FRAGTEX0 = 11, + NV40_STATE_FRAGTEX1 = 12, + NV40_STATE_FRAGTEX2 = 13, + NV40_STATE_FRAGTEX3 = 14, + NV40_STATE_FRAGTEX4 = 15, + NV40_STATE_FRAGTEX5 = 16, + NV40_STATE_FRAGTEX6 = 17, + NV40_STATE_FRAGTEX7 = 18, + NV40_STATE_FRAGTEX8 = 19, + NV40_STATE_FRAGTEX9 = 20, + NV40_STATE_FRAGTEX10 = 21, + NV40_STATE_FRAGTEX11 = 22, + NV40_STATE_FRAGTEX12 = 23, + NV40_STATE_FRAGTEX13 = 24, + NV40_STATE_FRAGTEX14 = 25, + NV40_STATE_FRAGTEX15 = 26, + NV40_STATE_VERTTEX0 = 27, + NV40_STATE_VERTTEX1 = 28, + NV40_STATE_VERTTEX2 = 29, + NV40_STATE_VERTTEX3 = 30, + NV40_STATE_VTXBUF = 31, + NV40_STATE_VTXFMT = 32, + NV40_STATE_VTXATTR = 33, + NV40_STATE_MAX = 34 +}; + +#include "nv40_screen.h" + +#define NV40_NEW_BLEND (1 << 0) +#define NV40_NEW_RAST (1 << 1) +#define NV40_NEW_ZSA (1 << 2) +#define NV40_NEW_SAMPLER (1 << 3) +#define NV40_NEW_FB (1 << 4) +#define NV40_NEW_STIPPLE (1 << 5) +#define NV40_NEW_SCISSOR (1 << 6) +#define NV40_NEW_VIEWPORT (1 << 7) +#define NV40_NEW_BCOL (1 << 8) +#define NV40_NEW_VERTPROG (1 << 9) +#define NV40_NEW_FRAGPROG (1 << 10) +#define NV40_NEW_ARRAYS (1 << 11) +#define NV40_NEW_UCP (1 << 12) + +struct nv40_rasterizer_state { + struct pipe_rasterizer_state pipe; + struct nouveau_stateobj *so; +}; + +struct nv40_zsa_state { + struct pipe_depth_stencil_alpha_state pipe; + struct nouveau_stateobj *so; +}; + +struct nv40_blend_state { + struct pipe_blend_state pipe; + struct nouveau_stateobj *so; +}; + + +struct nv40_state { + unsigned scissor_enabled; + unsigned stipple_enabled; + unsigned viewport_bypass; + unsigned fp_samplers; + + uint64_t dirty; + struct nouveau_stateobj *hw[NV40_STATE_MAX]; +}; + +struct nv40_context { + struct pipe_context pipe; + + struct nouveau_winsys *nvws; + struct nv40_screen *screen; + unsigned pctx_id; + + struct draw_context *draw; + + /* HW state derived from pipe states */ + struct nv40_state state; + struct { + struct nv40_vertex_program *vertprog; + + unsigned nr_attribs; + unsigned hw[PIPE_MAX_SHADER_INPUTS]; + unsigned draw[PIPE_MAX_SHADER_INPUTS]; + unsigned emit[PIPE_MAX_SHADER_INPUTS]; + } swtnl; + + enum { + HW, SWTNL, SWRAST + } render_mode; + unsigned fallback_swtnl; + unsigned fallback_swrast; + + /* Context state */ + unsigned dirty, draw_dirty; + struct pipe_scissor_state scissor; + unsigned stipple[32]; + struct pipe_clip_state clip; + struct nv40_vertex_program *vertprog; + struct nv40_fragment_program *fragprog; + struct pipe_buffer *constbuf[PIPE_SHADER_TYPES]; + unsigned constbuf_nr[PIPE_SHADER_TYPES]; + struct nv40_rasterizer_state *rasterizer; + struct nv40_zsa_state *zsa; + struct nv40_blend_state *blend; + struct pipe_blend_color blend_colour; + struct pipe_viewport_state viewport; + struct pipe_framebuffer_state framebuffer; + struct pipe_buffer *idxbuf; + unsigned idxbuf_format; + struct nv40_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS]; + struct nv40_miptree *tex_miptree[PIPE_MAX_SAMPLERS]; + unsigned nr_samplers; + unsigned nr_textures; + unsigned dirty_samplers; + struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS]; + unsigned vtxbuf_nr; + struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS]; + unsigned vtxelt_nr; + const unsigned *edgeflags; +}; + +static INLINE struct nv40_context * +nv40_context(struct pipe_context *pipe) +{ + return (struct nv40_context *)pipe; +} + +struct nv40_state_entry { + boolean (*validate)(struct nv40_context *nv40); + struct { + unsigned pipe; + unsigned hw; + } dirty; +}; + +extern void nv40_init_state_functions(struct nv40_context *nv40); +extern void nv40_init_surface_functions(struct nv40_context *nv40); +extern void nv40_init_query_functions(struct nv40_context *nv40); + +extern void nv40_screen_init_miptree_functions(struct pipe_screen *pscreen); + +/* nv40_draw.c */ +extern struct draw_stage *nv40_draw_render_stage(struct nv40_context *nv40); +extern boolean nv40_draw_elements_swtnl(struct pipe_context *pipe, + struct pipe_buffer *idxbuf, + unsigned ib_size, unsigned mode, + unsigned start, unsigned count); + +/* nv40_vertprog.c */ +extern void nv40_vertprog_destroy(struct nv40_context *, + struct nv40_vertex_program *); + +/* nv40_fragprog.c */ +extern void nv40_fragprog_destroy(struct nv40_context *, + struct nv40_fragment_program *); + +/* nv40_fragtex.c */ +extern void nv40_fragtex_bind(struct nv40_context *); + +/* nv40_state.c and friends */ +extern boolean nv40_state_validate(struct nv40_context *nv40); +extern boolean nv40_state_validate_swtnl(struct nv40_context *nv40); +extern void nv40_state_emit(struct nv40_context *nv40); +extern struct nv40_state_entry nv40_state_rasterizer; +extern struct nv40_state_entry nv40_state_scissor; +extern struct nv40_state_entry nv40_state_stipple; +extern struct nv40_state_entry nv40_state_fragprog; +extern struct nv40_state_entry nv40_state_vertprog; +extern struct nv40_state_entry nv40_state_blend; +extern struct nv40_state_entry nv40_state_blend_colour; +extern struct nv40_state_entry nv40_state_zsa; +extern struct nv40_state_entry nv40_state_viewport; +extern struct nv40_state_entry nv40_state_framebuffer; +extern struct nv40_state_entry nv40_state_fragtex; +extern struct nv40_state_entry nv40_state_vbo; +extern struct nv40_state_entry nv40_state_vtxfmt; + +/* nv40_vbo.c */ +extern boolean nv40_draw_arrays(struct pipe_context *, unsigned mode, + unsigned start, unsigned count); +extern boolean nv40_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, + unsigned count); + +/* nv40_clear.c */ +extern void nv40_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue); + +#endif diff --git a/src/gallium/drivers/nv40/nv40_draw.c b/src/gallium/drivers/nv40/nv40_draw.c new file mode 100644 index 0000000000..8e56cdc2fe --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_draw.c @@ -0,0 +1,349 @@ +#include "pipe/p_shader_tokens.h" + +#include "util/u_pack_color.h" + +#include "draw/draw_context.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pipe.h" + +#include "nv40_context.h" +#define NV40_SHADER_NO_FUCKEDNESS +#include "nv40_shader.h" + +/* Simple, but crappy, swtnl path, hopefully we wont need to hit this very + * often at all. Uses "quadro style" vertex submission + a fixed vertex + * layout to avoid the need to generate a vertex program or vtxfmt. + */ + +struct nv40_render_stage { + struct draw_stage stage; + struct nv40_context *nv40; + unsigned prim; +}; + +static INLINE struct nv40_render_stage * +nv40_render_stage(struct draw_stage *stage) +{ + return (struct nv40_render_stage *)stage; +} + +static INLINE void +nv40_render_vertex(struct nv40_context *nv40, const struct vertex_header *v) +{ + unsigned i; + + for (i = 0; i < nv40->swtnl.nr_attribs; i++) { + unsigned idx = nv40->swtnl.draw[i]; + unsigned hw = nv40->swtnl.hw[i]; + + switch (nv40->swtnl.emit[i]) { + case EMIT_OMIT: + break; + case EMIT_1F: + BEGIN_RING(curie, NV40TCL_VTX_ATTR_1F(hw), 1); + OUT_RING (fui(v->data[idx][0])); + break; + case EMIT_2F: + BEGIN_RING(curie, NV40TCL_VTX_ATTR_2F_X(hw), 2); + OUT_RING (fui(v->data[idx][0])); + OUT_RING (fui(v->data[idx][1])); + break; + case EMIT_3F: + BEGIN_RING(curie, NV40TCL_VTX_ATTR_3F_X(hw), 3); + OUT_RING (fui(v->data[idx][0])); + OUT_RING (fui(v->data[idx][1])); + OUT_RING (fui(v->data[idx][2])); + break; + case EMIT_4F: + BEGIN_RING(curie, NV40TCL_VTX_ATTR_4F_X(hw), 4); + OUT_RING (fui(v->data[idx][0])); + OUT_RING (fui(v->data[idx][1])); + OUT_RING (fui(v->data[idx][2])); + OUT_RING (fui(v->data[idx][3])); + break; + case EMIT_4UB: + BEGIN_RING(curie, NV40TCL_VTX_ATTR_4UB(hw), 1); + OUT_RING (pack_ub4(float_to_ubyte(v->data[idx][0]), + float_to_ubyte(v->data[idx][1]), + float_to_ubyte(v->data[idx][2]), + float_to_ubyte(v->data[idx][3]))); + break; + default: + assert(0); + break; + } + } +} + +static INLINE void +nv40_render_prim(struct draw_stage *stage, struct prim_header *prim, + unsigned mode, unsigned count) +{ + struct nv40_render_stage *rs = nv40_render_stage(stage); + struct nv40_context *nv40 = rs->nv40; + struct nouveau_pushbuf *pb = nv40->nvws->channel->pushbuf; + unsigned i; + + /* Ensure there's room for 4xfloat32 + potentially 3 begin/end */ + if (pb->remaining < ((count * 20) + 6)) { + if (rs->prim != NV40TCL_BEGIN_END_STOP) { + NOUVEAU_ERR("AIII, missed flush\n"); + assert(0); + } + FIRE_RING(NULL); + nv40_state_emit(nv40); + } + + /* Switch primitive modes if necessary */ + if (rs->prim != mode) { + if (rs->prim != NV40TCL_BEGIN_END_STOP) { + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (NV40TCL_BEGIN_END_STOP); + } + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (mode); + rs->prim = mode; + } + + /* Emit vertex data */ + for (i = 0; i < count; i++) + nv40_render_vertex(nv40, prim->v[i]); + + /* If it's likely we'll need to empty the push buffer soon, finish + * off the primitive now. + */ + if (pb->remaining < ((count * 20) + 6)) { + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (NV40TCL_BEGIN_END_STOP); + rs->prim = NV40TCL_BEGIN_END_STOP; + } +} + +static void +nv40_render_point(struct draw_stage *draw, struct prim_header *prim) +{ + nv40_render_prim(draw, prim, NV40TCL_BEGIN_END_POINTS, 1); +} + +static void +nv40_render_line(struct draw_stage *draw, struct prim_header *prim) +{ + nv40_render_prim(draw, prim, NV40TCL_BEGIN_END_LINES, 2); +} + +static void +nv40_render_tri(struct draw_stage *draw, struct prim_header *prim) +{ + nv40_render_prim(draw, prim, NV40TCL_BEGIN_END_TRIANGLES, 3); +} + +static void +nv40_render_flush(struct draw_stage *draw, unsigned flags) +{ + struct nv40_render_stage *rs = nv40_render_stage(draw); + struct nv40_context *nv40 = rs->nv40; + + if (rs->prim != NV40TCL_BEGIN_END_STOP) { + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (NV40TCL_BEGIN_END_STOP); + rs->prim = NV40TCL_BEGIN_END_STOP; + } +} + +static void +nv40_render_reset_stipple_counter(struct draw_stage *draw) +{ +} + +static void +nv40_render_destroy(struct draw_stage *draw) +{ + FREE(draw); +} + +static INLINE void +emit_mov(struct nv40_vertex_program *vp, + unsigned dst, unsigned src, unsigned vor, unsigned mask) +{ + struct nv40_vertex_program_exec *inst; + + vp->insns = realloc(vp->insns, + sizeof(struct nv40_vertex_program_exec) * + ++vp->nr_insns); + inst = &vp->insns[vp->nr_insns - 1]; + + inst->data[0] = 0x401f9c6c; + inst->data[1] = 0x0040000d | (src << 8); + inst->data[2] = 0x8106c083; + inst->data[3] = 0x6041ff80 | (dst << 2) | (mask << 13); + inst->const_index = -1; + inst->has_branch_offset = FALSE; + + vp->ir |= (1 << src); + if (vor != ~0) + vp->or |= (1 << vor); +} + +static struct nv40_vertex_program * +create_drawvp(struct nv40_context *nv40) +{ + struct nv40_vertex_program *vp = CALLOC_STRUCT(nv40_vertex_program); + unsigned i; + + emit_mov(vp, NV40_VP_INST_DEST_POS, 0, ~0, 0xf); + emit_mov(vp, NV40_VP_INST_DEST_COL0, 3, 0, 0xf); + emit_mov(vp, NV40_VP_INST_DEST_COL1, 4, 1, 0xf); + emit_mov(vp, NV40_VP_INST_DEST_BFC0, 3, 2, 0xf); + emit_mov(vp, NV40_VP_INST_DEST_BFC1, 4, 3, 0xf); + emit_mov(vp, NV40_VP_INST_DEST_FOGC, 5, 4, 0x8); + for (i = 0; i < 8; i++) + emit_mov(vp, NV40_VP_INST_DEST_TC(i), 8 + i, 14 + i, 0xf); + + vp->insns[vp->nr_insns - 1].data[3] |= 1; + vp->translated = TRUE; + return vp; +} + +struct draw_stage * +nv40_draw_render_stage(struct nv40_context *nv40) +{ + struct nv40_render_stage *render = CALLOC_STRUCT(nv40_render_stage); + + if (!nv40->swtnl.vertprog) + nv40->swtnl.vertprog = create_drawvp(nv40); + + render->nv40 = nv40; + render->stage.draw = nv40->draw; + render->stage.point = nv40_render_point; + render->stage.line = nv40_render_line; + render->stage.tri = nv40_render_tri; + render->stage.flush = nv40_render_flush; + render->stage.reset_stipple_counter = nv40_render_reset_stipple_counter; + render->stage.destroy = nv40_render_destroy; + + return &render->stage; +} + +boolean +nv40_draw_elements_swtnl(struct pipe_context *pipe, + struct pipe_buffer *idxbuf, unsigned idxbuf_size, + unsigned mode, unsigned start, unsigned count) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + unsigned i; + void *map; + + if (!nv40_state_validate_swtnl(nv40)) + return FALSE; + nv40->state.dirty &= ~(1ULL << NV40_STATE_VTXBUF); + nv40_state_emit(nv40); + + for (i = 0; i < nv40->vtxbuf_nr; i++) { + map = ws->buffer_map(ws, nv40->vtxbuf[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(nv40->draw, i, map); + } + + if (idxbuf) { + map = ws->buffer_map(ws, idxbuf, PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_element_buffer(nv40->draw, idxbuf_size, map); + } else { + draw_set_mapped_element_buffer(nv40->draw, 0, NULL); + } + + if (nv40->constbuf[PIPE_SHADER_VERTEX]) { + const unsigned nr = nv40->constbuf_nr[PIPE_SHADER_VERTEX]; + + map = ws->buffer_map(ws, nv40->constbuf[PIPE_SHADER_VERTEX], + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_constant_buffer(nv40->draw, map, nr); + } + + draw_arrays(nv40->draw, mode, start, count); + + for (i = 0; i < nv40->vtxbuf_nr; i++) + ws->buffer_unmap(ws, nv40->vtxbuf[i].buffer); + + if (idxbuf) + ws->buffer_unmap(ws, idxbuf); + + if (nv40->constbuf[PIPE_SHADER_VERTEX]) + ws->buffer_unmap(ws, nv40->constbuf[PIPE_SHADER_VERTEX]); + + draw_flush(nv40->draw); + pipe->flush(pipe, 0, NULL); + + return TRUE; +} + +static INLINE void +emit_attrib(struct nv40_context *nv40, unsigned hw, unsigned emit, + unsigned semantic, unsigned index) +{ + unsigned draw_out = draw_find_vs_output(nv40->draw, semantic, index); + unsigned a = nv40->swtnl.nr_attribs++; + + nv40->swtnl.hw[a] = hw; + nv40->swtnl.emit[a] = emit; + nv40->swtnl.draw[a] = draw_out; +} + +static boolean +nv40_state_vtxfmt_validate(struct nv40_context *nv40) +{ + struct nv40_fragment_program *fp = nv40->fragprog; + unsigned colour = 0, texcoords = 0, fog = 0, i; + + /* Determine needed fragprog inputs */ + for (i = 0; i < fp->info.num_inputs; i++) { + switch (fp->info.input_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + break; + case TGSI_SEMANTIC_COLOR: + colour |= (1 << fp->info.input_semantic_index[i]); + break; + case TGSI_SEMANTIC_GENERIC: + texcoords |= (1 << fp->info.input_semantic_index[i]); + break; + case TGSI_SEMANTIC_FOG: + fog = 1; + break; + default: + assert(0); + } + } + + nv40->swtnl.nr_attribs = 0; + + /* Map draw vtxprog output to hw attribute IDs */ + for (i = 0; i < 2; i++) { + if (!(colour & (1 << i))) + continue; + emit_attrib(nv40, 3 + i, EMIT_4UB, TGSI_SEMANTIC_COLOR, i); + } + + for (i = 0; i < 8; i++) { + if (!(texcoords & (1 << i))) + continue; + emit_attrib(nv40, 8 + i, EMIT_4F, TGSI_SEMANTIC_GENERIC, i); + } + + if (fog) { + emit_attrib(nv40, 5, EMIT_1F, TGSI_SEMANTIC_FOG, 0); + } + + emit_attrib(nv40, 0, EMIT_3F, TGSI_SEMANTIC_POSITION, 0); + + return FALSE; +} + +struct nv40_state_entry nv40_state_vtxfmt = { + .validate = nv40_state_vtxfmt_validate, + .dirty = { + .pipe = NV40_NEW_ARRAYS | NV40_NEW_FRAGPROG, + .hw = 0 + } +}; + diff --git a/src/gallium/drivers/nv40/nv40_fragprog.c b/src/gallium/drivers/nv40/nv40_fragprog.c new file mode 100644 index 0000000000..91dcbebda0 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_fragprog.c @@ -0,0 +1,991 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" + +#include "nv40_context.h" + +#define SWZ_X 0 +#define SWZ_Y 1 +#define SWZ_Z 2 +#define SWZ_W 3 +#define MASK_X 1 +#define MASK_Y 2 +#define MASK_Z 4 +#define MASK_W 8 +#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W) +#define DEF_SCALE NV40_FP_OP_DST_SCALE_1X +#define DEF_CTEST NV40_FP_OP_COND_TR +#include "nv40_shader.h" + +#define swz(s,x,y,z,w) nv40_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w) +#define neg(s) nv40_sr_neg((s)) +#define abs(s) nv40_sr_abs((s)) +#define scale(s,v) nv40_sr_scale((s), NV40_FP_OP_DST_SCALE_##v) + +#define MAX_CONSTS 128 +#define MAX_IMM 32 +struct nv40_fpc { + struct nv40_fragment_program *fp; + + uint attrib_map[PIPE_MAX_SHADER_INPUTS]; + + unsigned r_temps; + unsigned r_temps_discard; + struct nv40_sreg r_result[PIPE_MAX_SHADER_OUTPUTS]; + struct nv40_sreg *r_temp; + + int num_regs; + + unsigned inst_offset; + unsigned have_const; + + struct { + int pipe; + float vals[4]; + } consts[MAX_CONSTS]; + int nr_consts; + + struct nv40_sreg imm[MAX_IMM]; + unsigned nr_imm; +}; + +static INLINE struct nv40_sreg +temp(struct nv40_fpc *fpc) +{ + int idx = ffs(~fpc->r_temps) - 1; + + if (idx < 0) { + NOUVEAU_ERR("out of temps!!\n"); + assert(0); + return nv40_sr(NV40SR_TEMP, 0); + } + + fpc->r_temps |= (1 << idx); + fpc->r_temps_discard |= (1 << idx); + return nv40_sr(NV40SR_TEMP, idx); +} + +static INLINE void +release_temps(struct nv40_fpc *fpc) +{ + fpc->r_temps &= ~fpc->r_temps_discard; + fpc->r_temps_discard = 0; +} + +static INLINE struct nv40_sreg +constant(struct nv40_fpc *fpc, int pipe, float vals[4]) +{ + int idx; + + if (fpc->nr_consts == MAX_CONSTS) + assert(0); + idx = fpc->nr_consts++; + + fpc->consts[idx].pipe = pipe; + if (pipe == -1) + memcpy(fpc->consts[idx].vals, vals, 4 * sizeof(float)); + return nv40_sr(NV40SR_CONST, idx); +} + +#define arith(cc,s,o,d,m,s0,s1,s2) \ + nv40_fp_arith((cc), (s), NV40_FP_OP_OPCODE_##o, \ + (d), (m), (s0), (s1), (s2)) +#define tex(cc,s,o,u,d,m,s0,s1,s2) \ + nv40_fp_tex((cc), (s), NV40_FP_OP_OPCODE_##o, (u), \ + (d), (m), (s0), none, none) + +static void +grow_insns(struct nv40_fpc *fpc, int size) +{ + struct nv40_fragment_program *fp = fpc->fp; + + fp->insn_len += size; + fp->insn = realloc(fp->insn, sizeof(uint32_t) * fp->insn_len); +} + +static void +emit_src(struct nv40_fpc *fpc, int pos, struct nv40_sreg src) +{ + struct nv40_fragment_program *fp = fpc->fp; + uint32_t *hw = &fp->insn[fpc->inst_offset]; + uint32_t sr = 0; + + switch (src.type) { + case NV40SR_INPUT: + sr |= (NV40_FP_REG_TYPE_INPUT << NV40_FP_REG_TYPE_SHIFT); + hw[0] |= (src.index << NV40_FP_OP_INPUT_SRC_SHIFT); + break; + case NV40SR_OUTPUT: + sr |= NV40_FP_REG_SRC_HALF; + /* fall-through */ + case NV40SR_TEMP: + sr |= (NV40_FP_REG_TYPE_TEMP << NV40_FP_REG_TYPE_SHIFT); + sr |= (src.index << NV40_FP_REG_SRC_SHIFT); + break; + case NV40SR_CONST: + if (!fpc->have_const) { + grow_insns(fpc, 4); + fpc->have_const = 1; + } + + hw = &fp->insn[fpc->inst_offset]; + if (fpc->consts[src.index].pipe >= 0) { + struct nv40_fragment_program_data *fpd; + + fp->consts = realloc(fp->consts, ++fp->nr_consts * + sizeof(*fpd)); + fpd = &fp->consts[fp->nr_consts - 1]; + fpd->offset = fpc->inst_offset + 4; + fpd->index = fpc->consts[src.index].pipe; + memset(&fp->insn[fpd->offset], 0, sizeof(uint32_t) * 4); + } else { + memcpy(&fp->insn[fpc->inst_offset + 4], + fpc->consts[src.index].vals, + sizeof(uint32_t) * 4); + } + + sr |= (NV40_FP_REG_TYPE_CONST << NV40_FP_REG_TYPE_SHIFT); + break; + case NV40SR_NONE: + sr |= (NV40_FP_REG_TYPE_INPUT << NV40_FP_REG_TYPE_SHIFT); + break; + default: + assert(0); + } + + if (src.negate) + sr |= NV40_FP_REG_NEGATE; + + if (src.abs) + hw[1] |= (1 << (29 + pos)); + + sr |= ((src.swz[0] << NV40_FP_REG_SWZ_X_SHIFT) | + (src.swz[1] << NV40_FP_REG_SWZ_Y_SHIFT) | + (src.swz[2] << NV40_FP_REG_SWZ_Z_SHIFT) | + (src.swz[3] << NV40_FP_REG_SWZ_W_SHIFT)); + + hw[pos + 1] |= sr; +} + +static void +emit_dst(struct nv40_fpc *fpc, struct nv40_sreg dst) +{ + struct nv40_fragment_program *fp = fpc->fp; + uint32_t *hw = &fp->insn[fpc->inst_offset]; + + switch (dst.type) { + case NV40SR_TEMP: + if (fpc->num_regs < (dst.index + 1)) + fpc->num_regs = dst.index + 1; + break; + case NV40SR_OUTPUT: + if (dst.index == 1) { + fp->fp_control |= 0xe; + } else { + hw[0] |= NV40_FP_OP_OUT_REG_HALF; + } + break; + case NV40SR_NONE: + hw[0] |= (1 << 30); + break; + default: + assert(0); + } + + hw[0] |= (dst.index << NV40_FP_OP_OUT_REG_SHIFT); +} + +static void +nv40_fp_arith(struct nv40_fpc *fpc, int sat, int op, + struct nv40_sreg dst, int mask, + struct nv40_sreg s0, struct nv40_sreg s1, struct nv40_sreg s2) +{ + struct nv40_fragment_program *fp = fpc->fp; + uint32_t *hw; + + fpc->inst_offset = fp->insn_len; + fpc->have_const = 0; + grow_insns(fpc, 4); + hw = &fp->insn[fpc->inst_offset]; + memset(hw, 0, sizeof(uint32_t) * 4); + + if (op == NV40_FP_OP_OPCODE_KIL) + fp->fp_control |= NV40TCL_FP_CONTROL_KIL; + hw[0] |= (op << NV40_FP_OP_OPCODE_SHIFT); + hw[0] |= (mask << NV40_FP_OP_OUTMASK_SHIFT); + hw[2] |= (dst.dst_scale << NV40_FP_OP_DST_SCALE_SHIFT); + + if (sat) + hw[0] |= NV40_FP_OP_OUT_SAT; + + if (dst.cc_update) + hw[0] |= NV40_FP_OP_COND_WRITE_ENABLE; + hw[1] |= (dst.cc_test << NV40_FP_OP_COND_SHIFT); + hw[1] |= ((dst.cc_swz[0] << NV40_FP_OP_COND_SWZ_X_SHIFT) | + (dst.cc_swz[1] << NV40_FP_OP_COND_SWZ_Y_SHIFT) | + (dst.cc_swz[2] << NV40_FP_OP_COND_SWZ_Z_SHIFT) | + (dst.cc_swz[3] << NV40_FP_OP_COND_SWZ_W_SHIFT)); + + emit_dst(fpc, dst); + emit_src(fpc, 0, s0); + emit_src(fpc, 1, s1); + emit_src(fpc, 2, s2); +} + +static void +nv40_fp_tex(struct nv40_fpc *fpc, int sat, int op, int unit, + struct nv40_sreg dst, int mask, + struct nv40_sreg s0, struct nv40_sreg s1, struct nv40_sreg s2) +{ + struct nv40_fragment_program *fp = fpc->fp; + + nv40_fp_arith(fpc, sat, op, dst, mask, s0, s1, s2); + + fp->insn[fpc->inst_offset] |= (unit << NV40_FP_OP_TEX_UNIT_SHIFT); + fp->samplers |= (1 << unit); +} + +static INLINE struct nv40_sreg +tgsi_src(struct nv40_fpc *fpc, const struct tgsi_full_src_register *fsrc) +{ + struct nv40_sreg src; + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + src = nv40_sr(NV40SR_INPUT, + fpc->attrib_map[fsrc->SrcRegister.Index]); + break; + case TGSI_FILE_CONSTANT: + src = constant(fpc, fsrc->SrcRegister.Index, NULL); + break; + case TGSI_FILE_IMMEDIATE: + assert(fsrc->SrcRegister.Index < fpc->nr_imm); + src = fpc->imm[fsrc->SrcRegister.Index]; + break; + case TGSI_FILE_TEMPORARY: + src = fpc->r_temp[fsrc->SrcRegister.Index]; + break; + /* NV40 fragprog result regs are just temps, so this is simple */ + case TGSI_FILE_OUTPUT: + src = fpc->r_result[fsrc->SrcRegister.Index]; + break; + default: + NOUVEAU_ERR("bad src file\n"); + break; + } + + src.abs = fsrc->SrcRegisterExtMod.Absolute; + src.negate = fsrc->SrcRegister.Negate; + src.swz[0] = fsrc->SrcRegister.SwizzleX; + src.swz[1] = fsrc->SrcRegister.SwizzleY; + src.swz[2] = fsrc->SrcRegister.SwizzleZ; + src.swz[3] = fsrc->SrcRegister.SwizzleW; + return src; +} + +static INLINE struct nv40_sreg +tgsi_dst(struct nv40_fpc *fpc, const struct tgsi_full_dst_register *fdst) { + switch (fdst->DstRegister.File) { + case TGSI_FILE_OUTPUT: + return fpc->r_result[fdst->DstRegister.Index]; + case TGSI_FILE_TEMPORARY: + return fpc->r_temp[fdst->DstRegister.Index]; + case TGSI_FILE_NULL: + return nv40_sr(NV40SR_NONE, 0); + default: + NOUVEAU_ERR("bad dst file %d\n", fdst->DstRegister.File); + return nv40_sr(NV40SR_NONE, 0); + } +} + +static INLINE int +tgsi_mask(uint tgsi) +{ + int mask = 0; + + if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X; + if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y; + if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z; + if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W; + return mask; +} + +static boolean +src_native_swz(struct nv40_fpc *fpc, const struct tgsi_full_src_register *fsrc, + struct nv40_sreg *src) +{ + const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0); + struct nv40_sreg tgsi = tgsi_src(fpc, fsrc); + uint mask = 0, zero_mask = 0, one_mask = 0, neg_mask = 0; + uint neg[4] = { fsrc->SrcRegisterExtSwz.NegateX, + fsrc->SrcRegisterExtSwz.NegateY, + fsrc->SrcRegisterExtSwz.NegateZ, + fsrc->SrcRegisterExtSwz.NegateW }; + uint c; + + for (c = 0; c < 4; c++) { + switch (tgsi_util_get_full_src_register_extswizzle(fsrc, c)) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + mask |= (1 << c); + break; + case TGSI_EXTSWIZZLE_ZERO: + zero_mask |= (1 << c); + tgsi.swz[c] = SWZ_X; + break; + case TGSI_EXTSWIZZLE_ONE: + one_mask |= (1 << c); + tgsi.swz[c] = SWZ_X; + break; + default: + assert(0); + } + + if (!tgsi.negate && neg[c]) + neg_mask |= (1 << c); + } + + if (mask == MASK_ALL && !neg_mask) + return TRUE; + + *src = temp(fpc); + + if (mask) + arith(fpc, 0, MOV, *src, mask, tgsi, none, none); + + if (zero_mask) + arith(fpc, 0, SFL, *src, zero_mask, *src, none, none); + + if (one_mask) + arith(fpc, 0, STR, *src, one_mask, *src, none, none); + + if (neg_mask) { + struct nv40_sreg one = temp(fpc); + arith(fpc, 0, STR, one, neg_mask, one, none, none); + arith(fpc, 0, MUL, *src, neg_mask, *src, neg(one), none); + } + + return FALSE; +} + +static boolean +nv40_fragprog_parse_instruction(struct nv40_fpc *fpc, + const struct tgsi_full_instruction *finst) +{ + const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0); + struct nv40_sreg src[3], dst, tmp; + int mask, sat, unit; + int ai = -1, ci = -1, ii = -1; + int i; + + if (finst->Instruction.Opcode == TGSI_OPCODE_END) + return TRUE; + + for (i = 0; i < finst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *fsrc; + + fsrc = &finst->FullSrcRegisters[i]; + if (fsrc->SrcRegister.File == TGSI_FILE_TEMPORARY) { + src[i] = tgsi_src(fpc, fsrc); + } + } + + for (i = 0; i < finst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *fsrc; + + fsrc = &finst->FullSrcRegisters[i]; + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + case TGSI_FILE_CONSTANT: + case TGSI_FILE_TEMPORARY: + if (!src_native_swz(fpc, fsrc, &src[i])) + continue; + break; + default: + break; + } + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + if (ai == -1 || ai == fsrc->SrcRegister.Index) { + ai = fsrc->SrcRegister.Index; + src[i] = tgsi_src(fpc, fsrc); + } else { + src[i] = temp(fpc); + arith(fpc, 0, MOV, src[i], MASK_ALL, + tgsi_src(fpc, fsrc), none, none); + } + break; + case TGSI_FILE_CONSTANT: + if ((ci == -1 && ii == -1) || + ci == fsrc->SrcRegister.Index) { + ci = fsrc->SrcRegister.Index; + src[i] = tgsi_src(fpc, fsrc); + } else { + src[i] = temp(fpc); + arith(fpc, 0, MOV, src[i], MASK_ALL, + tgsi_src(fpc, fsrc), none, none); + } + break; + case TGSI_FILE_IMMEDIATE: + if ((ci == -1 && ii == -1) || + ii == fsrc->SrcRegister.Index) { + ii = fsrc->SrcRegister.Index; + src[i] = tgsi_src(fpc, fsrc); + } else { + src[i] = temp(fpc); + arith(fpc, 0, MOV, src[i], MASK_ALL, + tgsi_src(fpc, fsrc), none, none); + } + break; + case TGSI_FILE_TEMPORARY: + /* handled above */ + break; + case TGSI_FILE_SAMPLER: + unit = fsrc->SrcRegister.Index; + break; + case TGSI_FILE_OUTPUT: + break; + default: + NOUVEAU_ERR("bad src file\n"); + return FALSE; + } + } + + dst = tgsi_dst(fpc, &finst->FullDstRegisters[0]); + mask = tgsi_mask(finst->FullDstRegisters[0].DstRegister.WriteMask); + sat = (finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE); + + switch (finst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + arith(fpc, sat, MOV, dst, mask, abs(src[0]), none, none); + break; + case TGSI_OPCODE_ADD: + arith(fpc, sat, ADD, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_CMP: + tmp = temp(fpc); + arith(fpc, sat, MOV, dst, mask, src[2], none, none); + tmp.cc_update = 1; + arith(fpc, 0, MOV, tmp, 0xf, src[0], none, none); + dst.cc_test = NV40_VP_INST_COND_LT; + arith(fpc, sat, MOV, dst, mask, src[1], none, none); + break; + case TGSI_OPCODE_COS: + arith(fpc, sat, COS, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_DDX: + if (mask & (MASK_Z | MASK_W)) { + tmp = temp(fpc); + arith(fpc, sat, DDX, tmp, MASK_X | MASK_Y, + swz(src[0], Z, W, Z, W), none, none); + arith(fpc, 0, MOV, tmp, MASK_Z | MASK_W, + swz(tmp, X, Y, X, Y), none, none); + arith(fpc, sat, DDX, tmp, MASK_X | MASK_Y, src[0], + none, none); + arith(fpc, 0, MOV, dst, mask, tmp, none, none); + } else { + arith(fpc, sat, DDX, dst, mask, src[0], none, none); + } + break; + case TGSI_OPCODE_DDY: + if (mask & (MASK_Z | MASK_W)) { + tmp = temp(fpc); + arith(fpc, sat, DDY, tmp, MASK_X | MASK_Y, + swz(src[0], Z, W, Z, W), none, none); + arith(fpc, 0, MOV, tmp, MASK_Z | MASK_W, + swz(tmp, X, Y, X, Y), none, none); + arith(fpc, sat, DDY, tmp, MASK_X | MASK_Y, src[0], + none, none); + arith(fpc, 0, MOV, dst, mask, tmp, none, none); + } else { + arith(fpc, sat, DDY, dst, mask, src[0], none, none); + } + break; + case TGSI_OPCODE_DP3: + arith(fpc, sat, DP3, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DP4: + arith(fpc, sat, DP4, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DPH: + tmp = temp(fpc); + arith(fpc, 0, DP3, tmp, MASK_X, src[0], src[1], none); + arith(fpc, sat, ADD, dst, mask, swz(tmp, X, X, X, X), + swz(src[1], W, W, W, W), none); + break; + case TGSI_OPCODE_DST: + arith(fpc, sat, DST, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_EX2: + arith(fpc, sat, EX2, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_FLR: + arith(fpc, sat, FLR, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_FRC: + arith(fpc, sat, FRC, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_KILP: + arith(fpc, 0, KIL, none, 0, none, none, none); + break; + case TGSI_OPCODE_KIL: + dst = nv40_sr(NV40SR_NONE, 0); + dst.cc_update = 1; + arith(fpc, 0, MOV, dst, MASK_ALL, src[0], none, none); + dst.cc_update = 0; dst.cc_test = NV40_FP_OP_COND_LT; + arith(fpc, 0, KIL, dst, 0, none, none, none); + break; + case TGSI_OPCODE_LG2: + arith(fpc, sat, LG2, dst, mask, src[0], none, none); + break; +// case TGSI_OPCODE_LIT: + case TGSI_OPCODE_LRP: + tmp = temp(fpc); + arith(fpc, 0, MAD, tmp, mask, neg(src[0]), src[2], src[2]); + arith(fpc, sat, MAD, dst, mask, src[0], src[1], tmp); + break; + case TGSI_OPCODE_MAD: + arith(fpc, sat, MAD, dst, mask, src[0], src[1], src[2]); + break; + case TGSI_OPCODE_MAX: + arith(fpc, sat, MAX, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_MIN: + arith(fpc, sat, MIN, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_MOV: + arith(fpc, sat, MOV, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_MUL: + arith(fpc, sat, MUL, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_NOISE1: + case TGSI_OPCODE_NOISE2: + case TGSI_OPCODE_NOISE3: + case TGSI_OPCODE_NOISE4: + arith(fpc, sat, SFL, dst, mask, none, none, none); + break; + case TGSI_OPCODE_POW: + tmp = temp(fpc); + arith(fpc, 0, LG2, tmp, MASK_X, + swz(src[0], X, X, X, X), none, none); + arith(fpc, 0, MUL, tmp, MASK_X, swz(tmp, X, X, X, X), + swz(src[1], X, X, X, X), none); + arith(fpc, sat, EX2, dst, mask, + swz(tmp, X, X, X, X), none, none); + break; + case TGSI_OPCODE_RCP: + arith(fpc, sat, RCP, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_RET: + assert(0); + break; + case TGSI_OPCODE_RFL: + tmp = temp(fpc); + arith(fpc, 0, DP3, tmp, MASK_X, src[0], src[0], none); + arith(fpc, 0, DP3, tmp, MASK_Y, src[0], src[1], none); + arith(fpc, 0, DIV, scale(tmp, 2X), MASK_Z, + swz(tmp, Y, Y, Y, Y), swz(tmp, X, X, X, X), none); + arith(fpc, sat, MAD, dst, mask, + swz(tmp, Z, Z, Z, Z), src[0], neg(src[1])); + break; + case TGSI_OPCODE_RSQ: + tmp = temp(fpc); + arith(fpc, 0, LG2, scale(tmp, INV_2X), MASK_X, + abs(swz(src[0], X, X, X, X)), none, none); + arith(fpc, sat, EX2, dst, mask, + neg(swz(tmp, X, X, X, X)), none, none); + break; + case TGSI_OPCODE_SCS: + if (mask & MASK_X) { + arith(fpc, sat, COS, dst, MASK_X, + swz(src[0], X, X, X, X), none, none); + } + if (mask & MASK_Y) { + arith(fpc, sat, SIN, dst, MASK_Y, + swz(src[0], X, X, X, X), none, none); + } + break; + case TGSI_OPCODE_SEQ: + arith(fpc, sat, SEQ, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SFL: + arith(fpc, sat, SFL, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SGE: + arith(fpc, sat, SGE, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SGT: + arith(fpc, sat, SGT, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SIN: + arith(fpc, sat, SIN, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_SLE: + arith(fpc, sat, SLE, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SLT: + arith(fpc, sat, SLT, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SNE: + arith(fpc, sat, SNE, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_STR: + arith(fpc, sat, STR, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SUB: + arith(fpc, sat, ADD, dst, mask, src[0], neg(src[1]), none); + break; + case TGSI_OPCODE_TEX: + tex(fpc, sat, TEX, unit, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_TXB: + tex(fpc, sat, TXB, unit, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_TXP: + tex(fpc, sat, TXP, unit, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_XPD: + tmp = temp(fpc); + arith(fpc, 0, MUL, tmp, mask, + swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none); + arith(fpc, sat, MAD, dst, (mask & ~MASK_W), + swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y), + neg(tmp)); + break; + default: + NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode); + return FALSE; + } + + release_temps(fpc); + return TRUE; +} + +static boolean +nv40_fragprog_parse_decl_attrib(struct nv40_fpc *fpc, + const struct tgsi_full_declaration *fdec) +{ + int hw; + + switch (fdec->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: + hw = NV40_FP_OP_INPUT_SRC_POSITION; + break; + case TGSI_SEMANTIC_COLOR: + if (fdec->Semantic.SemanticIndex == 0) { + hw = NV40_FP_OP_INPUT_SRC_COL0; + } else + if (fdec->Semantic.SemanticIndex == 1) { + hw = NV40_FP_OP_INPUT_SRC_COL1; + } else { + NOUVEAU_ERR("bad colour semantic index\n"); + return FALSE; + } + break; + case TGSI_SEMANTIC_FOG: + hw = NV40_FP_OP_INPUT_SRC_FOGC; + break; + case TGSI_SEMANTIC_GENERIC: + if (fdec->Semantic.SemanticIndex <= 7) { + hw = NV40_FP_OP_INPUT_SRC_TC(fdec->Semantic. + SemanticIndex); + } else { + NOUVEAU_ERR("bad generic semantic index\n"); + return FALSE; + } + break; + default: + NOUVEAU_ERR("bad input semantic\n"); + return FALSE; + } + + fpc->attrib_map[fdec->DeclarationRange.First] = hw; + return TRUE; +} + +static boolean +nv40_fragprog_parse_decl_output(struct nv40_fpc *fpc, + const struct tgsi_full_declaration *fdec) +{ + unsigned idx = fdec->DeclarationRange.First; + unsigned hw; + + switch (fdec->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: + hw = 1; + break; + case TGSI_SEMANTIC_COLOR: + switch (fdec->Semantic.SemanticIndex) { + case 0: hw = 0; break; + case 1: hw = 2; break; + case 2: hw = 3; break; + case 3: hw = 4; break; + default: + NOUVEAU_ERR("bad rcol index\n"); + return FALSE; + } + break; + default: + NOUVEAU_ERR("bad output semantic\n"); + return FALSE; + } + + fpc->r_result[idx] = nv40_sr(NV40SR_OUTPUT, hw); + fpc->r_temps |= (1 << hw); + return TRUE; +} + +static boolean +nv40_fragprog_prepare(struct nv40_fpc *fpc) +{ + struct tgsi_parse_context p; + int high_temp = -1, i; + + tgsi_parse_init(&p, fpc->fp->pipe.tokens); + while (!tgsi_parse_end_of_tokens(&p)) { + const union tgsi_full_token *tok = &p.FullToken; + + tgsi_parse_token(&p); + switch(tok->Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + const struct tgsi_full_declaration *fdec; + fdec = &p.FullToken.FullDeclaration; + switch (fdec->Declaration.File) { + case TGSI_FILE_INPUT: + if (!nv40_fragprog_parse_decl_attrib(fpc, fdec)) + goto out_err; + break; + case TGSI_FILE_OUTPUT: + if (!nv40_fragprog_parse_decl_output(fpc, fdec)) + goto out_err; + break; + case TGSI_FILE_TEMPORARY: + if (fdec->DeclarationRange.Last > high_temp) { + high_temp = + fdec->DeclarationRange.Last; + } + break; + default: + break; + } + } + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *imm; + float vals[4]; + + imm = &p.FullToken.FullImmediate; + assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32); + assert(fpc->nr_imm < MAX_IMM); + + vals[0] = imm->u.ImmediateFloat32[0].Float; + vals[1] = imm->u.ImmediateFloat32[1].Float; + vals[2] = imm->u.ImmediateFloat32[2].Float; + vals[3] = imm->u.ImmediateFloat32[3].Float; + fpc->imm[fpc->nr_imm++] = constant(fpc, -1, vals); + } + break; + default: + break; + } + } + tgsi_parse_free(&p); + + if (++high_temp) { + fpc->r_temp = CALLOC(high_temp, sizeof(struct nv40_sreg)); + for (i = 0; i < high_temp; i++) + fpc->r_temp[i] = temp(fpc); + fpc->r_temps_discard = 0; + } + + return TRUE; + +out_err: + if (fpc->r_temp) + FREE(fpc->r_temp); + tgsi_parse_free(&p); + return FALSE; +} + +static void +nv40_fragprog_translate(struct nv40_context *nv40, + struct nv40_fragment_program *fp) +{ + struct tgsi_parse_context parse; + struct nv40_fpc *fpc = NULL; + + fpc = CALLOC(1, sizeof(struct nv40_fpc)); + if (!fpc) + return; + fpc->fp = fp; + fpc->num_regs = 2; + + if (!nv40_fragprog_prepare(fpc)) { + FREE(fpc); + return; + } + + tgsi_parse_init(&parse, fp->pipe.tokens); + + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + const struct tgsi_full_instruction *finst; + + finst = &parse.FullToken.FullInstruction; + if (!nv40_fragprog_parse_instruction(fpc, finst)) + goto out_err; + } + break; + default: + break; + } + } + + fp->fp_control |= fpc->num_regs << NV40TCL_FP_CONTROL_TEMP_COUNT_SHIFT; + + /* Terminate final instruction */ + fp->insn[fpc->inst_offset] |= 0x00000001; + + /* Append NOP + END instruction, may or may not be necessary. */ + fpc->inst_offset = fp->insn_len; + grow_insns(fpc, 4); + fp->insn[fpc->inst_offset + 0] = 0x00000001; + fp->insn[fpc->inst_offset + 1] = 0x00000000; + fp->insn[fpc->inst_offset + 2] = 0x00000000; + fp->insn[fpc->inst_offset + 3] = 0x00000000; + + fp->translated = TRUE; +out_err: + tgsi_parse_free(&parse); + if (fpc->r_temp) + FREE(fpc->r_temp); + FREE(fpc); +} + +static void +nv40_fragprog_upload(struct nv40_context *nv40, + struct nv40_fragment_program *fp) +{ + struct pipe_winsys *ws = nv40->pipe.winsys; + const uint32_t le = 1; + uint32_t *map; + int i; + + map = ws->buffer_map(ws, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); + +#if 0 + for (i = 0; i < fp->insn_len; i++) { + fflush(stdout); fflush(stderr); + NOUVEAU_ERR("%d 0x%08x\n", i, fp->insn[i]); + fflush(stdout); fflush(stderr); + } +#endif + + if ((*(const uint8_t *)&le)) { + for (i = 0; i < fp->insn_len; i++) { + map[i] = fp->insn[i]; + } + } else { + /* Weird swapping for big-endian chips */ + for (i = 0; i < fp->insn_len; i++) { + map[i] = ((fp->insn[i] & 0xffff) << 16) | + ((fp->insn[i] >> 16) & 0xffff); + } + } + + ws->buffer_unmap(ws, fp->buffer); +} + +static boolean +nv40_fragprog_validate(struct nv40_context *nv40) +{ + struct nv40_fragment_program *fp = nv40->fragprog; + struct pipe_buffer *constbuf = + nv40->constbuf[PIPE_SHADER_FRAGMENT]; + struct pipe_winsys *ws = nv40->pipe.winsys; + struct nouveau_stateobj *so; + boolean new_consts = FALSE; + int i; + + if (fp->translated) + goto update_constants; + + nv40->fallback_swrast &= ~NV40_NEW_FRAGPROG; + nv40_fragprog_translate(nv40, fp); + if (!fp->translated) { + nv40->fallback_swrast |= NV40_NEW_FRAGPROG; + return FALSE; + } + + fp->buffer = ws->buffer_create(ws, 0x100, 0, fp->insn_len * 4); + nv40_fragprog_upload(nv40, fp); + + so = so_new(4, 1); + so_method(so, nv40->screen->curie, NV40TCL_FP_ADDRESS, 1); + so_reloc (so, fp->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | + NOUVEAU_BO_RD | NOUVEAU_BO_LOW | NOUVEAU_BO_OR, + NV40TCL_FP_ADDRESS_DMA0, NV40TCL_FP_ADDRESS_DMA1); + so_method(so, nv40->screen->curie, NV40TCL_FP_CONTROL, 1); + so_data (so, fp->fp_control); + so_ref(so, &fp->so); + +update_constants: + if (fp->nr_consts) { + float *map; + + map = ws->buffer_map(ws, constbuf, PIPE_BUFFER_USAGE_CPU_READ); + for (i = 0; i < fp->nr_consts; i++) { + struct nv40_fragment_program_data *fpd = &fp->consts[i]; + uint32_t *p = &fp->insn[fpd->offset]; + uint32_t *cb = (uint32_t *)&map[fpd->index * 4]; + + if (!memcmp(p, cb, 4 * sizeof(float))) + continue; + memcpy(p, cb, 4 * sizeof(float)); + new_consts = TRUE; + } + ws->buffer_unmap(ws, constbuf); + + if (new_consts) + nv40_fragprog_upload(nv40, fp); + } + + if (new_consts || fp->so != nv40->state.hw[NV40_STATE_FRAGPROG]) { + so_ref(fp->so, &nv40->state.hw[NV40_STATE_FRAGPROG]); + return TRUE; + } + + return FALSE; +} + +void +nv40_fragprog_destroy(struct nv40_context *nv40, + struct nv40_fragment_program *fp) +{ + if (fp->insn_len) + FREE(fp->insn); +} + +struct nv40_state_entry nv40_state_fragprog = { + .validate = nv40_fragprog_validate, + .dirty = { + .pipe = NV40_NEW_FRAGPROG, + .hw = NV40_STATE_FRAGPROG + } +}; + diff --git a/src/gallium/drivers/nv40/nv40_fragtex.c b/src/gallium/drivers/nv40/nv40_fragtex.c new file mode 100644 index 0000000000..0227d22620 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_fragtex.c @@ -0,0 +1,168 @@ +#include "nv40_context.h" + +#define _(m,tf,ts0x,ts0y,ts0z,ts0w,ts1x,ts1y,ts1z,ts1w,sx,sy,sz,sw) \ +{ \ + TRUE, \ + PIPE_FORMAT_##m, \ + NV40TCL_TEX_FORMAT_FORMAT_##tf, \ + (NV40TCL_TEX_SWIZZLE_S0_X_##ts0x | NV40TCL_TEX_SWIZZLE_S0_Y_##ts0y | \ + NV40TCL_TEX_SWIZZLE_S0_Z_##ts0z | NV40TCL_TEX_SWIZZLE_S0_W_##ts0w | \ + NV40TCL_TEX_SWIZZLE_S1_X_##ts1x | NV40TCL_TEX_SWIZZLE_S1_Y_##ts1y | \ + NV40TCL_TEX_SWIZZLE_S1_Z_##ts1z | NV40TCL_TEX_SWIZZLE_S1_W_##ts1w), \ + ((NV40TCL_TEX_FILTER_SIGNED_RED*sx) | (NV40TCL_TEX_FILTER_SIGNED_GREEN*sy) | \ + (NV40TCL_TEX_FILTER_SIGNED_BLUE*sz) | (NV40TCL_TEX_FILTER_SIGNED_ALPHA*sw)) \ +} + +struct nv40_texture_format { + boolean defined; + uint pipe; + int format; + int swizzle; + int sign; +}; + +static struct nv40_texture_format +nv40_texture_formats[] = { + _(A8R8G8B8_UNORM, A8R8G8B8, S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + _(A1R5G5B5_UNORM, A1R5G5B5, S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + _(A4R4G4B4_UNORM, A4R4G4B4, S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + _(R5G6B5_UNORM , R5G6B5 , S1, S1, S1, ONE, X, Y, Z, W, 0, 0, 0, 0), + _(L8_UNORM , L8 , S1, S1, S1, ONE, X, X, X, X, 0, 0, 0, 0), + _(A8_UNORM , L8 , ZERO, ZERO, ZERO, S1, X, X, X, X, 0, 0, 0, 0), + _(R16_SNORM , A16 , ZERO, ZERO, S1, ONE, X, X, X, Y, 1, 1, 1, 1), + _(I8_UNORM , L8 , S1, S1, S1, S1, X, X, X, X, 0, 0, 0, 0), + _(A8L8_UNORM , A8L8 , S1, S1, S1, S1, X, X, X, Y, 0, 0, 0, 0), + _(Z16_UNORM , Z16 , S1, S1, S1, ONE, X, X, X, X, 0, 0, 0, 0), + _(Z24S8_UNORM , Z24 , S1, S1, S1, ONE, X, X, X, X, 0, 0, 0, 0), + _(DXT1_RGB , DXT1 , S1, S1, S1, ONE, X, Y, Z, W, 0, 0, 0, 0), + _(DXT1_RGBA , DXT1 , S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + _(DXT3_RGBA , DXT3 , S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + _(DXT5_RGBA , DXT5 , S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + {}, +}; + +static struct nv40_texture_format * +nv40_fragtex_format(uint pipe_format) +{ + struct nv40_texture_format *tf = nv40_texture_formats; + + while (tf->defined) { + if (tf->pipe == pipe_format) + return tf; + tf++; + } + + NOUVEAU_ERR("unknown texture format %s\n", pf_name(pipe_format)); + return NULL; +} + + +static struct nouveau_stateobj * +nv40_fragtex_build(struct nv40_context *nv40, int unit) +{ + struct nv40_sampler_state *ps = nv40->tex_sampler[unit]; + struct nv40_miptree *nv40mt = nv40->tex_miptree[unit]; + struct pipe_texture *pt = &nv40mt->base; + struct nv40_texture_format *tf; + struct nouveau_stateobj *so; + uint32_t txf, txs, txp; + unsigned tex_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD; + + tf = nv40_fragtex_format(pt->format); + if (!tf) + assert(0); + + txf = ps->fmt; + txf |= tf->format | 0x8000; + txf |= ((pt->last_level + 1) << NV40TCL_TEX_FORMAT_MIPMAP_COUNT_SHIFT); + + if (1) /* XXX */ + txf |= NV40TCL_TEX_FORMAT_NO_BORDER; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE: + txf |= NV40TCL_TEX_FORMAT_CUBIC; + /* fall-through */ + case PIPE_TEXTURE_2D: + txf |= NV40TCL_TEX_FORMAT_DIMS_2D; + break; + case PIPE_TEXTURE_3D: + txf |= NV40TCL_TEX_FORMAT_DIMS_3D; + break; + case PIPE_TEXTURE_1D: + txf |= NV40TCL_TEX_FORMAT_DIMS_1D; + break; + default: + NOUVEAU_ERR("Unknown target %d\n", pt->target); + return NULL; + } + + if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + txp = 0; + } else { + txp = nv40mt->level[0].pitch; + txf |= NV40TCL_TEX_FORMAT_LINEAR; + } + + txs = tf->swizzle; + + so = so_new(16, 2); + so_method(so, nv40->screen->curie, NV40TCL_TEX_OFFSET(unit), 8); + so_reloc (so, nv40mt->buffer, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, nv40mt->buffer, txf, tex_flags | NOUVEAU_BO_OR, + NV40TCL_TEX_FORMAT_DMA0, NV40TCL_TEX_FORMAT_DMA1); + so_data (so, ps->wrap); + so_data (so, NV40TCL_TEX_ENABLE_ENABLE | ps->en); + so_data (so, txs); + so_data (so, ps->filt | tf->sign | 0x2000 /*voodoo*/); + so_data (so, (pt->width[0] << NV40TCL_TEX_SIZE0_W_SHIFT) | + pt->height[0]); + so_data (so, ps->bcol); + so_method(so, nv40->screen->curie, NV40TCL_TEX_SIZE1(unit), 1); + so_data (so, (pt->depth[0] << NV40TCL_TEX_SIZE1_DEPTH_SHIFT) | txp); + + return so; +} + +static boolean +nv40_fragtex_validate(struct nv40_context *nv40) +{ + struct nv40_fragment_program *fp = nv40->fragprog; + struct nv40_state *state = &nv40->state; + struct nouveau_stateobj *so; + unsigned samplers, unit; + + samplers = state->fp_samplers & ~fp->samplers; + while (samplers) { + unit = ffs(samplers) - 1; + samplers &= ~(1 << unit); + + so = so_new(2, 0); + so_method(so, nv40->screen->curie, NV40TCL_TEX_ENABLE(unit), 1); + so_data (so, 0); + so_ref(so, &nv40->state.hw[NV40_STATE_FRAGTEX0 + unit]); + state->dirty |= (1ULL << (NV40_STATE_FRAGTEX0 + unit)); + } + + samplers = nv40->dirty_samplers & fp->samplers; + while (samplers) { + unit = ffs(samplers) - 1; + samplers &= ~(1 << unit); + + so = nv40_fragtex_build(nv40, unit); + so_ref(so, &nv40->state.hw[NV40_STATE_FRAGTEX0 + unit]); + state->dirty |= (1ULL << (NV40_STATE_FRAGTEX0 + unit)); + } + + nv40->state.fp_samplers = fp->samplers; + return FALSE; +} + +struct nv40_state_entry nv40_state_fragtex = { + .validate = nv40_fragtex_validate, + .dirty = { + .pipe = NV40_NEW_SAMPLER | NV40_NEW_FRAGPROG, + .hw = 0 + } +}; + diff --git a/src/gallium/drivers/nv40/nv40_miptree.c b/src/gallium/drivers/nv40/nv40_miptree.c new file mode 100644 index 0000000000..b68967c07f --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_miptree.c @@ -0,0 +1,194 @@ +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "nv40_context.h" + +static void +nv40_miptree_layout(struct nv40_miptree *mt) +{ + struct pipe_texture *pt = &mt->base; + uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; + uint offset = 0; + int nr_faces, l, f, pitch; + + if (pt->target == PIPE_TEXTURE_CUBE) { + nr_faces = 6; + } else + if (pt->target == PIPE_TEXTURE_3D) { + nr_faces = pt->depth[0]; + } else { + nr_faces = 1; + } + + pitch = pt->width[0]; + for (l = 0; l <= pt->last_level; l++) { + pt->width[l] = width; + pt->height[l] = height; + pt->depth[l] = depth; + pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); + pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); + + if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) + pitch = pt->nblocksx[l]; + pitch = align(pitch, 64); + + mt->level[l].pitch = pitch * pt->block.size; + mt->level[l].image_offset = + CALLOC(nr_faces, sizeof(unsigned)); + + width = MAX2(1, width >> 1); + height = MAX2(1, height >> 1); + depth = MAX2(1, depth >> 1); + } + + for (f = 0; f < nr_faces; f++) { + for (l = 0; l <= pt->last_level; l++) { + mt->level[l].image_offset[f] = offset; + offset += mt->level[l].pitch * pt->height[l]; + } + } + + mt->total_size = offset; +} + +static struct pipe_texture * +nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) +{ + struct pipe_winsys *ws = pscreen->winsys; + struct nv40_miptree *mt; + + mt = MALLOC(sizeof(struct nv40_miptree)); + if (!mt) + return NULL; + mt->base = *pt; + mt->base.refcount = 1; + mt->base.screen = pscreen; + mt->shadow_tex = NULL; + mt->shadow_surface = NULL; + + /* Swizzled textures must be POT */ + if (pt->width[0] & (pt->width[0] - 1) || + pt->height[0] & (pt->height[0] - 1)) + mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; + else + if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY | + PIPE_TEXTURE_USAGE_DISPLAY_TARGET)) + mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; + else { + switch (pt->format) { + /* TODO: Figure out which formats can be swizzled */ + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_X8R8G8B8_UNORM: + /* XXX: Re-enable when SIFM size limits are fixed */ + /*case PIPE_FORMAT_R16_SNORM:*/ + break; + default: + mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; + } + } + + nv40_miptree_layout(mt); + + mt->buffer = ws->buffer_create(ws, 256, + PIPE_BUFFER_USAGE_PIXEL | + NOUVEAU_BUFFER_USAGE_TEXTURE, + mt->total_size); + if (!mt->buffer) { + FREE(mt); + return NULL; + } + + return &mt->base; +} + +static void +nv40_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt) +{ + struct pipe_texture *pt = *ppt; + struct nv40_miptree *mt = (struct nv40_miptree *)pt; + int l; + + *ppt = NULL; + if (--pt->refcount) + return; + + pipe_buffer_reference(pscreen, &mt->buffer, NULL); + for (l = 0; l <= pt->last_level; l++) { + if (mt->level[l].image_offset) + FREE(mt->level[l].image_offset); + } + + if (mt->shadow_tex) { + assert(mt->shadow_surface); + pscreen->tex_surface_release(pscreen, &mt->shadow_surface); + nv40_miptree_release(pscreen, &mt->shadow_tex); + } + + FREE(mt); +} + +static struct pipe_surface * +nv40_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + unsigned flags) +{ + struct nv40_miptree *mt = (struct nv40_miptree *)pt; + struct pipe_surface *ps; + + ps = CALLOC_STRUCT(pipe_surface); + if (!ps) + return NULL; + pipe_texture_reference(&ps->texture, pt); + pipe_buffer_reference(pscreen, &ps->buffer, mt->buffer); + ps->format = pt->format; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->block = pt->block; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = mt->level[level].pitch; + ps->usage = flags; + ps->status = PIPE_SURFACE_STATUS_DEFINED; + ps->refcount = 1; + ps->winsys = pscreen->winsys; + ps->face = face; + ps->level = level; + ps->zslice = zslice; + + if (pt->target == PIPE_TEXTURE_CUBE) { + ps->offset = mt->level[level].image_offset[face]; + } else + if (pt->target == PIPE_TEXTURE_3D) { + ps->offset = mt->level[level].image_offset[zslice]; + } else { + ps->offset = mt->level[level].image_offset[0]; + } + + return ps; +} + +static void +nv40_miptree_surface_del(struct pipe_screen *pscreen, + struct pipe_surface **psurface) +{ + struct pipe_surface *ps = *psurface; + + *psurface = NULL; + if (--ps->refcount > 0) + return; + + pipe_texture_reference(&ps->texture, NULL); + pipe_buffer_reference(pscreen, &ps->buffer, NULL); + FREE(ps); +} + +void +nv40_screen_init_miptree_functions(struct pipe_screen *pscreen) +{ + pscreen->texture_create = nv40_miptree_create; + pscreen->texture_release = nv40_miptree_release; + pscreen->get_tex_surface = nv40_miptree_surface_new; + pscreen->tex_surface_release = nv40_miptree_surface_del; +} + diff --git a/src/gallium/drivers/nv40/nv40_query.c b/src/gallium/drivers/nv40/nv40_query.c new file mode 100644 index 0000000000..57f39cfab0 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_query.c @@ -0,0 +1,122 @@ +#include "pipe/p_context.h" + +#include "nv40_context.h" + +struct nv40_query { + struct nouveau_resource *object; + unsigned type; + boolean ready; + uint64_t result; +}; + +static INLINE struct nv40_query * +nv40_query(struct pipe_query *pipe) +{ + return (struct nv40_query *)pipe; +} + +static struct pipe_query * +nv40_query_create(struct pipe_context *pipe, unsigned query_type) +{ + struct nv40_query *q; + + q = CALLOC(1, sizeof(struct nv40_query)); + q->type = query_type; + + return (struct pipe_query *)q; +} + +static void +nv40_query_destroy(struct pipe_context *pipe, struct pipe_query *pq) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nv40_query *q = nv40_query(pq); + + if (q->object) + nv40->nvws->res_free(&q->object); + FREE(q); +} + +static void +nv40_query_begin(struct pipe_context *pipe, struct pipe_query *pq) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nv40_query *q = nv40_query(pq); + + assert(q->type == PIPE_QUERY_OCCLUSION_COUNTER); + + /* Happens when end_query() is called, then another begin_query() + * without querying the result in-between. For now we'll wait for + * the existing query to notify completion, but it could be better. + */ + if (q->object) { + uint64 tmp; + pipe->get_query_result(pipe, pq, 1, &tmp); + } + + if (nv40->nvws->res_alloc(nv40->screen->query_heap, 1, NULL, &q->object)) + assert(0); + nv40->nvws->notifier_reset(nv40->screen->query, q->object->start); + + BEGIN_RING(curie, NV40TCL_QUERY_RESET, 1); + OUT_RING (1); + BEGIN_RING(curie, NV40TCL_QUERY_UNK17CC, 1); + OUT_RING (1); + + q->ready = FALSE; +} + +static void +nv40_query_end(struct pipe_context *pipe, struct pipe_query *pq) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nv40_query *q = nv40_query(pq); + + BEGIN_RING(curie, NV40TCL_QUERY_GET, 1); + OUT_RING ((0x01 << NV40TCL_QUERY_GET_UNK24_SHIFT) | + ((q->object->start * 32) << NV40TCL_QUERY_GET_OFFSET_SHIFT)); + FIRE_RING(NULL); +} + +static boolean +nv40_query_result(struct pipe_context *pipe, struct pipe_query *pq, + boolean wait, uint64 *result) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nv40_query *q = nv40_query(pq); + struct nouveau_winsys *nvws = nv40->nvws; + + assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER); + + if (!q->ready) { + unsigned status; + + status = nvws->notifier_status(nv40->screen->query, + q->object->start); + if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) { + if (wait == FALSE) + return FALSE; + nvws->notifier_wait(nv40->screen->query, q->object->start, + NV_NOTIFY_STATE_STATUS_COMPLETED, + 0); + } + + q->result = nvws->notifier_retval(nv40->screen->query, + q->object->start); + q->ready = TRUE; + nvws->res_free(&q->object); + } + + *result = q->result; + return TRUE; +} + +void +nv40_init_query_functions(struct nv40_context *nv40) +{ + nv40->pipe.create_query = nv40_query_create; + nv40->pipe.destroy_query = nv40_query_destroy; + nv40->pipe.begin_query = nv40_query_begin; + nv40->pipe.end_query = nv40_query_end; + nv40->pipe.get_query_result = nv40_query_result; +} diff --git a/src/gallium/drivers/nv40/nv40_screen.c b/src/gallium/drivers/nv40/nv40_screen.c new file mode 100644 index 0000000000..995148e03f --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_screen.c @@ -0,0 +1,367 @@ +#include "pipe/p_screen.h" + +#include "nv40_context.h" +#include "nv40_screen.h" + +#define NV4X_GRCLASS4097_CHIPSETS 0x00000baf +#define NV4X_GRCLASS4497_CHIPSETS 0x00005450 +#define NV6X_GRCLASS4497_CHIPSETS 0x00000088 + +static const char * +nv40_screen_get_name(struct pipe_screen *pscreen) +{ + struct nv40_screen *screen = nv40_screen(pscreen); + struct nouveau_device *dev = screen->nvws->channel->device; + static char buffer[128]; + + snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); + return buffer; +} + +static const char * +nv40_screen_get_vendor(struct pipe_screen *pscreen) +{ + return "nouveau"; +} + +static int +nv40_screen_get_param(struct pipe_screen *pscreen, int param) +{ + struct nv40_screen *screen = nv40_screen(pscreen); + + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 16; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 1; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 1; + case PIPE_CAP_POINT_SPRITE: + return 1; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 4; + case PIPE_CAP_OCCLUSION_QUERY: + return 1; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 13; + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 10; + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 13; + case PIPE_CAP_TEXTURE_MIRROR_CLAMP: + case PIPE_CAP_TEXTURE_MIRROR_REPEAT: + return 1; + case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: + return 0; /* We have 4 - but unsupported currently */ + case NOUVEAU_CAP_HW_VTXBUF: + return 1; + case NOUVEAU_CAP_HW_IDXBUF: + if (screen->curie->grclass == NV40TCL) + return 1; + return 0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0; + } +} + +static float +nv40_screen_get_paramf(struct pipe_screen *pscreen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 10.0; + case PIPE_CAP_MAX_POINT_WIDTH: + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 64.0; + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 16.0; + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0.0; + } +} + +static boolean +nv40_screen_surface_format_supported(struct pipe_screen *pscreen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, unsigned geom_flags) +{ + if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z16_UNORM: + return TRUE; + default: + break; + } + } else { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_A1R5G5B5_UNORM: + case PIPE_FORMAT_A4R4G4B4_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_R16_SNORM: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_A8_UNORM: + case PIPE_FORMAT_I8_UNORM: + case PIPE_FORMAT_A8L8_UNORM: + case PIPE_FORMAT_Z16_UNORM: + case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_DXT1_RGB: + case PIPE_FORMAT_DXT1_RGBA: + case PIPE_FORMAT_DXT3_RGBA: + case PIPE_FORMAT_DXT5_RGBA: + return TRUE; + default: + break; + } + } + + return FALSE; +} + +static void * +nv40_surface_map(struct pipe_screen *screen, struct pipe_surface *surface, + unsigned flags ) +{ + struct pipe_winsys *ws = screen->winsys; + struct pipe_surface *surface_to_map; + void *map; + + if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture; + + if (!mt->shadow_tex) { + unsigned old_tex_usage = surface->texture->tex_usage; + surface->texture->tex_usage = NOUVEAU_TEXTURE_USAGE_LINEAR; + mt->shadow_tex = screen->texture_create(screen, surface->texture); + surface->texture->tex_usage = old_tex_usage; + + assert(mt->shadow_tex->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR); + mt->shadow_surface = screen->get_tex_surface + ( + screen, mt->shadow_tex, + surface->face, surface->level, surface->zslice, + surface->usage + ); + } + + surface_to_map = mt->shadow_surface; + } + else + surface_to_map = surface; + + assert(surface_to_map); + + map = ws->buffer_map(ws, surface_to_map->buffer, flags); + if (!map) + return NULL; + + return map + surface_to_map->offset; +} + +static void +nv40_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface) +{ + struct pipe_winsys *ws = screen->winsys; + struct pipe_surface *surface_to_unmap; + + /* TODO: Copy from shadow just before push buffer is flushed instead. + There are probably some programs that map/unmap excessively + before rendering. */ + if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture; + + assert(mt->shadow_tex); + + surface_to_unmap = mt->shadow_surface; + } + else + surface_to_unmap = surface; + + assert(surface_to_unmap); + + ws->buffer_unmap(ws, surface_to_unmap->buffer); + + if (surface_to_unmap != surface) { + struct nv40_screen *nvscreen = nv40_screen(screen); + + nvscreen->nvws->surface_copy(nvscreen->nvws, + surface, 0, 0, + surface_to_unmap, 0, 0, + surface->width, surface->height); + } +} + +static void +nv40_screen_destroy(struct pipe_screen *pscreen) +{ + struct nv40_screen *screen = nv40_screen(pscreen); + struct nouveau_winsys *nvws = screen->nvws; + + nvws->res_free(&screen->vp_exec_heap); + nvws->res_free(&screen->vp_data_heap); + nvws->res_free(&screen->query_heap); + nvws->notifier_free(&screen->query); + nvws->notifier_free(&screen->sync); + nvws->grobj_free(&screen->curie); + + FREE(pscreen); +} + +struct pipe_screen * +nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +{ + struct nv40_screen *screen = CALLOC_STRUCT(nv40_screen); + struct nouveau_stateobj *so; + unsigned curie_class; + unsigned chipset = nvws->channel->device->chipset; + int ret; + + if (!screen) + return NULL; + screen->nvws = nvws; + + /* 3D object */ + switch (chipset & 0xf0) { + case 0x40: + if (NV4X_GRCLASS4097_CHIPSETS & (1 << (chipset & 0x0f))) + curie_class = NV40TCL; + else + if (NV4X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f))) + curie_class = NV44TCL; + break; + case 0x60: + if (NV6X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f))) + curie_class = NV44TCL; + break; + default: + break; + } + + if (!curie_class) { + NOUVEAU_ERR("Unknown nv4x chipset: nv%02x\n", chipset); + return NULL; + } + + ret = nvws->grobj_alloc(nvws, curie_class, &screen->curie); + if (ret) { + NOUVEAU_ERR("Error creating 3D object: %d\n", ret); + return FALSE; + } + + /* Notifier for sync purposes */ + ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + if (ret) { + NOUVEAU_ERR("Error creating notifier object: %d\n", ret); + nv40_screen_destroy(&screen->pipe); + return NULL; + } + + /* Query objects */ + ret = nvws->notifier_alloc(nvws, 32, &screen->query); + if (ret) { + NOUVEAU_ERR("Error initialising query objects: %d\n", ret); + nv40_screen_destroy(&screen->pipe); + return NULL; + } + + ret = nvws->res_init(&screen->query_heap, 0, 32); + if (ret) { + NOUVEAU_ERR("Error initialising query object heap: %d\n", ret); + nv40_screen_destroy(&screen->pipe); + return NULL; + } + + /* Vtxprog resources */ + if (nvws->res_init(&screen->vp_exec_heap, 0, 512) || + nvws->res_init(&screen->vp_data_heap, 0, 256)) { + nv40_screen_destroy(&screen->pipe); + return NULL; + } + + /* Static curie initialisation */ + so = so_new(128, 0); + so_method(so, screen->curie, NV40TCL_DMA_NOTIFY, 1); + so_data (so, screen->sync->handle); + so_method(so, screen->curie, NV40TCL_DMA_TEXTURE0, 2); + so_data (so, nvws->channel->vram->handle); + so_data (so, nvws->channel->gart->handle); + so_method(so, screen->curie, NV40TCL_DMA_COLOR1, 1); + so_data (so, nvws->channel->vram->handle); + so_method(so, screen->curie, NV40TCL_DMA_COLOR0, 2); + so_data (so, nvws->channel->vram->handle); + so_data (so, nvws->channel->vram->handle); + so_method(so, screen->curie, NV40TCL_DMA_VTXBUF0, 2); + so_data (so, nvws->channel->vram->handle); + so_data (so, nvws->channel->gart->handle); + so_method(so, screen->curie, NV40TCL_DMA_FENCE, 2); + so_data (so, 0); + so_data (so, screen->query->handle); + so_method(so, screen->curie, NV40TCL_DMA_UNK01AC, 2); + so_data (so, nvws->channel->vram->handle); + so_data (so, nvws->channel->vram->handle); + so_method(so, screen->curie, NV40TCL_DMA_COLOR2, 2); + so_data (so, nvws->channel->vram->handle); + so_data (so, nvws->channel->vram->handle); + + so_method(so, screen->curie, 0x1ea4, 3); + so_data (so, 0x00000010); + so_data (so, 0x01000100); + so_data (so, 0xff800006); + + /* vtxprog output routing */ + so_method(so, screen->curie, 0x1fc4, 1); + so_data (so, 0x06144321); + so_method(so, screen->curie, 0x1fc8, 2); + so_data (so, 0xedcba987); + so_data (so, 0x00000021); + so_method(so, screen->curie, 0x1fd0, 1); + so_data (so, 0x00171615); + so_method(so, screen->curie, 0x1fd4, 1); + so_data (so, 0x001b1a19); + + so_method(so, screen->curie, 0x1ef8, 1); + so_data (so, 0x0020ffff); + so_method(so, screen->curie, 0x1d64, 1); + so_data (so, 0x00d30000); + so_method(so, screen->curie, 0x1e94, 1); + so_data (so, 0x00000001); + + so_emit(nvws, so); + so_ref(NULL, &so); + nvws->push_flush(nvws, 0, NULL); + + screen->pipe.winsys = ws; + screen->pipe.destroy = nv40_screen_destroy; + + screen->pipe.get_name = nv40_screen_get_name; + screen->pipe.get_vendor = nv40_screen_get_vendor; + screen->pipe.get_param = nv40_screen_get_param; + screen->pipe.get_paramf = nv40_screen_get_paramf; + + screen->pipe.is_format_supported = nv40_screen_surface_format_supported; + + screen->pipe.surface_map = nv40_surface_map; + screen->pipe.surface_unmap = nv40_surface_unmap; + + nv40_screen_init_miptree_functions(&screen->pipe); + + return &screen->pipe; +} + diff --git a/src/gallium/drivers/nv40/nv40_screen.h b/src/gallium/drivers/nv40/nv40_screen.h new file mode 100644 index 0000000000..c04a1275a0 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_screen.h @@ -0,0 +1,35 @@ +#ifndef __NV40_SCREEN_H__ +#define __NV40_SCREEN_H__ + +#include "pipe/p_screen.h" + +struct nv40_screen { + struct pipe_screen pipe; + + struct nouveau_winsys *nvws; + + unsigned cur_pctx; + + /* HW graphics objects */ + struct nouveau_grobj *curie; + struct nouveau_notifier *sync; + + /* Query object resources */ + struct nouveau_notifier *query; + struct nouveau_resource *query_heap; + + /* Vtxprog resources */ + struct nouveau_resource *vp_exec_heap; + struct nouveau_resource *vp_data_heap; + + /* Current 3D state of channel */ + struct nouveau_stateobj *state[NV40_STATE_MAX]; +}; + +static INLINE struct nv40_screen * +nv40_screen(struct pipe_screen *screen) +{ + return (struct nv40_screen *)screen; +} + +#endif diff --git a/src/gallium/drivers/nv40/nv40_shader.h b/src/gallium/drivers/nv40/nv40_shader.h new file mode 100644 index 0000000000..854dccf548 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_shader.h @@ -0,0 +1,556 @@ +#ifndef __NV40_SHADER_H__ +#define __NV40_SHADER_H__ + +/* Vertex programs instruction set + * + * The NV40 instruction set is very similar to NV30. Most fields are in + * a slightly different position in the instruction however. + * + * Merged instructions + * In some cases it is possible to put two instructions into one opcode + * slot. The rules for when this is OK is not entirely clear to me yet. + * + * There are separate writemasks and dest temp register fields for each + * grouping of instructions. There is however only one field with the + * ID of a result register. Writing to temp/result regs is selected by + * setting VEC_RESULT/SCA_RESULT. + * + * Temporary registers + * The source/dest temp register fields have been extended by 1 bit, to + * give a total of 32 temporary registers. + * + * Relative Addressing + * NV40 can use an address register to index into vertex attribute regs. + * This is done by putting the offset value into INPUT_SRC and setting + * the INDEX_INPUT flag. + * + * Conditional execution (see NV_vertex_program{2,3} for details) + * There is a second condition code register on NV40, it's use is enabled + * by setting the COND_REG_SELECT_1 flag. + * + * Texture lookup + * TODO + */ + +/* ---- OPCODE BITS 127:96 / data DWORD 0 --- */ +#define NV40_VP_INST_VEC_RESULT (1 << 30) +/* uncertain.. */ +#define NV40_VP_INST_COND_UPDATE_ENABLE ((1 << 14)|1<<29) +/* use address reg as index into attribs */ +#define NV40_VP_INST_INDEX_INPUT (1 << 27) +#define NV40_VP_INST_COND_REG_SELECT_1 (1 << 25) +#define NV40_VP_INST_ADDR_REG_SELECT_1 (1 << 24) +#define NV40_VP_INST_SRC2_ABS (1 << 23) +#define NV40_VP_INST_SRC1_ABS (1 << 22) +#define NV40_VP_INST_SRC0_ABS (1 << 21) +#define NV40_VP_INST_VEC_DEST_TEMP_SHIFT 15 +#define NV40_VP_INST_VEC_DEST_TEMP_MASK (0x1F << 15) +#define NV40_VP_INST_COND_TEST_ENABLE (1 << 13) +#define NV40_VP_INST_COND_SHIFT 10 +#define NV40_VP_INST_COND_MASK (0x7 << 10) +# define NV40_VP_INST_COND_FL 0 +# define NV40_VP_INST_COND_LT 1 +# define NV40_VP_INST_COND_EQ 2 +# define NV40_VP_INST_COND_LE 3 +# define NV40_VP_INST_COND_GT 4 +# define NV40_VP_INST_COND_NE 5 +# define NV40_VP_INST_COND_GE 6 +# define NV40_VP_INST_COND_TR 7 +#define NV40_VP_INST_COND_SWZ_X_SHIFT 8 +#define NV40_VP_INST_COND_SWZ_X_MASK (3 << 8) +#define NV40_VP_INST_COND_SWZ_Y_SHIFT 6 +#define NV40_VP_INST_COND_SWZ_Y_MASK (3 << 6) +#define NV40_VP_INST_COND_SWZ_Z_SHIFT 4 +#define NV40_VP_INST_COND_SWZ_Z_MASK (3 << 4) +#define NV40_VP_INST_COND_SWZ_W_SHIFT 2 +#define NV40_VP_INST_COND_SWZ_W_MASK (3 << 2) +#define NV40_VP_INST_COND_SWZ_ALL_SHIFT 2 +#define NV40_VP_INST_COND_SWZ_ALL_MASK (0xFF << 2) +#define NV40_VP_INST_ADDR_SWZ_SHIFT 0 +#define NV40_VP_INST_ADDR_SWZ_MASK (0x03 << 0) +#define NV40_VP_INST0_KNOWN ( \ + NV40_VP_INST_INDEX_INPUT | \ + NV40_VP_INST_COND_REG_SELECT_1 | \ + NV40_VP_INST_ADDR_REG_SELECT_1 | \ + NV40_VP_INST_SRC2_ABS | \ + NV40_VP_INST_SRC1_ABS | \ + NV40_VP_INST_SRC0_ABS | \ + NV40_VP_INST_VEC_DEST_TEMP_MASK | \ + NV40_VP_INST_COND_TEST_ENABLE | \ + NV40_VP_INST_COND_MASK | \ + NV40_VP_INST_COND_SWZ_ALL_MASK | \ + NV40_VP_INST_ADDR_SWZ_MASK) + +/* ---- OPCODE BITS 95:64 / data DWORD 1 --- */ +#define NV40_VP_INST_VEC_OPCODE_SHIFT 22 +#define NV40_VP_INST_VEC_OPCODE_MASK (0x1F << 22) +# define NV40_VP_INST_OP_NOP 0x00 +# define NV40_VP_INST_OP_MOV 0x01 +# define NV40_VP_INST_OP_MUL 0x02 +# define NV40_VP_INST_OP_ADD 0x03 +# define NV40_VP_INST_OP_MAD 0x04 +# define NV40_VP_INST_OP_DP3 0x05 +# define NV40_VP_INST_OP_DPH 0x06 +# define NV40_VP_INST_OP_DP4 0x07 +# define NV40_VP_INST_OP_DST 0x08 +# define NV40_VP_INST_OP_MIN 0x09 +# define NV40_VP_INST_OP_MAX 0x0A +# define NV40_VP_INST_OP_SLT 0x0B +# define NV40_VP_INST_OP_SGE 0x0C +# define NV40_VP_INST_OP_ARL 0x0D +# define NV40_VP_INST_OP_FRC 0x0E +# define NV40_VP_INST_OP_FLR 0x0F +# define NV40_VP_INST_OP_SEQ 0x10 +# define NV40_VP_INST_OP_SFL 0x11 +# define NV40_VP_INST_OP_SGT 0x12 +# define NV40_VP_INST_OP_SLE 0x13 +# define NV40_VP_INST_OP_SNE 0x14 +# define NV40_VP_INST_OP_STR 0x15 +# define NV40_VP_INST_OP_SSG 0x16 +# define NV40_VP_INST_OP_ARR 0x17 +# define NV40_VP_INST_OP_ARA 0x18 +# define NV40_VP_INST_OP_TXL 0x19 +#define NV40_VP_INST_SCA_OPCODE_SHIFT 27 +#define NV40_VP_INST_SCA_OPCODE_MASK (0x1F << 27) +# define NV40_VP_INST_OP_NOP 0x00 +# define NV40_VP_INST_OP_MOV 0x01 +# define NV40_VP_INST_OP_RCP 0x02 +# define NV40_VP_INST_OP_RCC 0x03 +# define NV40_VP_INST_OP_RSQ 0x04 +# define NV40_VP_INST_OP_EXP 0x05 +# define NV40_VP_INST_OP_LOG 0x06 +# define NV40_VP_INST_OP_LIT 0x07 +# define NV40_VP_INST_OP_BRA 0x09 +# define NV40_VP_INST_OP_CAL 0x0B +# define NV40_VP_INST_OP_RET 0x0C +# define NV40_VP_INST_OP_LG2 0x0D +# define NV40_VP_INST_OP_EX2 0x0E +# define NV40_VP_INST_OP_SIN 0x0F +# define NV40_VP_INST_OP_COS 0x10 +# define NV40_VP_INST_OP_PUSHA 0x13 +# define NV40_VP_INST_OP_POPA 0x14 +#define NV40_VP_INST_CONST_SRC_SHIFT 12 +#define NV40_VP_INST_CONST_SRC_MASK (0xFF << 12) +#define NV40_VP_INST_INPUT_SRC_SHIFT 8 +#define NV40_VP_INST_INPUT_SRC_MASK (0x0F << 8) +# define NV40_VP_INST_IN_POS 0 +# define NV40_VP_INST_IN_WEIGHT 1 +# define NV40_VP_INST_IN_NORMAL 2 +# define NV40_VP_INST_IN_COL0 3 +# define NV40_VP_INST_IN_COL1 4 +# define NV40_VP_INST_IN_FOGC 5 +# define NV40_VP_INST_IN_TC0 8 +# define NV40_VP_INST_IN_TC(n) (8+n) +#define NV40_VP_INST_SRC0H_SHIFT 0 +#define NV40_VP_INST_SRC0H_MASK (0xFF << 0) +#define NV40_VP_INST1_KNOWN ( \ + NV40_VP_INST_VEC_OPCODE_MASK | \ + NV40_VP_INST_SCA_OPCODE_MASK | \ + NV40_VP_INST_CONST_SRC_MASK | \ + NV40_VP_INST_INPUT_SRC_MASK | \ + NV40_VP_INST_SRC0H_MASK \ + ) + +/* ---- OPCODE BITS 63:32 / data DWORD 2 --- */ +#define NV40_VP_INST_SRC0L_SHIFT 23 +#define NV40_VP_INST_SRC0L_MASK (0x1FF << 23) +#define NV40_VP_INST_SRC1_SHIFT 6 +#define NV40_VP_INST_SRC1_MASK (0x1FFFF << 6) +#define NV40_VP_INST_SRC2H_SHIFT 0 +#define NV40_VP_INST_SRC2H_MASK (0x3F << 0) +#define NV40_VP_INST_IADDRH_SHIFT 0 +#define NV40_VP_INST_IADDRH_MASK (0x1F << 0) + +/* ---- OPCODE BITS 31:0 / data DWORD 3 --- */ +#define NV40_VP_INST_IADDRL_SHIFT 29 +#define NV40_VP_INST_IADDRL_MASK (7 << 29) +#define NV40_VP_INST_SRC2L_SHIFT 21 +#define NV40_VP_INST_SRC2L_MASK (0x7FF << 21) +#define NV40_VP_INST_SCA_WRITEMASK_SHIFT 17 +#define NV40_VP_INST_SCA_WRITEMASK_MASK (0xF << 17) +# define NV40_VP_INST_SCA_WRITEMASK_X (1 << 20) +# define NV40_VP_INST_SCA_WRITEMASK_Y (1 << 19) +# define NV40_VP_INST_SCA_WRITEMASK_Z (1 << 18) +# define NV40_VP_INST_SCA_WRITEMASK_W (1 << 17) +#define NV40_VP_INST_VEC_WRITEMASK_SHIFT 13 +#define NV40_VP_INST_VEC_WRITEMASK_MASK (0xF << 13) +# define NV40_VP_INST_VEC_WRITEMASK_X (1 << 16) +# define NV40_VP_INST_VEC_WRITEMASK_Y (1 << 15) +# define NV40_VP_INST_VEC_WRITEMASK_Z (1 << 14) +# define NV40_VP_INST_VEC_WRITEMASK_W (1 << 13) +#define NV40_VP_INST_SCA_RESULT (1 << 12) +#define NV40_VP_INST_SCA_DEST_TEMP_SHIFT 7 +#define NV40_VP_INST_SCA_DEST_TEMP_MASK (0x1F << 7) +#define NV40_VP_INST_DEST_SHIFT 2 +#define NV40_VP_INST_DEST_MASK (31 << 2) +# define NV40_VP_INST_DEST_POS 0 +# define NV40_VP_INST_DEST_COL0 1 +# define NV40_VP_INST_DEST_COL1 2 +# define NV40_VP_INST_DEST_BFC0 3 +# define NV40_VP_INST_DEST_BFC1 4 +# define NV40_VP_INST_DEST_FOGC 5 +# define NV40_VP_INST_DEST_PSZ 6 +# define NV40_VP_INST_DEST_TC0 7 +# define NV40_VP_INST_DEST_TC(n) (7+n) +# define NV40_VP_INST_DEST_TEMP 0x1F +#define NV40_VP_INST_INDEX_CONST (1 << 1) +#define NV40_VP_INST_LAST (1 << 0) +#define NV40_VP_INST3_KNOWN ( \ + NV40_VP_INST_SRC2L_MASK |\ + NV40_VP_INST_SCA_WRITEMASK_MASK |\ + NV40_VP_INST_VEC_WRITEMASK_MASK |\ + NV40_VP_INST_SCA_DEST_TEMP_MASK |\ + NV40_VP_INST_DEST_MASK |\ + NV40_VP_INST_INDEX_CONST) + +/* Useful to split the source selection regs into their pieces */ +#define NV40_VP_SRC0_HIGH_SHIFT 9 +#define NV40_VP_SRC0_HIGH_MASK 0x0001FE00 +#define NV40_VP_SRC0_LOW_MASK 0x000001FF +#define NV40_VP_SRC2_HIGH_SHIFT 11 +#define NV40_VP_SRC2_HIGH_MASK 0x0001F800 +#define NV40_VP_SRC2_LOW_MASK 0x000007FF + +/* Source selection - these are the bits you fill NV40_VP_INST_SRCn with */ +#define NV40_VP_SRC_NEGATE (1 << 16) +#define NV40_VP_SRC_SWZ_X_SHIFT 14 +#define NV40_VP_SRC_SWZ_X_MASK (3 << 14) +#define NV40_VP_SRC_SWZ_Y_SHIFT 12 +#define NV40_VP_SRC_SWZ_Y_MASK (3 << 12) +#define NV40_VP_SRC_SWZ_Z_SHIFT 10 +#define NV40_VP_SRC_SWZ_Z_MASK (3 << 10) +#define NV40_VP_SRC_SWZ_W_SHIFT 8 +#define NV40_VP_SRC_SWZ_W_MASK (3 << 8) +#define NV40_VP_SRC_SWZ_ALL_SHIFT 8 +#define NV40_VP_SRC_SWZ_ALL_MASK (0xFF << 8) +#define NV40_VP_SRC_TEMP_SRC_SHIFT 2 +#define NV40_VP_SRC_TEMP_SRC_MASK (0x1F << 2) +#define NV40_VP_SRC_REG_TYPE_SHIFT 0 +#define NV40_VP_SRC_REG_TYPE_MASK (3 << 0) +# define NV40_VP_SRC_REG_TYPE_UNK0 0 +# define NV40_VP_SRC_REG_TYPE_TEMP 1 +# define NV40_VP_SRC_REG_TYPE_INPUT 2 +# define NV40_VP_SRC_REG_TYPE_CONST 3 + + +/* + * Each fragment program opcode appears to be comprised of 4 32-bit values. + * + * 0 - Opcode, output reg/mask, ATTRIB source + * 1 - Source 0 + * 2 - Source 1 + * 3 - Source 2 + * + * There appears to be no special difference between result regs and temp regs. + * result.color == R0.xyzw + * result.depth == R1.z + * When the fragprog contains instructions to write depth, + * NV30_TCL_PRIMITIVE_3D_UNK1D78=0 otherwise it is set to 1. + * + * Constants are inserted directly after the instruction that uses them. + * + * It appears that it's not possible to use two input registers in one + * instruction as the input sourcing is done in the instruction dword + * and not the source selection dwords. As such instructions such as: + * + * ADD result.color, fragment.color, fragment.texcoord[0]; + * + * must be split into two MOV's and then an ADD (nvidia does this) but + * I'm not sure why it's not just one MOV and then source the second input + * in the ADD instruction.. + * + * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary + * negation requires multiplication with a const. + * + * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO and + * SWIZZLE_ONE. + * + * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as + * SWIZZLE_ZERO is implemented simply by not writing to the relevant components + * of the destination. + * + * Looping + * Loops appear to be fairly expensive on NV40 at least, the proprietary + * driver goes to a lot of effort to avoid using the native looping + * instructions. If the total number of *executed* instructions between + * REP/ENDREP or LOOP/ENDLOOP is <=500, the driver will unroll the loop. + * The maximum loop count is 255. + * + * Conditional execution + * TODO + * + * Non-native instructions: + * LIT + * LRP - MAD+MAD + * SUB - ADD, negate second source + * RSQ - LG2 + EX2 + * POW - LG2 + MUL + EX2 + * SCS - COS + SIN + * XPD + * DP2 - MUL + ADD + * NRM + */ + +//== Opcode / Destination selection == +#define NV40_FP_OP_PROGRAM_END (1 << 0) +#define NV40_FP_OP_OUT_REG_SHIFT 1 +#define NV40_FP_OP_OUT_REG_MASK (63 << 1) +/* Needs to be set when writing outputs to get expected result.. */ +#define NV40_FP_OP_OUT_REG_HALF (1 << 7) +#define NV40_FP_OP_COND_WRITE_ENABLE (1 << 8) +#define NV40_FP_OP_OUTMASK_SHIFT 9 +#define NV40_FP_OP_OUTMASK_MASK (0xF << 9) +# define NV40_FP_OP_OUT_X (1 << 9) +# define NV40_FP_OP_OUT_Y (1 <<10) +# define NV40_FP_OP_OUT_Z (1 <<11) +# define NV40_FP_OP_OUT_W (1 <<12) +/* Uncertain about these, especially the input_src values.. it's possible that + * they can be dynamically changed. + */ +#define NV40_FP_OP_INPUT_SRC_SHIFT 13 +#define NV40_FP_OP_INPUT_SRC_MASK (15 << 13) +# define NV40_FP_OP_INPUT_SRC_POSITION 0x0 +# define NV40_FP_OP_INPUT_SRC_COL0 0x1 +# define NV40_FP_OP_INPUT_SRC_COL1 0x2 +# define NV40_FP_OP_INPUT_SRC_FOGC 0x3 +# define NV40_FP_OP_INPUT_SRC_TC0 0x4 +# define NV40_FP_OP_INPUT_SRC_TC(n) (0x4 + n) +# define NV40_FP_OP_INPUT_SRC_FACING 0xE +#define NV40_FP_OP_TEX_UNIT_SHIFT 17 +#define NV40_FP_OP_TEX_UNIT_MASK (0xF << 17) +#define NV40_FP_OP_PRECISION_SHIFT 22 +#define NV40_FP_OP_PRECISION_MASK (3 << 22) +# define NV40_FP_PRECISION_FP32 0 +# define NV40_FP_PRECISION_FP16 1 +# define NV40_FP_PRECISION_FX12 2 +#define NV40_FP_OP_OPCODE_SHIFT 24 +#define NV40_FP_OP_OPCODE_MASK (0x3F << 24) +# define NV40_FP_OP_OPCODE_NOP 0x00 +# define NV40_FP_OP_OPCODE_MOV 0x01 +# define NV40_FP_OP_OPCODE_MUL 0x02 +# define NV40_FP_OP_OPCODE_ADD 0x03 +# define NV40_FP_OP_OPCODE_MAD 0x04 +# define NV40_FP_OP_OPCODE_DP3 0x05 +# define NV40_FP_OP_OPCODE_DP4 0x06 +# define NV40_FP_OP_OPCODE_DST 0x07 +# define NV40_FP_OP_OPCODE_MIN 0x08 +# define NV40_FP_OP_OPCODE_MAX 0x09 +# define NV40_FP_OP_OPCODE_SLT 0x0A +# define NV40_FP_OP_OPCODE_SGE 0x0B +# define NV40_FP_OP_OPCODE_SLE 0x0C +# define NV40_FP_OP_OPCODE_SGT 0x0D +# define NV40_FP_OP_OPCODE_SNE 0x0E +# define NV40_FP_OP_OPCODE_SEQ 0x0F +# define NV40_FP_OP_OPCODE_FRC 0x10 +# define NV40_FP_OP_OPCODE_FLR 0x11 +# define NV40_FP_OP_OPCODE_KIL 0x12 +# define NV40_FP_OP_OPCODE_PK4B 0x13 +# define NV40_FP_OP_OPCODE_UP4B 0x14 +/* DDX/DDY can only write to XY */ +# define NV40_FP_OP_OPCODE_DDX 0x15 +# define NV40_FP_OP_OPCODE_DDY 0x16 +# define NV40_FP_OP_OPCODE_TEX 0x17 +# define NV40_FP_OP_OPCODE_TXP 0x18 +# define NV40_FP_OP_OPCODE_TXD 0x19 +# define NV40_FP_OP_OPCODE_RCP 0x1A +# define NV40_FP_OP_OPCODE_EX2 0x1C +# define NV40_FP_OP_OPCODE_LG2 0x1D +# define NV40_FP_OP_OPCODE_STR 0x20 +# define NV40_FP_OP_OPCODE_SFL 0x21 +# define NV40_FP_OP_OPCODE_COS 0x22 +# define NV40_FP_OP_OPCODE_SIN 0x23 +# define NV40_FP_OP_OPCODE_PK2H 0x24 +# define NV40_FP_OP_OPCODE_UP2H 0x25 +# define NV40_FP_OP_OPCODE_PK4UB 0x27 +# define NV40_FP_OP_OPCODE_UP4UB 0x28 +# define NV40_FP_OP_OPCODE_PK2US 0x29 +# define NV40_FP_OP_OPCODE_UP2US 0x2A +# define NV40_FP_OP_OPCODE_DP2A 0x2E +# define NV40_FP_OP_OPCODE_TXL 0x2F +# define NV40_FP_OP_OPCODE_TXB 0x31 +# define NV40_FP_OP_OPCODE_DIV 0x3A +# define NV40_FP_OP_OPCODE_UNK_LIT 0x3C +/* The use of these instructions appears to be indicated by bit 31 of DWORD 2.*/ +# define NV40_FP_OP_BRA_OPCODE_BRK 0x0 +# define NV40_FP_OP_BRA_OPCODE_CAL 0x1 +# define NV40_FP_OP_BRA_OPCODE_IF 0x2 +# define NV40_FP_OP_BRA_OPCODE_LOOP 0x3 +# define NV40_FP_OP_BRA_OPCODE_REP 0x4 +# define NV40_FP_OP_BRA_OPCODE_RET 0x5 +#define NV40_FP_OP_OUT_SAT (1 << 31) + +/* high order bits of SRC0 */ +#define NV40_FP_OP_OUT_ABS (1 << 29) +#define NV40_FP_OP_COND_SWZ_W_SHIFT 27 +#define NV40_FP_OP_COND_SWZ_W_MASK (3 << 27) +#define NV40_FP_OP_COND_SWZ_Z_SHIFT 25 +#define NV40_FP_OP_COND_SWZ_Z_MASK (3 << 25) +#define NV40_FP_OP_COND_SWZ_Y_SHIFT 23 +#define NV40_FP_OP_COND_SWZ_Y_MASK (3 << 23) +#define NV40_FP_OP_COND_SWZ_X_SHIFT 21 +#define NV40_FP_OP_COND_SWZ_X_MASK (3 << 21) +#define NV40_FP_OP_COND_SWZ_ALL_SHIFT 21 +#define NV40_FP_OP_COND_SWZ_ALL_MASK (0xFF << 21) +#define NV40_FP_OP_COND_SHIFT 18 +#define NV40_FP_OP_COND_MASK (0x07 << 18) +# define NV40_FP_OP_COND_FL 0 +# define NV40_FP_OP_COND_LT 1 +# define NV40_FP_OP_COND_EQ 2 +# define NV40_FP_OP_COND_LE 3 +# define NV40_FP_OP_COND_GT 4 +# define NV40_FP_OP_COND_NE 5 +# define NV40_FP_OP_COND_GE 6 +# define NV40_FP_OP_COND_TR 7 + +/* high order bits of SRC1 */ +#define NV40_FP_OP_OPCODE_IS_BRANCH (1<<31) +#define NV40_FP_OP_DST_SCALE_SHIFT 28 +#define NV40_FP_OP_DST_SCALE_MASK (3 << 28) +#define NV40_FP_OP_DST_SCALE_1X 0 +#define NV40_FP_OP_DST_SCALE_2X 1 +#define NV40_FP_OP_DST_SCALE_4X 2 +#define NV40_FP_OP_DST_SCALE_8X 3 +#define NV40_FP_OP_DST_SCALE_INV_2X 5 +#define NV40_FP_OP_DST_SCALE_INV_4X 6 +#define NV40_FP_OP_DST_SCALE_INV_8X 7 + +/* SRC1 LOOP */ +#define NV40_FP_OP_LOOP_INCR_SHIFT 19 +#define NV40_FP_OP_LOOP_INCR_MASK (0xFF << 19) +#define NV40_FP_OP_LOOP_INDEX_SHIFT 10 +#define NV40_FP_OP_LOOP_INDEX_MASK (0xFF << 10) +#define NV40_FP_OP_LOOP_COUNT_SHIFT 2 +#define NV40_FP_OP_LOOP_COUNT_MASK (0xFF << 2) + +/* SRC1 IF */ +#define NV40_FP_OP_ELSE_ID_SHIFT 2 +#define NV40_FP_OP_ELSE_ID_MASK (0xFF << 2) + +/* SRC1 CAL */ +#define NV40_FP_OP_IADDR_SHIFT 2 +#define NV40_FP_OP_IADDR_MASK (0xFF << 2) + +/* SRC1 REP + * I have no idea why there are 3 count values here.. but they + * have always been filled with the same value in my tests so + * far.. + */ +#define NV40_FP_OP_REP_COUNT1_SHIFT 2 +#define NV40_FP_OP_REP_COUNT1_MASK (0xFF << 2) +#define NV40_FP_OP_REP_COUNT2_SHIFT 10 +#define NV40_FP_OP_REP_COUNT2_MASK (0xFF << 10) +#define NV40_FP_OP_REP_COUNT3_SHIFT 19 +#define NV40_FP_OP_REP_COUNT3_MASK (0xFF << 19) + +/* SRC2 REP/IF */ +#define NV40_FP_OP_END_ID_SHIFT 2 +#define NV40_FP_OP_END_ID_MASK (0xFF << 2) + +// SRC2 high-order +#define NV40_FP_OP_INDEX_INPUT (1 << 30) +#define NV40_FP_OP_ADDR_INDEX_SHIFT 19 +#define NV40_FP_OP_ADDR_INDEX_MASK (0xF << 19) + +//== Register selection == +#define NV40_FP_REG_TYPE_SHIFT 0 +#define NV40_FP_REG_TYPE_MASK (3 << 0) +# define NV40_FP_REG_TYPE_TEMP 0 +# define NV40_FP_REG_TYPE_INPUT 1 +# define NV40_FP_REG_TYPE_CONST 2 +#define NV40_FP_REG_SRC_SHIFT 2 +#define NV40_FP_REG_SRC_MASK (63 << 2) +#define NV40_FP_REG_SRC_HALF (1 << 8) +#define NV40_FP_REG_SWZ_ALL_SHIFT 9 +#define NV40_FP_REG_SWZ_ALL_MASK (255 << 9) +#define NV40_FP_REG_SWZ_X_SHIFT 9 +#define NV40_FP_REG_SWZ_X_MASK (3 << 9) +#define NV40_FP_REG_SWZ_Y_SHIFT 11 +#define NV40_FP_REG_SWZ_Y_MASK (3 << 11) +#define NV40_FP_REG_SWZ_Z_SHIFT 13 +#define NV40_FP_REG_SWZ_Z_MASK (3 << 13) +#define NV40_FP_REG_SWZ_W_SHIFT 15 +#define NV40_FP_REG_SWZ_W_MASK (3 << 15) +# define NV40_FP_SWIZZLE_X 0 +# define NV40_FP_SWIZZLE_Y 1 +# define NV40_FP_SWIZZLE_Z 2 +# define NV40_FP_SWIZZLE_W 3 +#define NV40_FP_REG_NEGATE (1 << 17) + +#ifndef NV40_SHADER_NO_FUCKEDNESS +#define NV40SR_NONE 0 +#define NV40SR_OUTPUT 1 +#define NV40SR_INPUT 2 +#define NV40SR_TEMP 3 +#define NV40SR_CONST 4 + +struct nv40_sreg { + int type; + int index; + + int dst_scale; + + int negate; + int abs; + int swz[4]; + + int cc_update; + int cc_update_reg; + int cc_test; + int cc_test_reg; + int cc_swz[4]; +}; + +static INLINE struct nv40_sreg +nv40_sr(int type, int index) +{ + struct nv40_sreg temp = { + .type = type, + .index = index, + .dst_scale = DEF_SCALE, + .abs = 0, + .negate = 0, + .swz = { 0, 1, 2, 3 }, + .cc_update = 0, + .cc_update_reg = 0, + .cc_test = DEF_CTEST, + .cc_test_reg = 0, + .cc_swz = { 0, 1, 2, 3 }, + }; + return temp; +} + +static INLINE struct nv40_sreg +nv40_sr_swz(struct nv40_sreg src, int x, int y, int z, int w) +{ + struct nv40_sreg dst = src; + + dst.swz[SWZ_X] = src.swz[x]; + dst.swz[SWZ_Y] = src.swz[y]; + dst.swz[SWZ_Z] = src.swz[z]; + dst.swz[SWZ_W] = src.swz[w]; + return dst; +} + +static INLINE struct nv40_sreg +nv40_sr_neg(struct nv40_sreg src) +{ + src.negate = !src.negate; + return src; +} + +static INLINE struct nv40_sreg +nv40_sr_abs(struct nv40_sreg src) +{ + src.abs = 1; + return src; +} + +static INLINE struct nv40_sreg +nv40_sr_scale(struct nv40_sreg src, int scale) +{ + src.dst_scale = scale; + return src; +} +#endif + +#endif diff --git a/src/gallium/drivers/nv40/nv40_state.c b/src/gallium/drivers/nv40/nv40_state.c new file mode 100644 index 0000000000..255c4b294d --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_state.c @@ -0,0 +1,740 @@ +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "draw/draw_context.h" + +#include "tgsi/tgsi_parse.h" + +#include "nv40_context.h" +#include "nv40_state.h" + +static void * +nv40_blend_state_create(struct pipe_context *pipe, + const struct pipe_blend_state *cso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nouveau_grobj *curie = nv40->screen->curie; + struct nv40_blend_state *bso = CALLOC(1, sizeof(*bso)); + struct nouveau_stateobj *so = so_new(16, 0); + + if (cso->blend_enable) { + so_method(so, curie, NV40TCL_BLEND_ENABLE, 3); + so_data (so, 1); + so_data (so, (nvgl_blend_func(cso->alpha_src_factor) << 16) | + nvgl_blend_func(cso->rgb_src_factor)); + so_data (so, nvgl_blend_func(cso->alpha_dst_factor) << 16 | + nvgl_blend_func(cso->rgb_dst_factor)); + so_method(so, curie, NV40TCL_BLEND_EQUATION, 1); + so_data (so, nvgl_blend_eqn(cso->alpha_func) << 16 | + nvgl_blend_eqn(cso->rgb_func)); + } else { + so_method(so, curie, NV40TCL_BLEND_ENABLE, 1); + so_data (so, 0); + } + + so_method(so, curie, NV40TCL_COLOR_MASK, 1); + so_data (so, (((cso->colormask & PIPE_MASK_A) ? (0x01 << 24) : 0) | + ((cso->colormask & PIPE_MASK_R) ? (0x01 << 16) : 0) | + ((cso->colormask & PIPE_MASK_G) ? (0x01 << 8) : 0) | + ((cso->colormask & PIPE_MASK_B) ? (0x01 << 0) : 0))); + + if (cso->logicop_enable) { + so_method(so, curie, NV40TCL_COLOR_LOGIC_OP_ENABLE, 2); + so_data (so, 1); + so_data (so, nvgl_logicop_func(cso->logicop_func)); + } else { + so_method(so, curie, NV40TCL_COLOR_LOGIC_OP_ENABLE, 1); + so_data (so, 0); + } + + so_method(so, curie, NV40TCL_DITHER_ENABLE, 1); + so_data (so, cso->dither ? 1 : 0); + + so_ref(so, &bso->so); + bso->pipe = *cso; + return (void *)bso; +} + +static void +nv40_blend_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->blend = hwcso; + nv40->dirty |= NV40_NEW_BLEND; +} + +static void +nv40_blend_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv40_blend_state *bso = hwcso; + + so_ref(NULL, &bso->so); + FREE(bso); +} + + +static INLINE unsigned +wrap_mode(unsigned wrap) { + unsigned ret; + + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + ret = NV40TCL_TEX_WRAP_S_REPEAT; + break; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + ret = NV40TCL_TEX_WRAP_S_MIRRORED_REPEAT; + break; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + ret = NV40TCL_TEX_WRAP_S_CLAMP_TO_EDGE; + break; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + ret = NV40TCL_TEX_WRAP_S_CLAMP_TO_BORDER; + break; + case PIPE_TEX_WRAP_CLAMP: + ret = NV40TCL_TEX_WRAP_S_CLAMP; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP_TO_EDGE; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP_TO_BORDER; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP; + break; + default: + NOUVEAU_ERR("unknown wrap mode: %d\n", wrap); + ret = NV40TCL_TEX_WRAP_S_REPEAT; + break; + } + + return ret >> NV40TCL_TEX_WRAP_S_SHIFT; +} + +static void * +nv40_sampler_state_create(struct pipe_context *pipe, + const struct pipe_sampler_state *cso) +{ + struct nv40_sampler_state *ps; + uint32_t filter = 0; + + ps = MALLOC(sizeof(struct nv40_sampler_state)); + + ps->fmt = 0; + if (!cso->normalized_coords) + ps->fmt |= NV40TCL_TEX_FORMAT_RECT; + + ps->wrap = ((wrap_mode(cso->wrap_s) << NV40TCL_TEX_WRAP_S_SHIFT) | + (wrap_mode(cso->wrap_t) << NV40TCL_TEX_WRAP_T_SHIFT) | + (wrap_mode(cso->wrap_r) << NV40TCL_TEX_WRAP_R_SHIFT)); + + ps->en = 0; + if (cso->max_anisotropy >= 2.0) { + /* no idea, binary driver sets it, works without it.. meh.. */ + ps->wrap |= (1 << 5); + + if (cso->max_anisotropy >= 16.0) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_16X; + } else + if (cso->max_anisotropy >= 12.0) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_12X; + } else + if (cso->max_anisotropy >= 10.0) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_10X; + } else + if (cso->max_anisotropy >= 8.0) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_8X; + } else + if (cso->max_anisotropy >= 6.0) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_6X; + } else + if (cso->max_anisotropy >= 4.0) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_4X; + } else { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_2X; + } + } + + switch (cso->mag_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + filter |= NV40TCL_TEX_FILTER_MAG_LINEAR; + break; + case PIPE_TEX_FILTER_NEAREST: + default: + filter |= NV40TCL_TEX_FILTER_MAG_NEAREST; + break; + } + + switch (cso->min_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + filter |= NV40TCL_TEX_FILTER_MIN_LINEAR_MIPMAP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + filter |= NV40TCL_TEX_FILTER_MIN_LINEAR_MIPMAP_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + filter |= NV40TCL_TEX_FILTER_MIN_LINEAR; + break; + } + break; + case PIPE_TEX_FILTER_NEAREST: + default: + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + filter |= NV40TCL_TEX_FILTER_MIN_NEAREST_MIPMAP_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + filter |= NV40TCL_TEX_FILTER_MIN_NEAREST_MIPMAP_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + filter |= NV40TCL_TEX_FILTER_MIN_NEAREST; + break; + } + break; + } + + ps->filt = filter; + + { + float limit; + + limit = CLAMP(cso->lod_bias, -16.0, 15.0); + ps->filt |= (int)(cso->lod_bias * 256.0) & 0x1fff; + + limit = CLAMP(cso->max_lod, 0.0, 15.0); + ps->en |= (int)(limit * 256.0) << 7; + + limit = CLAMP(cso->min_lod, 0.0, 15.0); + ps->en |= (int)(limit * 256.0) << 19; + } + + + if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + switch (cso->compare_func) { + case PIPE_FUNC_NEVER: + ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_NEVER; + break; + case PIPE_FUNC_GREATER: + ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_GREATER; + break; + case PIPE_FUNC_EQUAL: + ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_EQUAL; + break; + case PIPE_FUNC_GEQUAL: + ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_GEQUAL; + break; + case PIPE_FUNC_LESS: + ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_LESS; + break; + case PIPE_FUNC_NOTEQUAL: + ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_NOTEQUAL; + break; + case PIPE_FUNC_LEQUAL: + ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_LEQUAL; + break; + case PIPE_FUNC_ALWAYS: + ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_ALWAYS; + break; + default: + break; + } + } + + ps->bcol = ((float_to_ubyte(cso->border_color[3]) << 24) | + (float_to_ubyte(cso->border_color[0]) << 16) | + (float_to_ubyte(cso->border_color[1]) << 8) | + (float_to_ubyte(cso->border_color[2]) << 0)); + + return (void *)ps; +} + +static void +nv40_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler) +{ + struct nv40_context *nv40 = nv40_context(pipe); + unsigned unit; + + for (unit = 0; unit < nr; unit++) { + nv40->tex_sampler[unit] = sampler[unit]; + nv40->dirty_samplers |= (1 << unit); + } + + for (unit = nr; unit < nv40->nr_samplers; unit++) { + nv40->tex_sampler[unit] = NULL; + nv40->dirty_samplers |= (1 << unit); + } + + nv40->nr_samplers = nr; + nv40->dirty |= NV40_NEW_SAMPLER; +} + +static void +nv40_sampler_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + +static void +nv40_set_sampler_texture(struct pipe_context *pipe, unsigned nr, + struct pipe_texture **miptree) +{ + struct nv40_context *nv40 = nv40_context(pipe); + unsigned unit; + + for (unit = 0; unit < nr; unit++) { + pipe_texture_reference((struct pipe_texture **) + &nv40->tex_miptree[unit], miptree[unit]); + nv40->dirty_samplers |= (1 << unit); + } + + for (unit = nr; unit < nv40->nr_textures; unit++) { + pipe_texture_reference((struct pipe_texture **) + &nv40->tex_miptree[unit], NULL); + nv40->dirty_samplers |= (1 << unit); + } + + nv40->nr_textures = nr; + nv40->dirty |= NV40_NEW_SAMPLER; +} + +static void * +nv40_rasterizer_state_create(struct pipe_context *pipe, + const struct pipe_rasterizer_state *cso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nv40_rasterizer_state *rsso = CALLOC(1, sizeof(*rsso)); + struct nouveau_stateobj *so = so_new(32, 0); + struct nouveau_grobj *curie = nv40->screen->curie; + + /*XXX: ignored: + * light_twoside + * point_smooth -nohw + * multisample + */ + + so_method(so, curie, NV40TCL_SHADE_MODEL, 1); + so_data (so, cso->flatshade ? NV40TCL_SHADE_MODEL_FLAT : + NV40TCL_SHADE_MODEL_SMOOTH); + + so_method(so, curie, NV40TCL_LINE_WIDTH, 2); + so_data (so, (unsigned char)(cso->line_width * 8.0) & 0xff); + so_data (so, cso->line_smooth ? 1 : 0); + so_method(so, curie, NV40TCL_LINE_STIPPLE_ENABLE, 2); + so_data (so, cso->line_stipple_enable ? 1 : 0); + so_data (so, (cso->line_stipple_pattern << 16) | + cso->line_stipple_factor); + + so_method(so, curie, NV40TCL_POINT_SIZE, 1); + so_data (so, fui(cso->point_size)); + + so_method(so, curie, NV40TCL_POLYGON_MODE_FRONT, 6); + if (cso->front_winding == PIPE_WINDING_CCW) { + so_data(so, nvgl_polygon_mode(cso->fill_ccw)); + so_data(so, nvgl_polygon_mode(cso->fill_cw)); + switch (cso->cull_mode) { + case PIPE_WINDING_CCW: + so_data(so, NV40TCL_CULL_FACE_FRONT); + break; + case PIPE_WINDING_CW: + so_data(so, NV40TCL_CULL_FACE_BACK); + break; + case PIPE_WINDING_BOTH: + so_data(so, NV40TCL_CULL_FACE_FRONT_AND_BACK); + break; + default: + so_data(so, NV40TCL_CULL_FACE_BACK); + break; + } + so_data(so, NV40TCL_FRONT_FACE_CCW); + } else { + so_data(so, nvgl_polygon_mode(cso->fill_cw)); + so_data(so, nvgl_polygon_mode(cso->fill_ccw)); + switch (cso->cull_mode) { + case PIPE_WINDING_CCW: + so_data(so, NV40TCL_CULL_FACE_BACK); + break; + case PIPE_WINDING_CW: + so_data(so, NV40TCL_CULL_FACE_FRONT); + break; + case PIPE_WINDING_BOTH: + so_data(so, NV40TCL_CULL_FACE_FRONT_AND_BACK); + break; + default: + so_data(so, NV40TCL_CULL_FACE_BACK); + break; + } + so_data(so, NV40TCL_FRONT_FACE_CW); + } + so_data(so, cso->poly_smooth ? 1 : 0); + so_data(so, (cso->cull_mode != PIPE_WINDING_NONE) ? 1 : 0); + + so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1); + so_data (so, cso->poly_stipple_enable ? 1 : 0); + + so_method(so, curie, NV40TCL_POLYGON_OFFSET_POINT_ENABLE, 3); + if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) || + (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT)) + so_data(so, 1); + else + so_data(so, 0); + if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) || + (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE)) + so_data(so, 1); + else + so_data(so, 0); + if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) || + (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL)) + so_data(so, 1); + else + so_data(so, 0); + if (cso->offset_cw || cso->offset_ccw) { + so_method(so, curie, NV40TCL_POLYGON_OFFSET_FACTOR, 2); + so_data (so, fui(cso->offset_scale)); + so_data (so, fui(cso->offset_units * 2)); + } + + so_method(so, curie, NV40TCL_POINT_SPRITE, 1); + if (cso->point_sprite) { + unsigned psctl = (1 << 0), i; + + for (i = 0; i < 8; i++) { + if (cso->sprite_coord_mode[i] != PIPE_SPRITE_COORD_NONE) + psctl |= (1 << (8 + i)); + } + + so_data(so, psctl); + } else { + so_data(so, 0); + } + + so_ref(so, &rsso->so); + rsso->pipe = *cso; + return (void *)rsso; +} + +static void +nv40_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->rasterizer = hwcso; + nv40->dirty |= NV40_NEW_RAST; + nv40->draw_dirty |= NV40_NEW_RAST; +} + +static void +nv40_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv40_rasterizer_state *rsso = hwcso; + + so_ref(NULL, &rsso->so); + FREE(rsso); +} + +static void * +nv40_depth_stencil_alpha_state_create(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *cso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nv40_zsa_state *zsaso = CALLOC(1, sizeof(*zsaso)); + struct nouveau_stateobj *so = so_new(32, 0); + struct nouveau_grobj *curie = nv40->screen->curie; + + so_method(so, curie, NV40TCL_DEPTH_FUNC, 3); + so_data (so, nvgl_comparison_op(cso->depth.func)); + so_data (so, cso->depth.writemask ? 1 : 0); + so_data (so, cso->depth.enabled ? 1 : 0); + + so_method(so, curie, NV40TCL_ALPHA_TEST_ENABLE, 3); + so_data (so, cso->alpha.enabled ? 1 : 0); + so_data (so, nvgl_comparison_op(cso->alpha.func)); + so_data (so, float_to_ubyte(cso->alpha.ref)); + + if (cso->stencil[0].enabled) { + so_method(so, curie, NV40TCL_STENCIL_FRONT_ENABLE, 8); + so_data (so, cso->stencil[0].enabled ? 1 : 0); + so_data (so, cso->stencil[0].write_mask); + so_data (so, nvgl_comparison_op(cso->stencil[0].func)); + so_data (so, cso->stencil[0].ref_value); + so_data (so, cso->stencil[0].value_mask); + so_data (so, nvgl_stencil_op(cso->stencil[0].fail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[0].zfail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[0].zpass_op)); + } else { + so_method(so, curie, NV40TCL_STENCIL_FRONT_ENABLE, 1); + so_data (so, 0); + } + + if (cso->stencil[1].enabled) { + so_method(so, curie, NV40TCL_STENCIL_BACK_ENABLE, 8); + so_data (so, cso->stencil[1].enabled ? 1 : 0); + so_data (so, cso->stencil[1].write_mask); + so_data (so, nvgl_comparison_op(cso->stencil[1].func)); + so_data (so, cso->stencil[1].ref_value); + so_data (so, cso->stencil[1].value_mask); + so_data (so, nvgl_stencil_op(cso->stencil[1].fail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[1].zfail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[1].zpass_op)); + } else { + so_method(so, curie, NV40TCL_STENCIL_BACK_ENABLE, 1); + so_data (so, 0); + } + + so_ref(so, &zsaso->so); + zsaso->pipe = *cso; + return (void *)zsaso; +} + +static void +nv40_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->zsa = hwcso; + nv40->dirty |= NV40_NEW_ZSA; +} + +static void +nv40_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv40_zsa_state *zsaso = hwcso; + + so_ref(NULL, &zsaso->so); + FREE(zsaso); +} + +static void * +nv40_vp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *cso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nv40_vertex_program *vp; + + vp = CALLOC(1, sizeof(struct nv40_vertex_program)); + vp->pipe.tokens = tgsi_dup_tokens(cso->tokens); + vp->draw = draw_create_vertex_shader(nv40->draw, &vp->pipe); + + return (void *)vp; +} + +static void +nv40_vp_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->vertprog = hwcso; + nv40->dirty |= NV40_NEW_VERTPROG; + nv40->draw_dirty |= NV40_NEW_VERTPROG; +} + +static void +nv40_vp_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nv40_vertex_program *vp = hwcso; + + draw_delete_vertex_shader(nv40->draw, vp->draw); + nv40_vertprog_destroy(nv40, vp); + FREE((void*)vp->pipe.tokens); + FREE(vp); +} + +static void * +nv40_fp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *cso) +{ + struct nv40_fragment_program *fp; + + fp = CALLOC(1, sizeof(struct nv40_fragment_program)); + fp->pipe.tokens = tgsi_dup_tokens(cso->tokens); + + tgsi_scan_shader(fp->pipe.tokens, &fp->info); + + return (void *)fp; +} + +static void +nv40_fp_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->fragprog = hwcso; + nv40->dirty |= NV40_NEW_FRAGPROG; +} + +static void +nv40_fp_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nv40_fragment_program *fp = hwcso; + + nv40_fragprog_destroy(nv40, fp); + FREE((void*)fp->pipe.tokens); + FREE(fp); +} + +static void +nv40_set_blend_color(struct pipe_context *pipe, + const struct pipe_blend_color *bcol) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->blend_colour = *bcol; + nv40->dirty |= NV40_NEW_BCOL; +} + +static void +nv40_set_clip_state(struct pipe_context *pipe, + const struct pipe_clip_state *clip) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->clip = *clip; + nv40->dirty |= NV40_NEW_UCP; + nv40->draw_dirty |= NV40_NEW_UCP; +} + +static void +nv40_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, + const struct pipe_constant_buffer *buf ) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->constbuf[shader] = buf->buffer; + nv40->constbuf_nr[shader] = buf->size / (4 * sizeof(float)); + + if (shader == PIPE_SHADER_VERTEX) { + nv40->dirty |= NV40_NEW_VERTPROG; + } else + if (shader == PIPE_SHADER_FRAGMENT) { + nv40->dirty |= NV40_NEW_FRAGPROG; + } +} + +static void +nv40_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->framebuffer = *fb; + nv40->dirty |= NV40_NEW_FB; +} + +static void +nv40_set_polygon_stipple(struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + memcpy(nv40->stipple, stipple->stipple, 4 * 32); + nv40->dirty |= NV40_NEW_STIPPLE; +} + +static void +nv40_set_scissor_state(struct pipe_context *pipe, + const struct pipe_scissor_state *s) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->scissor = *s; + nv40->dirty |= NV40_NEW_SCISSOR; +} + +static void +nv40_set_viewport_state(struct pipe_context *pipe, + const struct pipe_viewport_state *vpt) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->viewport = *vpt; + nv40->dirty |= NV40_NEW_VIEWPORT; + nv40->draw_dirty |= NV40_NEW_VIEWPORT; +} + +static void +nv40_set_vertex_buffers(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_buffer *vb) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + memcpy(nv40->vtxbuf, vb, sizeof(*vb) * count); + nv40->vtxbuf_nr = count; + + nv40->dirty |= NV40_NEW_ARRAYS; + nv40->draw_dirty |= NV40_NEW_ARRAYS; +} + +static void +nv40_set_vertex_elements(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_element *ve) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + memcpy(nv40->vtxelt, ve, sizeof(*ve) * count); + nv40->vtxelt_nr = count; + + nv40->dirty |= NV40_NEW_ARRAYS; + nv40->draw_dirty |= NV40_NEW_ARRAYS; +} + +static void +nv40_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield) +{ + struct nv40_context *nv40 = nv40_context(pipe); + + nv40->edgeflags = bitfield; + nv40->dirty |= NV40_NEW_ARRAYS; + nv40->draw_dirty |= NV40_NEW_ARRAYS; +} + +void +nv40_init_state_functions(struct nv40_context *nv40) +{ + nv40->pipe.create_blend_state = nv40_blend_state_create; + nv40->pipe.bind_blend_state = nv40_blend_state_bind; + nv40->pipe.delete_blend_state = nv40_blend_state_delete; + + nv40->pipe.create_sampler_state = nv40_sampler_state_create; + nv40->pipe.bind_sampler_states = nv40_sampler_state_bind; + nv40->pipe.delete_sampler_state = nv40_sampler_state_delete; + nv40->pipe.set_sampler_textures = nv40_set_sampler_texture; + + nv40->pipe.create_rasterizer_state = nv40_rasterizer_state_create; + nv40->pipe.bind_rasterizer_state = nv40_rasterizer_state_bind; + nv40->pipe.delete_rasterizer_state = nv40_rasterizer_state_delete; + + nv40->pipe.create_depth_stencil_alpha_state = + nv40_depth_stencil_alpha_state_create; + nv40->pipe.bind_depth_stencil_alpha_state = + nv40_depth_stencil_alpha_state_bind; + nv40->pipe.delete_depth_stencil_alpha_state = + nv40_depth_stencil_alpha_state_delete; + + nv40->pipe.create_vs_state = nv40_vp_state_create; + nv40->pipe.bind_vs_state = nv40_vp_state_bind; + nv40->pipe.delete_vs_state = nv40_vp_state_delete; + + nv40->pipe.create_fs_state = nv40_fp_state_create; + nv40->pipe.bind_fs_state = nv40_fp_state_bind; + nv40->pipe.delete_fs_state = nv40_fp_state_delete; + + nv40->pipe.set_blend_color = nv40_set_blend_color; + nv40->pipe.set_clip_state = nv40_set_clip_state; + nv40->pipe.set_constant_buffer = nv40_set_constant_buffer; + nv40->pipe.set_framebuffer_state = nv40_set_framebuffer_state; + nv40->pipe.set_polygon_stipple = nv40_set_polygon_stipple; + nv40->pipe.set_scissor_state = nv40_set_scissor_state; + nv40->pipe.set_viewport_state = nv40_set_viewport_state; + + nv40->pipe.set_edgeflags = nv40_set_edgeflags; + nv40->pipe.set_vertex_buffers = nv40_set_vertex_buffers; + nv40->pipe.set_vertex_elements = nv40_set_vertex_elements; +} + diff --git a/src/gallium/drivers/nv40/nv40_state.h b/src/gallium/drivers/nv40/nv40_state.h new file mode 100644 index 0000000000..9c55903ae3 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_state.h @@ -0,0 +1,91 @@ +#ifndef __NV40_STATE_H__ +#define __NV40_STATE_H__ + +#include "pipe/p_state.h" +#include "tgsi/tgsi_scan.h" + +struct nv40_sampler_state { + uint32_t fmt; + uint32_t wrap; + uint32_t en; + uint32_t filt; + uint32_t bcol; +}; + +struct nv40_vertex_program_exec { + uint32_t data[4]; + boolean has_branch_offset; + int const_index; +}; + +struct nv40_vertex_program_data { + int index; /* immediates == -1 */ + float value[4]; +}; + +struct nv40_vertex_program { + struct pipe_shader_state pipe; + + struct draw_vertex_shader *draw; + + boolean translated; + + struct pipe_clip_state ucp; + + struct nv40_vertex_program_exec *insns; + unsigned nr_insns; + struct nv40_vertex_program_data *consts; + unsigned nr_consts; + + struct nouveau_resource *exec; + unsigned exec_start; + struct nouveau_resource *data; + unsigned data_start; + unsigned data_start_min; + + uint32_t ir; + uint32_t or; + uint32_t clip_ctrl; + struct nouveau_stateobj *so; +}; + +struct nv40_fragment_program_data { + unsigned offset; + unsigned index; +}; + +struct nv40_fragment_program { + struct pipe_shader_state pipe; + struct tgsi_shader_info info; + + boolean translated; + unsigned samplers; + + uint32_t *insn; + int insn_len; + + struct nv40_fragment_program_data *consts; + unsigned nr_consts; + + struct pipe_buffer *buffer; + + uint32_t fp_control; + struct nouveau_stateobj *so; +}; + +struct nv40_miptree { + struct pipe_texture base; + + struct pipe_buffer *buffer; + uint total_size; + + struct pipe_texture *shadow_tex; + struct pipe_surface *shadow_surface; + + struct { + uint pitch; + uint *image_offset; + } level[PIPE_MAX_TEXTURE_LEVELS]; +}; + +#endif diff --git a/src/gallium/drivers/nv40/nv40_state_blend.c b/src/gallium/drivers/nv40/nv40_state_blend.c new file mode 100644 index 0000000000..95e6d7394f --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_state_blend.c @@ -0,0 +1,40 @@ +#include "nv40_context.h" + +static boolean +nv40_state_blend_validate(struct nv40_context *nv40) +{ + so_ref(nv40->blend->so, &nv40->state.hw[NV40_STATE_BLEND]); + return TRUE; +} + +struct nv40_state_entry nv40_state_blend = { + .validate = nv40_state_blend_validate, + .dirty = { + .pipe = NV40_NEW_BLEND, + .hw = NV40_STATE_BLEND + } +}; + +static boolean +nv40_state_blend_colour_validate(struct nv40_context *nv40) +{ + struct nouveau_stateobj *so = so_new(2, 0); + struct pipe_blend_color *bcol = &nv40->blend_colour; + + so_method(so, nv40->screen->curie, NV40TCL_BLEND_COLOR, 1); + so_data (so, ((float_to_ubyte(bcol->color[3]) << 24) | + (float_to_ubyte(bcol->color[0]) << 16) | + (float_to_ubyte(bcol->color[1]) << 8) | + (float_to_ubyte(bcol->color[2]) << 0))); + + so_ref(so, &nv40->state.hw[NV40_STATE_BCOL]); + return TRUE; +} + +struct nv40_state_entry nv40_state_blend_colour = { + .validate = nv40_state_blend_colour_validate, + .dirty = { + .pipe = NV40_NEW_BCOL, + .hw = NV40_STATE_BCOL + } +}; diff --git a/src/gallium/drivers/nv40/nv40_state_emit.c b/src/gallium/drivers/nv40/nv40_state_emit.c new file mode 100644 index 0000000000..ab88dc416e --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_state_emit.c @@ -0,0 +1,184 @@ +#include "nv40_context.h" +#include "nv40_state.h" +#include "draw/draw_context.h" + +static struct nv40_state_entry *render_states[] = { + &nv40_state_framebuffer, + &nv40_state_rasterizer, + &nv40_state_scissor, + &nv40_state_stipple, + &nv40_state_fragprog, + &nv40_state_fragtex, + &nv40_state_vertprog, + &nv40_state_blend, + &nv40_state_blend_colour, + &nv40_state_zsa, + &nv40_state_viewport, + &nv40_state_vbo, + NULL +}; + +static struct nv40_state_entry *swtnl_states[] = { + &nv40_state_framebuffer, + &nv40_state_rasterizer, + &nv40_state_scissor, + &nv40_state_stipple, + &nv40_state_fragprog, + &nv40_state_fragtex, + &nv40_state_vertprog, + &nv40_state_blend, + &nv40_state_blend_colour, + &nv40_state_zsa, + &nv40_state_viewport, + &nv40_state_vtxfmt, + NULL +}; + +static void +nv40_state_do_validate(struct nv40_context *nv40, + struct nv40_state_entry **states) +{ + const struct pipe_framebuffer_state *fb = &nv40->framebuffer; + unsigned i; + + for (i = 0; i < fb->num_cbufs; i++) + fb->cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED; + if (fb->zsbuf) + fb->zsbuf->status = PIPE_SURFACE_STATUS_DEFINED; + + while (*states) { + struct nv40_state_entry *e = *states; + + if (nv40->dirty & e->dirty.pipe) { + if (e->validate(nv40)) + nv40->state.dirty |= (1ULL << e->dirty.hw); + } + + states++; + } + nv40->dirty = 0; +} + +void +nv40_state_emit(struct nv40_context *nv40) +{ + struct nv40_state *state = &nv40->state; + struct nv40_screen *screen = nv40->screen; + unsigned i, samplers; + uint64 states; + + if (nv40->pctx_id != screen->cur_pctx) { + for (i = 0; i < NV40_STATE_MAX; i++) { + if (state->hw[i] && screen->state[i] != state->hw[i]) + state->dirty |= (1ULL << i); + } + + screen->cur_pctx = nv40->pctx_id; + } + + for (i = 0, states = state->dirty; states; i++) { + if (!(states & (1ULL << i))) + continue; + so_ref (state->hw[i], &nv40->screen->state[i]); + if (state->hw[i]) + so_emit(nv40->nvws, nv40->screen->state[i]); + states &= ~(1ULL << i); + } + + if (state->dirty & ((1ULL << NV40_STATE_FRAGPROG) | + (1ULL << NV40_STATE_FRAGTEX0))) { + BEGIN_RING(curie, NV40TCL_TEX_CACHE_CTL, 1); + OUT_RING (2); + BEGIN_RING(curie, NV40TCL_TEX_CACHE_CTL, 1); + OUT_RING (1); + } + + state->dirty = 0; + + so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_FB]); + for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) { + if (!(samplers & (1 << i))) + continue; + so_emit_reloc_markers(nv40->nvws, + state->hw[NV40_STATE_FRAGTEX0+i]); + samplers &= ~(1ULL << i); + } + so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_FRAGPROG]); + if (state->hw[NV40_STATE_VTXBUF] && nv40->render_mode == HW) + so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_VTXBUF]); +} + +boolean +nv40_state_validate(struct nv40_context *nv40) +{ + boolean was_sw = nv40->fallback_swtnl ? TRUE : FALSE; + + if (nv40->render_mode != HW) { + /* Don't even bother trying to go back to hw if none + * of the states that caused swtnl previously have changed. + */ + if ((nv40->fallback_swtnl & nv40->dirty) + != nv40->fallback_swtnl) + return FALSE; + + /* Attempt to go to hwtnl again */ + nv40->pipe.flush(&nv40->pipe, 0, NULL); + nv40->dirty |= (NV40_NEW_VIEWPORT | + NV40_NEW_VERTPROG | + NV40_NEW_ARRAYS); + nv40->render_mode = HW; + } + + nv40_state_do_validate(nv40, render_states); + if (nv40->fallback_swtnl || nv40->fallback_swrast) + return FALSE; + + if (was_sw) + NOUVEAU_ERR("swtnl->hw\n"); + + return TRUE; +} + +boolean +nv40_state_validate_swtnl(struct nv40_context *nv40) +{ + struct draw_context *draw = nv40->draw; + + /* Setup for swtnl */ + if (nv40->render_mode == HW) { + NOUVEAU_ERR("hw->swtnl 0x%08x\n", nv40->fallback_swtnl); + nv40->pipe.flush(&nv40->pipe, 0, NULL); + nv40->dirty |= (NV40_NEW_VIEWPORT | + NV40_NEW_VERTPROG | + NV40_NEW_ARRAYS); + nv40->render_mode = SWTNL; + } + + if (nv40->draw_dirty & NV40_NEW_VERTPROG) + draw_bind_vertex_shader(draw, nv40->vertprog->draw); + + if (nv40->draw_dirty & NV40_NEW_RAST) + draw_set_rasterizer_state(draw, &nv40->rasterizer->pipe); + + if (nv40->draw_dirty & NV40_NEW_UCP) + draw_set_clip_state(draw, &nv40->clip); + + if (nv40->draw_dirty & NV40_NEW_VIEWPORT) + draw_set_viewport_state(draw, &nv40->viewport); + + if (nv40->draw_dirty & NV40_NEW_ARRAYS) { + draw_set_edgeflags(draw, nv40->edgeflags); + draw_set_vertex_buffers(draw, nv40->vtxbuf_nr, nv40->vtxbuf); + draw_set_vertex_elements(draw, nv40->vtxelt_nr, nv40->vtxelt); + } + + nv40_state_do_validate(nv40, swtnl_states); + if (nv40->fallback_swrast) { + NOUVEAU_ERR("swtnl->swrast 0x%08x\n", nv40->fallback_swrast); + return FALSE; + } + + nv40->draw_dirty = 0; + return TRUE; +} + diff --git a/src/gallium/drivers/nv40/nv40_state_fb.c b/src/gallium/drivers/nv40/nv40_state_fb.c new file mode 100644 index 0000000000..28592d71c3 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_state_fb.c @@ -0,0 +1,155 @@ +#include "nv40_context.h" +#include "nouveau/nouveau_util.h" + +static boolean +nv40_state_framebuffer_validate(struct nv40_context *nv40) +{ + struct pipe_framebuffer_state *fb = &nv40->framebuffer; + struct pipe_surface *rt[4], *zeta; + uint32_t rt_enable, rt_format; + int i, colour_format = 0, zeta_format = 0; + struct nouveau_stateobj *so = so_new(64, 10); + unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM; + unsigned w = fb->width; + unsigned h = fb->height; + + rt_enable = 0; + for (i = 0; i < fb->num_cbufs; i++) { + if (colour_format) { + assert(colour_format == fb->cbufs[i]->format); + } else { + colour_format = fb->cbufs[i]->format; + rt_enable |= (NV40TCL_RT_ENABLE_COLOR0 << i); + rt[i] = fb->cbufs[i]; + } + } + + if (rt_enable & (NV40TCL_RT_ENABLE_COLOR1 | NV40TCL_RT_ENABLE_COLOR2 | + NV40TCL_RT_ENABLE_COLOR3)) + rt_enable |= NV40TCL_RT_ENABLE_MRT; + + if (fb->zsbuf) { + zeta_format = fb->zsbuf->format; + zeta = fb->zsbuf; + } + + if (!(rt[0]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1))); + for (i = 1; i < fb->num_cbufs; i++) + assert(!(rt[i]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)); + + rt_format = NV40TCL_RT_FORMAT_TYPE_SWIZZLED | + log2i(fb->width) << NV40TCL_RT_FORMAT_LOG2_WIDTH_SHIFT | + log2i(fb->height) << NV40TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT; + } + else + rt_format = NV40TCL_RT_FORMAT_TYPE_LINEAR; + + switch (colour_format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case 0: + rt_format |= NV40TCL_RT_FORMAT_COLOR_A8R8G8B8; + break; + case PIPE_FORMAT_R5G6B5_UNORM: + rt_format |= NV40TCL_RT_FORMAT_COLOR_R5G6B5; + break; + default: + assert(0); + } + + switch (zeta_format) { + case PIPE_FORMAT_Z16_UNORM: + rt_format |= NV40TCL_RT_FORMAT_ZETA_Z16; + break; + case PIPE_FORMAT_Z24S8_UNORM: + case 0: + rt_format |= NV40TCL_RT_FORMAT_ZETA_Z24S8; + break; + default: + assert(0); + } + + if (rt_enable & NV40TCL_RT_ENABLE_COLOR0) { + so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR0, 1); + so_reloc (so, rt[0]->buffer, 0, rt_flags | NOUVEAU_BO_OR, + nv40->nvws->channel->vram->handle, + nv40->nvws->channel->gart->handle); + so_method(so, nv40->screen->curie, NV40TCL_COLOR0_PITCH, 2); + so_data (so, rt[0]->stride); + so_reloc (so, rt[0]->buffer, rt[0]->offset, rt_flags | + NOUVEAU_BO_LOW, 0, 0); + } + + if (rt_enable & NV40TCL_RT_ENABLE_COLOR1) { + so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR1, 1); + so_reloc (so, rt[1]->buffer, 0, rt_flags | NOUVEAU_BO_OR, + nv40->nvws->channel->vram->handle, + nv40->nvws->channel->gart->handle); + so_method(so, nv40->screen->curie, NV40TCL_COLOR1_OFFSET, 2); + so_reloc (so, rt[1]->buffer, rt[1]->offset, rt_flags | + NOUVEAU_BO_LOW, 0, 0); + so_data (so, rt[1]->stride); + } + + if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) { + so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR2, 1); + so_reloc (so, rt[2]->buffer, 0, rt_flags | NOUVEAU_BO_OR, + nv40->nvws->channel->vram->handle, + nv40->nvws->channel->gart->handle); + so_method(so, nv40->screen->curie, NV40TCL_COLOR2_OFFSET, 1); + so_reloc (so, rt[2]->buffer, rt[2]->offset, rt_flags | + NOUVEAU_BO_LOW, 0, 0); + so_method(so, nv40->screen->curie, NV40TCL_COLOR2_PITCH, 1); + so_data (so, rt[2]->stride); + } + + if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) { + so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR3, 1); + so_reloc (so, rt[3]->buffer, 0, rt_flags | NOUVEAU_BO_OR, + nv40->nvws->channel->vram->handle, + nv40->nvws->channel->gart->handle); + so_method(so, nv40->screen->curie, NV40TCL_COLOR3_OFFSET, 1); + so_reloc (so, rt[3]->buffer, rt[3]->offset, rt_flags | + NOUVEAU_BO_LOW, 0, 0); + so_method(so, nv40->screen->curie, NV40TCL_COLOR3_PITCH, 1); + so_data (so, rt[3]->stride); + } + + if (zeta_format) { + so_method(so, nv40->screen->curie, NV40TCL_DMA_ZETA, 1); + so_reloc (so, zeta->buffer, 0, rt_flags | NOUVEAU_BO_OR, + nv40->nvws->channel->vram->handle, + nv40->nvws->channel->gart->handle); + so_method(so, nv40->screen->curie, NV40TCL_ZETA_OFFSET, 1); + so_reloc (so, zeta->buffer, zeta->offset, rt_flags | + NOUVEAU_BO_LOW, 0, 0); + so_method(so, nv40->screen->curie, NV40TCL_ZETA_PITCH, 1); + so_data (so, zeta->stride); + } + + so_method(so, nv40->screen->curie, NV40TCL_RT_ENABLE, 1); + so_data (so, rt_enable); + so_method(so, nv40->screen->curie, NV40TCL_RT_HORIZ, 3); + so_data (so, (w << 16) | 0); + so_data (so, (h << 16) | 0); + so_data (so, rt_format); + so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_HORIZ, 2); + so_data (so, (w << 16) | 0); + so_data (so, (h << 16) | 0); + so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_CLIP_HORIZ(0), 2); + so_data (so, ((w - 1) << 16) | 0); + so_data (so, ((h - 1) << 16) | 0); + so_method(so, nv40->screen->curie, 0x1d88, 1); + so_data (so, (1 << 12) | h); + + so_ref(so, &nv40->state.hw[NV40_STATE_FB]); + return TRUE; +} + +struct nv40_state_entry nv40_state_framebuffer = { + .validate = nv40_state_framebuffer_validate, + .dirty = { + .pipe = NV40_NEW_FB, + .hw = NV40_STATE_FB + } +}; diff --git a/src/gallium/drivers/nv40/nv40_state_rasterizer.c b/src/gallium/drivers/nv40/nv40_state_rasterizer.c new file mode 100644 index 0000000000..9ecda5990f --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_state_rasterizer.c @@ -0,0 +1,17 @@ +#include "nv40_context.h" + +static boolean +nv40_state_rasterizer_validate(struct nv40_context *nv40) +{ + so_ref(nv40->rasterizer->so, + &nv40->state.hw[NV40_STATE_RAST]); + return TRUE; +} + +struct nv40_state_entry nv40_state_rasterizer = { + .validate = nv40_state_rasterizer_validate, + .dirty = { + .pipe = NV40_NEW_RAST, + .hw = NV40_STATE_RAST + } +}; diff --git a/src/gallium/drivers/nv40/nv40_state_scissor.c b/src/gallium/drivers/nv40/nv40_state_scissor.c new file mode 100644 index 0000000000..285239ef41 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_state_scissor.c @@ -0,0 +1,35 @@ +#include "nv40_context.h" + +static boolean +nv40_state_scissor_validate(struct nv40_context *nv40) +{ + struct pipe_rasterizer_state *rast = &nv40->rasterizer->pipe; + struct pipe_scissor_state *s = &nv40->scissor; + struct nouveau_stateobj *so; + + if (nv40->state.hw[NV40_STATE_SCISSOR] && + (rast->scissor == 0 && nv40->state.scissor_enabled == 0)) + return FALSE; + nv40->state.scissor_enabled = rast->scissor; + + so = so_new(3, 0); + so_method(so, nv40->screen->curie, NV40TCL_SCISSOR_HORIZ, 2); + if (nv40->state.scissor_enabled) { + so_data (so, ((s->maxx - s->minx) << 16) | s->minx); + so_data (so, ((s->maxy - s->miny) << 16) | s->miny); + } else { + so_data (so, 4096 << 16); + so_data (so, 4096 << 16); + } + + so_ref(so, &nv40->state.hw[NV40_STATE_SCISSOR]); + return TRUE; +} + +struct nv40_state_entry nv40_state_scissor = { + .validate = nv40_state_scissor_validate, + .dirty = { + .pipe = NV40_NEW_SCISSOR | NV40_NEW_RAST, + .hw = NV40_STATE_SCISSOR + } +}; diff --git a/src/gallium/drivers/nv40/nv40_state_stipple.c b/src/gallium/drivers/nv40/nv40_state_stipple.c new file mode 100644 index 0000000000..b51024ad9b --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_state_stipple.c @@ -0,0 +1,39 @@ +#include "nv40_context.h" + +static boolean +nv40_state_stipple_validate(struct nv40_context *nv40) +{ + struct pipe_rasterizer_state *rast = &nv40->rasterizer->pipe; + struct nouveau_grobj *curie = nv40->screen->curie; + struct nouveau_stateobj *so; + + if (nv40->state.hw[NV40_STATE_STIPPLE] && + (rast->poly_stipple_enable == 0 && nv40->state.stipple_enabled == 0)) + return FALSE; + + if (rast->poly_stipple_enable) { + unsigned i; + + so = so_new(35, 0); + so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1); + so_data (so, 1); + so_method(so, curie, NV40TCL_POLYGON_STIPPLE_PATTERN(0), 32); + for (i = 0; i < 32; i++) + so_data(so, nv40->stipple[i]); + } else { + so = so_new(2, 0); + so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1); + so_data (so, 0); + } + + so_ref(so, &nv40->state.hw[NV40_STATE_STIPPLE]); + return TRUE; +} + +struct nv40_state_entry nv40_state_stipple = { + .validate = nv40_state_stipple_validate, + .dirty = { + .pipe = NV40_NEW_STIPPLE | NV40_NEW_RAST, + .hw = NV40_STATE_STIPPLE, + } +}; diff --git a/src/gallium/drivers/nv40/nv40_state_viewport.c b/src/gallium/drivers/nv40/nv40_state_viewport.c new file mode 100644 index 0000000000..869a55b405 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_state_viewport.c @@ -0,0 +1,67 @@ +#include "nv40_context.h" + +static boolean +nv40_state_viewport_validate(struct nv40_context *nv40) +{ + struct pipe_viewport_state *vpt = &nv40->viewport; + struct nouveau_stateobj *so; + unsigned bypass; + + if (nv40->render_mode == HW && !nv40->rasterizer->pipe.bypass_clipping) + bypass = 0; + else + bypass = 1; + + if (nv40->state.hw[NV40_STATE_VIEWPORT] && + (bypass || !(nv40->dirty & NV40_NEW_VIEWPORT)) && + nv40->state.viewport_bypass == bypass) + return FALSE; + nv40->state.viewport_bypass = bypass; + + so = so_new(11, 0); + if (!bypass) { + so_method(so, nv40->screen->curie, + NV40TCL_VIEWPORT_TRANSLATE_X, 8); + so_data (so, fui(vpt->translate[0])); + so_data (so, fui(vpt->translate[1])); + so_data (so, fui(vpt->translate[2])); + so_data (so, fui(vpt->translate[3])); + so_data (so, fui(vpt->scale[0])); + so_data (so, fui(vpt->scale[1])); + so_data (so, fui(vpt->scale[2])); + so_data (so, fui(vpt->scale[3])); + so_method(so, nv40->screen->curie, 0x1d78, 1); + so_data (so, 1); + } else { + so_method(so, nv40->screen->curie, + NV40TCL_VIEWPORT_TRANSLATE_X, 8); + so_data (so, fui(0.0)); + so_data (so, fui(0.0)); + so_data (so, fui(0.0)); + so_data (so, fui(0.0)); + so_data (so, fui(1.0)); + so_data (so, fui(1.0)); + so_data (so, fui(1.0)); + so_data (so, fui(0.0)); + /* Not entirely certain what this is yet. The DDX uses this + * value also as it fixes rendering when you pass + * pre-transformed vertices to the GPU. My best gusss is that + * this bypasses some culling/clipping stage. Might be worth + * noting that points/lines are uneffected by whatever this + * value fixes, only filled polygons are effected. + */ + so_method(so, nv40->screen->curie, 0x1d78, 1); + so_data (so, 0x110); + } + + so_ref(so, &nv40->state.hw[NV40_STATE_VIEWPORT]); + return TRUE; +} + +struct nv40_state_entry nv40_state_viewport = { + .validate = nv40_state_viewport_validate, + .dirty = { + .pipe = NV40_NEW_VIEWPORT | NV40_NEW_RAST, + .hw = NV40_STATE_VIEWPORT + } +}; diff --git a/src/gallium/drivers/nv40/nv40_state_zsa.c b/src/gallium/drivers/nv40/nv40_state_zsa.c new file mode 100644 index 0000000000..fb760677c8 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_state_zsa.c @@ -0,0 +1,17 @@ +#include "nv40_context.h" + +static boolean +nv40_state_zsa_validate(struct nv40_context *nv40) +{ + so_ref(nv40->zsa->so, + &nv40->state.hw[NV40_STATE_ZSA]); + return TRUE; +} + +struct nv40_state_entry nv40_state_zsa = { + .validate = nv40_state_zsa_validate, + .dirty = { + .pipe = NV40_NEW_ZSA, + .hw = NV40_STATE_ZSA + } +}; diff --git a/src/gallium/drivers/nv40/nv40_surface.c b/src/gallium/drivers/nv40/nv40_surface.c new file mode 100644 index 0000000000..576af7c59e --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_surface.c @@ -0,0 +1,77 @@ + +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "nv40_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/p_inlines.h" + +#include "util/u_tile.h" + +static void +nv40_surface_copy(struct pipe_context *pipe, boolean do_flip, + struct pipe_surface *dest, unsigned destx, unsigned desty, + struct pipe_surface *src, unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nouveau_winsys *nvws = nv40->nvws; + + if (do_flip) { + /*XXX: This dodgyness will do for now for correctness. But, + * need to investigate whether the 2D engine is able to + * manage a flip (perhaps SIFM?), if not, use the 3D engine + */ + desty += height; + while (height--) { + nvws->surface_copy(nvws, dest, destx, desty--, src, + srcx, srcy++, width, 1); + } + } else { + nvws->surface_copy(nvws, dest, destx, desty, src, srcx, srcy, + width, height); + } +} + +static void +nv40_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest, + unsigned destx, unsigned desty, unsigned width, + unsigned height, unsigned value) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nouveau_winsys *nvws = nv40->nvws; + + nvws->surface_fill(nvws, dest, destx, desty, width, height, value); +} + +void +nv40_init_surface_functions(struct nv40_context *nv40) +{ + nv40->pipe.surface_copy = nv40_surface_copy; + nv40->pipe.surface_fill = nv40_surface_fill; +} diff --git a/src/gallium/drivers/nv40/nv40_vbo.c b/src/gallium/drivers/nv40/nv40_vbo.c new file mode 100644 index 0000000000..09f6e79d32 --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_vbo.c @@ -0,0 +1,555 @@ +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +#include "nv40_context.h" +#include "nv40_state.h" + +#include "nouveau/nouveau_channel.h" +#include "nouveau/nouveau_pushbuf.h" +#include "nouveau/nouveau_util.h" + +#define FORCE_SWTNL 0 + +static INLINE int +nv40_vbo_format_to_hw(enum pipe_format pipe, unsigned *fmt, unsigned *ncomp) +{ + switch (pipe) { + case PIPE_FORMAT_R32_FLOAT: + case PIPE_FORMAT_R32G32_FLOAT: + case PIPE_FORMAT_R32G32B32_FLOAT: + case PIPE_FORMAT_R32G32B32A32_FLOAT: + *fmt = NV40TCL_VTXFMT_TYPE_FLOAT; + break; + case PIPE_FORMAT_R8_UNORM: + case PIPE_FORMAT_R8G8_UNORM: + case PIPE_FORMAT_R8G8B8_UNORM: + case PIPE_FORMAT_R8G8B8A8_UNORM: + *fmt = NV40TCL_VTXFMT_TYPE_UBYTE; + break; + case PIPE_FORMAT_R16_SSCALED: + case PIPE_FORMAT_R16G16_SSCALED: + case PIPE_FORMAT_R16G16B16_SSCALED: + case PIPE_FORMAT_R16G16B16A16_SSCALED: + *fmt = NV40TCL_VTXFMT_TYPE_USHORT; + break; + default: + NOUVEAU_ERR("Unknown format %s\n", pf_name(pipe)); + return 1; + } + + switch (pipe) { + case PIPE_FORMAT_R8_UNORM: + case PIPE_FORMAT_R32_FLOAT: + case PIPE_FORMAT_R16_SSCALED: + *ncomp = 1; + break; + case PIPE_FORMAT_R8G8_UNORM: + case PIPE_FORMAT_R32G32_FLOAT: + case PIPE_FORMAT_R16G16_SSCALED: + *ncomp = 2; + break; + case PIPE_FORMAT_R8G8B8_UNORM: + case PIPE_FORMAT_R32G32B32_FLOAT: + case PIPE_FORMAT_R16G16B16_SSCALED: + *ncomp = 3; + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + case PIPE_FORMAT_R32G32B32A32_FLOAT: + case PIPE_FORMAT_R16G16B16A16_SSCALED: + *ncomp = 4; + break; + default: + NOUVEAU_ERR("Unknown format %s\n", pf_name(pipe)); + return 1; + } + + return 0; +} + +static boolean +nv40_vbo_set_idxbuf(struct nv40_context *nv40, struct pipe_buffer *ib, + unsigned ib_size) +{ + struct pipe_screen *pscreen = &nv40->screen->pipe; + unsigned type; + + if (!ib) { + nv40->idxbuf = NULL; + nv40->idxbuf_format = 0xdeadbeef; + return FALSE; + } + + if (!pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF) || ib_size == 1) + return FALSE; + + switch (ib_size) { + case 2: + type = NV40TCL_IDXBUF_FORMAT_TYPE_U16; + break; + case 4: + type = NV40TCL_IDXBUF_FORMAT_TYPE_U32; + break; + default: + return FALSE; + } + + if (ib != nv40->idxbuf || + type != nv40->idxbuf_format) { + nv40->dirty |= NV40_NEW_ARRAYS; + nv40->idxbuf = ib; + nv40->idxbuf_format = type; + } + + return TRUE; +} + +static boolean +nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so, + int attrib, struct pipe_vertex_element *ve, + struct pipe_vertex_buffer *vb) +{ + struct pipe_winsys *ws = nv40->pipe.winsys; + struct nouveau_grobj *curie = nv40->screen->curie; + unsigned type, ncomp; + void *map; + + if (nv40_vbo_format_to_hw(ve->src_format, &type, &ncomp)) + return FALSE; + + map = ws->buffer_map(ws, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ); + map += vb->buffer_offset + ve->src_offset; + + switch (type) { + case NV40TCL_VTXFMT_TYPE_FLOAT: + { + float *v = map; + + switch (ncomp) { + case 4: + so_method(so, curie, NV40TCL_VTX_ATTR_4F_X(attrib), 4); + so_data (so, fui(v[0])); + so_data (so, fui(v[1])); + so_data (so, fui(v[2])); + so_data (so, fui(v[3])); + break; + case 3: + so_method(so, curie, NV40TCL_VTX_ATTR_3F_X(attrib), 3); + so_data (so, fui(v[0])); + so_data (so, fui(v[1])); + so_data (so, fui(v[2])); + break; + case 2: + so_method(so, curie, NV40TCL_VTX_ATTR_2F_X(attrib), 2); + so_data (so, fui(v[0])); + so_data (so, fui(v[1])); + break; + case 1: + so_method(so, curie, NV40TCL_VTX_ATTR_1F(attrib), 1); + so_data (so, fui(v[0])); + break; + default: + ws->buffer_unmap(ws, vb->buffer); + return FALSE; + } + } + break; + default: + ws->buffer_unmap(ws, vb->buffer); + return FALSE; + } + + ws->buffer_unmap(ws, vb->buffer); + + return TRUE; +} + +boolean +nv40_draw_arrays(struct pipe_context *pipe, + unsigned mode, unsigned start, unsigned count) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nouveau_channel *chan = nv40->nvws->channel; + unsigned restart; + + nv40_vbo_set_idxbuf(nv40, NULL, 0); + if (FORCE_SWTNL || !nv40_state_validate(nv40)) { + return nv40_draw_elements_swtnl(pipe, NULL, 0, + mode, start, count); + } + + while (count) { + unsigned vc, nr; + + nv40_state_emit(nv40); + + vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 256, + mode, start, count, &restart); + if (!vc) { + FIRE_RING(NULL); + continue; + } + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (nvgl_primitive(mode)); + + nr = (vc & 0xff); + if (nr) { + BEGIN_RING(curie, NV40TCL_VB_VERTEX_BATCH, 1); + OUT_RING (((nr - 1) << 24) | start); + start += nr; + } + + nr = vc >> 8; + while (nr) { + unsigned push = nr > 2047 ? 2047 : nr; + + nr -= push; + + BEGIN_RING_NI(curie, NV40TCL_VB_VERTEX_BATCH, push); + while (push--) { + OUT_RING(((0x100 - 1) << 24) | start); + start += 0x100; + } + } + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (0); + + count -= vc; + start = restart; + } + + pipe->flush(pipe, 0, NULL); + return TRUE; +} + +static INLINE void +nv40_draw_elements_u08(struct nv40_context *nv40, void *ib, + unsigned mode, unsigned start, unsigned count) +{ + struct nouveau_channel *chan = nv40->nvws->channel; + + while (count) { + uint8_t *elts = (uint8_t *)ib + start; + unsigned vc, push, restart; + + nv40_state_emit(nv40); + + vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 2, + mode, start, count, &restart); + if (vc == 0) { + FIRE_RING(NULL); + continue; + } + count -= vc; + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (nvgl_primitive(mode)); + + if (vc & 1) { + BEGIN_RING(curie, NV40TCL_VB_ELEMENT_U32, 1); + OUT_RING (elts[0]); + elts++; vc--; + } + + while (vc) { + unsigned i; + + push = MIN2(vc, 2047 * 2); + + BEGIN_RING_NI(curie, NV40TCL_VB_ELEMENT_U16, push >> 1); + for (i = 0; i < push; i+=2) + OUT_RING((elts[i+1] << 16) | elts[i]); + + vc -= push; + elts += push; + } + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (0); + + start = restart; + } +} + +static INLINE void +nv40_draw_elements_u16(struct nv40_context *nv40, void *ib, + unsigned mode, unsigned start, unsigned count) +{ + struct nouveau_channel *chan = nv40->nvws->channel; + + while (count) { + uint16_t *elts = (uint16_t *)ib + start; + unsigned vc, push, restart; + + nv40_state_emit(nv40); + + vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 2, + mode, start, count, &restart); + if (vc == 0) { + FIRE_RING(NULL); + continue; + } + count -= vc; + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (nvgl_primitive(mode)); + + if (vc & 1) { + BEGIN_RING(curie, NV40TCL_VB_ELEMENT_U32, 1); + OUT_RING (elts[0]); + elts++; vc--; + } + + while (vc) { + unsigned i; + + push = MIN2(vc, 2047 * 2); + + BEGIN_RING_NI(curie, NV40TCL_VB_ELEMENT_U16, push >> 1); + for (i = 0; i < push; i+=2) + OUT_RING((elts[i+1] << 16) | elts[i]); + + vc -= push; + elts += push; + } + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (0); + + start = restart; + } +} + +static INLINE void +nv40_draw_elements_u32(struct nv40_context *nv40, void *ib, + unsigned mode, unsigned start, unsigned count) +{ + struct nouveau_channel *chan = nv40->nvws->channel; + + while (count) { + uint32_t *elts = (uint32_t *)ib + start; + unsigned vc, push, restart; + + nv40_state_emit(nv40); + + vc = nouveau_vbuf_split(chan->pushbuf->remaining, 5, 1, + mode, start, count, &restart); + if (vc == 0) { + FIRE_RING(NULL); + continue; + } + count -= vc; + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (nvgl_primitive(mode)); + + while (vc) { + push = MIN2(vc, 2047); + + BEGIN_RING_NI(curie, NV40TCL_VB_ELEMENT_U32, push); + OUT_RINGp (elts, push); + + vc -= push; + elts += push; + } + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (0); + + start = restart; + } +} + +static boolean +nv40_draw_elements_inline(struct pipe_context *pipe, + struct pipe_buffer *ib, unsigned ib_size, + unsigned mode, unsigned start, unsigned count) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + void *map; + + map = ws->buffer_map(ws, ib, PIPE_BUFFER_USAGE_CPU_READ); + if (!ib) { + NOUVEAU_ERR("failed mapping ib\n"); + return FALSE; + } + + switch (ib_size) { + case 1: + nv40_draw_elements_u08(nv40, map, mode, start, count); + break; + case 2: + nv40_draw_elements_u16(nv40, map, mode, start, count); + break; + case 4: + nv40_draw_elements_u32(nv40, map, mode, start, count); + break; + default: + NOUVEAU_ERR("invalid idxbuf fmt %d\n", ib_size); + break; + } + + ws->buffer_unmap(ws, ib); + return TRUE; +} + +static boolean +nv40_draw_elements_vbo(struct pipe_context *pipe, + unsigned mode, unsigned start, unsigned count) +{ + struct nv40_context *nv40 = nv40_context(pipe); + struct nouveau_channel *chan = nv40->nvws->channel; + unsigned restart; + + while (count) { + unsigned nr, vc; + + nv40_state_emit(nv40); + + vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 256, + mode, start, count, &restart); + if (!vc) { + FIRE_RING(NULL); + continue; + } + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (nvgl_primitive(mode)); + + nr = (vc & 0xff); + if (nr) { + BEGIN_RING(curie, NV40TCL_VB_INDEX_BATCH, 1); + OUT_RING (((nr - 1) << 24) | start); + start += nr; + } + + nr = vc >> 8; + while (nr) { + unsigned push = nr > 2047 ? 2047 : nr; + + nr -= push; + + BEGIN_RING_NI(curie, NV40TCL_VB_INDEX_BATCH, push); + while (push--) { + OUT_RING(((0x100 - 1) << 24) | start); + start += 0x100; + } + } + + BEGIN_RING(curie, NV40TCL_BEGIN_END, 1); + OUT_RING (0); + + count -= vc; + start = restart; + } + + return TRUE; +} + +boolean +nv40_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, unsigned indexSize, + unsigned mode, unsigned start, unsigned count) +{ + struct nv40_context *nv40 = nv40_context(pipe); + boolean idxbuf; + + idxbuf = nv40_vbo_set_idxbuf(nv40, indexBuffer, indexSize); + if (FORCE_SWTNL || !nv40_state_validate(nv40)) { + return nv40_draw_elements_swtnl(pipe, NULL, 0, + mode, start, count); + } + + if (idxbuf) { + nv40_draw_elements_vbo(pipe, mode, start, count); + } else { + nv40_draw_elements_inline(pipe, indexBuffer, indexSize, + mode, start, count); + } + + pipe->flush(pipe, 0, NULL); + return TRUE; +} + +static boolean +nv40_vbo_validate(struct nv40_context *nv40) +{ + struct nouveau_stateobj *vtxbuf, *vtxfmt, *sattr = NULL; + struct nouveau_grobj *curie = nv40->screen->curie; + struct pipe_buffer *ib = nv40->idxbuf; + unsigned ib_format = nv40->idxbuf_format; + unsigned vb_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD; + int hw; + + if (nv40->edgeflags) { + nv40->fallback_swtnl |= NV40_NEW_ARRAYS; + return FALSE; + } + + vtxbuf = so_new(20, 18); + so_method(vtxbuf, curie, NV40TCL_VTXBUF_ADDRESS(0), nv40->vtxelt_nr); + vtxfmt = so_new(17, 0); + so_method(vtxfmt, curie, NV40TCL_VTXFMT(0), nv40->vtxelt_nr); + + for (hw = 0; hw < nv40->vtxelt_nr; hw++) { + struct pipe_vertex_element *ve; + struct pipe_vertex_buffer *vb; + unsigned type, ncomp; + + ve = &nv40->vtxelt[hw]; + vb = &nv40->vtxbuf[ve->vertex_buffer_index]; + + if (!vb->pitch) { + if (!sattr) + sattr = so_new(16 * 5, 0); + + if (nv40_vbo_static_attrib(nv40, sattr, hw, ve, vb)) { + so_data(vtxbuf, 0); + so_data(vtxfmt, NV40TCL_VTXFMT_TYPE_FLOAT); + continue; + } + } + + if (nv40_vbo_format_to_hw(ve->src_format, &type, &ncomp)) { + nv40->fallback_swtnl |= NV40_NEW_ARRAYS; + so_ref(NULL, &vtxbuf); + so_ref(NULL, &vtxfmt); + return FALSE; + } + + so_reloc(vtxbuf, vb->buffer, vb->buffer_offset + ve->src_offset, + vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR, + 0, NV40TCL_VTXBUF_ADDRESS_DMA1); + so_data (vtxfmt, ((vb->pitch << NV40TCL_VTXFMT_STRIDE_SHIFT) | + (ncomp << NV40TCL_VTXFMT_SIZE_SHIFT) | type)); + } + + if (ib) { + so_method(vtxbuf, curie, NV40TCL_IDXBUF_ADDRESS, 2); + so_reloc (vtxbuf, ib, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (vtxbuf, ib, ib_format, vb_flags | NOUVEAU_BO_OR, + 0, NV40TCL_IDXBUF_FORMAT_DMA1); + } + + so_method(vtxbuf, curie, 0x1710, 1); + so_data (vtxbuf, 0); + + so_ref(vtxbuf, &nv40->state.hw[NV40_STATE_VTXBUF]); + nv40->state.dirty |= (1ULL << NV40_STATE_VTXBUF); + so_ref(vtxfmt, &nv40->state.hw[NV40_STATE_VTXFMT]); + nv40->state.dirty |= (1ULL << NV40_STATE_VTXFMT); + so_ref(sattr, &nv40->state.hw[NV40_STATE_VTXATTR]); + nv40->state.dirty |= (1ULL << NV40_STATE_VTXATTR); + return FALSE; +} + +struct nv40_state_entry nv40_state_vbo = { + .validate = nv40_vbo_validate, + .dirty = { + .pipe = NV40_NEW_ARRAYS, + .hw = 0, + } +}; + diff --git a/src/gallium/drivers/nv40/nv40_vertprog.c b/src/gallium/drivers/nv40/nv40_vertprog.c new file mode 100644 index 0000000000..1392fe956f --- /dev/null +++ b/src/gallium/drivers/nv40/nv40_vertprog.c @@ -0,0 +1,1070 @@ +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" + +#include "nv40_context.h" +#include "nv40_state.h" + +/* TODO (at least...): + * 1. Indexed consts + ARL + * 3. NV_vp11, NV_vp2, NV_vp3 features + * - extra arith opcodes + * - branching + * - texture sampling + * - indexed attribs + * - indexed results + * 4. bugs + */ + +#define SWZ_X 0 +#define SWZ_Y 1 +#define SWZ_Z 2 +#define SWZ_W 3 +#define MASK_X 8 +#define MASK_Y 4 +#define MASK_Z 2 +#define MASK_W 1 +#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W) +#define DEF_SCALE 0 +#define DEF_CTEST 0 +#include "nv40_shader.h" + +#define swz(s,x,y,z,w) nv40_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w) +#define neg(s) nv40_sr_neg((s)) +#define abs(s) nv40_sr_abs((s)) + +#define NV40_VP_INST_DEST_CLIP(n) ((~0 - 6) + (n)) + +struct nv40_vpc { + struct nv40_vertex_program *vp; + + struct nv40_vertex_program_exec *vpi; + + unsigned r_temps; + unsigned r_temps_discard; + struct nv40_sreg r_result[PIPE_MAX_SHADER_OUTPUTS]; + struct nv40_sreg *r_address; + struct nv40_sreg *r_temp; + + struct nv40_sreg *imm; + unsigned nr_imm; + + unsigned hpos_idx; +}; + +static struct nv40_sreg +temp(struct nv40_vpc *vpc) +{ + int idx = ffs(~vpc->r_temps) - 1; + + if (idx < 0) { + NOUVEAU_ERR("out of temps!!\n"); + assert(0); + return nv40_sr(NV40SR_TEMP, 0); + } + + vpc->r_temps |= (1 << idx); + vpc->r_temps_discard |= (1 << idx); + return nv40_sr(NV40SR_TEMP, idx); +} + +static INLINE void +release_temps(struct nv40_vpc *vpc) +{ + vpc->r_temps &= ~vpc->r_temps_discard; + vpc->r_temps_discard = 0; +} + +static struct nv40_sreg +constant(struct nv40_vpc *vpc, int pipe, float x, float y, float z, float w) +{ + struct nv40_vertex_program *vp = vpc->vp; + struct nv40_vertex_program_data *vpd; + int idx; + + if (pipe >= 0) { + for (idx = 0; idx < vp->nr_consts; idx++) { + if (vp->consts[idx].index == pipe) + return nv40_sr(NV40SR_CONST, idx); + } + } + + idx = vp->nr_consts++; + vp->consts = realloc(vp->consts, sizeof(*vpd) * vp->nr_consts); + vpd = &vp->consts[idx]; + + vpd->index = pipe; + vpd->value[0] = x; + vpd->value[1] = y; + vpd->value[2] = z; + vpd->value[3] = w; + return nv40_sr(NV40SR_CONST, idx); +} + +#define arith(cc,s,o,d,m,s0,s1,s2) \ + nv40_vp_arith((cc), (s), NV40_VP_INST_##o, (d), (m), (s0), (s1), (s2)) + +static void +emit_src(struct nv40_vpc *vpc, uint32_t *hw, int pos, struct nv40_sreg src) +{ + struct nv40_vertex_program *vp = vpc->vp; + uint32_t sr = 0; + + switch (src.type) { + case NV40SR_TEMP: + sr |= (NV40_VP_SRC_REG_TYPE_TEMP << NV40_VP_SRC_REG_TYPE_SHIFT); + sr |= (src.index << NV40_VP_SRC_TEMP_SRC_SHIFT); + break; + case NV40SR_INPUT: + sr |= (NV40_VP_SRC_REG_TYPE_INPUT << + NV40_VP_SRC_REG_TYPE_SHIFT); + vp->ir |= (1 << src.index); + hw[1] |= (src.index << NV40_VP_INST_INPUT_SRC_SHIFT); + break; + case NV40SR_CONST: + sr |= (NV40_VP_SRC_REG_TYPE_CONST << + NV40_VP_SRC_REG_TYPE_SHIFT); + assert(vpc->vpi->const_index == -1 || + vpc->vpi->const_index == src.index); + vpc->vpi->const_index = src.index; + break; + case NV40SR_NONE: + sr |= (NV40_VP_SRC_REG_TYPE_INPUT << + NV40_VP_SRC_REG_TYPE_SHIFT); + break; + default: + assert(0); + } + + if (src.negate) + sr |= NV40_VP_SRC_NEGATE; + + if (src.abs) + hw[0] |= (1 << (21 + pos)); + + sr |= ((src.swz[0] << NV40_VP_SRC_SWZ_X_SHIFT) | + (src.swz[1] << NV40_VP_SRC_SWZ_Y_SHIFT) | + (src.swz[2] << NV40_VP_SRC_SWZ_Z_SHIFT) | + (src.swz[3] << NV40_VP_SRC_SWZ_W_SHIFT)); + + switch (pos) { + case 0: + hw[1] |= ((sr & NV40_VP_SRC0_HIGH_MASK) >> + NV40_VP_SRC0_HIGH_SHIFT) << NV40_VP_INST_SRC0H_SHIFT; + hw[2] |= (sr & NV40_VP_SRC0_LOW_MASK) << + NV40_VP_INST_SRC0L_SHIFT; + break; + case 1: + hw[2] |= sr << NV40_VP_INST_SRC1_SHIFT; + break; + case 2: + hw[2] |= ((sr & NV40_VP_SRC2_HIGH_MASK) >> + NV40_VP_SRC2_HIGH_SHIFT) << NV40_VP_INST_SRC2H_SHIFT; + hw[3] |= (sr & NV40_VP_SRC2_LOW_MASK) << + NV40_VP_INST_SRC2L_SHIFT; + break; + default: + assert(0); + } +} + +static void +emit_dst(struct nv40_vpc *vpc, uint32_t *hw, int slot, struct nv40_sreg dst) +{ + struct nv40_vertex_program *vp = vpc->vp; + + switch (dst.type) { + case NV40SR_TEMP: + hw[3] |= NV40_VP_INST_DEST_MASK; + if (slot == 0) { + hw[0] |= (dst.index << + NV40_VP_INST_VEC_DEST_TEMP_SHIFT); + } else { + hw[3] |= (dst.index << + NV40_VP_INST_SCA_DEST_TEMP_SHIFT); + } + break; + case NV40SR_OUTPUT: + switch (dst.index) { + case NV40_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break; + case NV40_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break; + case NV40_VP_INST_DEST_BFC0 : vp->or |= (1 << 2); break; + case NV40_VP_INST_DEST_BFC1 : vp->or |= (1 << 3); break; + case NV40_VP_INST_DEST_FOGC : vp->or |= (1 << 4); break; + case NV40_VP_INST_DEST_PSZ : vp->or |= (1 << 5); break; + case NV40_VP_INST_DEST_TC(0): vp->or |= (1 << 14); break; + case NV40_VP_INST_DEST_TC(1): vp->or |= (1 << 15); break; + case NV40_VP_INST_DEST_TC(2): vp->or |= (1 << 16); break; + case NV40_VP_INST_DEST_TC(3): vp->or |= (1 << 17); break; + case NV40_VP_INST_DEST_TC(4): vp->or |= (1 << 18); break; + case NV40_VP_INST_DEST_TC(5): vp->or |= (1 << 19); break; + case NV40_VP_INST_DEST_TC(6): vp->or |= (1 << 20); break; + case NV40_VP_INST_DEST_TC(7): vp->or |= (1 << 21); break; + case NV40_VP_INST_DEST_CLIP(0): + vp->or |= (1 << 6); + vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE0; + dst.index = NV40_VP_INST_DEST_FOGC; + break; + case NV40_VP_INST_DEST_CLIP(1): + vp->or |= (1 << 7); + vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE1; + dst.index = NV40_VP_INST_DEST_FOGC; + break; + case NV40_VP_INST_DEST_CLIP(2): + vp->or |= (1 << 8); + vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE2; + dst.index = NV40_VP_INST_DEST_FOGC; + break; + case NV40_VP_INST_DEST_CLIP(3): + vp->or |= (1 << 9); + vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE3; + dst.index = NV40_VP_INST_DEST_PSZ; + break; + case NV40_VP_INST_DEST_CLIP(4): + vp->or |= (1 << 10); + vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE4; + dst.index = NV40_VP_INST_DEST_PSZ; + break; + case NV40_VP_INST_DEST_CLIP(5): + vp->or |= (1 << 11); + vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE5; + dst.index = NV40_VP_INST_DEST_PSZ; + break; + default: + break; + } + + hw[3] |= (dst.index << NV40_VP_INST_DEST_SHIFT); + if (slot == 0) { + hw[0] |= NV40_VP_INST_VEC_RESULT; + hw[0] |= NV40_VP_INST_VEC_DEST_TEMP_MASK | (1<<20); + } else { + hw[3] |= NV40_VP_INST_SCA_RESULT; + hw[3] |= NV40_VP_INST_SCA_DEST_TEMP_MASK; + } + break; + default: + assert(0); + } +} + +static void +nv40_vp_arith(struct nv40_vpc *vpc, int slot, int op, + struct nv40_sreg dst, int mask, + struct nv40_sreg s0, struct nv40_sreg s1, + struct nv40_sreg s2) +{ + struct nv40_vertex_program *vp = vpc->vp; + uint32_t *hw; + + vp->insns = realloc(vp->insns, ++vp->nr_insns * sizeof(*vpc->vpi)); + vpc->vpi = &vp->insns[vp->nr_insns - 1]; + memset(vpc->vpi, 0, sizeof(*vpc->vpi)); + vpc->vpi->const_index = -1; + + hw = vpc->vpi->data; + + hw[0] |= (NV40_VP_INST_COND_TR << NV40_VP_INST_COND_SHIFT); + hw[0] |= ((0 << NV40_VP_INST_COND_SWZ_X_SHIFT) | + (1 << NV40_VP_INST_COND_SWZ_Y_SHIFT) | + (2 << NV40_VP_INST_COND_SWZ_Z_SHIFT) | + (3 << NV40_VP_INST_COND_SWZ_W_SHIFT)); + + if (slot == 0) { + hw[1] |= (op << NV40_VP_INST_VEC_OPCODE_SHIFT); + hw[3] |= NV40_VP_INST_SCA_DEST_TEMP_MASK; + hw[3] |= (mask << NV40_VP_INST_VEC_WRITEMASK_SHIFT); + } else { + hw[1] |= (op << NV40_VP_INST_SCA_OPCODE_SHIFT); + hw[0] |= (NV40_VP_INST_VEC_DEST_TEMP_MASK | (1 << 20)); + hw[3] |= (mask << NV40_VP_INST_SCA_WRITEMASK_SHIFT); + } + + emit_dst(vpc, hw, slot, dst); + emit_src(vpc, hw, 0, s0); + emit_src(vpc, hw, 1, s1); + emit_src(vpc, hw, 2, s2); +} + +static INLINE struct nv40_sreg +tgsi_src(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc) { + struct nv40_sreg src; + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + src = nv40_sr(NV40SR_INPUT, fsrc->SrcRegister.Index); + break; + case TGSI_FILE_CONSTANT: + src = constant(vpc, fsrc->SrcRegister.Index, 0, 0, 0, 0); + break; + case TGSI_FILE_IMMEDIATE: + src = vpc->imm[fsrc->SrcRegister.Index]; + break; + case TGSI_FILE_TEMPORARY: + src = vpc->r_temp[fsrc->SrcRegister.Index]; + break; + default: + NOUVEAU_ERR("bad src file\n"); + break; + } + + src.abs = fsrc->SrcRegisterExtMod.Absolute; + src.negate = fsrc->SrcRegister.Negate; + src.swz[0] = fsrc->SrcRegister.SwizzleX; + src.swz[1] = fsrc->SrcRegister.SwizzleY; + src.swz[2] = fsrc->SrcRegister.SwizzleZ; + src.swz[3] = fsrc->SrcRegister.SwizzleW; + return src; +} + +static INLINE struct nv40_sreg +tgsi_dst(struct nv40_vpc *vpc, const struct tgsi_full_dst_register *fdst) { + struct nv40_sreg dst; + + switch (fdst->DstRegister.File) { + case TGSI_FILE_OUTPUT: + dst = vpc->r_result[fdst->DstRegister.Index]; + break; + case TGSI_FILE_TEMPORARY: + dst = vpc->r_temp[fdst->DstRegister.Index]; + break; + case TGSI_FILE_ADDRESS: + dst = vpc->r_address[fdst->DstRegister.Index]; + break; + default: + NOUVEAU_ERR("bad dst file\n"); + break; + } + + return dst; +} + +static INLINE int +tgsi_mask(uint tgsi) +{ + int mask = 0; + + if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X; + if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y; + if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z; + if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W; + return mask; +} + +static boolean +src_native_swz(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc, + struct nv40_sreg *src) +{ + const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0); + struct nv40_sreg tgsi = tgsi_src(vpc, fsrc); + uint mask = 0, zero_mask = 0, one_mask = 0, neg_mask = 0; + uint neg[4] = { fsrc->SrcRegisterExtSwz.NegateX, + fsrc->SrcRegisterExtSwz.NegateY, + fsrc->SrcRegisterExtSwz.NegateZ, + fsrc->SrcRegisterExtSwz.NegateW }; + uint c; + + for (c = 0; c < 4; c++) { + switch (tgsi_util_get_full_src_register_extswizzle(fsrc, c)) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + mask |= tgsi_mask(1 << c); + break; + case TGSI_EXTSWIZZLE_ZERO: + zero_mask |= tgsi_mask(1 << c); + tgsi.swz[c] = SWZ_X; + break; + case TGSI_EXTSWIZZLE_ONE: + one_mask |= tgsi_mask(1 << c); + tgsi.swz[c] = SWZ_X; + break; + default: + assert(0); + } + + if (!tgsi.negate && neg[c]) + neg_mask |= tgsi_mask(1 << c); + } + + if (mask == MASK_ALL && !neg_mask) + return TRUE; + + *src = temp(vpc); + + if (mask) + arith(vpc, 0, OP_MOV, *src, mask, tgsi, none, none); + + if (zero_mask) + arith(vpc, 0, OP_SFL, *src, zero_mask, *src, none, none); + + if (one_mask) + arith(vpc, 0, OP_STR, *src, one_mask, *src, none, none); + + if (neg_mask) { + struct nv40_sreg one = temp(vpc); + arith(vpc, 0, OP_STR, one, neg_mask, one, none, none); + arith(vpc, 0, OP_MUL, *src, neg_mask, *src, neg(one), none); + } + + return FALSE; +} + +static boolean +nv40_vertprog_parse_instruction(struct nv40_vpc *vpc, + const struct tgsi_full_instruction *finst) +{ + struct nv40_sreg src[3], dst, tmp; + struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0); + int mask; + int ai = -1, ci = -1, ii = -1; + int i; + + if (finst->Instruction.Opcode == TGSI_OPCODE_END) + return TRUE; + + for (i = 0; i < finst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *fsrc; + + fsrc = &finst->FullSrcRegisters[i]; + if (fsrc->SrcRegister.File == TGSI_FILE_TEMPORARY) { + src[i] = tgsi_src(vpc, fsrc); + } + } + + for (i = 0; i < finst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *fsrc; + + fsrc = &finst->FullSrcRegisters[i]; + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + case TGSI_FILE_CONSTANT: + case TGSI_FILE_TEMPORARY: + if (!src_native_swz(vpc, fsrc, &src[i])) + continue; + break; + default: + break; + } + + switch (fsrc->SrcRegister.File) { + case TGSI_FILE_INPUT: + if (ai == -1 || ai == fsrc->SrcRegister.Index) { + ai = fsrc->SrcRegister.Index; + src[i] = tgsi_src(vpc, fsrc); + } else { + src[i] = temp(vpc); + arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + tgsi_src(vpc, fsrc), none, none); + } + break; + case TGSI_FILE_CONSTANT: + if ((ci == -1 && ii == -1) || + ci == fsrc->SrcRegister.Index) { + ci = fsrc->SrcRegister.Index; + src[i] = tgsi_src(vpc, fsrc); + } else { + src[i] = temp(vpc); + arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + tgsi_src(vpc, fsrc), none, none); + } + break; + case TGSI_FILE_IMMEDIATE: + if ((ci == -1 && ii == -1) || + ii == fsrc->SrcRegister.Index) { + ii = fsrc->SrcRegister.Index; + src[i] = tgsi_src(vpc, fsrc); + } else { + src[i] = temp(vpc); + arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + tgsi_src(vpc, fsrc), none, none); + } + break; + case TGSI_FILE_TEMPORARY: + /* handled above */ + break; + default: + NOUVEAU_ERR("bad src file\n"); + return FALSE; + } + } + + dst = tgsi_dst(vpc, &finst->FullDstRegisters[0]); + mask = tgsi_mask(finst->FullDstRegisters[0].DstRegister.WriteMask); + + switch (finst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + arith(vpc, 0, OP_MOV, dst, mask, abs(src[0]), none, none); + break; + case TGSI_OPCODE_ADD: + arith(vpc, 0, OP_ADD, dst, mask, src[0], none, src[1]); + break; + case TGSI_OPCODE_ARL: + arith(vpc, 0, OP_ARL, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_DP3: + arith(vpc, 0, OP_DP3, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DP4: + arith(vpc, 0, OP_DP4, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DPH: + arith(vpc, 0, OP_DPH, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_DST: + arith(vpc, 0, OP_DST, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_EX2: + arith(vpc, 1, OP_EX2, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_EXP: + arith(vpc, 1, OP_EXP, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_FLR: + arith(vpc, 0, OP_FLR, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_FRC: + arith(vpc, 0, OP_FRC, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_LG2: + arith(vpc, 1, OP_LG2, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_LIT: + arith(vpc, 1, OP_LIT, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_LOG: + arith(vpc, 1, OP_LOG, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_MAD: + arith(vpc, 0, OP_MAD, dst, mask, src[0], src[1], src[2]); + break; + case TGSI_OPCODE_MAX: + arith(vpc, 0, OP_MAX, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_MIN: + arith(vpc, 0, OP_MIN, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_MOV: + arith(vpc, 0, OP_MOV, dst, mask, src[0], none, none); + break; + case TGSI_OPCODE_MUL: + arith(vpc, 0, OP_MUL, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_POW: + tmp = temp(vpc); + arith(vpc, 1, OP_LG2, tmp, MASK_X, none, none, + swz(src[0], X, X, X, X)); + arith(vpc, 0, OP_MUL, tmp, MASK_X, swz(tmp, X, X, X, X), + swz(src[1], X, X, X, X), none); + arith(vpc, 1, OP_EX2, dst, mask, none, none, + swz(tmp, X, X, X, X)); + break; + case TGSI_OPCODE_RCP: + arith(vpc, 1, OP_RCP, dst, mask, none, none, src[0]); + break; + case TGSI_OPCODE_RET: + break; + case TGSI_OPCODE_RSQ: + arith(vpc, 1, OP_RSQ, dst, mask, none, none, abs(src[0])); + break; + case TGSI_OPCODE_SGE: + arith(vpc, 0, OP_SGE, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SLT: + arith(vpc, 0, OP_SLT, dst, mask, src[0], src[1], none); + break; + case TGSI_OPCODE_SUB: + arith(vpc, 0, OP_ADD, dst, mask, src[0], none, neg(src[1])); + break; + case TGSI_OPCODE_XPD: + tmp = temp(vpc); + arith(vpc, 0, OP_MUL, tmp, mask, + swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none); + arith(vpc, 0, OP_MAD, dst, (mask & ~MASK_W), + swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y), + neg(tmp)); + break; + default: + NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode); + return FALSE; + } + + release_temps(vpc); + return TRUE; +} + +static boolean +nv40_vertprog_parse_decl_output(struct nv40_vpc *vpc, + const struct tgsi_full_declaration *fdec) +{ + unsigned idx = fdec->DeclarationRange.First; + int hw; + + switch (fdec->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: + hw = NV40_VP_INST_DEST_POS; + vpc->hpos_idx = idx; + break; + case TGSI_SEMANTIC_COLOR: + if (fdec->Semantic.SemanticIndex == 0) { + hw = NV40_VP_INST_DEST_COL0; + } else + if (fdec->Semantic.SemanticIndex == 1) { + hw = NV40_VP_INST_DEST_COL1; + } else { + NOUVEAU_ERR("bad colour semantic index\n"); + return FALSE; + } + break; + case TGSI_SEMANTIC_BCOLOR: + if (fdec->Semantic.SemanticIndex == 0) { + hw = NV40_VP_INST_DEST_BFC0; + } else + if (fdec->Semantic.SemanticIndex == 1) { + hw = NV40_VP_INST_DEST_BFC1; + } else { + NOUVEAU_ERR("bad bcolour semantic index\n"); + return FALSE; + } + break; + case TGSI_SEMANTIC_FOG: + hw = NV40_VP_INST_DEST_FOGC; + break; + case TGSI_SEMANTIC_PSIZE: + hw = NV40_VP_INST_DEST_PSZ; + break; + case TGSI_SEMANTIC_GENERIC: + if (fdec->Semantic.SemanticIndex <= 7) { + hw = NV40_VP_INST_DEST_TC(fdec->Semantic.SemanticIndex); + } else { + NOUVEAU_ERR("bad generic semantic index\n"); + return FALSE; + } + break; + default: + NOUVEAU_ERR("bad output semantic\n"); + return FALSE; + } + + vpc->r_result[idx] = nv40_sr(NV40SR_OUTPUT, hw); + return TRUE; +} + +static boolean +nv40_vertprog_prepare(struct nv40_vpc *vpc) +{ + struct tgsi_parse_context p; + int high_temp = -1, high_addr = -1, nr_imm = 0, i; + + tgsi_parse_init(&p, vpc->vp->pipe.tokens); + while (!tgsi_parse_end_of_tokens(&p)) { + const union tgsi_full_token *tok = &p.FullToken; + + tgsi_parse_token(&p); + switch(tok->Token.Type) { + case TGSI_TOKEN_TYPE_IMMEDIATE: + nr_imm++; + break; + case TGSI_TOKEN_TYPE_DECLARATION: + { + const struct tgsi_full_declaration *fdec; + + fdec = &p.FullToken.FullDeclaration; + switch (fdec->Declaration.File) { + case TGSI_FILE_TEMPORARY: + if (fdec->DeclarationRange.Last > high_temp) { + high_temp = + fdec->DeclarationRange.Last; + } + break; +#if 0 /* this would be nice.. except gallium doesn't track it */ + case TGSI_FILE_ADDRESS: + if (fdec->DeclarationRange.Last > high_addr) { + high_addr = + fdec->DeclarationRange.Last; + } + break; +#endif + case TGSI_FILE_OUTPUT: + if (!nv40_vertprog_parse_decl_output(vpc, fdec)) + return FALSE; + break; + default: + break; + } + } + break; +#if 1 /* yay, parse instructions looking for address regs instead */ + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + const struct tgsi_full_instruction *finst; + const struct tgsi_full_dst_register *fdst; + + finst = &p.FullToken.FullInstruction; + fdst = &finst->FullDstRegisters[0]; + + if (fdst->DstRegister.File == TGSI_FILE_ADDRESS) { + if (fdst->DstRegister.Index > high_addr) + high_addr = fdst->DstRegister.Index; + } + + } + break; +#endif + default: + break; + } + } + tgsi_parse_free(&p); + + if (nr_imm) { + vpc->imm = CALLOC(nr_imm, sizeof(struct nv40_sreg)); + assert(vpc->imm); + } + + if (++high_temp) { + vpc->r_temp = CALLOC(high_temp, sizeof(struct nv40_sreg)); + for (i = 0; i < high_temp; i++) + vpc->r_temp[i] = temp(vpc); + } + + if (++high_addr) { + vpc->r_address = CALLOC(high_addr, sizeof(struct nv40_sreg)); + for (i = 0; i < high_addr; i++) + vpc->r_address[i] = temp(vpc); + } + + vpc->r_temps_discard = 0; + return TRUE; +} + +static void +nv40_vertprog_translate(struct nv40_context *nv40, + struct nv40_vertex_program *vp) +{ + struct tgsi_parse_context parse; + struct nv40_vpc *vpc = NULL; + struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0); + int i; + + vpc = CALLOC(1, sizeof(struct nv40_vpc)); + if (!vpc) + return; + vpc->vp = vp; + + if (!nv40_vertprog_prepare(vpc)) { + FREE(vpc); + return; + } + + /* Redirect post-transform vertex position to a temp if user clip + * planes are enabled. We need to append code the the vtxprog + * to handle clip planes later. + */ + if (vp->ucp.nr) { + vpc->r_result[vpc->hpos_idx] = temp(vpc); + vpc->r_temps_discard = 0; + } + + tgsi_parse_init(&parse, vp->pipe.tokens); + + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + const struct tgsi_full_immediate *imm; + + imm = &parse.FullToken.FullImmediate; + assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32); +// assert(imm->Immediate.Size == 4); + vpc->imm[vpc->nr_imm++] = + constant(vpc, -1, + imm->u.ImmediateFloat32[0].Float, + imm->u.ImmediateFloat32[1].Float, + imm->u.ImmediateFloat32[2].Float, + imm->u.ImmediateFloat32[3].Float); + } + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + const struct tgsi_full_instruction *finst; + finst = &parse.FullToken.FullInstruction; + if (!nv40_vertprog_parse_instruction(vpc, finst)) + goto out_err; + } + break; + default: + break; + } + } + + /* Write out HPOS if it was redirected to a temp earlier */ + if (vpc->r_result[vpc->hpos_idx].type != NV40SR_OUTPUT) { + struct nv40_sreg hpos = nv40_sr(NV40SR_OUTPUT, + NV40_VP_INST_DEST_POS); + struct nv40_sreg htmp = vpc->r_result[vpc->hpos_idx]; + + arith(vpc, 0, OP_MOV, hpos, MASK_ALL, htmp, none, none); + } + + /* Insert code to handle user clip planes */ + for (i = 0; i < vp->ucp.nr; i++) { + struct nv40_sreg cdst = nv40_sr(NV40SR_OUTPUT, + NV40_VP_INST_DEST_CLIP(i)); + struct nv40_sreg ceqn = constant(vpc, -1, + nv40->clip.ucp[i][0], + nv40->clip.ucp[i][1], + nv40->clip.ucp[i][2], + nv40->clip.ucp[i][3]); + struct nv40_sreg htmp = vpc->r_result[vpc->hpos_idx]; + unsigned mask; + + switch (i) { + case 0: case 3: mask = MASK_Y; break; + case 1: case 4: mask = MASK_Z; break; + case 2: case 5: mask = MASK_W; break; + default: + NOUVEAU_ERR("invalid clip dist #%d\n", i); + goto out_err; + } + + arith(vpc, 0, OP_DP4, cdst, mask, htmp, ceqn, none); + } + + vp->insns[vp->nr_insns - 1].data[3] |= NV40_VP_INST_LAST; + vp->translated = TRUE; +out_err: + tgsi_parse_free(&parse); + if (vpc->r_temp) + FREE(vpc->r_temp); + if (vpc->r_address) + FREE(vpc->r_address); + if (vpc->imm) + FREE(vpc->imm); + FREE(vpc); +} + +static boolean +nv40_vertprog_validate(struct nv40_context *nv40) +{ + struct nouveau_winsys *nvws = nv40->nvws; + struct pipe_winsys *ws = nv40->pipe.winsys; + struct nouveau_grobj *curie = nv40->screen->curie; + struct nv40_vertex_program *vp; + struct pipe_buffer *constbuf; + boolean upload_code = FALSE, upload_data = FALSE; + int i; + + if (nv40->render_mode == HW) { + vp = nv40->vertprog; + constbuf = nv40->constbuf[PIPE_SHADER_VERTEX]; + + if ((nv40->dirty & NV40_NEW_UCP) || + memcmp(&nv40->clip, &vp->ucp, sizeof(vp->ucp))) { + nv40_vertprog_destroy(nv40, vp); + memcpy(&vp->ucp, &nv40->clip, sizeof(vp->ucp)); + } + } else { + vp = nv40->swtnl.vertprog; + constbuf = NULL; + } + + /* Translate TGSI shader into hw bytecode */ + if (vp->translated) + goto check_gpu_resources; + + nv40->fallback_swtnl &= ~NV40_NEW_VERTPROG; + nv40_vertprog_translate(nv40, vp); + if (!vp->translated) { + nv40->fallback_swtnl |= NV40_NEW_VERTPROG; + return FALSE; + } + +check_gpu_resources: + /* Allocate hw vtxprog exec slots */ + if (!vp->exec) { + struct nouveau_resource *heap = nv40->screen->vp_exec_heap; + struct nouveau_stateobj *so; + uint vplen = vp->nr_insns; + + if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) { + while (heap->next && heap->size < vplen) { + struct nv40_vertex_program *evict; + + evict = heap->next->priv; + nvws->res_free(&evict->exec); + } + + if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) + assert(0); + } + + so = so_new(7, 0); + so_method(so, curie, NV40TCL_VP_START_FROM_ID, 1); + so_data (so, vp->exec->start); + so_method(so, curie, NV40TCL_VP_ATTRIB_EN, 2); + so_data (so, vp->ir); + so_data (so, vp->or); + so_method(so, curie, NV40TCL_CLIP_PLANE_ENABLE, 1); + so_data (so, vp->clip_ctrl); + so_ref(so, &vp->so); + + upload_code = TRUE; + } + + /* Allocate hw vtxprog const slots */ + if (vp->nr_consts && !vp->data) { + struct nouveau_resource *heap = nv40->screen->vp_data_heap; + + if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) { + while (heap->next && heap->size < vp->nr_consts) { + struct nv40_vertex_program *evict; + + evict = heap->next->priv; + nvws->res_free(&evict->data); + } + + if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) + assert(0); + } + + /*XXX: handle this some day */ + assert(vp->data->start >= vp->data_start_min); + + upload_data = TRUE; + if (vp->data_start != vp->data->start) + upload_code = TRUE; + } + + /* If exec or data segments moved we need to patch the program to + * fixup offsets and register IDs. + */ + if (vp->exec_start != vp->exec->start) { + for (i = 0; i < vp->nr_insns; i++) { + struct nv40_vertex_program_exec *vpi = &vp->insns[i]; + + if (vpi->has_branch_offset) { + assert(0); + } + } + + vp->exec_start = vp->exec->start; + } + + if (vp->nr_consts && vp->data_start != vp->data->start) { + for (i = 0; i < vp->nr_insns; i++) { + struct nv40_vertex_program_exec *vpi = &vp->insns[i]; + + if (vpi->const_index >= 0) { + vpi->data[1] &= ~NV40_VP_INST_CONST_SRC_MASK; + vpi->data[1] |= + (vpi->const_index + vp->data->start) << + NV40_VP_INST_CONST_SRC_SHIFT; + + } + } + + vp->data_start = vp->data->start; + } + + /* Update + Upload constant values */ + if (vp->nr_consts) { + float *map = NULL; + + if (constbuf) { + map = ws->buffer_map(ws, constbuf, + PIPE_BUFFER_USAGE_CPU_READ); + } + + for (i = 0; i < vp->nr_consts; i++) { + struct nv40_vertex_program_data *vpd = &vp->consts[i]; + + if (vpd->index >= 0) { + if (!upload_data && + !memcmp(vpd->value, &map[vpd->index * 4], + 4 * sizeof(float))) + continue; + memcpy(vpd->value, &map[vpd->index * 4], + 4 * sizeof(float)); + } + + BEGIN_RING(curie, NV40TCL_VP_UPLOAD_CONST_ID, 5); + OUT_RING (i + vp->data->start); + OUT_RINGp ((uint32_t *)vpd->value, 4); + } + + if (constbuf) + ws->buffer_unmap(ws, constbuf); + } + + /* Upload vtxprog */ + if (upload_code) { +#if 0 + for (i = 0; i < vp->nr_insns; i++) { + NOUVEAU_MSG("VP %d: 0x%08x\n", i, vp->insns[i].data[0]); + NOUVEAU_MSG("VP %d: 0x%08x\n", i, vp->insns[i].data[1]); + NOUVEAU_MSG("VP %d: 0x%08x\n", i, vp->insns[i].data[2]); + NOUVEAU_MSG("VP %d: 0x%08x\n", i, vp->insns[i].data[3]); + } +#endif + BEGIN_RING(curie, NV40TCL_VP_UPLOAD_FROM_ID, 1); + OUT_RING (vp->exec->start); + for (i = 0; i < vp->nr_insns; i++) { + BEGIN_RING(curie, NV40TCL_VP_UPLOAD_INST(0), 4); + OUT_RINGp (vp->insns[i].data, 4); + } + } + + if (vp->so != nv40->state.hw[NV40_STATE_VERTPROG]) { + so_ref(vp->so, &nv40->state.hw[NV40_STATE_VERTPROG]); + return TRUE; + } + + return FALSE; +} + +void +nv40_vertprog_destroy(struct nv40_context *nv40, struct nv40_vertex_program *vp) +{ + struct nouveau_winsys *nvws = nv40->screen->nvws; + + vp->translated = FALSE; + + if (vp->nr_insns) { + FREE(vp->insns); + vp->insns = NULL; + vp->nr_insns = 0; + } + + if (vp->nr_consts) { + FREE(vp->consts); + vp->consts = NULL; + vp->nr_consts = 0; + } + + nvws->res_free(&vp->exec); + vp->exec_start = 0; + nvws->res_free(&vp->data); + vp->data_start = 0; + vp->data_start_min = 0; + + vp->ir = vp->or = vp->clip_ctrl = 0; + so_ref(NULL, &vp->so); +} + +struct nv40_state_entry nv40_state_vertprog = { + .validate = nv40_vertprog_validate, + .dirty = { + .pipe = NV40_NEW_VERTPROG | NV40_NEW_UCP, + .hw = NV40_STATE_VERTPROG, + } +}; + diff --git a/src/gallium/drivers/nv50/Makefile b/src/gallium/drivers/nv50/Makefile new file mode 100644 index 0000000000..be30400c03 --- /dev/null +++ b/src/gallium/drivers/nv50/Makefile @@ -0,0 +1,29 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = nv50 + +DRIVER_SOURCES = \ + nv50_clear.c \ + nv50_context.c \ + nv50_draw.c \ + nv50_miptree.c \ + nv50_query.c \ + nv50_program.c \ + nv50_screen.c \ + nv50_state.c \ + nv50_state_validate.c \ + nv50_surface.c \ + nv50_tex.c \ + nv50_vbo.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/drivers/nv50/nv50_clear.c b/src/gallium/drivers/nv50/nv50_clear.c new file mode 100644 index 0000000000..a31a42d6b5 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_clear.c @@ -0,0 +1,90 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "nv50_context.h" + +void +nv50_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + struct nv50_context *nv50 = nv50_context(pipe); + struct pipe_framebuffer_state fb, s_fb = nv50->framebuffer; + struct pipe_scissor_state sc, s_sc = nv50->scissor; + unsigned dirty = nv50->dirty; + + nv50->dirty = 0; + + if (ps->format == PIPE_FORMAT_Z24S8_UNORM || + ps->format == PIPE_FORMAT_Z16_UNORM) { + fb.num_cbufs = 0; + fb.zsbuf = ps; + } else { + fb.num_cbufs = 1; + fb.cbufs[0] = ps; + fb.zsbuf = NULL; + } + fb.width = ps->width; + fb.height = ps->height; + pipe->set_framebuffer_state(pipe, &fb); + + sc.minx = sc.miny = 0; + sc.maxx = fb.width; + sc.maxy = fb.height; + pipe->set_scissor_state(pipe, &sc); + + nv50_state_validate(nv50); + + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + BEGIN_RING(tesla, 0x0d80, 4); + OUT_RINGf (ubyte_to_float((clearValue >> 16) & 0xff)); + OUT_RINGf (ubyte_to_float((clearValue >> 8) & 0xff)); + OUT_RINGf (ubyte_to_float((clearValue >> 0) & 0xff)); + OUT_RINGf (ubyte_to_float((clearValue >> 24) & 0xff)); + BEGIN_RING(tesla, 0x19d0, 1); + OUT_RING (0x3c); + break; + case PIPE_FORMAT_Z24S8_UNORM: + BEGIN_RING(tesla, 0x0d90, 1); + OUT_RINGf ((float)(clearValue >> 8) * (1.0 / 16777215.0)); + BEGIN_RING(tesla, 0x0da0, 1); + OUT_RING (clearValue & 0xff); + BEGIN_RING(tesla, 0x19d0, 1); + OUT_RING (0x03); + break; + default: + pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, + clearValue); + break; + } + + pipe->set_framebuffer_state(pipe, &s_fb); + pipe->set_scissor_state(pipe, &s_sc); + nv50->dirty |= dirty; + + ps->status = PIPE_SURFACE_STATUS_CLEAR; +} + diff --git a/src/gallium/drivers/nv50/nv50_context.c b/src/gallium/drivers/nv50/nv50_context.c new file mode 100644 index 0000000000..b02c53f209 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_context.c @@ -0,0 +1,90 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" + +#include "nv50_context.h" +#include "nv50_screen.h" + +static void +nv50_flush(struct pipe_context *pipe, unsigned flags, + struct pipe_fence_handle **fence) +{ + struct nv50_context *nv50 = (struct nv50_context *)pipe; + + FIRE_RING(fence); +} + +static void +nv50_destroy(struct pipe_context *pipe) +{ + struct nv50_context *nv50 = (struct nv50_context *)pipe; + + draw_destroy(nv50->draw); + FREE(nv50); +} + + +static void +nv50_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield) +{ +} + +struct pipe_context * +nv50_create(struct pipe_screen *pscreen, unsigned pctx_id) +{ + struct pipe_winsys *pipe_winsys = pscreen->winsys; + struct nv50_screen *screen = nv50_screen(pscreen); + struct nv50_context *nv50; + + nv50 = CALLOC_STRUCT(nv50_context); + if (!nv50) + return NULL; + nv50->screen = screen; + nv50->pctx_id = pctx_id; + + nv50->pipe.winsys = pipe_winsys; + nv50->pipe.screen = pscreen; + + nv50->pipe.destroy = nv50_destroy; + + nv50->pipe.set_edgeflags = nv50_set_edgeflags; + nv50->pipe.draw_arrays = nv50_draw_arrays; + nv50->pipe.draw_elements = nv50_draw_elements; + nv50->pipe.clear = nv50_clear; + + nv50->pipe.flush = nv50_flush; + + nv50_init_surface_functions(nv50); + nv50_init_state_functions(nv50); + nv50_init_query_functions(nv50); + + nv50->draw = draw_create(); + assert(nv50->draw); + draw_set_rasterize_stage(nv50->draw, nv50_draw_render_stage(nv50)); + + return &nv50->pipe; +} + + diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h new file mode 100644 index 0000000000..5d377f2d06 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_context.h @@ -0,0 +1,185 @@ +#ifndef __NV50_CONTEXT_H__ +#define __NV50_CONTEXT_H__ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_compiler.h" + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "draw/draw_vertex.h" + +#include "nouveau/nouveau_winsys.h" +#include "nouveau/nouveau_gldefs.h" +#include "nouveau/nouveau_stateobj.h" + +#define NOUVEAU_PUSH_CONTEXT(ctx) \ + struct nv50_screen *ctx = nv50->screen +#include "nouveau/nouveau_push.h" + +#include "nv50_screen.h" +#include "nv50_program.h" + +#define NOUVEAU_ERR(fmt, args...) \ + fprintf(stderr, "%s:%d - "fmt, __func__, __LINE__, ##args); +#define NOUVEAU_MSG(fmt, args...) \ + fprintf(stderr, "nouveau: "fmt, ##args); + +/* Constant buffer assignment */ +#define NV50_CB_PMISC 0 +#define NV50_CB_PVP 1 +#define NV50_CB_PFP 2 +#define NV50_CB_PGP 3 +#define NV50_CB_TIC 4 +#define NV50_CB_TSC 5 +#define NV50_CB_PUPLOAD 6 + +#define NV50_NEW_BLEND (1 << 0) +#define NV50_NEW_ZSA (1 << 1) +#define NV50_NEW_BLEND_COLOUR (1 << 2) +#define NV50_NEW_STIPPLE (1 << 3) +#define NV50_NEW_SCISSOR (1 << 4) +#define NV50_NEW_VIEWPORT (1 << 5) +#define NV50_NEW_RASTERIZER (1 << 6) +#define NV50_NEW_FRAMEBUFFER (1 << 7) +#define NV50_NEW_VERTPROG (1 << 8) +#define NV50_NEW_VERTPROG_CB (1 << 9) +#define NV50_NEW_FRAGPROG (1 << 10) +#define NV50_NEW_FRAGPROG_CB (1 << 11) +#define NV50_NEW_ARRAYS (1 << 12) +#define NV50_NEW_SAMPLER (1 << 13) +#define NV50_NEW_TEXTURE (1 << 14) + +struct nv50_blend_stateobj { + struct pipe_blend_state pipe; + struct nouveau_stateobj *so; +}; + +struct nv50_zsa_stateobj { + struct pipe_depth_stencil_alpha_state pipe; + struct nouveau_stateobj *so; +}; + +struct nv50_rasterizer_stateobj { + struct pipe_rasterizer_state pipe; + struct nouveau_stateobj *so; +}; + +struct nv50_miptree { + struct pipe_texture base; + struct pipe_buffer *buffer; +}; + +static INLINE struct nv50_miptree * +nv50_miptree(struct pipe_texture *pt) +{ + return (struct nv50_miptree *)pt; +} + +struct nv50_surface { + struct pipe_surface base; + struct pipe_buffer *untiled; +}; + +static INLINE struct nv50_surface * +nv50_surface(struct pipe_surface *pt) +{ + return (struct nv50_surface *)pt; +} + +struct nv50_state { + unsigned dirty; + + struct nouveau_stateobj *fb; + struct nouveau_stateobj *blend; + struct nouveau_stateobj *blend_colour; + struct nouveau_stateobj *zsa; + struct nouveau_stateobj *rast; + struct nouveau_stateobj *stipple; + struct nouveau_stateobj *scissor; + unsigned scissor_enabled; + struct nouveau_stateobj *viewport; + unsigned viewport_bypass; + struct nouveau_stateobj *tsc_upload; + struct nouveau_stateobj *tic_upload; + struct nouveau_stateobj *vertprog; + struct nouveau_stateobj *fragprog; + struct nouveau_stateobj *vtxfmt; + struct nouveau_stateobj *vtxbuf; +}; + +struct nv50_context { + struct pipe_context pipe; + + struct nv50_screen *screen; + unsigned pctx_id; + + struct draw_context *draw; + + struct nv50_state state; + + unsigned dirty; + struct nv50_blend_stateobj *blend; + struct nv50_zsa_stateobj *zsa; + struct nv50_rasterizer_stateobj *rasterizer; + struct pipe_blend_color blend_colour; + struct pipe_poly_stipple stipple; + struct pipe_scissor_state scissor; + struct pipe_viewport_state viewport; + struct pipe_framebuffer_state framebuffer; + struct nv50_program *vertprog; + struct nv50_program *fragprog; + struct pipe_buffer *constbuf[PIPE_SHADER_TYPES]; + struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS]; + unsigned vtxbuf_nr; + struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS]; + unsigned vtxelt_nr; + unsigned *sampler[PIPE_MAX_SAMPLERS]; + unsigned sampler_nr; + struct nv50_miptree *miptree[PIPE_MAX_SAMPLERS]; + unsigned miptree_nr; +}; + +static INLINE struct nv50_context * +nv50_context(struct pipe_context *pipe) +{ + return (struct nv50_context *)pipe; +} + +extern void nv50_init_surface_functions(struct nv50_context *nv50); +extern void nv50_init_state_functions(struct nv50_context *nv50); +extern void nv50_init_query_functions(struct nv50_context *nv50); + +extern void nv50_screen_init_miptree_functions(struct pipe_screen *pscreen); + +/* nv50_draw.c */ +extern struct draw_stage *nv50_draw_render_stage(struct nv50_context *nv50); + +/* nv50_vbo.c */ +extern boolean nv50_draw_arrays(struct pipe_context *, unsigned mode, + unsigned start, unsigned count); +extern boolean nv50_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, + unsigned count); +extern void nv50_vbo_validate(struct nv50_context *nv50); + +/* nv50_clear.c */ +extern void nv50_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue); + +/* nv50_program.c */ +extern void nv50_vertprog_validate(struct nv50_context *nv50); +extern void nv50_fragprog_validate(struct nv50_context *nv50); +extern void nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p); + +/* nv50_state_validate.c */ +extern boolean nv50_state_validate(struct nv50_context *nv50); + +/* nv50_tex.c */ +extern void nv50_tex_validate(struct nv50_context *); + +#endif diff --git a/src/gallium/drivers/nv50/nv50_draw.c b/src/gallium/drivers/nv50/nv50_draw.c new file mode 100644 index 0000000000..2f6f607261 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_draw.c @@ -0,0 +1,89 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "draw/draw_pipe.h" + +#include "nv50_context.h" + +struct nv50_render_stage { + struct draw_stage stage; + struct nv50_context *nv50; +}; + +static INLINE struct nv50_render_stage * +nv50_render_stage(struct draw_stage *stage) +{ + return (struct nv50_render_stage *)stage; +} + +static void +nv50_render_point(struct draw_stage *stage, struct prim_header *prim) +{ + NOUVEAU_ERR("\n"); +} + +static void +nv50_render_line(struct draw_stage *stage, struct prim_header *prim) +{ + NOUVEAU_ERR("\n"); +} + +static void +nv50_render_tri(struct draw_stage *stage, struct prim_header *prim) +{ + NOUVEAU_ERR("\n"); +} + +static void +nv50_render_flush(struct draw_stage *stage, unsigned flags) +{ +} + +static void +nv50_render_reset_stipple_counter(struct draw_stage *stage) +{ + NOUVEAU_ERR("\n"); +} + +static void +nv50_render_destroy(struct draw_stage *stage) +{ + FREE(stage); +} + +struct draw_stage * +nv50_draw_render_stage(struct nv50_context *nv50) +{ + struct nv50_render_stage *rs = CALLOC_STRUCT(nv50_render_stage); + + rs->nv50 = nv50; + rs->stage.draw = nv50->draw; + rs->stage.destroy = nv50_render_destroy; + rs->stage.point = nv50_render_point; + rs->stage.line = nv50_render_line; + rs->stage.tri = nv50_render_tri; + rs->stage.flush = nv50_render_flush; + rs->stage.reset_stipple_counter = nv50_render_reset_stipple_counter; + + return &rs->stage; +} + diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c new file mode 100644 index 0000000000..28a8bdc0fa --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_miptree.c @@ -0,0 +1,135 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "nv50_context.h" + +static struct pipe_texture * +nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) +{ + struct pipe_winsys *ws = pscreen->winsys; + struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree); + unsigned usage, pitch; + + mt->base = *pt; + mt->base.refcount = 1; + mt->base.screen = pscreen; + + usage = PIPE_BUFFER_USAGE_PIXEL; + switch (pt->format) { + case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z16_UNORM: + usage |= NOUVEAU_BUFFER_USAGE_ZETA; + break; + default: + break; + } + + pitch = ((pt->width[0] + 63) & ~63) * pt->block.size; + /*XXX*/ + pitch *= 2; + + mt->buffer = ws->buffer_create(ws, 256, usage, pitch * pt->height[0]); + if (!mt->buffer) { + FREE(mt); + return NULL; + } + + return &mt->base; +} + +static void +nv50_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt) +{ + struct pipe_texture *pt = *ppt; + + *ppt = NULL; + + if (--pt->refcount <= 0) { + struct nv50_miptree *mt = nv50_miptree(pt); + + pipe_buffer_reference(pscreen, &mt->buffer, NULL); + FREE(mt); + } +} + +static struct pipe_surface * +nv50_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + unsigned flags) +{ + struct nv50_miptree *mt = nv50_miptree(pt); + struct nv50_surface *s; + struct pipe_surface *ps; + + s = CALLOC_STRUCT(nv50_surface); + if (!s) + return NULL; + ps = &s->base; + + ps->refcount = 1; + ps->winsys = pscreen->winsys; + ps->format = pt->format; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->block = pt->block; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = ps->width * ps->block.size; + ps->offset = 0; + ps->usage = flags; + ps->status = PIPE_SURFACE_STATUS_DEFINED; + + pipe_texture_reference(&ps->texture, pt); + pipe_buffer_reference(pscreen, &ps->buffer, mt->buffer); + + return ps; +} + +static void +nv50_miptree_surface_del(struct pipe_screen *pscreen, + struct pipe_surface **psurface) +{ + struct pipe_surface *ps = *psurface; + struct nv50_surface *s = nv50_surface(ps); + + *psurface = NULL; + + if (--ps->refcount <= 0) { + pipe_texture_reference(&ps->texture, NULL); + pipe_buffer_reference(pscreen, &ps->buffer, NULL); + FREE(s); + } +} + +void +nv50_screen_init_miptree_functions(struct pipe_screen *pscreen) +{ + pscreen->texture_create = nv50_miptree_create; + pscreen->texture_release = nv50_miptree_release; + pscreen->get_tex_surface = nv50_miptree_surface_new; + pscreen->tex_surface_release = nv50_miptree_surface_del; +} + diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c new file mode 100644 index 0000000000..d6fbdd1824 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -0,0 +1,1708 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" + +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" + +#include "nv50_context.h" + +#define NV50_SU_MAX_TEMP 64 +#define NV50_PROGRAM_DUMP + +/* ARL - gallium craps itself on progs/vp/arl.txt + * + * MSB - Like MAD, but MUL+SUB + * - Fuck it off, introduce a way to negate args for ops that + * support it. + * + * Look into inlining IMMD for ops other than MOV (make it general?) + * - Maybe even relax restrictions a bit, can't do P_RESULT + P_IMMD, + * but can emit to P_TEMP first - then MOV later. NVIDIA does this + * + * In ops such as ADD it's possible to construct a bad opcode in the !is_long() + * case, if the emit_src() causes the inst to suddenly become long. + * + * Verify half-insns work where expected - and force disable them where they + * don't work - MUL has it forcibly disabled atm as it fixes POW.. + * + * FUCK! watch dst==src vectors, can overwrite components that are needed. + * ie. SUB R0, R0.yzxw, R0 + * + * Things to check with renouveau: + * FP attr/result assignment - how? + * attrib + * - 0x16bc maps vp output onto fp hpos + * - 0x16c0 maps vp output onto fp col0 + * result + * - colr always 0-3 + * - depr always 4 + * 0x16bc->0x16e8 --> some binding between vp/fp regs + * 0x16b8 --> VP output count + * + * 0x1298 --> "MOV rcol.x, fcol.y" "MOV depr, fcol.y" = 0x00000005 + * "MOV rcol.x, fcol.y" = 0x00000004 + * 0x19a8 --> as above but 0x00000100 and 0x00000000 + * - 0x00100000 used when KIL used + * 0x196c --> as above but 0x00000011 and 0x00000000 + * + * 0x1988 --> 0xXXNNNNNN + * - XX == FP high something + */ +struct nv50_reg { + enum { + P_TEMP, + P_ATTR, + P_RESULT, + P_CONST, + P_IMMD + } type; + int index; + + int hw; + int neg; +}; + +struct nv50_pc { + struct nv50_program *p; + + /* hw resources */ + struct nv50_reg *r_temp[NV50_SU_MAX_TEMP]; + + /* tgsi resources */ + struct nv50_reg *temp; + int temp_nr; + struct nv50_reg *attr; + int attr_nr; + struct nv50_reg *result; + int result_nr; + struct nv50_reg *param; + int param_nr; + struct nv50_reg *immd; + float *immd_buf; + int immd_nr; + + struct nv50_reg *temp_temp[16]; + unsigned temp_temp_nr; +}; + +static void +alloc_reg(struct nv50_pc *pc, struct nv50_reg *reg) +{ + int i; + + if (reg->type == P_RESULT) { + if (pc->p->cfg.high_result < (reg->hw + 1)) + pc->p->cfg.high_result = reg->hw + 1; + } + + if (reg->type != P_TEMP) + return; + + if (reg->hw >= 0) { + /*XXX: do this here too to catch FP temp-as-attr usage.. + * not clean, but works */ + if (pc->p->cfg.high_temp < (reg->hw + 1)) + pc->p->cfg.high_temp = reg->hw + 1; + return; + } + + for (i = 0; i < NV50_SU_MAX_TEMP; i++) { + if (!(pc->r_temp[i])) { + pc->r_temp[i] = reg; + reg->hw = i; + if (pc->p->cfg.high_temp < (i + 1)) + pc->p->cfg.high_temp = i + 1; + return; + } + } + + assert(0); +} + +static struct nv50_reg * +alloc_temp(struct nv50_pc *pc, struct nv50_reg *dst) +{ + struct nv50_reg *r; + int i; + + if (dst && dst->type == P_TEMP && dst->hw == -1) + return dst; + + for (i = 0; i < NV50_SU_MAX_TEMP; i++) { + if (!pc->r_temp[i]) { + r = CALLOC_STRUCT(nv50_reg); + r->type = P_TEMP; + r->index = -1; + r->hw = i; + pc->r_temp[i] = r; + return r; + } + } + + assert(0); + return NULL; +} + +static void +free_temp(struct nv50_pc *pc, struct nv50_reg *r) +{ + if (r->index == -1) { + unsigned hw = r->hw; + + FREE(pc->r_temp[hw]); + pc->r_temp[hw] = NULL; + } +} + +static struct nv50_reg * +temp_temp(struct nv50_pc *pc) +{ + if (pc->temp_temp_nr >= 16) + assert(0); + + pc->temp_temp[pc->temp_temp_nr] = alloc_temp(pc, NULL); + return pc->temp_temp[pc->temp_temp_nr++]; +} + +static void +kill_temp_temp(struct nv50_pc *pc) +{ + int i; + + for (i = 0; i < pc->temp_temp_nr; i++) + free_temp(pc, pc->temp_temp[i]); + pc->temp_temp_nr = 0; +} + +static int +ctor_immd(struct nv50_pc *pc, float x, float y, float z, float w) +{ + pc->immd_buf = REALLOC(pc->immd_buf, (pc->immd_nr * r * sizeof(float)), + (pc->immd_nr + 1) * 4 * sizeof(float)); + pc->immd_buf[(pc->immd_nr * 4) + 0] = x; + pc->immd_buf[(pc->immd_nr * 4) + 1] = y; + pc->immd_buf[(pc->immd_nr * 4) + 2] = z; + pc->immd_buf[(pc->immd_nr * 4) + 3] = w; + + return pc->immd_nr++; +} + +static struct nv50_reg * +alloc_immd(struct nv50_pc *pc, float f) +{ + struct nv50_reg *r = CALLOC_STRUCT(nv50_reg); + unsigned hw; + + hw = ctor_immd(pc, f, 0, 0, 0) * 4; + r->type = P_IMMD; + r->hw = hw; + r->index = -1; + return r; +} + +static struct nv50_program_exec * +exec(struct nv50_pc *pc) +{ + struct nv50_program_exec *e = CALLOC_STRUCT(nv50_program_exec); + + e->param.index = -1; + return e; +} + +static void +emit(struct nv50_pc *pc, struct nv50_program_exec *e) +{ + struct nv50_program *p = pc->p; + + if (p->exec_tail) + p->exec_tail->next = e; + if (!p->exec_head) + p->exec_head = e; + p->exec_tail = e; + p->exec_size += (e->inst[0] & 1) ? 2 : 1; +} + +static INLINE void set_long(struct nv50_pc *, struct nv50_program_exec *); + +static boolean +is_long(struct nv50_program_exec *e) +{ + if (e->inst[0] & 1) + return TRUE; + return FALSE; +} + +static boolean +is_immd(struct nv50_program_exec *e) +{ + if (is_long(e) && (e->inst[1] & 3) == 3) + return TRUE; + return FALSE; +} + +static INLINE void +set_pred(struct nv50_pc *pc, unsigned pred, unsigned idx, + struct nv50_program_exec *e) +{ + set_long(pc, e); + e->inst[1] &= ~((0x1f << 7) | (0x3 << 12)); + e->inst[1] |= (pred << 7) | (idx << 12); +} + +static INLINE void +set_pred_wr(struct nv50_pc *pc, unsigned on, unsigned idx, + struct nv50_program_exec *e) +{ + set_long(pc, e); + e->inst[1] &= ~((0x3 << 4) | (1 << 6)); + e->inst[1] |= (idx << 4) | (on << 6); +} + +static INLINE void +set_long(struct nv50_pc *pc, struct nv50_program_exec *e) +{ + if (is_long(e)) + return; + + e->inst[0] |= 1; + set_pred(pc, 0xf, 0, e); + set_pred_wr(pc, 0, 0, e); +} + +static INLINE void +set_dst(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_program_exec *e) +{ + if (dst->type == P_RESULT) { + set_long(pc, e); + e->inst[1] |= 0x00000008; + } + + alloc_reg(pc, dst); + e->inst[0] |= (dst->hw << 2); +} + +static INLINE void +set_immd(struct nv50_pc *pc, struct nv50_reg *imm, struct nv50_program_exec *e) +{ + unsigned val = fui(pc->immd_buf[imm->hw]); /* XXX */ + + set_long(pc, e); + /*XXX: can't be predicated - bits overlap.. catch cases where both + * are required and avoid them. */ + set_pred(pc, 0, 0, e); + set_pred_wr(pc, 0, 0, e); + + e->inst[1] |= 0x00000002 | 0x00000001; + e->inst[0] |= (val & 0x3f) << 16; + e->inst[1] |= (val >> 6) << 2; +} + +static void +emit_interp(struct nv50_pc *pc, struct nv50_reg *dst, + struct nv50_reg *src, struct nv50_reg *iv) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] |= 0x80000000; + set_dst(pc, dst, e); + alloc_reg(pc, src); + e->inst[0] |= (src->hw << 16); + if (iv) { + e->inst[0] |= (1 << 25); + alloc_reg(pc, iv); + e->inst[0] |= (iv->hw << 9); + } + + emit(pc, e); +} + +static void +set_data(struct nv50_pc *pc, struct nv50_reg *src, unsigned m, unsigned s, + struct nv50_program_exec *e) +{ + set_long(pc, e); +#if 1 + e->inst[1] |= (1 << 22); +#else + if (src->type == P_IMMD) { + e->inst[1] |= (NV50_CB_PMISC << 22); + } else { + if (pc->p->type == PIPE_SHADER_VERTEX) + e->inst[1] |= (NV50_CB_PVP << 22); + else + e->inst[1] |= (NV50_CB_PFP << 22); + } +#endif + + e->param.index = src->hw; + e->param.shift = s; + e->param.mask = m << (s % 32); +} + +static void +emit_mov(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] |= 0x10000000; + + set_dst(pc, dst, e); + + if (0 && dst->type != P_RESULT && src->type == P_IMMD) { + set_immd(pc, src, e); + /*XXX: 32-bit, but steals part of "half" reg space - need to + * catch and handle this case if/when we do half-regs + */ + e->inst[0] |= 0x00008000; + } else + if (src->type == P_IMMD || src->type == P_CONST) { + set_long(pc, e); + set_data(pc, src, 0x7f, 9, e); + e->inst[1] |= 0x20000000; /* src0 const? */ + } else { + if (src->type == P_ATTR) { + set_long(pc, e); + e->inst[1] |= 0x00200000; + } + + alloc_reg(pc, src); + e->inst[0] |= (src->hw << 9); + } + + /* We really should support "half" instructions here at some point, + * but I don't feel confident enough about them yet. + */ + set_long(pc, e); + if (is_long(e) && !is_immd(e)) { + e->inst[1] |= 0x04000000; /* 32-bit */ + e->inst[1] |= 0x0003c000; /* "subsubop" 0xf == mov */ + } + + emit(pc, e); +} + +static boolean +check_swap_src_0_1(struct nv50_pc *pc, + struct nv50_reg **s0, struct nv50_reg **s1) +{ + struct nv50_reg *src0 = *s0, *src1 = *s1; + + if (src0->type == P_CONST) { + if (src1->type != P_CONST) { + *s0 = src1; + *s1 = src0; + return TRUE; + } + } else + if (src1->type == P_ATTR) { + if (src0->type != P_ATTR) { + *s0 = src1; + *s1 = src0; + return TRUE; + } + } + + return FALSE; +} + +static void +set_src_0(struct nv50_pc *pc, struct nv50_reg *src, struct nv50_program_exec *e) +{ + if (src->type == P_ATTR) { + set_long(pc, e); + e->inst[1] |= 0x00200000; + } else + if (src->type == P_CONST || src->type == P_IMMD) { + struct nv50_reg *temp = temp_temp(pc); + + emit_mov(pc, temp, src); + src = temp; + } + + alloc_reg(pc, src); + e->inst[0] |= (src->hw << 9); +} + +static void +set_src_1(struct nv50_pc *pc, struct nv50_reg *src, struct nv50_program_exec *e) +{ + if (src->type == P_ATTR) { + struct nv50_reg *temp = temp_temp(pc); + + emit_mov(pc, temp, src); + src = temp; + } else + if (src->type == P_CONST || src->type == P_IMMD) { + assert(!(e->inst[0] & 0x00800000)); + if (e->inst[0] & 0x01000000) { + struct nv50_reg *temp = temp_temp(pc); + + emit_mov(pc, temp, src); + src = temp; + } else { + set_data(pc, src, 0x7f, 16, e); + e->inst[0] |= 0x00800000; + } + } + + alloc_reg(pc, src); + e->inst[0] |= (src->hw << 16); +} + +static void +set_src_2(struct nv50_pc *pc, struct nv50_reg *src, struct nv50_program_exec *e) +{ + set_long(pc, e); + + if (src->type == P_ATTR) { + struct nv50_reg *temp = temp_temp(pc); + + emit_mov(pc, temp, src); + src = temp; + } else + if (src->type == P_CONST || src->type == P_IMMD) { + assert(!(e->inst[0] & 0x01000000)); + if (e->inst[0] & 0x00800000) { + struct nv50_reg *temp = temp_temp(pc); + + emit_mov(pc, temp, src); + src = temp; + } else { + set_data(pc, src, 0x7f, 32+14, e); + e->inst[0] |= 0x01000000; + } + } + + alloc_reg(pc, src); + e->inst[1] |= (src->hw << 14); +} + +static void +emit_mul(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, + struct nv50_reg *src1) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] |= 0xc0000000; + set_long(pc, e); + + check_swap_src_0_1(pc, &src0, &src1); + set_dst(pc, dst, e); + set_src_0(pc, src0, e); + set_src_1(pc, src1, e); + + emit(pc, e); +} + +static void +emit_add(struct nv50_pc *pc, struct nv50_reg *dst, + struct nv50_reg *src0, struct nv50_reg *src1) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] |= 0xb0000000; + + check_swap_src_0_1(pc, &src0, &src1); + set_dst(pc, dst, e); + set_src_0(pc, src0, e); + if (is_long(e)) + set_src_2(pc, src1, e); + else + set_src_1(pc, src1, e); + + emit(pc, e); +} + +static void +emit_minmax(struct nv50_pc *pc, unsigned sub, struct nv50_reg *dst, + struct nv50_reg *src0, struct nv50_reg *src1) +{ + struct nv50_program_exec *e = exec(pc); + + set_long(pc, e); + e->inst[0] |= 0xb0000000; + e->inst[1] |= (sub << 29); + + check_swap_src_0_1(pc, &src0, &src1); + set_dst(pc, dst, e); + set_src_0(pc, src0, e); + set_src_1(pc, src1, e); + + emit(pc, e); +} + +static void +emit_sub(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, + struct nv50_reg *src1) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] |= 0xb0000000; + + set_long(pc, e); + if (check_swap_src_0_1(pc, &src0, &src1)) + e->inst[1] |= 0x04000000; + else + e->inst[1] |= 0x08000000; + + set_dst(pc, dst, e); + set_src_0(pc, src0, e); + set_src_2(pc, src1, e); + + emit(pc, e); +} + +static void +emit_mad(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, + struct nv50_reg *src1, struct nv50_reg *src2) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] |= 0xe0000000; + + check_swap_src_0_1(pc, &src0, &src1); + set_dst(pc, dst, e); + set_src_0(pc, src0, e); + set_src_1(pc, src1, e); + set_src_2(pc, src2, e); + + emit(pc, e); +} + +static void +emit_msb(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, + struct nv50_reg *src1, struct nv50_reg *src2) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] |= 0xe0000000; + set_long(pc, e); + e->inst[1] |= 0x08000000; /* src0 * src1 - src2 */ + + check_swap_src_0_1(pc, &src0, &src1); + set_dst(pc, dst, e); + set_src_0(pc, src0, e); + set_src_1(pc, src1, e); + set_src_2(pc, src2, e); + + emit(pc, e); +} + +static void +emit_flop(struct nv50_pc *pc, unsigned sub, + struct nv50_reg *dst, struct nv50_reg *src) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] |= 0x90000000; + if (sub) { + set_long(pc, e); + e->inst[1] |= (sub << 29); + } + + set_dst(pc, dst, e); + set_src_0(pc, src, e); + + emit(pc, e); +} + +static void +emit_preex2(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] |= 0xb0000000; + + set_dst(pc, dst, e); + set_src_0(pc, src, e); + set_long(pc, e); + e->inst[1] |= (6 << 29) | 0x00004000; + + emit(pc, e); +} + +static void +emit_precossin(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] |= 0xb0000000; + + set_dst(pc, dst, e); + set_src_0(pc, src, e); + set_long(pc, e); + e->inst[1] |= (6 << 29); + + emit(pc, e); +} + +static void +emit_set(struct nv50_pc *pc, unsigned c_op, struct nv50_reg *dst, + struct nv50_reg *src0, struct nv50_reg *src1) +{ + struct nv50_program_exec *e = exec(pc); + unsigned inv_cop[8] = { 0, 4, 2, 6, 1, 5, 3, 7 }; + struct nv50_reg *rdst; + + assert(c_op <= 7); + if (check_swap_src_0_1(pc, &src0, &src1)) + c_op = inv_cop[c_op]; + + rdst = dst; + if (dst->type != P_TEMP) + dst = alloc_temp(pc, NULL); + + /* set.u32 */ + set_long(pc, e); + e->inst[0] |= 0xb0000000; + e->inst[1] |= (3 << 29); + e->inst[1] |= (c_op << 14); + /*XXX: breaks things, .u32 by default? + * decuda will disasm as .u16 and use .lo/.hi regs, but this + * doesn't seem to match what the hw actually does. + inst[1] |= 0x04000000; << breaks things.. .u32 by default? + */ + set_dst(pc, dst, e); + set_src_0(pc, src0, e); + set_src_1(pc, src1, e); + emit(pc, e); + + /* cvt.f32.u32 */ + e = exec(pc); + e->inst[0] = 0xa0000001; + e->inst[1] = 0x64014780; + set_dst(pc, rdst, e); + set_src_0(pc, dst, e); + emit(pc, e); + + if (dst != rdst) + free_temp(pc, dst); +} + +static void +emit_flr(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] = 0xa0000000; /* cvt */ + set_long(pc, e); + e->inst[1] |= (6 << 29); /* cvt */ + e->inst[1] |= 0x08000000; /* integer mode */ + e->inst[1] |= 0x04000000; /* 32 bit */ + e->inst[1] |= ((0x1 << 3)) << 14; /* .rn */ + e->inst[1] |= (1 << 14); /* src .f32 */ + set_dst(pc, dst, e); + set_src_0(pc, src, e); + + emit(pc, e); +} + +static void +emit_pow(struct nv50_pc *pc, struct nv50_reg *dst, + struct nv50_reg *v, struct nv50_reg *e) +{ + struct nv50_reg *temp = alloc_temp(pc, NULL); + + emit_flop(pc, 3, temp, v); + emit_mul(pc, temp, temp, e); + emit_preex2(pc, temp, temp); + emit_flop(pc, 6, dst, temp); + + free_temp(pc, temp); +} + +static void +emit_abs(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] = 0xa0000000; /* cvt */ + set_long(pc, e); + e->inst[1] |= (6 << 29); /* cvt */ + e->inst[1] |= 0x04000000; /* 32 bit */ + e->inst[1] |= (1 << 14); /* src .f32 */ + e->inst[1] |= ((1 << 6) << 14); /* .abs */ + set_dst(pc, dst, e); + set_src_0(pc, src, e); + + emit(pc, e); +} + +static void +emit_lit(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask, + struct nv50_reg **src) +{ + struct nv50_reg *one = alloc_immd(pc, 1.0); + struct nv50_reg *zero = alloc_immd(pc, 0.0); + struct nv50_reg *neg128 = alloc_immd(pc, -127.999999); + struct nv50_reg *pos128 = alloc_immd(pc, 127.999999); + struct nv50_reg *tmp[4]; + + if (mask & (1 << 0)) + emit_mov(pc, dst[0], one); + + if (mask & (1 << 3)) + emit_mov(pc, dst[3], one); + + if (mask & (3 << 1)) { + if (mask & (1 << 1)) + tmp[0] = dst[1]; + else + tmp[0] = temp_temp(pc); + emit_minmax(pc, 4, tmp[0], src[0], zero); + } + + if (mask & (1 << 2)) { + set_pred_wr(pc, 1, 0, pc->p->exec_tail); + + tmp[1] = temp_temp(pc); + emit_minmax(pc, 4, tmp[1], src[1], zero); + + tmp[3] = temp_temp(pc); + emit_minmax(pc, 4, tmp[3], src[3], neg128); + emit_minmax(pc, 5, tmp[3], tmp[3], pos128); + + emit_pow(pc, dst[2], tmp[1], tmp[3]); + emit_mov(pc, dst[2], zero); + set_pred(pc, 3, 0, pc->p->exec_tail); + } +} + +static void +emit_neg(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) +{ + struct nv50_program_exec *e = exec(pc); + + set_long(pc, e); + e->inst[0] |= 0xa0000000; /* delta */ + e->inst[1] |= (7 << 29); /* delta */ + e->inst[1] |= 0x04000000; /* negate arg0? probably not */ + e->inst[1] |= (1 << 14); /* src .f32 */ + set_dst(pc, dst, e); + set_src_0(pc, src, e); + + emit(pc, e); +} + +static struct nv50_reg * +tgsi_dst(struct nv50_pc *pc, int c, const struct tgsi_full_dst_register *dst) +{ + switch (dst->DstRegister.File) { + case TGSI_FILE_TEMPORARY: + return &pc->temp[dst->DstRegister.Index * 4 + c]; + case TGSI_FILE_OUTPUT: + return &pc->result[dst->DstRegister.Index * 4 + c]; + case TGSI_FILE_NULL: + return NULL; + default: + break; + } + + return NULL; +} + +static struct nv50_reg * +tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src) +{ + struct nv50_reg *r = NULL; + struct nv50_reg *temp; + unsigned c; + + c = tgsi_util_get_full_src_register_extswizzle(src, chan); + switch (c) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch (src->SrcRegister.File) { + case TGSI_FILE_INPUT: + r = &pc->attr[src->SrcRegister.Index * 4 + c]; + break; + case TGSI_FILE_TEMPORARY: + r = &pc->temp[src->SrcRegister.Index * 4 + c]; + break; + case TGSI_FILE_CONSTANT: + r = &pc->param[src->SrcRegister.Index * 4 + c]; + break; + case TGSI_FILE_IMMEDIATE: + r = &pc->immd[src->SrcRegister.Index * 4 + c]; + break; + case TGSI_FILE_SAMPLER: + break; + default: + assert(0); + break; + } + break; + case TGSI_EXTSWIZZLE_ZERO: + r = alloc_immd(pc, 0.0); + break; + case TGSI_EXTSWIZZLE_ONE: + r = alloc_immd(pc, 1.0); + break; + default: + assert(0); + break; + } + + switch (tgsi_util_get_full_src_register_sign_mode(src, chan)) { + case TGSI_UTIL_SIGN_KEEP: + break; + case TGSI_UTIL_SIGN_CLEAR: + temp = temp_temp(pc); + emit_abs(pc, temp, r); + r = temp; + break; + case TGSI_UTIL_SIGN_TOGGLE: + temp = temp_temp(pc); + emit_neg(pc, temp, r); + r = temp; + break; + case TGSI_UTIL_SIGN_SET: + temp = temp_temp(pc); + emit_abs(pc, temp, r); + emit_neg(pc, temp, r); + r = temp; + break; + default: + assert(0); + break; + } + + return r; +} + +static boolean +nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) +{ + const struct tgsi_full_instruction *inst = &tok->FullInstruction; + struct nv50_reg *rdst[4], *dst[4], *src[3][4], *temp; + unsigned mask, sat; + int i, c; + + mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + sat = inst->Instruction.Saturate == TGSI_SAT_ZERO_ONE; + + for (c = 0; c < 4; c++) { + if (mask & (1 << c)) + dst[c] = tgsi_dst(pc, c, &inst->FullDstRegisters[0]); + else + dst[c] = NULL; + } + + for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { + for (c = 0; c < 4; c++) + src[i][c] = tgsi_src(pc, c, &inst->FullSrcRegisters[i]); + } + + if (sat) { + for (c = 0; c < 4; c++) { + rdst[c] = dst[c]; + dst[c] = temp_temp(pc); + } + } + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_abs(pc, dst[c], src[0][c]); + } + break; + case TGSI_OPCODE_ADD: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_add(pc, dst[c], src[0][c], src[1][c]); + } + break; + case TGSI_OPCODE_COS: + temp = alloc_temp(pc, NULL); + emit_precossin(pc, temp, src[0][0]); + emit_flop(pc, 5, temp, temp); + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mov(pc, dst[c], temp); + } + break; + case TGSI_OPCODE_DP3: + temp = alloc_temp(pc, NULL); + emit_mul(pc, temp, src[0][0], src[1][0]); + emit_mad(pc, temp, src[0][1], src[1][1], temp); + emit_mad(pc, temp, src[0][2], src[1][2], temp); + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mov(pc, dst[c], temp); + } + free_temp(pc, temp); + break; + case TGSI_OPCODE_DP4: + temp = alloc_temp(pc, NULL); + emit_mul(pc, temp, src[0][0], src[1][0]); + emit_mad(pc, temp, src[0][1], src[1][1], temp); + emit_mad(pc, temp, src[0][2], src[1][2], temp); + emit_mad(pc, temp, src[0][3], src[1][3], temp); + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mov(pc, dst[c], temp); + } + free_temp(pc, temp); + break; + case TGSI_OPCODE_DPH: + temp = alloc_temp(pc, NULL); + emit_mul(pc, temp, src[0][0], src[1][0]); + emit_mad(pc, temp, src[0][1], src[1][1], temp); + emit_mad(pc, temp, src[0][2], src[1][2], temp); + emit_add(pc, temp, src[1][3], temp); + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mov(pc, dst[c], temp); + } + free_temp(pc, temp); + break; + case TGSI_OPCODE_DST: + { + struct nv50_reg *one = alloc_immd(pc, 1.0); + if (mask & (1 << 0)) + emit_mov(pc, dst[0], one); + if (mask & (1 << 1)) + emit_mul(pc, dst[1], src[0][1], src[1][1]); + if (mask & (1 << 2)) + emit_mov(pc, dst[2], src[0][2]); + if (mask & (1 << 3)) + emit_mov(pc, dst[3], src[1][3]); + FREE(one); + } + break; + case TGSI_OPCODE_EX2: + temp = alloc_temp(pc, NULL); + emit_preex2(pc, temp, src[0][0]); + emit_flop(pc, 6, temp, temp); + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mov(pc, dst[c], temp); + } + free_temp(pc, temp); + break; + case TGSI_OPCODE_FLR: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_flr(pc, dst[c], src[0][c]); + } + break; + case TGSI_OPCODE_FRC: + temp = alloc_temp(pc, NULL); + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_flr(pc, temp, src[0][c]); + emit_sub(pc, dst[c], src[0][c], temp); + } + free_temp(pc, temp); + break; + case TGSI_OPCODE_LIT: + emit_lit(pc, &dst[0], mask, &src[0][0]); + break; + case TGSI_OPCODE_LG2: + temp = alloc_temp(pc, NULL); + emit_flop(pc, 3, temp, src[0][0]); + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mov(pc, dst[c], temp); + } + break; + case TGSI_OPCODE_LRP: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + /*XXX: we can do better than this */ + temp = alloc_temp(pc, NULL); + emit_neg(pc, temp, src[0][c]); + emit_mad(pc, temp, temp, src[2][c], src[2][c]); + emit_mad(pc, dst[c], src[0][c], src[1][c], temp); + free_temp(pc, temp); + } + break; + case TGSI_OPCODE_MAD: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mad(pc, dst[c], src[0][c], src[1][c], src[2][c]); + } + break; + case TGSI_OPCODE_MAX: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_minmax(pc, 4, dst[c], src[0][c], src[1][c]); + } + break; + case TGSI_OPCODE_MIN: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_minmax(pc, 5, dst[c], src[0][c], src[1][c]); + } + break; + case TGSI_OPCODE_MOV: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mov(pc, dst[c], src[0][c]); + } + break; + case TGSI_OPCODE_MUL: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mul(pc, dst[c], src[0][c], src[1][c]); + } + break; + case TGSI_OPCODE_POW: + temp = alloc_temp(pc, NULL); + emit_pow(pc, temp, src[0][0], src[1][0]); + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mov(pc, dst[c], temp); + } + free_temp(pc, temp); + break; + case TGSI_OPCODE_RCP: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_flop(pc, 0, dst[c], src[0][0]); + } + break; + case TGSI_OPCODE_RSQ: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_flop(pc, 2, dst[c], src[0][0]); + } + break; + case TGSI_OPCODE_SCS: + temp = alloc_temp(pc, NULL); + emit_precossin(pc, temp, src[0][0]); + if (mask & (1 << 0)) + emit_flop(pc, 5, dst[0], temp); + if (mask & (1 << 1)) + emit_flop(pc, 4, dst[1], temp); + break; + case TGSI_OPCODE_SGE: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_set(pc, 6, dst[c], src[0][c], src[1][c]); + } + break; + case TGSI_OPCODE_SIN: + temp = alloc_temp(pc, NULL); + emit_precossin(pc, temp, src[0][0]); + emit_flop(pc, 4, temp, temp); + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_mov(pc, dst[c], temp); + } + break; + case TGSI_OPCODE_SLT: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_set(pc, 1, dst[c], src[0][c], src[1][c]); + } + break; + case TGSI_OPCODE_SUB: + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + emit_sub(pc, dst[c], src[0][c], src[1][c]); + } + break; + case TGSI_OPCODE_TEX: + { + struct nv50_reg *t0, *t1, *t2, *t3; + struct nv50_program_exec *e; + + t0 = alloc_temp(pc, NULL); + t0 = alloc_temp(pc, NULL); + t1 = alloc_temp(pc, NULL); + t2 = alloc_temp(pc, NULL); + t3 = alloc_temp(pc, NULL); + emit_mov(pc, t0, src[0][0]); + emit_mov(pc, t1, src[0][1]); + + e = exec(pc); + e->inst[0] = 0xf6400000; + set_long(pc, e); + e->inst[1] |= 0x0000c004; + set_dst(pc, t0, e); + emit(pc, e); + + if (mask & (1 << 0)) emit_mov(pc, dst[0], t0); + if (mask & (1 << 1)) emit_mov(pc, dst[1], t1); + if (mask & (1 << 2)) emit_mov(pc, dst[2], t2); + if (mask & (1 << 3)) emit_mov(pc, dst[3], t3); + + free_temp(pc, t0); + free_temp(pc, t1); + free_temp(pc, t2); + free_temp(pc, t3); + } + break; + case TGSI_OPCODE_XPD: + temp = alloc_temp(pc, NULL); + if (mask & (1 << 0)) { + emit_mul(pc, temp, src[0][2], src[1][1]); + emit_msb(pc, dst[0], src[0][1], src[1][2], temp); + } + if (mask & (1 << 1)) { + emit_mul(pc, temp, src[0][0], src[1][2]); + emit_msb(pc, dst[1], src[0][2], src[1][0], temp); + } + if (mask & (1 << 2)) { + emit_mul(pc, temp, src[0][1], src[1][0]); + emit_msb(pc, dst[2], src[0][0], src[1][1], temp); + } + free_temp(pc, temp); + break; + case TGSI_OPCODE_END: + break; + default: + NOUVEAU_ERR("invalid opcode %d\n", inst->Instruction.Opcode); + return FALSE; + } + + if (sat) { + for (c = 0; c < 4; c++) { + struct nv50_program_exec *e; + + if (!(mask & (1 << c))) + continue; + e = exec(pc); + + e->inst[0] = 0xa0000000; /* cvt */ + set_long(pc, e); + e->inst[1] |= (6 << 29); /* cvt */ + e->inst[1] |= 0x04000000; /* 32 bit */ + e->inst[1] |= (1 << 14); /* src .f32 */ + e->inst[1] |= ((1 << 5) << 14); /* .sat */ + set_dst(pc, rdst[c], e); + set_src_0(pc, dst[c], e); + emit(pc, e); + } + } + + kill_temp_temp(pc); + return TRUE; +} + +static boolean +nv50_program_tx_prep(struct nv50_pc *pc) +{ + struct tgsi_parse_context p; + boolean ret = FALSE; + unsigned i, c; + + tgsi_parse_init(&p, pc->p->pipe.tokens); + while (!tgsi_parse_end_of_tokens(&p)) { + const union tgsi_full_token *tok = &p.FullToken; + + tgsi_parse_token(&p); + switch (tok->Token.Type) { + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + const struct tgsi_full_immediate *imm = + &p.FullToken.FullImmediate; + + ctor_immd(pc, imm->u.ImmediateFloat32[0].Float, + imm->u.ImmediateFloat32[1].Float, + imm->u.ImmediateFloat32[2].Float, + imm->u.ImmediateFloat32[3].Float); + } + break; + case TGSI_TOKEN_TYPE_DECLARATION: + { + const struct tgsi_full_declaration *d; + unsigned last; + + d = &p.FullToken.FullDeclaration; + last = d->DeclarationRange.Last; + + switch (d->Declaration.File) { + case TGSI_FILE_TEMPORARY: + if (pc->temp_nr < (last + 1)) + pc->temp_nr = last + 1; + break; + case TGSI_FILE_OUTPUT: + if (pc->result_nr < (last + 1)) + pc->result_nr = last + 1; + break; + case TGSI_FILE_INPUT: + if (pc->attr_nr < (last + 1)) + pc->attr_nr = last + 1; + break; + case TGSI_FILE_CONSTANT: + if (pc->param_nr < (last + 1)) + pc->param_nr = last + 1; + break; + case TGSI_FILE_SAMPLER: + break; + default: + NOUVEAU_ERR("bad decl file %d\n", + d->Declaration.File); + goto out_err; + } + } + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + break; + default: + break; + } + } + + if (pc->temp_nr) { + pc->temp = CALLOC(pc->temp_nr * 4, sizeof(struct nv50_reg)); + if (!pc->temp) + goto out_err; + + for (i = 0; i < pc->temp_nr; i++) { + for (c = 0; c < 4; c++) { + pc->temp[i*4+c].type = P_TEMP; + pc->temp[i*4+c].hw = -1; + pc->temp[i*4+c].index = i; + } + } + } + + if (pc->attr_nr) { + struct nv50_reg *iv = NULL; + int aid = 0; + + pc->attr = CALLOC(pc->attr_nr * 4, sizeof(struct nv50_reg)); + if (!pc->attr) + goto out_err; + + if (pc->p->type == PIPE_SHADER_FRAGMENT) { + iv = alloc_temp(pc, NULL); + emit_interp(pc, iv, iv, NULL); + emit_flop(pc, 0, iv, iv); + aid++; + } + + for (i = 0; i < pc->attr_nr; i++) { + struct nv50_reg *a = &pc->attr[i*4]; + + for (c = 0; c < 4; c++) { + if (pc->p->type == PIPE_SHADER_FRAGMENT) { + struct nv50_reg *at = + alloc_temp(pc, NULL); + pc->attr[i*4+c].type = at->type; + pc->attr[i*4+c].hw = at->hw; + pc->attr[i*4+c].index = at->index; + } else { + pc->p->cfg.vp.attr[aid/32] |= + (1 << (aid % 32)); + pc->attr[i*4+c].type = P_ATTR; + pc->attr[i*4+c].hw = aid++; + pc->attr[i*4+c].index = i; + } + } + + if (pc->p->type != PIPE_SHADER_FRAGMENT) + continue; + + emit_interp(pc, &a[0], &a[0], iv); + emit_interp(pc, &a[1], &a[1], iv); + emit_interp(pc, &a[2], &a[2], iv); + emit_interp(pc, &a[3], &a[3], iv); + } + + if (iv) + free_temp(pc, iv); + } + + if (pc->result_nr) { + int rid = 0; + + pc->result = CALLOC(pc->result_nr * 4, sizeof(struct nv50_reg)); + if (!pc->result) + goto out_err; + + for (i = 0; i < pc->result_nr; i++) { + for (c = 0; c < 4; c++) { + if (pc->p->type == PIPE_SHADER_FRAGMENT) { + pc->result[i*4+c].type = P_TEMP; + pc->result[i*4+c].hw = -1; + } else { + pc->result[i*4+c].type = P_RESULT; + pc->result[i*4+c].hw = rid++; + } + pc->result[i*4+c].index = i; + } + } + } + + if (pc->param_nr) { + int rid = 0; + + pc->param = CALLOC(pc->param_nr * 4, sizeof(struct nv50_reg)); + if (!pc->param) + goto out_err; + + for (i = 0; i < pc->param_nr; i++) { + for (c = 0; c < 4; c++) { + pc->param[i*4+c].type = P_CONST; + pc->param[i*4+c].hw = rid++; + pc->param[i*4+c].index = i; + } + } + } + + if (pc->immd_nr) { + int rid = pc->param_nr * 4; + + pc->immd = CALLOC(pc->immd_nr * 4, sizeof(struct nv50_reg)); + if (!pc->immd) + goto out_err; + + for (i = 0; i < pc->immd_nr; i++) { + for (c = 0; c < 4; c++) { + pc->immd[i*4+c].type = P_IMMD; + pc->immd[i*4+c].hw = rid++; + pc->immd[i*4+c].index = i; + } + } + } + + ret = TRUE; +out_err: + tgsi_parse_free(&p); + return ret; +} + +static boolean +nv50_program_tx(struct nv50_program *p) +{ + struct tgsi_parse_context parse; + struct nv50_pc *pc; + boolean ret; + + pc = CALLOC_STRUCT(nv50_pc); + if (!pc) + return FALSE; + pc->p = p; + pc->p->cfg.high_temp = 4; + + ret = nv50_program_tx_prep(pc); + if (ret == FALSE) + goto out_cleanup; + + tgsi_parse_init(&parse, pc->p->pipe.tokens); + while (!tgsi_parse_end_of_tokens(&parse)) { + const union tgsi_full_token *tok = &parse.FullToken; + + tgsi_parse_token(&parse); + + switch (tok->Token.Type) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + ret = nv50_program_tx_insn(pc, tok); + if (ret == FALSE) + goto out_err; + break; + default: + break; + } + } + + if (p->type == PIPE_SHADER_FRAGMENT) { + struct nv50_reg out; + + out.type = P_TEMP; + for (out.hw = 0; out.hw < pc->result_nr * 4; out.hw++) + emit_mov(pc, &out, &pc->result[out.hw]); + } + + assert(is_long(pc->p->exec_tail) && !is_immd(pc->p->exec_head)); + pc->p->exec_tail->inst[1] |= 0x00000001; + + p->param_nr = pc->param_nr * 4; + p->immd_nr = pc->immd_nr * 4; + p->immd = pc->immd_buf; + +out_err: + tgsi_parse_free(&parse); + +out_cleanup: + return ret; +} + +static void +nv50_program_validate(struct nv50_context *nv50, struct nv50_program *p) +{ + if (nv50_program_tx(p) == FALSE) + assert(0); + p->translated = TRUE; +} + +static void +nv50_program_upload_data(struct nv50_context *nv50, float *map, + unsigned start, unsigned count) +{ + while (count) { + unsigned nr = count > 2047 ? 2047 : count; + + BEGIN_RING(tesla, 0x00000f00, 1); + OUT_RING ((NV50_CB_PMISC << 0) | (start << 8)); + BEGIN_RING(tesla, 0x40000f04, nr); + OUT_RINGp (map, nr); + + map += nr; + start += nr; + count -= nr; + } +} + +static void +nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p) +{ + struct nouveau_winsys *nvws = nv50->screen->nvws; + struct pipe_winsys *ws = nv50->pipe.winsys; + unsigned nr = p->param_nr + p->immd_nr; + + if (!p->data && nr) { + struct nouveau_resource *heap = nv50->screen->vp_data_heap; + + if (nvws->res_alloc(heap, nr, p, &p->data)) { + while (heap->next && heap->size < nr) { + struct nv50_program *evict = heap->next->priv; + nvws->res_free(&evict->data); + } + + if (nvws->res_alloc(heap, nr, p, &p->data)) + assert(0); + } + } + + if (p->param_nr) { + float *map = ws->buffer_map(ws, nv50->constbuf[p->type], + PIPE_BUFFER_USAGE_CPU_READ); + nv50_program_upload_data(nv50, map, p->data->start, + p->param_nr); + ws->buffer_unmap(ws, nv50->constbuf[p->type]); + } + + if (p->immd_nr) { + nv50_program_upload_data(nv50, p->immd, + p->data->start + p->param_nr, + p->immd_nr); + } +} + +static void +nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p) +{ + struct pipe_winsys *ws = nv50->pipe.winsys; + struct nv50_program_exec *e; + struct nouveau_stateobj *so; + const unsigned flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_WR; + unsigned start, count, *up, *ptr; + boolean upload = FALSE; + + if (!p->buffer) { + p->buffer = ws->buffer_create(ws, 0x100, 0, p->exec_size * 4); + upload = TRUE; + } + + if (p->data && p->data->start != p->data_start) { + for (e = p->exec_head; e; e = e->next) { + unsigned ei, ci; + + if (e->param.index < 0) + continue; + ei = e->param.shift >> 5; + ci = e->param.index + p->data->start; + + e->inst[ei] &= ~e->param.mask; + e->inst[ei] |= (ci << e->param.shift); + } + + p->data_start = p->data->start; + upload = TRUE; + } + + if (!upload) + return; + + up = ptr = MALLOC(p->exec_size * 4); + for (e = p->exec_head; e; e = e->next) { + *(ptr++) = e->inst[0]; + if (is_long(e)) + *(ptr++) = e->inst[1]; + } + + so = so_new(4,2); + so_method(so, nv50->screen->tesla, 0x1280, 3); + so_reloc (so, p->buffer, 0, flags | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, p->buffer, 0, flags | NOUVEAU_BO_LOW, 0, 0); + so_data (so, (NV50_CB_PUPLOAD << 16) | 0x0800); //(p->exec_size * 4)); + + start = 0; count = p->exec_size; + while (count) { + struct nouveau_winsys *nvws = nv50->screen->nvws; + unsigned nr; + + so_emit(nvws, so); + + nr = MIN2(count, 2047); + nr = MIN2(nvws->channel->pushbuf->remaining, nr); + if (nvws->channel->pushbuf->remaining < (nr + 3)) { + FIRE_RING(NULL); + continue; + } + + BEGIN_RING(tesla, 0x0f00, 1); + OUT_RING ((start << 8) | NV50_CB_PUPLOAD); + BEGIN_RING(tesla, 0x40000f04, nr); + OUT_RINGp (up + start, nr); + + start += nr; + count -= nr; + } + + FREE(up); + so_ref(NULL, &so); +} + +void +nv50_vertprog_validate(struct nv50_context *nv50) +{ + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nv50_program *p = nv50->vertprog; + struct nouveau_stateobj *so; + + if (!p->translated) { + nv50_program_validate(nv50, p); + if (!p->translated) + assert(0); + } + + nv50_program_validate_data(nv50, p); + nv50_program_validate_code(nv50, p); + + so = so_new(13, 2); + so_method(so, tesla, NV50TCL_VP_ADDRESS_HIGH, 2); + so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | + NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | + NOUVEAU_BO_LOW, 0, 0); + so_method(so, tesla, 0x1650, 2); + so_data (so, p->cfg.vp.attr[0]); + so_data (so, p->cfg.vp.attr[1]); + so_method(so, tesla, 0x16b8, 1); + so_data (so, p->cfg.high_result); + so_method(so, tesla, 0x16ac, 2); + so_data (so, p->cfg.high_result); //8); + so_data (so, p->cfg.high_temp); + so_method(so, tesla, 0x140c, 1); + so_data (so, 0); /* program start offset */ + so_ref(so, &nv50->state.vertprog); +} + +void +nv50_fragprog_validate(struct nv50_context *nv50) +{ + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nv50_program *p = nv50->fragprog; + struct nouveau_stateobj *so; + + if (!p->translated) { + nv50_program_validate(nv50, p); + if (!p->translated) + assert(0); + } + + nv50_program_validate_data(nv50, p); + nv50_program_validate_code(nv50, p); + + so = so_new(64, 2); + so_method(so, tesla, NV50TCL_FP_ADDRESS_HIGH, 2); + so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | + NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | + NOUVEAU_BO_LOW, 0, 0); + so_method(so, tesla, 0x1904, 4); + so_data (so, 0x01040404); /* p: 0x01000404 */ + so_data (so, 0x00000004); + so_data (so, 0x00000000); + so_data (so, 0x00000000); + so_method(so, tesla, 0x16bc, 3); /*XXX: fixme */ + so_data (so, 0x03020100); + so_data (so, 0x07060504); + so_data (so, 0x0b0a0908); + so_method(so, tesla, 0x1988, 2); + so_data (so, 0x08080408); //0x08040404); /* p: 0x0f000401 */ + so_data (so, p->cfg.high_temp); + so_method(so, tesla, 0x1414, 1); + so_data (so, 0); /* program start offset */ + so_ref(so, &nv50->state.fragprog); +} + +void +nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p) +{ + struct pipe_winsys *ws = nv50->pipe.winsys; + + while (p->exec_head) { + struct nv50_program_exec *e = p->exec_head; + + p->exec_head = e->next; + FREE(e); + } + p->exec_tail = NULL; + p->exec_size = 0; + + if (p->buffer) + pipe_buffer_reference(ws, &p->buffer, NULL); + + nv50->screen->nvws->res_free(&p->data); + + p->translated = 0; +} + diff --git a/src/gallium/drivers/nv50/nv50_program.h b/src/gallium/drivers/nv50/nv50_program.h new file mode 100644 index 0000000000..78deed6a38 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_program.h @@ -0,0 +1,45 @@ +#ifndef __NV50_PROGRAM_H__ +#define __NV50_PROGRAM_H__ + +#include "pipe/p_state.h" +#include "tgsi/tgsi_scan.h" + +struct nv50_program_exec { + struct nv50_program_exec *next; + + unsigned inst[2]; + struct { + int index; + unsigned mask; + unsigned shift; + } param; +}; + +struct nv50_program { + struct pipe_shader_state pipe; + struct tgsi_shader_info info; + boolean translated; + + unsigned type; + struct nv50_program_exec *exec_head; + struct nv50_program_exec *exec_tail; + unsigned exec_size; + struct nouveau_resource *data; + unsigned data_start; + + struct pipe_buffer *buffer; + + float *immd; + unsigned immd_nr; + unsigned param_nr; + + struct { + unsigned high_temp; + unsigned high_result; + struct { + unsigned attr[2]; + } vp; + } cfg; +}; + +#endif diff --git a/src/gallium/drivers/nv50/nv50_query.c b/src/gallium/drivers/nv50/nv50_query.c new file mode 100644 index 0000000000..26bd90ccc5 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_query.c @@ -0,0 +1,69 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "pipe/p_context.h" + +#include "nv50_context.h" + +static struct pipe_query * +nv50_query_create(struct pipe_context *pipe, unsigned type) +{ + NOUVEAU_ERR("unimplemented\n"); + return NULL; +} + +static void +nv50_query_destroy(struct pipe_context *pipe, struct pipe_query *q) +{ + NOUVEAU_ERR("unimplemented\n"); +} + +static void +nv50_query_begin(struct pipe_context *pipe, struct pipe_query *q) +{ + NOUVEAU_ERR("unimplemented\n"); +} + +static void +nv50_query_end(struct pipe_context *pipe, struct pipe_query *q) +{ + NOUVEAU_ERR("unimplemented\n"); +} + +static boolean +nv50_query_result(struct pipe_context *pipe, struct pipe_query *q, + boolean wait, uint64 *result) +{ + NOUVEAU_ERR("unimplemented\n"); + *result = 0xdeadcafe; + return TRUE; +} + +void +nv50_init_query_functions(struct nv50_context *nv50) +{ + nv50->pipe.create_query = nv50_query_create; + nv50->pipe.destroy_query = nv50_query_destroy; + nv50->pipe.begin_query = nv50_query_begin; + nv50->pipe.end_query = nv50_query_end; + nv50->pipe.get_query_result = nv50_query_result; +} diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c new file mode 100644 index 0000000000..52f6a40688 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_screen.c @@ -0,0 +1,325 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "pipe/p_screen.h" + +#include "nv50_context.h" +#include "nv50_screen.h" + +#include "nouveau/nouveau_stateobj.h" + +#define NV5X_GRCLASS5097_CHIPSETS 0x00000001 +#define NV8X_GRCLASS8297_CHIPSETS 0x00000050 +#define NV9X_GRCLASS8297_CHIPSETS 0x00000014 + +static boolean +nv50_screen_is_format_supported(struct pipe_screen *pscreen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, unsigned geom_flags) +{ + if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z16_UNORM: + return TRUE; + default: + break; + } + } else { + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_A1R5G5B5_UNORM: + case PIPE_FORMAT_A4R4G4B4_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_A8_UNORM: + case PIPE_FORMAT_I8_UNORM: + case PIPE_FORMAT_A8L8_UNORM: + return TRUE; + default: + break; + } + } + + return FALSE; +} + +static const char * +nv50_screen_get_name(struct pipe_screen *pscreen) +{ + struct nv50_screen *screen = nv50_screen(pscreen); + struct nouveau_device *dev = screen->nvws->channel->device; + static char buffer[128]; + + snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); + return buffer; +} + +static const char * +nv50_screen_get_vendor(struct pipe_screen *pscreen) +{ + return "nouveau"; +} + +static int +nv50_screen_get_param(struct pipe_screen *pscreen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 32; + case PIPE_CAP_NPOT_TEXTURES: + return 0; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 8; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 0; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 13; + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 10; + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 13; + case PIPE_CAP_TEXTURE_MIRROR_CLAMP: + case PIPE_CAP_TEXTURE_MIRROR_REPEAT: + return 1; + case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: + return 0; + case NOUVEAU_CAP_HW_VTXBUF: + return 1; + case NOUVEAU_CAP_HW_IDXBUF: + return 0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0; + } +} + +static float +nv50_screen_get_paramf(struct pipe_screen *pscreen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 10.0; + case PIPE_CAP_MAX_POINT_WIDTH: + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 64.0; + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 16.0; + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 4.0; + default: + NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param); + return 0.0; + } +} + +static void +nv50_screen_destroy(struct pipe_screen *pscreen) +{ + FREE(pscreen); +} + +struct pipe_screen * +nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +{ + struct nv50_screen *screen = CALLOC_STRUCT(nv50_screen); + struct nouveau_stateobj *so; + unsigned tesla_class = 0, ret; + unsigned chipset = nvws->channel->device->chipset; + int i; + + if (!screen) + return NULL; + screen->nvws = nvws; + + /* 3D object */ + if ((chipset & 0xf0) != 0x50 && (chipset & 0xf0) != 0x80) { + NOUVEAU_ERR("Not a G8x chipset\n"); + nv50_screen_destroy(&screen->pipe); + return NULL; + } + + switch (chipset & 0xf0) { + case 0x50: + if (NV5X_GRCLASS5097_CHIPSETS & (1 << (chipset & 0x0f))) + tesla_class = 0x5097; + break; + case 0x80: + if (NV8X_GRCLASS8297_CHIPSETS & (1 << (chipset & 0x0f))) + tesla_class = 0x8297; + break; + case 0x90: + if (NV9X_GRCLASS8297_CHIPSETS & (1 << (chipset & 0x0f))) + tesla_class = 0x8297; + break; + default: + break; + } + + if (tesla_class == 0) { + NOUVEAU_ERR("Unknown G8x chipset: NV%02x\n", chipset); + nv50_screen_destroy(&screen->pipe); + return NULL; + } + + ret = nvws->grobj_alloc(nvws, tesla_class, &screen->tesla); + if (ret) { + NOUVEAU_ERR("Error creating 3D object: %d\n", ret); + nv50_screen_destroy(&screen->pipe); + return NULL; + } + + /* Sync notifier */ + ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + if (ret) { + NOUVEAU_ERR("Error creating notifier object: %d\n", ret); + nv50_screen_destroy(&screen->pipe); + return NULL; + } + + /* Static tesla init */ + so = so_new(256, 20); + + so_method(so, screen->tesla, 0x1558, 1); + so_data (so, 1); + so_method(so, screen->tesla, NV50TCL_DMA_NOTIFY, 1); + so_data (so, screen->sync->handle); + so_method(so, screen->tesla, NV50TCL_DMA_UNK0(0), + NV50TCL_DMA_UNK0__SIZE); + for (i = 0; i < NV50TCL_DMA_UNK0__SIZE; i++) + so_data(so, nvws->channel->vram->handle); + so_method(so, screen->tesla, NV50TCL_DMA_UNK1(0), + NV50TCL_DMA_UNK1__SIZE); + for (i = 0; i < NV50TCL_DMA_UNK1__SIZE; i++) + so_data(so, nvws->channel->vram->handle); + so_method(so, screen->tesla, 0x121c, 1); + so_data (so, 1); + + so_method(so, screen->tesla, 0x13bc, 1); + so_data (so, 0x54); + so_method(so, screen->tesla, 0x13ac, 1); + so_data (so, 1); + so_method(so, screen->tesla, 0x16b8, 1); + so_data (so, 8); + + /* Shared constant buffer */ + screen->constbuf = ws->buffer_create(ws, 0, 0, 128 * 4 * 4); + if (nvws->res_init(&screen->vp_data_heap, 0, 128)) { + NOUVEAU_ERR("Error initialising constant buffer\n"); + nv50_screen_destroy(&screen->pipe); + return NULL; + } + + so_method(so, screen->tesla, 0x1280, 3); + so_reloc (so, screen->constbuf, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, screen->constbuf, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); + so_data (so, (NV50_CB_PMISC << 16) | 0x00001000); + + /* Texture sampler/image unit setup - we abuse the constant buffer + * upload mechanism for the moment to upload data to the tex config + * blocks. At some point we *may* want to go the NVIDIA way of doing + * things? + */ + screen->tic = ws->buffer_create(ws, 0, 0, 32 * 8 * 4); + so_method(so, screen->tesla, 0x1280, 3); + so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); + so_data (so, (NV50_CB_TIC << 16) | 0x0800); + so_method(so, screen->tesla, 0x1574, 3); + so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); + so_data (so, 0x00000800); + + screen->tsc = ws->buffer_create(ws, 0, 0, 32 * 8 * 4); + so_method(so, screen->tesla, 0x1280, 3); + so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); + so_data (so, (NV50_CB_TSC << 16) | 0x0800); + so_method(so, screen->tesla, 0x155c, 3); + so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); + so_data (so, 0x00000800); + + + /* Vertex array limits - max them out */ + for (i = 0; i < 16; i++) { + so_method(so, screen->tesla, 0x1080 + (i * 8), 2); + so_data (so, 0x000000ff); + so_data (so, 0xffffffff); + } + + so_method(so, screen->tesla, NV50TCL_DEPTH_RANGE_NEAR, 2); + so_data (so, fui(0.0)); + so_data (so, fui(1.0)); + + so_method(so, screen->tesla, 0x1234, 1); + so_data (so, 1); + so_method(so, screen->tesla, 0x1458, 1); + so_data (so, 1); + + so_emit(nvws, so); + so_ref(so, &screen->static_init); + nvws->push_flush(nvws, 0, NULL); + + screen->pipe.winsys = ws; + + screen->pipe.destroy = nv50_screen_destroy; + + screen->pipe.get_name = nv50_screen_get_name; + screen->pipe.get_vendor = nv50_screen_get_vendor; + screen->pipe.get_param = nv50_screen_get_param; + screen->pipe.get_paramf = nv50_screen_get_paramf; + + screen->pipe.is_format_supported = nv50_screen_is_format_supported; + + nv50_screen_init_miptree_functions(&screen->pipe); + nv50_surface_init_screen_functions(&screen->pipe); + + return &screen->pipe; +} + diff --git a/src/gallium/drivers/nv50/nv50_screen.h b/src/gallium/drivers/nv50/nv50_screen.h new file mode 100644 index 0000000000..400ddcef06 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_screen.h @@ -0,0 +1,33 @@ +#ifndef __NV50_SCREEN_H__ +#define __NV50_SCREEN_H__ + +#include "pipe/p_screen.h" + +struct nv50_screen { + struct pipe_screen pipe; + + struct nouveau_winsys *nvws; + + unsigned cur_pctx; + + struct nouveau_grobj *tesla; + struct nouveau_notifier *sync; + + struct pipe_buffer *constbuf; + struct nouveau_resource *vp_data_heap; + + struct pipe_buffer *tic; + struct pipe_buffer *tsc; + + struct nouveau_stateobj *static_init; +}; + +static INLINE struct nv50_screen * +nv50_screen(struct pipe_screen *screen) +{ + return (struct nv50_screen *)screen; +} + +void nv50_surface_init_screen_functions(struct pipe_screen *); + +#endif diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c new file mode 100644 index 0000000000..95f9d408b5 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_state.c @@ -0,0 +1,638 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "tgsi/tgsi_parse.h" + +#include "nv50_context.h" +#include "nv50_texture.h" + +#include "nouveau/nouveau_stateobj.h" + +static void * +nv50_blend_state_create(struct pipe_context *pipe, + const struct pipe_blend_state *cso) +{ + struct nouveau_stateobj *so = so_new(64, 0); + struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla; + struct nv50_blend_stateobj *bso = CALLOC_STRUCT(nv50_blend_stateobj); + unsigned cmask = 0, i; + + /*XXX ignored: + * - dither + */ + + if (cso->blend_enable == 0) { + so_method(so, tesla, NV50TCL_BLEND_ENABLE(0), 8); + for (i = 0; i < 8; i++) + so_data(so, 0); + } else { + so_method(so, tesla, NV50TCL_BLEND_ENABLE(0), 8); + for (i = 0; i < 8; i++) + so_data(so, 1); + so_method(so, tesla, NV50TCL_BLEND_EQUATION_RGB, 5); + so_data (so, nvgl_blend_eqn(cso->rgb_func)); + so_data (so, 0x4000 | nvgl_blend_func(cso->rgb_src_factor)); + so_data (so, 0x4000 | nvgl_blend_func(cso->rgb_dst_factor)); + so_data (so, nvgl_blend_eqn(cso->alpha_func)); + so_data (so, 0x4000 | nvgl_blend_func(cso->alpha_src_factor)); + so_method(so, tesla, NV50TCL_BLEND_FUNC_DST_ALPHA, 1); + so_data (so, 0x4000 | nvgl_blend_func(cso->alpha_dst_factor)); + } + + if (cso->logicop_enable == 0 ) { + so_method(so, tesla, NV50TCL_LOGIC_OP_ENABLE, 1); + so_data (so, 0); + } else { + so_method(so, tesla, NV50TCL_LOGIC_OP_ENABLE, 2); + so_data (so, 1); + so_data (so, nvgl_logicop_func(cso->logicop_func)); + } + + if (cso->colormask & PIPE_MASK_R) + cmask |= (1 << 0); + if (cso->colormask & PIPE_MASK_G) + cmask |= (1 << 4); + if (cso->colormask & PIPE_MASK_B) + cmask |= (1 << 8); + if (cso->colormask & PIPE_MASK_A) + cmask |= (1 << 12); + so_method(so, tesla, NV50TCL_COLOR_MASK(0), 8); + for (i = 0; i < 8; i++) + so_data(so, cmask); + + bso->pipe = *cso; + so_ref(so, &bso->so); + return (void *)bso; +} + +static void +nv50_blend_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50->blend = hwcso; + nv50->dirty |= NV50_NEW_BLEND; +} + +static void +nv50_blend_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv50_blend_stateobj *bso = hwcso; + + so_ref(NULL, &bso->so); + FREE(bso); +} + +static INLINE unsigned +wrap_mode(unsigned wrap) +{ + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + return NV50TSC_1_0_WRAPS_REPEAT; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + return NV50TSC_1_0_WRAPS_MIRROR_REPEAT; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return NV50TSC_1_0_WRAPS_CLAMP_TO_EDGE; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return NV50TSC_1_0_WRAPS_CLAMP_TO_BORDER; + case PIPE_TEX_WRAP_CLAMP: + return NV50TSC_1_0_WRAPS_CLAMP; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + return NV50TSC_1_0_WRAPS_MIRROR_CLAMP_TO_EDGE; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + return NV50TSC_1_0_WRAPS_MIRROR_CLAMP_TO_BORDER; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + return NV50TSC_1_0_WRAPS_MIRROR_CLAMP; + default: + NOUVEAU_ERR("unknown wrap mode: %d\n", wrap); + return NV50TSC_1_0_WRAPS_REPEAT; + } +} +static void * +nv50_sampler_state_create(struct pipe_context *pipe, + const struct pipe_sampler_state *cso) +{ + unsigned *tsc = CALLOC(8, sizeof(unsigned)); + + tsc[0] = (0x00024000 | + (wrap_mode(cso->wrap_s) << 0) | + (wrap_mode(cso->wrap_t) << 3) | + (wrap_mode(cso->wrap_r) << 6)); + + switch (cso->mag_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + tsc[1] |= NV50TSC_1_1_MAGF_LINEAR; + break; + case PIPE_TEX_FILTER_NEAREST: + default: + tsc[1] |= NV50TSC_1_1_MAGF_NEAREST; + break; + } + + switch (cso->min_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + tsc[1] |= NV50TSC_1_1_MINF_LINEAR; + break; + case PIPE_TEX_FILTER_NEAREST: + default: + tsc[1] |= NV50TSC_1_1_MINF_NEAREST; + break; + } + + switch (cso->min_mip_filter) { + case PIPE_TEX_MIPFILTER_LINEAR: + tsc[1] |= NV50TSC_1_1_MIPF_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NEAREST: + tsc[1] |= NV50TSC_1_1_MIPF_NEAREST; + break; + case PIPE_TEX_MIPFILTER_NONE: + default: + tsc[1] |= NV50TSC_1_1_MIPF_NONE; + break; + } + + return (void *)tsc; +} + +static void +nv50_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler) +{ + struct nv50_context *nv50 = nv50_context(pipe); + int i; + + nv50->sampler_nr = nr; + for (i = 0; i < nv50->sampler_nr; i++) + nv50->sampler[i] = sampler[i]; + + nv50->dirty |= NV50_NEW_SAMPLER; +} + +static void +nv50_sampler_state_delete(struct pipe_context *pipe, void *hwcso) +{ + FREE(hwcso); +} + +static void +nv50_set_sampler_texture(struct pipe_context *pipe, unsigned nr, + struct pipe_texture **pt) +{ + struct nv50_context *nv50 = nv50_context(pipe); + int i; + + for (i = 0; i < nr; i++) + pipe_texture_reference((void *)&nv50->miptree[i], pt[i]); + for (i = nr; i < nv50->miptree_nr; i++) + pipe_texture_reference((void *)&nv50->miptree[i], NULL); + + nv50->miptree_nr = nr; + nv50->dirty |= NV50_NEW_TEXTURE; +} + +static void * +nv50_rasterizer_state_create(struct pipe_context *pipe, + const struct pipe_rasterizer_state *cso) +{ + struct nouveau_stateobj *so = so_new(64, 0); + struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla; + struct nv50_rasterizer_stateobj *rso = + CALLOC_STRUCT(nv50_rasterizer_stateobj); + + /*XXX: ignored + * - light_twosize + * - point_smooth + * - multisample + * - point_sprite / sprite_coord_mode + */ + + so_method(so, tesla, NV50TCL_SHADE_MODEL, 1); + so_data (so, cso->flatshade ? NV50TCL_SHADE_MODEL_FLAT : + NV50TCL_SHADE_MODEL_SMOOTH); + + so_method(so, tesla, NV50TCL_LINE_WIDTH, 1); + so_data (so, fui(cso->line_width)); + so_method(so, tesla, NV50TCL_LINE_SMOOTH_ENABLE, 1); + so_data (so, cso->line_smooth ? 1 : 0); + if (cso->line_stipple_enable) { + so_method(so, tesla, NV50TCL_LINE_STIPPLE_ENABLE, 1); + so_data (so, 1); + so_method(so, tesla, NV50TCL_LINE_STIPPLE_PATTERN, 1); + so_data (so, (cso->line_stipple_pattern << 8) | + cso->line_stipple_factor); + } else { + so_method(so, tesla, NV50TCL_LINE_STIPPLE_ENABLE, 1); + so_data (so, 0); + } + + so_method(so, tesla, NV50TCL_POINT_SIZE, 1); + so_data (so, fui(cso->point_size)); + + so_method(so, tesla, NV50TCL_POLYGON_MODE_FRONT, 3); + if (cso->front_winding == PIPE_WINDING_CCW) { + so_data(so, nvgl_polygon_mode(cso->fill_ccw)); + so_data(so, nvgl_polygon_mode(cso->fill_cw)); + } else { + so_data(so, nvgl_polygon_mode(cso->fill_cw)); + so_data(so, nvgl_polygon_mode(cso->fill_ccw)); + } + so_data(so, cso->poly_smooth ? 1 : 0); + + so_method(so, tesla, NV50TCL_CULL_FACE_ENABLE, 3); + so_data (so, cso->cull_mode != PIPE_WINDING_NONE); + if (cso->front_winding == PIPE_WINDING_CCW) { + so_data(so, NV50TCL_FRONT_FACE_CCW); + switch (cso->cull_mode) { + case PIPE_WINDING_CCW: + so_data(so, NV50TCL_CULL_FACE_FRONT); + break; + case PIPE_WINDING_CW: + so_data(so, NV50TCL_CULL_FACE_BACK); + break; + case PIPE_WINDING_BOTH: + so_data(so, NV50TCL_CULL_FACE_FRONT_AND_BACK); + break; + default: + so_data(so, NV50TCL_CULL_FACE_BACK); + break; + } + } else { + so_data(so, NV50TCL_FRONT_FACE_CW); + switch (cso->cull_mode) { + case PIPE_WINDING_CCW: + so_data(so, NV50TCL_CULL_FACE_BACK); + break; + case PIPE_WINDING_CW: + so_data(so, NV50TCL_CULL_FACE_FRONT); + break; + case PIPE_WINDING_BOTH: + so_data(so, NV50TCL_CULL_FACE_FRONT_AND_BACK); + break; + default: + so_data(so, NV50TCL_CULL_FACE_BACK); + break; + } + } + + so_method(so, tesla, NV50TCL_POLYGON_STIPPLE_ENABLE, 1); + so_data (so, cso->poly_stipple_enable ? 1 : 0); + + so_method(so, tesla, NV50TCL_POLYGON_OFFSET_POINT_ENABLE, 3); + if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) || + (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT)) + so_data(so, 1); + else + so_data(so, 0); + if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) || + (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE)) + so_data(so, 1); + else + so_data(so, 0); + if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) || + (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL)) + so_data(so, 1); + else + so_data(so, 0); + + if (cso->offset_cw || cso->offset_ccw) { + so_method(so, tesla, NV50TCL_POLYGON_OFFSET_FACTOR, 1); + so_data (so, fui(cso->offset_scale)); + so_method(so, tesla, NV50TCL_POLYGON_OFFSET_UNITS, 1); + so_data (so, fui(cso->offset_units)); + } + + rso->pipe = *cso; + so_ref(so, &rso->so); + return (void *)rso; +} + +static void +nv50_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50->rasterizer = hwcso; + nv50->dirty |= NV50_NEW_RASTERIZER; +} + +static void +nv50_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv50_rasterizer_stateobj *rso = hwcso; + + so_ref(NULL, &rso->so); + FREE(rso); +} + +static void * +nv50_depth_stencil_alpha_state_create(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *cso) +{ + struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla; + struct nv50_zsa_stateobj *zsa = CALLOC_STRUCT(nv50_zsa_stateobj); + struct nouveau_stateobj *so = so_new(64, 0); + + so_method(so, tesla, NV50TCL_DEPTH_WRITE_ENABLE, 1); + so_data (so, cso->depth.writemask ? 1 : 0); + if (cso->depth.enabled) { + so_method(so, tesla, NV50TCL_DEPTH_TEST_ENABLE, 1); + so_data (so, 1); + so_method(so, tesla, NV50TCL_DEPTH_TEST_FUNC, 1); + so_data (so, nvgl_comparison_op(cso->depth.func)); + } else { + so_method(so, tesla, NV50TCL_DEPTH_TEST_ENABLE, 1); + so_data (so, 0); + } + + /*XXX: yes, I know they're backwards.. header needs fixing */ + if (cso->stencil[0].enabled) { + so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 5); + so_data (so, 1); + so_data (so, nvgl_stencil_op(cso->stencil[0].fail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[0].zfail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[0].zpass_op)); + so_data (so, nvgl_comparison_op(cso->stencil[0].func)); + so_method(so, tesla, NV50TCL_STENCIL_BACK_FUNC_REF, 3); + so_data (so, cso->stencil[0].ref_value); + so_data (so, cso->stencil[0].write_mask); + so_data (so, cso->stencil[0].value_mask); + } else { + so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 1); + so_data (so, 0); + } + + if (cso->stencil[1].enabled) { + so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 8); + so_data (so, 1); + so_data (so, nvgl_stencil_op(cso->stencil[1].fail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[1].zfail_op)); + so_data (so, nvgl_stencil_op(cso->stencil[1].zpass_op)); + so_data (so, nvgl_comparison_op(cso->stencil[1].func)); + so_data (so, cso->stencil[1].ref_value); + so_data (so, cso->stencil[1].write_mask); + so_data (so, cso->stencil[1].value_mask); + } else { + so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 1); + so_data (so, 0); + } + + if (cso->alpha.enabled) { + so_method(so, tesla, NV50TCL_ALPHA_TEST_ENABLE, 1); + so_data (so, 1); + so_method(so, tesla, NV50TCL_ALPHA_TEST_REF, 2); + so_data (so, fui(cso->alpha.ref)); + so_data (so, nvgl_comparison_op(cso->alpha.func)); + } else { + so_method(so, tesla, NV50TCL_ALPHA_TEST_ENABLE, 1); + so_data (so, 0); + } + + zsa->pipe = *cso; + so_ref(so, &zsa->so); + return (void *)zsa; +} + +static void +nv50_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50->zsa = hwcso; + nv50->dirty |= NV50_NEW_ZSA; +} + +static void +nv50_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv50_zsa_stateobj *zsa = hwcso; + + so_ref(NULL, &zsa->so); + FREE(zsa); +} + +static void * +nv50_vp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *cso) +{ + struct nv50_program *p = CALLOC_STRUCT(nv50_program); + + p->pipe.tokens = tgsi_dup_tokens(cso->tokens); + p->type = PIPE_SHADER_VERTEX; + tgsi_scan_shader(p->pipe.tokens, &p->info); + return (void *)p; +} + +static void +nv50_vp_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50->vertprog = hwcso; + nv50->dirty |= NV50_NEW_VERTPROG; +} + +static void +nv50_vp_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv50_context *nv50 = nv50_context(pipe); + struct nv50_program *p = hwcso; + + nv50_program_destroy(nv50, p); + FREE((void*)p->pipe.tokens); + FREE(p); +} + +static void * +nv50_fp_state_create(struct pipe_context *pipe, + const struct pipe_shader_state *cso) +{ + struct nv50_program *p = CALLOC_STRUCT(nv50_program); + + p->pipe.tokens = tgsi_dup_tokens(cso->tokens); + p->type = PIPE_SHADER_FRAGMENT; + tgsi_scan_shader(p->pipe.tokens, &p->info); + return (void *)p; +} + +static void +nv50_fp_state_bind(struct pipe_context *pipe, void *hwcso) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50->fragprog = hwcso; + nv50->dirty |= NV50_NEW_FRAGPROG; +} + +static void +nv50_fp_state_delete(struct pipe_context *pipe, void *hwcso) +{ + struct nv50_context *nv50 = nv50_context(pipe); + struct nv50_program *p = hwcso; + + nv50_program_destroy(nv50, p); + FREE((void*)p->pipe.tokens); + FREE(p); +} + +static void +nv50_set_blend_color(struct pipe_context *pipe, + const struct pipe_blend_color *bcol) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50->blend_colour = *bcol; + nv50->dirty |= NV50_NEW_BLEND_COLOUR; +} + +static void +nv50_set_clip_state(struct pipe_context *pipe, + const struct pipe_clip_state *clip) +{ +} + +static void +nv50_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, + const struct pipe_constant_buffer *buf ) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + if (shader == PIPE_SHADER_VERTEX) { + nv50->constbuf[PIPE_SHADER_VERTEX] = buf->buffer; + nv50->dirty |= NV50_NEW_VERTPROG_CB; + } else + if (shader == PIPE_SHADER_FRAGMENT) { + nv50->constbuf[PIPE_SHADER_FRAGMENT] = buf->buffer; + nv50->dirty |= NV50_NEW_FRAGPROG_CB; + } +} + +static void +nv50_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50->framebuffer = *fb; + nv50->dirty |= NV50_NEW_FRAMEBUFFER; +} + +static void +nv50_set_polygon_stipple(struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50->stipple = *stipple; + nv50->dirty |= NV50_NEW_STIPPLE; +} + +static void +nv50_set_scissor_state(struct pipe_context *pipe, + const struct pipe_scissor_state *s) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50->scissor = *s; + nv50->dirty |= NV50_NEW_SCISSOR; +} + +static void +nv50_set_viewport_state(struct pipe_context *pipe, + const struct pipe_viewport_state *vpt) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50->viewport = *vpt; + nv50->dirty |= NV50_NEW_VIEWPORT; +} + +static void +nv50_set_vertex_buffers(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_buffer *vb) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + memcpy(nv50->vtxbuf, vb, sizeof(*vb) * count); + nv50->vtxbuf_nr = count; + + nv50->dirty |= NV50_NEW_ARRAYS; +} + +static void +nv50_set_vertex_elements(struct pipe_context *pipe, unsigned count, + const struct pipe_vertex_element *ve) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + memcpy(nv50->vtxelt, ve, sizeof(*ve) * count); + nv50->vtxelt_nr = count; + + nv50->dirty |= NV50_NEW_ARRAYS; +} + +void +nv50_init_state_functions(struct nv50_context *nv50) +{ + nv50->pipe.create_blend_state = nv50_blend_state_create; + nv50->pipe.bind_blend_state = nv50_blend_state_bind; + nv50->pipe.delete_blend_state = nv50_blend_state_delete; + + nv50->pipe.create_sampler_state = nv50_sampler_state_create; + nv50->pipe.bind_sampler_states = nv50_sampler_state_bind; + nv50->pipe.delete_sampler_state = nv50_sampler_state_delete; + nv50->pipe.set_sampler_textures = nv50_set_sampler_texture; + + nv50->pipe.create_rasterizer_state = nv50_rasterizer_state_create; + nv50->pipe.bind_rasterizer_state = nv50_rasterizer_state_bind; + nv50->pipe.delete_rasterizer_state = nv50_rasterizer_state_delete; + + nv50->pipe.create_depth_stencil_alpha_state = + nv50_depth_stencil_alpha_state_create; + nv50->pipe.bind_depth_stencil_alpha_state = + nv50_depth_stencil_alpha_state_bind; + nv50->pipe.delete_depth_stencil_alpha_state = + nv50_depth_stencil_alpha_state_delete; + + nv50->pipe.create_vs_state = nv50_vp_state_create; + nv50->pipe.bind_vs_state = nv50_vp_state_bind; + nv50->pipe.delete_vs_state = nv50_vp_state_delete; + + nv50->pipe.create_fs_state = nv50_fp_state_create; + nv50->pipe.bind_fs_state = nv50_fp_state_bind; + nv50->pipe.delete_fs_state = nv50_fp_state_delete; + + nv50->pipe.set_blend_color = nv50_set_blend_color; + nv50->pipe.set_clip_state = nv50_set_clip_state; + nv50->pipe.set_constant_buffer = nv50_set_constant_buffer; + nv50->pipe.set_framebuffer_state = nv50_set_framebuffer_state; + nv50->pipe.set_polygon_stipple = nv50_set_polygon_stipple; + nv50->pipe.set_scissor_state = nv50_set_scissor_state; + nv50->pipe.set_viewport_state = nv50_set_viewport_state; + + nv50->pipe.set_vertex_buffers = nv50_set_vertex_buffers; + nv50->pipe.set_vertex_elements = nv50_set_vertex_elements; +} + diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c new file mode 100644 index 0000000000..198e25f448 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -0,0 +1,309 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "nv50_context.h" +#include "nouveau/nouveau_stateobj.h" + +static void +nv50_state_validate_fb(struct nv50_context *nv50) +{ + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nouveau_stateobj *so = so_new(128, 18); + struct pipe_framebuffer_state *fb = &nv50->framebuffer; + unsigned i, w, h, gw = 0; + + for (i = 0; i < fb->num_cbufs; i++) { + if (!gw) { + w = fb->cbufs[i]->width; + h = fb->cbufs[i]->height; + gw = 1; + } else { + assert(w == fb->cbufs[i]->width); + assert(h == fb->cbufs[i]->height); + } + + so_method(so, tesla, NV50TCL_RT_HORIZ(i), 2); + so_data (so, fb->cbufs[i]->width); + so_data (so, fb->cbufs[i]->height); + + so_method(so, tesla, NV50TCL_RT_ADDRESS_HIGH(i), 5); + so_reloc (so, fb->cbufs[i]->buffer, fb->cbufs[i]->offset, + NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, fb->cbufs[i]->buffer, fb->cbufs[i]->offset, + NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW, 0, 0); + switch (fb->cbufs[i]->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + so_data(so, 0xcf); + break; + case PIPE_FORMAT_R5G6B5_UNORM: + so_data(so, 0xe8); + break; + default: + NOUVEAU_ERR("AIIII unknown format %s\n", + pf_name(fb->cbufs[i]->format)); + so_data(so, 0xe6); + break; + } + so_data(so, 0x00000000); + so_data(so, 0x00000000); + + so_method(so, tesla, 0x1224, 1); + so_data (so, 1); + } + + if (fb->zsbuf) { + if (!gw) { + w = fb->zsbuf->width; + h = fb->zsbuf->height; + gw = 1; + } else { + assert(w == fb->zsbuf->width); + assert(h == fb->zsbuf->height); + } + + so_method(so, tesla, NV50TCL_ZETA_ADDRESS_HIGH, 5); + so_reloc (so, fb->zsbuf->buffer, fb->zsbuf->offset, + NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, fb->zsbuf->buffer, fb->zsbuf->offset, + NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW, 0, 0); + switch (fb->zsbuf->format) { + case PIPE_FORMAT_Z24S8_UNORM: + so_data(so, 0x16); + break; + case PIPE_FORMAT_Z16_UNORM: + so_data(so, 0x15); + break; + default: + NOUVEAU_ERR("AIIII unknown format %s\n", + pf_name(fb->zsbuf->format)); + so_data(so, 0x16); + break; + } + so_data(so, 0x00000000); + so_data(so, 0x00000000); + + so_method(so, tesla, 0x1538, 1); + so_data (so, 1); + so_method(so, tesla, 0x1228, 3); + so_data (so, fb->zsbuf->width); + so_data (so, fb->zsbuf->height); + so_data (so, 0x00010001); + } + + so_method(so, tesla, NV50TCL_VIEWPORT_HORIZ, 2); + so_data (so, w << 16); + so_data (so, h << 16); + so_method(so, tesla, 0x0e04, 2); + so_data (so, w << 16); + so_data (so, h << 16); + so_method(so, tesla, 0xdf8, 2); + so_data (so, 0); + so_data (so, h); + + so_ref(so, &nv50->state.fb); +} + +static void +nv50_state_emit(struct nv50_context *nv50) +{ + struct nv50_screen *screen = nv50->screen; + struct nouveau_winsys *nvws = screen->nvws; + + if (nv50->pctx_id != screen->cur_pctx) { + nv50->state.dirty |= 0xffffffff; + screen->cur_pctx = nv50->pctx_id; + } + + if (nv50->state.dirty & NV50_NEW_FRAMEBUFFER) + so_emit(nvws, nv50->state.fb); + if (nv50->state.dirty & NV50_NEW_BLEND) + so_emit(nvws, nv50->state.blend); + if (nv50->state.dirty & NV50_NEW_ZSA) + so_emit(nvws, nv50->state.zsa); + if (nv50->state.dirty & NV50_NEW_VERTPROG) + so_emit(nvws, nv50->state.vertprog); + if (nv50->state.dirty & NV50_NEW_FRAGPROG) + so_emit(nvws, nv50->state.fragprog); + if (nv50->state.dirty & NV50_NEW_RASTERIZER) + so_emit(nvws, nv50->state.rast); + if (nv50->state.dirty & NV50_NEW_BLEND_COLOUR) + so_emit(nvws, nv50->state.blend_colour); + if (nv50->state.dirty & NV50_NEW_STIPPLE) + so_emit(nvws, nv50->state.stipple); + if (nv50->state.dirty & NV50_NEW_SCISSOR) + so_emit(nvws, nv50->state.scissor); + if (nv50->state.dirty & NV50_NEW_VIEWPORT) + so_emit(nvws, nv50->state.viewport); + if (nv50->state.dirty & NV50_NEW_SAMPLER) + so_emit(nvws, nv50->state.tsc_upload); + if (nv50->state.dirty & NV50_NEW_TEXTURE) + so_emit(nvws, nv50->state.tic_upload); + if (nv50->state.dirty & NV50_NEW_ARRAYS) { + so_emit(nvws, nv50->state.vtxfmt); + so_emit(nvws, nv50->state.vtxbuf); + } + nv50->state.dirty = 0; + + so_emit_reloc_markers(nvws, nv50->state.fb); + so_emit_reloc_markers(nvws, nv50->state.vertprog); + so_emit_reloc_markers(nvws, nv50->state.fragprog); + so_emit_reloc_markers(nvws, nv50->state.vtxbuf); + so_emit_reloc_markers(nvws, nv50->screen->static_init); +} + +boolean +nv50_state_validate(struct nv50_context *nv50) +{ + const struct pipe_framebuffer_state *fb = &nv50->framebuffer; + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nouveau_stateobj *so; + unsigned i; + + for (i = 0; i < fb->num_cbufs; i++) + fb->cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED; + + if (fb->zsbuf) + fb->zsbuf->status = PIPE_SURFACE_STATUS_DEFINED; + + if (nv50->dirty & NV50_NEW_FRAMEBUFFER) + nv50_state_validate_fb(nv50); + + if (nv50->dirty & NV50_NEW_BLEND) + so_ref(nv50->blend->so, &nv50->state.blend); + + if (nv50->dirty & NV50_NEW_ZSA) + so_ref(nv50->zsa->so, &nv50->state.zsa); + + if (nv50->dirty & (NV50_NEW_VERTPROG | NV50_NEW_VERTPROG_CB)) + nv50_vertprog_validate(nv50); + + if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_FRAGPROG_CB)) + nv50_fragprog_validate(nv50); + + if (nv50->dirty & NV50_NEW_RASTERIZER) + so_ref(nv50->rasterizer->so, &nv50->state.rast); + + if (nv50->dirty & NV50_NEW_BLEND_COLOUR) { + so = so_new(5, 0); + so_method(so, tesla, NV50TCL_BLEND_COLOR(0), 4); + so_data (so, fui(nv50->blend_colour.color[0])); + so_data (so, fui(nv50->blend_colour.color[1])); + so_data (so, fui(nv50->blend_colour.color[2])); + so_data (so, fui(nv50->blend_colour.color[3])); + so_ref(so, &nv50->state.blend_colour); + } + + if (nv50->dirty & NV50_NEW_STIPPLE) { + so = so_new(33, 0); + so_method(so, tesla, NV50TCL_POLYGON_STIPPLE_PATTERN(0), 32); + for (i = 0; i < 32; i++) + so_data(so, nv50->stipple.stipple[i]); + so_ref(so, &nv50->state.stipple); + } + + if (nv50->dirty & (NV50_NEW_SCISSOR | NV50_NEW_RASTERIZER)) { + struct pipe_rasterizer_state *rast = &nv50->rasterizer->pipe; + struct pipe_scissor_state *s = &nv50->scissor; + + if (nv50->state.scissor && + (rast->scissor == 0 && nv50->state.scissor_enabled == 0)) + goto scissor_uptodate; + nv50->state.scissor_enabled = rast->scissor; + + so = so_new(3, 0); + so_method(so, tesla, 0x0ff4, 2); + if (nv50->state.scissor_enabled) { + so_data(so, ((s->maxx - s->minx) << 16) | s->minx); + so_data(so, ((s->maxy - s->miny) << 16) | s->miny); + } else { + so_data(so, (8192 << 16)); + so_data(so, (8192 << 16)); + } + so_ref(so, &nv50->state.scissor); + nv50->state.dirty |= NV50_NEW_SCISSOR; + } +scissor_uptodate: + + if (nv50->dirty & NV50_NEW_VIEWPORT) { + unsigned bypass; + + if (!nv50->rasterizer->pipe.bypass_clipping) + bypass = 0; + else + bypass = 1; + + if (nv50->state.viewport && + (bypass || !(nv50->dirty & NV50_NEW_VIEWPORT)) && + nv50->state.viewport_bypass == bypass) + goto viewport_uptodate; + nv50->state.viewport_bypass = bypass; + + so = so_new(12, 0); + if (!bypass) { + so_method(so, tesla, NV50TCL_VIEWPORT_UNK1(0), 3); + so_data (so, fui(nv50->viewport.translate[0])); + so_data (so, fui(nv50->viewport.translate[1])); + so_data (so, fui(nv50->viewport.translate[2])); + so_method(so, tesla, NV50TCL_VIEWPORT_UNK0(0), 3); + so_data (so, fui(nv50->viewport.scale[0])); + so_data (so, fui(-nv50->viewport.scale[1])); + so_data (so, fui(nv50->viewport.scale[2])); + so_method(so, tesla, 0x192c, 1); + so_data (so, 1); + so_method(so, tesla, 0x0f90, 1); + so_data (so, 0); + } else { + so_method(so, tesla, 0x192c, 1); + so_data (so, 0); + so_method(so, tesla, 0x0f90, 1); + so_data (so, 1); + } + + so_ref(so, &nv50->state.viewport); + } +viewport_uptodate: + + if (nv50->dirty & NV50_NEW_SAMPLER) { + int i; + + so = so_new(nv50->sampler_nr * 8 + 3, 0); + so_method(so, tesla, 0x0f00, 1); + so_data (so, NV50_CB_TSC); + so_method(so, tesla, 0x40000f04, nv50->sampler_nr * 8); + for (i = 0; i < nv50->sampler_nr; i++) + so_datap (so, nv50->sampler[i], 8); + so_ref(so, &nv50->state.tsc_upload); + } + + if (nv50->dirty & NV50_NEW_TEXTURE) + nv50_tex_validate(nv50); + + if (nv50->dirty & NV50_NEW_ARRAYS) + nv50_vbo_validate(nv50); + + nv50->state.dirty |= nv50->dirty; + nv50->dirty = 0; + nv50_state_emit(nv50); + + return TRUE; +} + diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c new file mode 100644 index 0000000000..5bf97d3a6b --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_surface.c @@ -0,0 +1,123 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "nv50_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/p_inlines.h" + +#include "util/u_tile.h" + +static void +nv50_surface_copy(struct pipe_context *pipe, boolean flip, + struct pipe_surface *dest, unsigned destx, unsigned desty, + struct pipe_surface *src, unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + struct nv50_context *nv50 = (struct nv50_context *)pipe; + struct nouveau_winsys *nvws = nv50->screen->nvws; + + if (flip) { + desty += height; + while (height--) { + nvws->surface_copy(nvws, dest, destx, desty--, src, + srcx, srcy++, width, 1); + } + } else { + nvws->surface_copy(nvws, dest, destx, desty, src, srcx, srcy, + width, height); + } +} + +static void +nv50_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest, + unsigned destx, unsigned desty, unsigned width, + unsigned height, unsigned value) +{ + struct nv50_context *nv50 = (struct nv50_context *)pipe; + struct nouveau_winsys *nvws = nv50->screen->nvws; + + nvws->surface_fill(nvws, dest, destx, desty, width, height, value); +} + +static void * +nv50_surface_map(struct pipe_screen *screen, struct pipe_surface *ps, + unsigned flags ) +{ + struct nouveau_winsys *nvws = nv50_screen(screen)->nvws; + struct pipe_winsys *ws = screen->winsys; + struct nv50_surface *s = nv50_surface(ps); + struct nv50_surface m = *s; + void *map; + + if (!s->untiled) { + s->untiled = ws->buffer_create(ws, 0, 0, ps->buffer->size); + + m.base.buffer = s->untiled; + nvws->surface_copy(nvws, &m.base, 0, 0, &s->base, 0, 0, + ps->width, ps->height); + } + + /* Map original tiled surface to disallow it being validated while + * untiled mirror is mapped. + */ + ws->buffer_map(ws, ps->buffer, flags); + + map = ws->buffer_map(ws, s->untiled, flags); + if (!map) + return NULL; + + return map; +} + +static void +nv50_surface_unmap(struct pipe_screen *pscreen, struct pipe_surface *ps) +{ + struct nouveau_winsys *nvws = nv50_screen(pscreen)->nvws; + struct pipe_winsys *ws = pscreen->winsys; + struct nv50_surface *s = nv50_surface(ps); + struct nv50_surface m = *s; + + ws->buffer_unmap(ws, s->untiled); + ws->buffer_unmap(ws, ps->buffer); + + m.base.buffer = s->untiled; + nvws->surface_copy(nvws, &s->base, 0, 0, &m.base, 0, 0, + ps->width, ps->height); + + pipe_buffer_reference(pscreen, &s->untiled, NULL); +} + +void +nv50_init_surface_functions(struct nv50_context *nv50) +{ + nv50->pipe.surface_copy = nv50_surface_copy; + nv50->pipe.surface_fill = nv50_surface_fill; +} + +void +nv50_surface_init_screen_functions(struct pipe_screen *pscreen) +{ + pscreen->surface_map = nv50_surface_map; + pscreen->surface_unmap = nv50_surface_unmap; +} + diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c new file mode 100644 index 0000000000..fde3c97c05 --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -0,0 +1,124 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "nv50_context.h" +#include "nv50_texture.h" + +#include "nouveau/nouveau_stateobj.h" + +static int +nv50_tex_construct(struct nouveau_stateobj *so, struct nv50_miptree *mt) +{ + switch (mt->base.format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM | + NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM | + NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_FMT_8_8_8_8); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM | + NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM | + NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_FMT_1_5_5_5); + break; + case PIPE_FORMAT_A4R4G4B4_UNORM: + so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM | + NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM | + NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_FMT_4_4_4_4); + break; + case PIPE_FORMAT_R5G6B5_UNORM: + so_data(so, NV50TIC_0_0_MAPA_ONE | NV50TIC_0_0_TYPEA_UNORM | + NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM | + NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_FMT_5_6_5); + break; + case PIPE_FORMAT_L8_UNORM: + so_data(so, NV50TIC_0_0_MAPA_ONE | NV50TIC_0_0_TYPEA_UNORM | + NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPG_C0 | NV50TIC_0_0_TYPEG_UNORM | + NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_FMT_8); + break; + case PIPE_FORMAT_A8_UNORM: + so_data(so, NV50TIC_0_0_MAPA_C0 | NV50TIC_0_0_TYPEA_UNORM | + NV50TIC_0_0_MAPR_ZERO | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPG_ZERO | NV50TIC_0_0_TYPEG_UNORM | + NV50TIC_0_0_MAPB_ZERO | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_FMT_8); + break; + case PIPE_FORMAT_I8_UNORM: + so_data(so, NV50TIC_0_0_MAPA_C0 | NV50TIC_0_0_TYPEA_UNORM | + NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPG_C0 | NV50TIC_0_0_TYPEG_UNORM | + NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_FMT_8); + break; + case PIPE_FORMAT_A8L8_UNORM: + so_data(so, NV50TIC_0_0_MAPA_C1 | NV50TIC_0_0_TYPEA_UNORM | + NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPG_C0 | NV50TIC_0_0_TYPEG_UNORM | + NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_FMT_8_8); + break; + default: + return 1; + } + + so_reloc(so, mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW, 0, 0); + so_data (so, 0xd0005000); + so_data (so, 0x00300000); + so_data (so, mt->base.width[0]); + so_data (so, (mt->base.depth[0] << 16) | mt->base.height[0]); + so_data (so, 0x03000000); + so_reloc(so, mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH, 0, 0); + + return 0; +} + +void +nv50_tex_validate(struct nv50_context *nv50) +{ + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nouveau_stateobj *so; + int i; + + so = so_new(nv50->miptree_nr * 8 + 3, nv50->miptree_nr * 2); + so_method(so, tesla, 0x0f00, 1); + so_data (so, NV50_CB_TIC); + so_method(so, tesla, 0x40000f04, nv50->miptree_nr * 8); + for (i = 0; i < nv50->miptree_nr; i++) { + if (nv50_tex_construct(so, nv50->miptree[i])) { + NOUVEAU_ERR("failed tex validate\n"); + so_ref(NULL, &so); + return; + } + } + + so_ref(so, &nv50->state.tic_upload); +} + diff --git a/src/gallium/drivers/nv50/nv50_texture.h b/src/gallium/drivers/nv50/nv50_texture.h new file mode 100644 index 0000000000..6861d67b4d --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_texture.h @@ -0,0 +1,126 @@ +#ifndef __NV50_TEXTURE_H__ +#define __NV50_TEXTURE_H__ + +/* It'd be really nice to have these in nouveau_class.h generated by + * renouveau like the rest of the object header - but not sure it can + * handle non-object stuff nicely - need to look into it. + */ + +/* Texture image control block */ +#define NV50TIC_0_0_MAPA_MASK 0x38000000 +#define NV50TIC_0_0_MAPA_ZERO 0x00000000 +#define NV50TIC_0_0_MAPA_C0 0x10000000 +#define NV50TIC_0_0_MAPA_C1 0x18000000 +#define NV50TIC_0_0_MAPA_C2 0x20000000 +#define NV50TIC_0_0_MAPA_C3 0x28000000 +#define NV50TIC_0_0_MAPA_ONE 0x38000000 +#define NV50TIC_0_0_MAPR_MASK 0x07000000 +#define NV50TIC_0_0_MAPR_ZERO 0x00000000 +#define NV50TIC_0_0_MAPR_C0 0x02000000 +#define NV50TIC_0_0_MAPR_C1 0x03000000 +#define NV50TIC_0_0_MAPR_C2 0x04000000 +#define NV50TIC_0_0_MAPR_C3 0x05000000 +#define NV50TIC_0_0_MAPR_ONE 0x07000000 +#define NV50TIC_0_0_MAPG_MASK 0x00e00000 +#define NV50TIC_0_0_MAPG_ZERO 0x00000000 +#define NV50TIC_0_0_MAPG_C0 0x00400000 +#define NV50TIC_0_0_MAPG_C1 0x00600000 +#define NV50TIC_0_0_MAPG_C2 0x00800000 +#define NV50TIC_0_0_MAPG_C3 0x00a00000 +#define NV50TIC_0_0_MAPG_ONE 0x00e00000 +#define NV50TIC_0_0_MAPB_MASK 0x001c0000 +#define NV50TIC_0_0_MAPB_ZERO 0x00000000 +#define NV50TIC_0_0_MAPB_C0 0x00080000 +#define NV50TIC_0_0_MAPB_C1 0x000c0000 +#define NV50TIC_0_0_MAPB_C2 0x00100000 +#define NV50TIC_0_0_MAPB_C3 0x00140000 +#define NV50TIC_0_0_MAPB_ONE 0x001c0000 +#define NV50TIC_0_0_TYPEA_MASK 0x00038000 +#define NV50TIC_0_0_TYPEA_UNORM 0x00010000 +#define NV50TIC_0_0_TYPER_MASK 0x00007000 +#define NV50TIC_0_0_TYPER_UNORM 0x00002000 +#define NV50TIC_0_0_TYPEG_MASK 0x00000e00 +#define NV50TIC_0_0_TYPEG_UNORM 0x00000400 +#define NV50TIC_0_0_TYPEB_MASK 0x000001c0 +#define NV50TIC_0_0_TYPEB_UNORM 0x00000080 +#define NV50TIC_0_0_FMT_MASK 0x0000003c +#define NV50TIC_0_0_FMT_8_8_8_8 0x00000008 +#define NV50TIC_0_0_FMT_4_4_4_4 0x00000012 +#define NV50TIC_0_0_FMT_1_5_5_5 0x00000013 +#define NV50TIC_0_0_FMT_5_6_5 0x00000015 +#define NV50TIC_0_0_FMT_8_8 0x00000018 +#define NV50TIC_0_0_FMT_8 0x0000001d + +#define NV50TIC_0_1_OFFSET_LOW_MASK 0xffffffff +#define NV50TIC_0_1_OFFSET_LOW_SHIFT 0 + +#define NV50TIC_0_2_UNKNOWN_MASK 0xffffffff + +#define NV50TIC_0_3_UNKNOWN_MASK 0xffffffff + +#define NV50TIC_0_4_WIDTH_MASK 0x0000ffff +#define NV50TIC_0_4_WIDTH_SHIFT 0 + +#define NV50TIC_0_5_DEPTH_MASK 0xffff0000 +#define NV50TIC_0_5_DEPTH_SHIFT 16 +#define NV50TIC_0_5_HEIGHT_MASK 0x0000ffff +#define NV50TIC_0_5_HEIGHT_SHIFT 0 + +#define NV50TIC_0_6_UNKNOWN_MASK 0xffffffff + +#define NV50TIC_0_7_OFFSET_HIGH_MASK 0xffffffff +#define NV50TIC_0_7_OFFSET_HIGH_SHIFT 0 + +/* Texture sampler control block */ +#define NV50TSC_1_0_WRAPS_MASK 0x00000007 +#define NV50TSC_1_0_WRAPS_REPEAT 0x00000000 +#define NV50TSC_1_0_WRAPS_MIRROR_REPEAT 0x00000001 +#define NV50TSC_1_0_WRAPS_CLAMP_TO_EDGE 0x00000002 +#define NV50TSC_1_0_WRAPS_CLAMP_TO_BORDER 0x00000003 +#define NV50TSC_1_0_WRAPS_CLAMP 0x00000004 +#define NV50TSC_1_0_WRAPS_MIRROR_CLAMP_TO_EDGE 0x00000005 +#define NV50TSC_1_0_WRAPS_MIRROR_CLAMP_TO_BORDER 0x00000006 +#define NV50TSC_1_0_WRAPS_MIRROR_CLAMP 0x00000007 +#define NV50TSC_1_0_WRAPT_MASK 0x00000038 +#define NV50TSC_1_0_WRAPT_REPEAT 0x00000000 +#define NV50TSC_1_0_WRAPT_MIRROR_REPEAT 0x00000008 +#define NV50TSC_1_0_WRAPT_CLAMP_TO_EDGE 0x00000010 +#define NV50TSC_1_0_WRAPT_CLAMP_TO_BORDER 0x00000018 +#define NV50TSC_1_0_WRAPT_CLAMP 0x00000020 +#define NV50TSC_1_0_WRAPT_MIRROR_CLAMP_TO_EDGE 0x00000028 +#define NV50TSC_1_0_WRAPT_MIRROR_CLAMP_TO_BORDER 0x00000030 +#define NV50TSC_1_0_WRAPT_MIRROR_CLAMP 0x00000038 +#define NV50TSC_1_0_WRAPR_MASK 0x000001c0 +#define NV50TSC_1_0_WRAPR_REPEAT 0x00000000 +#define NV50TSC_1_0_WRAPR_MIRROR_REPEAT 0x00000040 +#define NV50TSC_1_0_WRAPR_CLAMP_TO_EDGE 0x00000080 +#define NV50TSC_1_0_WRAPR_CLAMP_TO_BORDER 0x000000c0 +#define NV50TSC_1_0_WRAPR_CLAMP 0x00000100 +#define NV50TSC_1_0_WRAPR_MIRROR_CLAMP_TO_EDGE 0x00000140 +#define NV50TSC_1_0_WRAPR_MIRROR_CLAMP_TO_BORDER 0x00000180 +#define NV50TSC_1_0_WRAPR_MIRROR_CLAMP 0x000001c0 + +#define NV50TSC_1_1_MAGF_MASK 0x00000003 +#define NV50TSC_1_1_MAGF_NEAREST 0x00000001 +#define NV50TSC_1_1_MAGF_LINEAR 0x00000002 +#define NV50TSC_1_1_MINF_MASK 0x00000030 +#define NV50TSC_1_1_MINF_NEAREST 0x00000010 +#define NV50TSC_1_1_MINF_LINEAR 0x00000020 +#define NV50TSC_1_1_MIPF_MASK 0x000000c0 +#define NV50TSC_1_1_MIPF_NONE 0x00000040 +#define NV50TSC_1_1_MIPF_NEAREST 0x00000080 +#define NV50TSC_1_1_MIPF_LINEAR 0x000000c0 + +#define NV50TSC_1_2_UNKNOWN_MASK 0xffffffff + +#define NV50TSC_1_3_UNKNOWN_MASK 0xffffffff + +#define NV50TSC_1_4_UNKNOWN_MASK 0xffffffff + +#define NV50TSC_1_5_UNKNOWN_MASK 0xffffffff + +#define NV50TSC_1_6_UNKNOWN_MASK 0xffffffff + +#define NV50TSC_1_7_UNKNOWN_MASK 0xffffffff + +#endif diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c new file mode 100644 index 0000000000..584336682e --- /dev/null +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -0,0 +1,237 @@ +/* + * Copyright 2008 Ben Skeggs + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +#include "nv50_context.h" + +static INLINE unsigned +nv50_prim(unsigned mode) +{ + switch (mode) { + case PIPE_PRIM_POINTS: return NV50TCL_VERTEX_BEGIN_POINTS; + case PIPE_PRIM_LINES: return NV50TCL_VERTEX_BEGIN_LINES; + case PIPE_PRIM_LINE_LOOP: return NV50TCL_VERTEX_BEGIN_LINE_LOOP; + case PIPE_PRIM_LINE_STRIP: return NV50TCL_VERTEX_BEGIN_LINE_STRIP; + case PIPE_PRIM_TRIANGLES: return NV50TCL_VERTEX_BEGIN_TRIANGLES; + case PIPE_PRIM_TRIANGLE_STRIP: + return NV50TCL_VERTEX_BEGIN_TRIANGLE_STRIP; + case PIPE_PRIM_TRIANGLE_FAN: return NV50TCL_VERTEX_BEGIN_TRIANGLE_FAN; + case PIPE_PRIM_QUADS: return NV50TCL_VERTEX_BEGIN_QUADS; + case PIPE_PRIM_QUAD_STRIP: return NV50TCL_VERTEX_BEGIN_QUAD_STRIP; + case PIPE_PRIM_POLYGON: return NV50TCL_VERTEX_BEGIN_POLYGON; + default: + break; + } + + NOUVEAU_ERR("invalid primitive type %d\n", mode); + return NV50TCL_VERTEX_BEGIN_POINTS; +} + +boolean +nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, + unsigned count) +{ + struct nv50_context *nv50 = nv50_context(pipe); + + nv50_state_validate(nv50); + + BEGIN_RING(tesla, 0x142c, 1); + OUT_RING (0); + BEGIN_RING(tesla, 0x142c, 1); + OUT_RING (0); + + BEGIN_RING(tesla, NV50TCL_VERTEX_BEGIN, 1); + OUT_RING (nv50_prim(mode)); + BEGIN_RING(tesla, NV50TCL_VERTEX_BUFFER_FIRST, 2); + OUT_RING (start); + OUT_RING (count); + BEGIN_RING(tesla, NV50TCL_VERTEX_END, 1); + OUT_RING (0); + + pipe->flush(pipe, 0, NULL); + return TRUE; +} + +static INLINE void +nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map, + unsigned start, unsigned count) +{ + map += start; + + if (count & 1) { + BEGIN_RING(tesla, 0x15e8, 1); + OUT_RING (map[0]); + map++; + count--; + } + + while (count) { + unsigned nr = count > 2046 ? 2046 : count; + int i; + + BEGIN_RING(tesla, 0x400015f0, nr >> 1); + for (i = 0; i < nr; i += 2) + OUT_RING ((map[1] << 16) | map[0]); + + count -= nr; + map += nr; + } +} + +static INLINE void +nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map, + unsigned start, unsigned count) +{ + map += start; + + if (count & 1) { + BEGIN_RING(tesla, 0x15e8, 1); + OUT_RING (map[0]); + map++; + count--; + } + + while (count) { + unsigned nr = count > 2046 ? 2046 : count; + int i; + + BEGIN_RING(tesla, 0x400015f0, nr >> 1); + for (i = 0; i < nr; i += 2) + OUT_RING ((map[1] << 16) | map[0]); + + count -= nr; + map += nr; + } +} + +static INLINE void +nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint8_t *map, + unsigned start, unsigned count) +{ + map += start; + + while (count) { + unsigned nr = count > 2047 ? 2047 : count; + + BEGIN_RING(tesla, 0x400015e8, nr); + OUT_RINGp (map, nr); + + count -= nr; + map += nr; + } +} + +boolean +nv50_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, unsigned indexSize, + unsigned mode, unsigned start, unsigned count) +{ + struct nv50_context *nv50 = nv50_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + void *map = ws->buffer_map(ws, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ); + + nv50_state_validate(nv50); + + BEGIN_RING(tesla, 0x142c, 1); + OUT_RING (0); + BEGIN_RING(tesla, 0x142c, 1); + OUT_RING (0); + + BEGIN_RING(tesla, NV50TCL_VERTEX_BEGIN, 1); + OUT_RING (nv50_prim(mode)); + switch (indexSize) { + case 1: + nv50_draw_elements_inline_u08(nv50, map, start, count); + break; + case 2: + nv50_draw_elements_inline_u16(nv50, map, start, count); + break; + case 4: + nv50_draw_elements_inline_u32(nv50, map, start, count); + break; + default: + assert(0); + } + BEGIN_RING(tesla, NV50TCL_VERTEX_END, 1); + OUT_RING (0); + + pipe->flush(pipe, 0, NULL); + return TRUE; +} + +void +nv50_vbo_validate(struct nv50_context *nv50) +{ + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nouveau_stateobj *vtxbuf, *vtxfmt; + int i, vpi = 0; + + vtxbuf = so_new(nv50->vtxelt_nr * 4, nv50->vtxelt_nr * 2); + vtxfmt = so_new(nv50->vtxelt_nr + 1, 0); + so_method(vtxfmt, tesla, 0x1ac0, nv50->vtxelt_nr); + + for (i = 0; i < nv50->vtxelt_nr; i++) { + struct pipe_vertex_element *ve = &nv50->vtxelt[i]; + struct pipe_vertex_buffer *vb = + &nv50->vtxbuf[ve->vertex_buffer_index]; + + switch (ve->src_format) { + case PIPE_FORMAT_R32G32B32A32_FLOAT: + so_data(vtxfmt, 0x7e080000 | i); + break; + case PIPE_FORMAT_R32G32B32_FLOAT: + so_data(vtxfmt, 0x7e100000 | i); + break; + case PIPE_FORMAT_R32G32_FLOAT: + so_data(vtxfmt, 0x7e200000 | i); + break; + case PIPE_FORMAT_R32_FLOAT: + so_data(vtxfmt, 0x7e900000 | i); + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + so_data(vtxfmt, 0x24500000 | i); + break; + default: + { + NOUVEAU_ERR("invalid vbo format %s\n", + pf_name(ve->src_format)); + assert(0); + return; + } + } + + so_method(vtxbuf, tesla, 0x900 + (i * 16), 3); + so_data (vtxbuf, 0x20000000 | vb->pitch); + so_reloc (vtxbuf, vb->buffer, vb->buffer_offset + + ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | + NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (vtxbuf, vb->buffer, vb->buffer_offset + + ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | + NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); + } + + so_ref (vtxfmt, &nv50->state.vtxfmt); + so_ref (vtxbuf, &nv50->state.vtxbuf); +} + diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index cd1e6663d8..800f944838 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -2,6 +2,7 @@ * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. + * Copyright 2008 VMware, Inc. All rights reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the @@ -221,6 +222,24 @@ softpipe_create( struct pipe_screen *screen, softpipe->quad[i].output = sp_quad_output_stage(softpipe); } + /* vertex shader samplers */ + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + softpipe->tgsi.vert_samplers[i].base.get_samples = sp_get_samples_vertex; + softpipe->tgsi.vert_samplers[i].unit = i; + softpipe->tgsi.vert_samplers[i].sp = softpipe; + softpipe->tgsi.vert_samplers[i].cache = softpipe->tex_cache[i]; + softpipe->tgsi.vert_samplers_list[i] = &softpipe->tgsi.vert_samplers[i]; + } + + /* fragment shader samplers */ + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + softpipe->tgsi.frag_samplers[i].base.get_samples = sp_get_samples_fragment; + softpipe->tgsi.frag_samplers[i].unit = i; + softpipe->tgsi.frag_samplers[i].sp = softpipe; + softpipe->tgsi.frag_samplers[i].cache = softpipe->tex_cache[i]; + softpipe->tgsi.frag_samplers_list[i] = &softpipe->tgsi.frag_samplers[i]; + } + /* * Create drawing context and plug our rendering stage into it. */ @@ -228,6 +247,11 @@ softpipe_create( struct pipe_screen *screen, if (!softpipe->draw) goto fail; + draw_texture_samplers(softpipe->draw, + PIPE_MAX_SAMPLERS, + (struct tgsi_sampler **) + softpipe->tgsi.vert_samplers_list); + softpipe->setup = sp_draw_render_stage(softpipe); if (!softpipe->setup) goto fail; diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index 2b9a2a8ee5..7ab12a6d70 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -37,6 +37,7 @@ #include "draw/draw_vertex.h" #include "sp_quad.h" +#include "sp_tex_sample.h" /** @@ -139,6 +140,14 @@ struct softpipe_context { struct quad_stage *first; /**< points to one of the above stages */ } quad[SP_NUM_QUAD_THREADS]; + /** TGSI exec things */ + struct { + struct sp_shader_sampler vert_samplers[PIPE_MAX_SAMPLERS]; + struct sp_shader_sampler *vert_samplers_list[PIPE_MAX_SAMPLERS]; + struct sp_shader_sampler frag_samplers[PIPE_MAX_SAMPLERS]; + struct sp_shader_sampler *frag_samplers_list[PIPE_MAX_SAMPLERS]; + } tgsi; + /** The primitive drawing context */ struct draw_context *draw; struct draw_stage *setup; diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c index 701ee4c72f..453b0373f0 100644 --- a/src/gallium/drivers/softpipe/sp_fs_exec.c +++ b/src/gallium/drivers/softpipe/sp_fs_exec.c @@ -39,11 +39,19 @@ #include "tgsi/tgsi_exec.h" #include "tgsi/tgsi_parse.h" -struct sp_exec_fragment_shader { +struct sp_exec_fragment_shader +{ struct sp_fragment_shader base; }; +/** cast wrapper */ +static INLINE struct sp_exec_fragment_shader * +sp_exec_fragment_shader(const struct sp_fragment_shader *base) +{ + return (struct sp_exec_fragment_shader *) base; +} + /** * Compute quad X,Y,Z,W for the four fragments in a quad. @@ -84,12 +92,18 @@ sp_setup_pos_vector(const struct tgsi_interp_coef *coef, static void exec_prepare( const struct sp_fragment_shader *base, struct tgsi_exec_machine *machine, - struct tgsi_sampler *samplers ) + struct tgsi_sampler **samplers ) { - tgsi_exec_machine_bind_shader( machine, - base->shader.tokens, - PIPE_MAX_SAMPLERS, - samplers ); + /* + * Bind tokens/shader to the interpreter's machine state. + * Avoid redundant binding. + */ + if (machine->Tokens != base->shader.tokens) { + tgsi_exec_machine_bind_shader( machine, + base->shader.tokens, + PIPE_MAX_SAMPLERS, + samplers ); + } } diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 50eb2c07bc..9a273c8764 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -40,7 +40,7 @@ #include "tgsi/tgsi_sse2.h" -#ifdef PIPE_ARCH_X86 +#if defined(PIPE_ARCH_X86) #include "rtasm/rtasm_x86sse.h" @@ -69,7 +69,7 @@ struct sp_sse_fragment_shader { static void fs_sse_prepare( const struct sp_fragment_shader *base, struct tgsi_exec_machine *machine, - struct tgsi_sampler *samplers ) + struct tgsi_sampler **samplers ) { } diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 425e13cd28..9cd5784e5b 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -26,10 +26,10 @@ **************************************************************************/ /** - * Post-transform vertex buffering. This is an optional part of the - * softpipe rendering pipeline. - * Probably not desired in general, but useful for testing/debuggin. - * Enabled/Disabled with SP_VBUF env var. + * Interface between 'draw' module's output and the softpipe rasterizer/setup + * code. When the 'draw' module has finished filling a vertex buffer, the + * draw_arrays() functions below will be called. Loop over the vertices and + * call the point/line/tri setup functions. * * Authors * Brian Paul @@ -131,21 +131,23 @@ static INLINE cptrf4 get_vert( const void *vertex_buffer, } +/** + * draw elements / indexed primitives + */ static void sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; - unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float); - unsigned i; + const unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float); const void *vertex_buffer = cvbr->vertex_buffer; + unsigned i; /* XXX: break this dependency - make setup_context live under * softpipe, rename the old "setup" draw stage to something else. */ struct draw_stage *setup = softpipe->setup; - struct setup_context *setup_ctx = sp_draw_setup_context(softpipe->setup); - + struct setup_context *setup_ctx = sp_draw_setup_context(setup); switch (cvbr->prim) { case PIPE_PRIM_POINTS: @@ -258,13 +260,16 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; - struct draw_stage *setup = softpipe->setup; - const void *vertex_buffer = NULL; const unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float); + const void *vertex_buffer = + (void *) get_vert(cvbr->vertex_buffer, start, stride); unsigned i; - struct setup_context *setup_ctx = sp_draw_setup_context(setup); - vertex_buffer = (void *)get_vert(cvbr->vertex_buffer, start, stride); + /* XXX: break this dependency - make setup_context live under + * softpipe, rename the old "setup" draw stage to something else. + */ + struct draw_stage *setup = softpipe->setup; + struct setup_context *setup_ctx = sp_draw_setup_context(setup); switch (cvbr->prim) { case PIPE_PRIM_POINTS: diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 1f0cb3e035..40329a9562 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -2,6 +2,7 @@ * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. + * Copyright 2008 VMware, Inc. All rights reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the @@ -50,8 +51,7 @@ struct quad_shade_stage { - struct quad_stage stage; - struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; + struct quad_stage stage; /**< base class */ struct tgsi_exec_machine machine; struct tgsi_exec_vector *inputs, *outputs; }; @@ -147,18 +147,11 @@ static void shade_begin(struct quad_stage *qs) { struct quad_shade_stage *qss = quad_shade_stage(qs); struct softpipe_context *softpipe = qs->softpipe; - unsigned i; - unsigned num = MAX2(softpipe->num_textures, softpipe->num_samplers); - - /* set TGSI sampler state that varies */ - for (i = 0; i < num; i++) { - qss->samplers[i].state = softpipe->sampler[i]; - qss->samplers[i].texture = softpipe->texture[i]; - } softpipe->fs->prepare( softpipe->fs, &qss->machine, - qss->samplers ); + (struct tgsi_sampler **) + softpipe->tgsi.frag_samplers_list ); qs->next->begin(qs->next); } @@ -191,14 +184,6 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) qss->stage.run = shade_quad; qss->stage.destroy = shade_destroy; - /* set TGSI sampler state that's constant */ - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - assert(softpipe->tex_cache[i]); - qss->samplers[i].get_samples = sp_get_samples; - qss->samplers[i].pipe = &softpipe->pipe; - qss->samplers[i].cache = softpipe->tex_cache[i]; - } - tgsi_exec_machine_init( &qss->machine ); return &qss->stage; diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index 9644dbd168..12f98c32f5 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -56,6 +56,8 @@ softpipe_get_param(struct pipe_screen *screen, int param) switch (param) { case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: return 8; + case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: + return 8; case PIPE_CAP_NPOT_TEXTURES: return 1; case PIPE_CAP_TWO_SIDED_STENCIL: diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 476ef3dc8f..3eff41ffa5 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -69,7 +69,7 @@ struct sp_fragment_shader { void (*prepare)( const struct sp_fragment_shader *shader, struct tgsi_exec_machine *machine, - struct tgsi_sampler *samplers); + struct tgsi_sampler **samplers); /* Run the shader - this interface will get cleaned up in the * future: diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 49250ec084..32aa5025e4 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -2,6 +2,7 @@ * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. + * Copyright 2008 VMware, Inc. All rights reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the @@ -35,15 +36,16 @@ #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" +#include "sp_texture.h" #include "sp_tex_sample.h" #include "sp_tile_cache.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "tgsi/tgsi_exec.h" #include "util/u_math.h" #include "util/u_memory.h" + /* * Note, the FRAC macro has to work perfectly. Otherwise you'll sometimes * see 1-pixel bands of improperly weighted linear-filtered textures. @@ -57,7 +59,11 @@ /** * Linear interpolation macro */ -#define LERP(T, A, B) ( (A) + (T) * ((B) - (A)) ) +static INLINE float +lerp(float a, float v0, float v1) +{ + return v0 + a * (v1 - v0); +} /** @@ -72,13 +78,28 @@ static INLINE float lerp_2d(float a, float b, float v00, float v10, float v01, float v11) { - const float temp0 = LERP(a, v00, v10); - const float temp1 = LERP(a, v01, v11); - return LERP(b, temp0, temp1); + const float temp0 = lerp(a, v00, v10); + const float temp1 = lerp(a, v01, v11); + return lerp(b, temp0, temp1); } /** + * As above, but 3D interpolation of 8 values. + */ +static INLINE float +lerp_3d(float a, float b, float c, + float v000, float v100, float v010, float v110, + float v001, float v101, float v011, float v111) +{ + const float temp0 = lerp_2d(a, b, v000, v100, v010, v110); + const float temp1 = lerp_2d(a, b, v001, v101, v011, v111); + return lerp(c, temp0, temp1); +} + + + +/** * If A is a signed integer, A % B doesn't give the right value for A < 0 * (in terms of texture repeat). Just casting to unsigned fixes that. */ @@ -86,250 +107,275 @@ lerp_2d(float a, float b, /** - * Apply texture coord wrapping mode and return integer texture index. + * Apply texture coord wrapping mode and return integer texture indexes + * for a vector of four texcoords (S or T or P). * \param wrapMode PIPE_TEX_WRAP_x - * \param s the texcoord + * \param s the incoming texcoords * \param size the texture image size + * \param icoord returns the integer texcoords * \return integer texture index */ -static INLINE int -nearest_texcoord(unsigned wrapMode, float s, unsigned size) +static INLINE void +nearest_texcoord_4(unsigned wrapMode, const float s[4], unsigned size, + int icoord[4]) { - int i; + uint ch; switch (wrapMode) { case PIPE_TEX_WRAP_REPEAT: /* s limited to [0,1) */ /* i limited to [0,size-1] */ - i = util_ifloor(s * size); - i = REMAINDER(i, size); - return i; + for (ch = 0; ch < 4; ch++) { + int i = util_ifloor(s[ch] * size); + icoord[ch] = REMAINDER(i, size); + } + return; case PIPE_TEX_WRAP_CLAMP: /* s limited to [0,1] */ /* i limited to [0,size-1] */ - if (s <= 0.0F) - i = 0; - else if (s >= 1.0F) - i = size - 1; - else - i = util_ifloor(s * size); - return i; + for (ch = 0; ch < 4; ch++) { + if (s[ch] <= 0.0F) + icoord[ch] = 0; + else if (s[ch] >= 1.0F) + icoord[ch] = size - 1; + else + icoord[ch] = util_ifloor(s[ch] * size); + } + return; case PIPE_TEX_WRAP_CLAMP_TO_EDGE: { /* s limited to [min,max] */ /* i limited to [0, size-1] */ const float min = 1.0F / (2.0F * size); const float max = 1.0F - min; - if (s < min) - i = 0; - else if (s > max) - i = size - 1; - else - i = util_ifloor(s * size); + for (ch = 0; ch < 4; ch++) { + if (s[ch] < min) + icoord[ch] = 0; + else if (s[ch] > max) + icoord[ch] = size - 1; + else + icoord[ch] = util_ifloor(s[ch] * size); + } } - return i; + return; case PIPE_TEX_WRAP_CLAMP_TO_BORDER: { /* s limited to [min,max] */ /* i limited to [-1, size] */ const float min = -1.0F / (2.0F * size); const float max = 1.0F - min; - if (s <= min) - i = -1; - else if (s >= max) - i = size; - else - i = util_ifloor(s * size); + for (ch = 0; ch < 4; ch++) { + if (s[ch] <= min) + icoord[ch] = -1; + else if (s[ch] >= max) + icoord[ch] = size; + else + icoord[ch] = util_ifloor(s[ch] * size); + } } - return i; + return; case PIPE_TEX_WRAP_MIRROR_REPEAT: { const float min = 1.0F / (2.0F * size); const float max = 1.0F - min; - const int flr = util_ifloor(s); - float u; - if (flr & 1) - u = 1.0F - (s - (float) flr); - else - u = s - (float) flr; - if (u < min) - i = 0; - else if (u > max) - i = size - 1; - else - i = util_ifloor(u * size); + for (ch = 0; ch < 4; ch++) { + const int flr = util_ifloor(s[ch]); + float u; + if (flr & 1) + u = 1.0F - (s[ch] - (float) flr); + else + u = s[ch] - (float) flr; + if (u < min) + icoord[ch] = 0; + else if (u > max) + icoord[ch] = size - 1; + else + icoord[ch] = util_ifloor(u * size); + } } - return i; + return; case PIPE_TEX_WRAP_MIRROR_CLAMP: - { + for (ch = 0; ch < 4; ch++) { /* s limited to [0,1] */ /* i limited to [0,size-1] */ - const float u = fabsf(s); + const float u = fabsf(s[ch]); if (u <= 0.0F) - i = 0; + icoord[ch] = 0; else if (u >= 1.0F) - i = size - 1; + icoord[ch] = size - 1; else - i = util_ifloor(u * size); + icoord[ch] = util_ifloor(u * size); } - return i; + return; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: { /* s limited to [min,max] */ /* i limited to [0, size-1] */ const float min = 1.0F / (2.0F * size); const float max = 1.0F - min; - const float u = fabsf(s); - if (u < min) - i = 0; - else if (u > max) - i = size - 1; - else - i = util_ifloor(u * size); + for (ch = 0; ch < 4; ch++) { + const float u = fabsf(s[ch]); + if (u < min) + icoord[ch] = 0; + else if (u > max) + icoord[ch] = size - 1; + else + icoord[ch] = util_ifloor(u * size); + } } - return i; + return; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: { /* s limited to [min,max] */ /* i limited to [0, size-1] */ const float min = -1.0F / (2.0F * size); const float max = 1.0F - min; - const float u = fabsf(s); - if (u < min) - i = -1; - else if (u > max) - i = size; - else - i = util_ifloor(u * size); + for (ch = 0; ch < 4; ch++) { + const float u = fabsf(s[ch]); + if (u < min) + icoord[ch] = -1; + else if (u > max) + icoord[ch] = size; + else + icoord[ch] = util_ifloor(u * size); + } } - return i; + return; default: assert(0); - return 0; } } /** - * Used to compute texel locations for linear sampling. + * Used to compute texel locations for linear sampling for four texcoords. * \param wrapMode PIPE_TEX_WRAP_x - * \param s the texcoord + * \param s the texcoords * \param size the texture image size - * \param i0 returns first texture index - * \param i1 returns second texture index (usually *i0 + 1) - * \param a returns blend factor/weight between texture indexes + * \param icoord0 returns first texture indexes + * \param icoord1 returns second texture indexes (usually icoord0 + 1) + * \param w returns blend factor/weight between texture indexes + * \param icoord returns the computed integer texture coords */ static INLINE void -linear_texcoord(unsigned wrapMode, float s, unsigned size, - int *i0, int *i1, float *a) +linear_texcoord_4(unsigned wrapMode, const float s[4], unsigned size, + int icoord0[4], int icoord1[4], float w[4]) { - float u; + uint ch; + switch (wrapMode) { case PIPE_TEX_WRAP_REPEAT: - u = s * size - 0.5F; - *i0 = REMAINDER(util_ifloor(u), size); - *i1 = REMAINDER(*i0 + 1, size); - break; + for (ch = 0; ch < 4; ch++) { + float u = s[ch] * size - 0.5F; + icoord0[ch] = REMAINDER(util_ifloor(u), size); + icoord1[ch] = REMAINDER(icoord0[ch] + 1, size); + w[ch] = FRAC(u); + } + break;; case PIPE_TEX_WRAP_CLAMP: - if (s <= 0.0F) - u = 0.0F; - else if (s >= 1.0F) - u = (float) size; - else - u = s * size; - u -= 0.5F; - *i0 = util_ifloor(u); - *i1 = *i0 + 1; - break; + for (ch = 0; ch < 4; ch++) { + float u = CLAMP(s[ch], 0.0F, 1.0F); + u = u * size - 0.5f; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + w[ch] = FRAC(u); + } + break;; case PIPE_TEX_WRAP_CLAMP_TO_EDGE: - if (s <= 0.0F) - u = 0.0F; - else if (s >= 1.0F) - u = (float) size; - else - u = s * size; - u -= 0.5F; - *i0 = util_ifloor(u); - *i1 = *i0 + 1; - if (*i0 < 0) - *i0 = 0; - if (*i1 >= (int) size) - *i1 = size - 1; - break; + for (ch = 0; ch < 4; ch++) { + float u = CLAMP(s[ch], 0.0F, 1.0F); + u = u * size - 0.5f; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + if (icoord0[ch] < 0) + icoord0[ch] = 0; + if (icoord1[ch] >= (int) size) + icoord1[ch] = size - 1; + w[ch] = FRAC(u); + } + break;; case PIPE_TEX_WRAP_CLAMP_TO_BORDER: { const float min = -1.0F / (2.0F * size); const float max = 1.0F - min; - if (s <= min) - u = min * size; - else if (s >= max) - u = max * size; - else - u = s * size; - u -= 0.5F; - *i0 = util_ifloor(u); - *i1 = *i0 + 1; + for (ch = 0; ch < 4; ch++) { + float u = CLAMP(s[ch], min, max); + u = u * size - 0.5f; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + w[ch] = FRAC(u); + } } - break; + break;; case PIPE_TEX_WRAP_MIRROR_REPEAT: - { - const int flr = util_ifloor(s); + for (ch = 0; ch < 4; ch++) { + const int flr = util_ifloor(s[ch]); + float u; if (flr & 1) - u = 1.0F - (s - (float) flr); + u = 1.0F - (s[ch] - (float) flr); else - u = s - (float) flr; - u = (u * size) - 0.5F; - *i0 = util_ifloor(u); - *i1 = *i0 + 1; - if (*i0 < 0) - *i0 = 0; - if (*i1 >= (int) size) - *i1 = size - 1; + u = s[ch] - (float) flr; + u = u * size - 0.5F; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + if (icoord0[ch] < 0) + icoord0[ch] = 0; + if (icoord1[ch] >= (int) size) + icoord1[ch] = size - 1; + w[ch] = FRAC(u); } - break; + break;; case PIPE_TEX_WRAP_MIRROR_CLAMP: - u = fabsf(s); - if (u >= 1.0F) - u = (float) size; - else - u *= size; - u -= 0.5F; - *i0 = util_ifloor(u); - *i1 = *i0 + 1; - break; + for (ch = 0; ch < 4; ch++) { + float u = fabsf(s[ch]); + if (u >= 1.0F) + u = (float) size; + else + u *= size; + u -= 0.5F; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + w[ch] = FRAC(u); + } + break;; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: - u = fabsf(s); - if (u >= 1.0F) - u = (float) size; - else - u *= size; - u -= 0.5F; - *i0 = util_ifloor(u); - *i1 = *i0 + 1; - if (*i0 < 0) - *i0 = 0; - if (*i1 >= (int) size) - *i1 = size - 1; - break; + for (ch = 0; ch < 4; ch++) { + float u = fabsf(s[ch]); + if (u >= 1.0F) + u = (float) size; + else + u *= size; + u -= 0.5F; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + if (icoord0[ch] < 0) + icoord0[ch] = 0; + if (icoord1[ch] >= (int) size) + icoord1[ch] = size - 1; + w[ch] = FRAC(u); + } + break;; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: { const float min = -1.0F / (2.0F * size); const float max = 1.0F - min; - u = fabsf(s); - if (u <= min) - u = min * size; - else if (u >= max) - u = max * size; - else - u *= size; - u -= 0.5F; - *i0 = util_ifloor(u); - *i1 = *i0 + 1; + for (ch = 0; ch < 4; ch++) { + float u = fabsf(s[ch]); + if (u <= min) + u = min * size; + else if (u >= max) + u = max * size; + else + u *= size; + u -= 0.5F; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + w[ch] = FRAC(u); + } } - break; + break;; default: assert(0); } - *a = FRAC(u); } @@ -337,21 +383,27 @@ linear_texcoord(unsigned wrapMode, float s, unsigned size, * For RECT textures / unnormalized texcoords * Only a subset of wrap modes supported. */ -static INLINE int -nearest_texcoord_unnorm(unsigned wrapMode, float s, unsigned size) +static INLINE void +nearest_texcoord_unnorm_4(unsigned wrapMode, const float s[4], unsigned size, + int icoord[4]) { - int i; + uint ch; switch (wrapMode) { case PIPE_TEX_WRAP_CLAMP: - i = util_ifloor(s); - return CLAMP(i, 0, (int) size-1); + for (ch = 0; ch < 4; ch++) { + int i = util_ifloor(s[ch]); + icoord[ch]= CLAMP(i, 0, (int) size-1); + } + return; case PIPE_TEX_WRAP_CLAMP_TO_EDGE: /* fall-through */ case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - return util_ifloor( CLAMP(s, 0.5F, (float) size - 0.5F) ); + for (ch = 0; ch < 4; ch++) { + icoord[ch]= util_ifloor( CLAMP(s[ch], 0.5F, (float) size - 0.5F) ); + } + return; default: assert(0); - return 0; } } @@ -361,30 +413,36 @@ nearest_texcoord_unnorm(unsigned wrapMode, float s, unsigned size) * Only a subset of wrap modes supported. */ static INLINE void -linear_texcoord_unnorm(unsigned wrapMode, float s, unsigned size, - int *i0, int *i1, float *a) +linear_texcoord_unnorm_4(unsigned wrapMode, const float s[4], unsigned size, + int icoord0[4], int icoord1[4], float w[4]) { + uint ch; switch (wrapMode) { case PIPE_TEX_WRAP_CLAMP: - /* Not exactly what the spec says, but it matches NVIDIA output */ - s = CLAMP(s - 0.5F, 0.0f, (float) size - 1.0f); - *i0 = util_ifloor(s); - *i1 = *i0 + 1; - break; + for (ch = 0; ch < 4; ch++) { + /* Not exactly what the spec says, but it matches NVIDIA output */ + float u = CLAMP(s[ch] - 0.5F, 0.0f, (float) size - 1.0f); + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + w[ch] = FRAC(u); + } + return; case PIPE_TEX_WRAP_CLAMP_TO_EDGE: /* fall-through */ case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - s = CLAMP(s, 0.5F, (float) size - 0.5F); - s -= 0.5F; - *i0 = util_ifloor(s); - *i1 = *i0 + 1; - if (*i1 > (int) size - 1) - *i1 = size - 1; + for (ch = 0; ch < 4; ch++) { + float u = CLAMP(s[ch], 0.5F, (float) size - 0.5F); + u -= 0.5F; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + if (icoord1[ch] > (int) size - 1) + icoord1[ch] = size - 1; + w[ch] = FRAC(u); + } break; default: assert(0); } - *a = FRAC(s); } @@ -463,7 +521,8 @@ choose_cube_face(float rx, float ry, float rz, float *newS, float *newT) * This is only done for fragment shaders, not vertex shaders. */ static float -compute_lambda(struct tgsi_sampler *sampler, +compute_lambda(const struct pipe_texture *tex, + const struct pipe_sampler_state *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], @@ -471,7 +530,7 @@ compute_lambda(struct tgsi_sampler *sampler, { float rho, lambda; - assert(sampler->state->normalized_coords); + assert(sampler->normalized_coords); assert(s); { @@ -479,7 +538,7 @@ compute_lambda(struct tgsi_sampler *sampler, float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]; dsdx = fabsf(dsdx); dsdy = fabsf(dsdy); - rho = MAX2(dsdx, dsdy) * sampler->texture->width[0]; + rho = MAX2(dsdx, dsdy) * tex->width[0]; } if (t) { float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]; @@ -487,7 +546,7 @@ compute_lambda(struct tgsi_sampler *sampler, float max; dtdx = fabsf(dtdx); dtdy = fabsf(dtdy); - max = MAX2(dtdx, dtdy) * sampler->texture->height[0]; + max = MAX2(dtdx, dtdy) * tex->height[0]; rho = MAX2(rho, max); } if (p) { @@ -496,13 +555,13 @@ compute_lambda(struct tgsi_sampler *sampler, float max; dpdx = fabsf(dpdx); dpdy = fabsf(dpdy); - max = MAX2(dpdx, dpdy) * sampler->texture->depth[0]; + max = MAX2(dpdx, dpdy) * tex->depth[0]; rho = MAX2(rho, max); } lambda = util_fast_log2(rho); - lambda += lodbias + sampler->state->lod_bias; - lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); + lambda += lodbias + sampler->lod_bias; + lambda = CLAMP(lambda, sampler->min_lod, sampler->max_lod); return lambda; } @@ -514,68 +573,74 @@ compute_lambda(struct tgsi_sampler *sampler, * 2. Determine if we're minifying or magnifying * 3. If minifying, choose mipmap levels * 4. Return image filter to use within mipmap images + * \param level0 Returns first mipmap level to sample from + * \param level1 Returns second mipmap level to sample from + * \param levelBlend Returns blend factor between levels, in [0,1] + * \param imgFilter Returns either the min or mag filter, depending on lambda */ static void -choose_mipmap_levels(struct tgsi_sampler *sampler, +choose_mipmap_levels(const struct pipe_texture *texture, + const struct pipe_sampler_state *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + boolean computeLambda, float lodbias, unsigned *level0, unsigned *level1, float *levelBlend, unsigned *imgFilter) { - if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { + if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { /* no mipmap selection needed */ - *level0 = *level1 = CLAMP((int) sampler->state->min_lod, - 0, (int) sampler->texture->last_level); + *level0 = *level1 = CLAMP((int) sampler->min_lod, + 0, (int) texture->last_level); - if (sampler->state->min_img_filter != sampler->state->mag_img_filter) { + if (sampler->min_img_filter != sampler->mag_img_filter) { /* non-mipmapped texture, but still need to determine if doing * minification or magnification. */ - float lambda = compute_lambda(sampler, s, t, p, lodbias); + float lambda = compute_lambda(texture, sampler, s, t, p, lodbias); if (lambda <= 0.0) { - *imgFilter = sampler->state->mag_img_filter; + *imgFilter = sampler->mag_img_filter; } else { - *imgFilter = sampler->state->min_img_filter; + *imgFilter = sampler->min_img_filter; } } else { - *imgFilter = sampler->state->mag_img_filter; + *imgFilter = sampler->mag_img_filter; } } else { float lambda; - if (1) + if (computeLambda) /* fragment shader */ - lambda = compute_lambda(sampler, s, t, p, lodbias); + lambda = compute_lambda(texture, sampler, s, t, p, lodbias); else /* vertex shader */ lambda = lodbias; /* not really a bias, but absolute LOD */ if (lambda <= 0.0) { /* XXX threshold depends on the filter */ /* magnifying */ - *imgFilter = sampler->state->mag_img_filter; + *imgFilter = sampler->mag_img_filter; *level0 = *level1 = 0; } else { /* minifying */ - *imgFilter = sampler->state->min_img_filter; + *imgFilter = sampler->min_img_filter; /* choose mipmap level(s) and compute the blend factor between them */ - if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NEAREST) { + if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NEAREST) { /* Nearest mipmap level */ const int lvl = (int) (lambda + 0.5); *level0 = - *level1 = CLAMP(lvl, 0, (int) sampler->texture->last_level); + *level1 = CLAMP(lvl, 0, (int) texture->last_level); } else { /* Linear interpolation between mipmap levels */ const int lvl = (int) lambda; - *level0 = CLAMP(lvl, 0, (int) sampler->texture->last_level); - *level1 = CLAMP(lvl + 1, 0, (int) sampler->texture->last_level); + *level0 = CLAMP(lvl, 0, (int) texture->last_level); + *level1 = CLAMP(lvl + 1, 0, (int) texture->last_level); *levelBlend = FRAC(lambda); /* blending weight between levels */ } } @@ -598,23 +663,29 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, * sp_get_cached_tile_tex() function. Also, get 4 texels instead of 1... */ static void -get_texel(struct tgsi_sampler *sampler, +get_texel(const struct tgsi_sampler *tgsi_sampler, unsigned face, unsigned level, int x, int y, int z, float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j) { - if (x < 0 || x >= (int) sampler->texture->width[level] || - y < 0 || y >= (int) sampler->texture->height[level] || - z < 0 || z >= (int) sampler->texture->depth[level]) { - rgba[0][j] = sampler->state->border_color[0]; - rgba[1][j] = sampler->state->border_color[1]; - rgba[2][j] = sampler->state->border_color[2]; - rgba[3][j] = sampler->state->border_color[3]; + const struct sp_shader_sampler *samp = sp_shader_sampler(tgsi_sampler); + struct softpipe_context *sp = samp->sp; + const uint unit = samp->unit; + const struct pipe_texture *texture = sp->texture[unit]; + const struct pipe_sampler_state *sampler = sp->sampler[unit]; + + if (x < 0 || x >= (int) texture->width[level] || + y < 0 || y >= (int) texture->height[level] || + z < 0 || z >= (int) texture->depth[level]) { + rgba[0][j] = sampler->border_color[0]; + rgba[1][j] = sampler->border_color[1]; + rgba[2][j] = sampler->border_color[2]; + rgba[3][j] = sampler->border_color[3]; } else { const int tx = x % TILE_SIZE; const int ty = y % TILE_SIZE; const struct softpipe_cached_tile *tile - = sp_get_cached_tile_tex(sampler->pipe, sampler->cache, + = sp_get_cached_tile_tex(sp, samp->cache, x, y, z, face, level); rgba[0][j] = tile->data.color[ty][tx][0]; rgba[1][j] = tile->data.color[ty][tx][1]; @@ -624,7 +695,7 @@ get_texel(struct tgsi_sampler *sampler, { debug_printf("Get texel %f %f %f %f from %s\n", rgba[0][j], rgba[1][j], rgba[2][j], rgba[3][j], - pf_name(sampler->texture->format)); + pf_name(texture->format)); } } } @@ -682,103 +753,124 @@ shadow_compare(uint compare_func, * Could probably extend for 3D... */ static void -sp_get_samples_2d_common(struct tgsi_sampler *sampler, +sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + boolean computeLambda, float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE], const unsigned faces[4]) { - const uint compare_func = sampler->state->compare_func; + const struct sp_shader_sampler *samp = sp_shader_sampler(tgsi_sampler); + const struct softpipe_context *sp = samp->sp; + const uint unit = samp->unit; + const struct pipe_texture *texture = sp->texture[unit]; + const struct pipe_sampler_state *sampler = sp->sampler[unit]; + const uint compare_func = sampler->compare_func; unsigned level0, level1, j, imgFilter; int width, height; float levelBlend; - choose_mipmap_levels(sampler, s, t, p, lodbias, + choose_mipmap_levels(texture, sampler, s, t, p, computeLambda, lodbias, &level0, &level1, &levelBlend, &imgFilter); - assert(sampler->state->normalized_coords); + assert(sampler->normalized_coords); - width = sampler->texture->width[level0]; - height = sampler->texture->height[level0]; + width = texture->width[level0]; + height = texture->height[level0]; assert(width > 0); switch (imgFilter) { case PIPE_TEX_FILTER_NEAREST: - for (j = 0; j < QUAD_SIZE; j++) { - int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); - int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); - get_texel(sampler, faces[j], level0, x, y, 0, rgba, j); - if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - shadow_compare(compare_func, rgba, p, j); - } - - if (level0 != level1) { - /* get texels from second mipmap level and blend */ - float rgba2[4][4]; - unsigned c; - x = x / 2; - y = y / 2; - get_texel(sampler, faces[j], level1, x, y, 0, rgba2, j); - if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ - shadow_compare(compare_func, rgba2, p, j); + { + int x[4], y[4]; + nearest_texcoord_4(sampler->wrap_s, s, width, x); + nearest_texcoord_4(sampler->wrap_t, t, height, y); + + for (j = 0; j < QUAD_SIZE; j++) { + get_texel(tgsi_sampler, faces[j], level0, x[j], y[j], 0, rgba, j); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(compare_func, rgba, p, j); } - for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + unsigned c; + x[j] /= 2; + y[j] /= 2; + get_texel(tgsi_sampler, faces[j], level1, x[j], y[j], 0, + rgba2, j); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ + shadow_compare(compare_func, rgba2, p, j); + } + + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]); + } } } } break; case PIPE_TEX_FILTER_LINEAR: case PIPE_TEX_FILTER_ANISO: - for (j = 0; j < QUAD_SIZE; j++) { - float tx[4][4], a, b; - int x0, y0, x1, y1, c; - linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &a); - linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &b); - get_texel(sampler, faces[j], level0, x0, y0, 0, tx, 0); - get_texel(sampler, faces[j], level0, x1, y0, 0, tx, 1); - get_texel(sampler, faces[j], level0, x0, y1, 0, tx, 2); - get_texel(sampler, faces[j], level0, x1, y1, 0, tx, 3); - if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - shadow_compare(compare_func, tx, p, 0); - shadow_compare(compare_func, tx, p, 1); - shadow_compare(compare_func, tx, p, 2); - shadow_compare(compare_func, tx, p, 3); - } - - for (c = 0; c < 4; c++) { - rgba[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); - } - - if (level0 != level1) { - /* get texels from second mipmap level and blend */ - float rgba2[4][4]; - x0 = x0 / 2; - y0 = y0 / 2; - x1 = x1 / 2; - y1 = y1 / 2; - get_texel(sampler, faces[j], level1, x0, y0, 0, tx, 0); - get_texel(sampler, faces[j], level1, x1, y0, 0, tx, 1); - get_texel(sampler, faces[j], level1, x0, y1, 0, tx, 2); - get_texel(sampler, faces[j], level1, x1, y1, 0, tx, 3); - if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ + { + int x0[4], y0[4], x1[4], y1[4]; + float xw[4], yw[4]; /* weights */ + + linear_texcoord_4(sampler->wrap_s, s, width, x0, x1, xw); + linear_texcoord_4(sampler->wrap_t, t, height, y0, y1, yw); + + for (j = 0; j < QUAD_SIZE; j++) { + float tx[4][4]; /* texels */ + int c; + get_texel(tgsi_sampler, faces[j], level0, x0[j], y0[j], 0, tx, 0); + get_texel(tgsi_sampler, faces[j], level0, x1[j], y0[j], 0, tx, 1); + get_texel(tgsi_sampler, faces[j], level0, x0[j], y1[j], 0, tx, 2); + get_texel(tgsi_sampler, faces[j], level0, x1[j], y1[j], 0, tx, 3); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { shadow_compare(compare_func, tx, p, 0); shadow_compare(compare_func, tx, p, 1); shadow_compare(compare_func, tx, p, 2); shadow_compare(compare_func, tx, p, 3); } + /* interpolate R, G, B, A */ for (c = 0; c < 4; c++) { - rgba2[c][j] = lerp_2d(a, b, - tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + rgba[c][j] = lerp_2d(xw[j], yw[j], + tx[c][0], tx[c][1], + tx[c][2], tx[c][3]); } - for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + x0[j] /= 2; + y0[j] /= 2; + x1[j] /= 2; + y1[j] /= 2; + get_texel(tgsi_sampler, faces[j], level1, x0[j], y0[j], 0, tx, 0); + get_texel(tgsi_sampler, faces[j], level1, x1[j], y0[j], 0, tx, 1); + get_texel(tgsi_sampler, faces[j], level1, x0[j], y1[j], 0, tx, 2); + get_texel(tgsi_sampler, faces[j], level1, x1[j], y1[j], 0, tx, 3); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ + shadow_compare(compare_func, tx, p, 0); + shadow_compare(compare_func, tx, p, 1); + shadow_compare(compare_func, tx, p, 2); + shadow_compare(compare_func, tx, p, 3); + } + + /* interpolate R, G, B, A */ + for (c = 0; c < 4; c++) { + rgba2[c][j] = lerp_2d(xw[j], yw[j], + tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + } + + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]); + } } } } @@ -789,55 +881,65 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, } -static void -sp_get_samples_1d(struct tgsi_sampler *sampler, +static INLINE void +sp_get_samples_1d(const struct tgsi_sampler *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + boolean computeLambda, float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { static const unsigned faces[4] = {0, 0, 0, 0}; static const float tzero[4] = {0, 0, 0, 0}; - sp_get_samples_2d_common(sampler, s, tzero, NULL, lodbias, rgba, faces); + sp_get_samples_2d_common(sampler, s, tzero, NULL, + computeLambda, lodbias, rgba, faces); } -static void -sp_get_samples_2d(struct tgsi_sampler *sampler, +static INLINE void +sp_get_samples_2d(const struct tgsi_sampler *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + boolean computeLambda, float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { static const unsigned faces[4] = {0, 0, 0, 0}; - sp_get_samples_2d_common(sampler, s, t, p, lodbias, rgba, faces); + sp_get_samples_2d_common(sampler, s, t, p, + computeLambda, lodbias, rgba, faces); } -static void -sp_get_samples_3d(struct tgsi_sampler *sampler, +static INLINE void +sp_get_samples_3d(const struct tgsi_sampler *tgsi_sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + boolean computeLambda, float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { + const struct sp_shader_sampler *samp = sp_shader_sampler(tgsi_sampler); + const struct softpipe_context *sp = samp->sp; + const uint unit = samp->unit; + const struct pipe_texture *texture = sp->texture[unit]; + const struct pipe_sampler_state *sampler = sp->sampler[unit]; /* get/map pipe_surfaces corresponding to 3D tex slices */ unsigned level0, level1, j, imgFilter; int width, height, depth; float levelBlend; const uint face = 0; - choose_mipmap_levels(sampler, s, t, p, lodbias, + choose_mipmap_levels(texture, sampler, s, t, p, computeLambda, lodbias, &level0, &level1, &levelBlend, &imgFilter); - assert(sampler->state->normalized_coords); + assert(sampler->normalized_coords); - width = sampler->texture->width[level0]; - height = sampler->texture->height[level0]; - depth = sampler->texture->depth[level0]; + width = texture->width[level0]; + height = texture->height[level0]; + depth = texture->depth[level0]; assert(width > 0); assert(height > 0); @@ -845,89 +947,89 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, switch (imgFilter) { case PIPE_TEX_FILTER_NEAREST: - for (j = 0; j < QUAD_SIZE; j++) { - int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); - int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); - int z = nearest_texcoord(sampler->state->wrap_r, p[j], depth); - get_texel(sampler, face, level0, x, y, z, rgba, j); - - if (level0 != level1) { - /* get texels from second mipmap level and blend */ - float rgba2[4][4]; - unsigned c; - x /= 2; - y /= 2; - z /= 2; - get_texel(sampler, face, level1, x, y, z, rgba2, j); - for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(levelBlend, rgba2[c][j], rgba[c][j]); + { + int x[4], y[4], z[4]; + nearest_texcoord_4(sampler->wrap_s, s, width, x); + nearest_texcoord_4(sampler->wrap_t, t, height, y); + nearest_texcoord_4(sampler->wrap_r, p, depth, z); + for (j = 0; j < QUAD_SIZE; j++) { + get_texel(tgsi_sampler, face, level0, x[j], y[j], z[j], rgba, j); + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + unsigned c; + x[j] /= 2; + y[j] /= 2; + z[j] /= 2; + get_texel(tgsi_sampler, face, level1, x[j], y[j], z[j], rgba2, j); + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = lerp(levelBlend, rgba2[c][j], rgba[c][j]); + } } } } break; case PIPE_TEX_FILTER_LINEAR: case PIPE_TEX_FILTER_ANISO: - for (j = 0; j < QUAD_SIZE; j++) { - float texel0[4][4], texel1[4][4]; - float xw, yw, zw; /* interpolation weights */ - int x0, x1, y0, y1, z0, z1, c; - linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &xw); - linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &yw); - linear_texcoord(sampler->state->wrap_r, p[j], depth, &z0, &z1, &zw); - get_texel(sampler, face, level0, x0, y0, z0, texel0, 0); - get_texel(sampler, face, level0, x1, y0, z0, texel0, 1); - get_texel(sampler, face, level0, x0, y1, z0, texel0, 2); - get_texel(sampler, face, level0, x1, y1, z0, texel0, 3); - get_texel(sampler, face, level0, x0, y0, z1, texel1, 0); - get_texel(sampler, face, level0, x1, y0, z1, texel1, 1); - get_texel(sampler, face, level0, x0, y1, z1, texel1, 2); - get_texel(sampler, face, level0, x1, y1, z1, texel1, 3); - - /* 3D lerp */ - for (c = 0; c < 4; c++) { - float ctemp0[4][4], ctemp1[4][4]; - ctemp0[c][j] = lerp_2d(xw, yw, - texel0[c][0], texel0[c][1], - texel0[c][2], texel0[c][3]); - ctemp1[c][j] = lerp_2d(xw, yw, - texel1[c][0], texel1[c][1], - texel1[c][2], texel1[c][3]); - rgba[c][j] = LERP(zw, ctemp0[c][j], ctemp1[c][j]); - } - - if (level0 != level1) { - /* get texels from second mipmap level and blend */ - float rgba2[4][4]; - x0 /= 2; - y0 /= 2; - z0 /= 2; - x1 /= 2; - y1 /= 2; - z1 /= 2; - get_texel(sampler, face, level1, x0, y0, z0, texel0, 0); - get_texel(sampler, face, level1, x1, y0, z0, texel0, 1); - get_texel(sampler, face, level1, x0, y1, z0, texel0, 2); - get_texel(sampler, face, level1, x1, y1, z0, texel0, 3); - get_texel(sampler, face, level1, x0, y0, z1, texel1, 0); - get_texel(sampler, face, level1, x1, y0, z1, texel1, 1); - get_texel(sampler, face, level1, x0, y1, z1, texel1, 2); - get_texel(sampler, face, level1, x1, y1, z1, texel1, 3); - - /* 3D lerp */ + { + int x0[4], x1[4], y0[4], y1[4], z0[4], z1[4]; + float xw[4], yw[4], zw[4]; /* interpolation weights */ + linear_texcoord_4(sampler->wrap_s, s, width, x0, x1, xw); + linear_texcoord_4(sampler->wrap_t, t, height, y0, y1, yw); + linear_texcoord_4(sampler->wrap_r, p, depth, z0, z1, zw); + + for (j = 0; j < QUAD_SIZE; j++) { + int c; + float tx0[4][4], tx1[4][4]; + get_texel(tgsi_sampler, face, level0, x0[j], y0[j], z0[j], tx0, 0); + get_texel(tgsi_sampler, face, level0, x1[j], y0[j], z0[j], tx0, 1); + get_texel(tgsi_sampler, face, level0, x0[j], y1[j], z0[j], tx0, 2); + get_texel(tgsi_sampler, face, level0, x1[j], y1[j], z0[j], tx0, 3); + get_texel(tgsi_sampler, face, level0, x0[j], y0[j], z1[j], tx1, 0); + get_texel(tgsi_sampler, face, level0, x1[j], y0[j], z1[j], tx1, 1); + get_texel(tgsi_sampler, face, level0, x0[j], y1[j], z1[j], tx1, 2); + get_texel(tgsi_sampler, face, level0, x1[j], y1[j], z1[j], tx1, 3); + + /* interpolate R, G, B, A */ for (c = 0; c < 4; c++) { - float ctemp0[4][4], ctemp1[4][4]; - ctemp0[c][j] = lerp_2d(xw, yw, - texel0[c][0], texel0[c][1], - texel0[c][2], texel0[c][3]); - ctemp1[c][j] = lerp_2d(xw, yw, - texel1[c][0], texel1[c][1], - texel1[c][2], texel1[c][3]); - rgba2[c][j] = LERP(zw, ctemp0[c][j], ctemp1[c][j]); + rgba[c][j] = lerp_3d(xw[j], yw[j], zw[j], + tx0[c][0], tx0[c][1], + tx0[c][2], tx0[c][3], + tx1[c][0], tx1[c][1], + tx1[c][2], tx1[c][3]); } - /* blend mipmap levels */ - for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + x0[j] /= 2; + y0[j] /= 2; + z0[j] /= 2; + x1[j] /= 2; + y1[j] /= 2; + z1[j] /= 2; + get_texel(tgsi_sampler, face, level1, x0[j], y0[j], z0[j], tx0, 0); + get_texel(tgsi_sampler, face, level1, x1[j], y0[j], z0[j], tx0, 1); + get_texel(tgsi_sampler, face, level1, x0[j], y1[j], z0[j], tx0, 2); + get_texel(tgsi_sampler, face, level1, x1[j], y1[j], z0[j], tx0, 3); + get_texel(tgsi_sampler, face, level1, x0[j], y0[j], z1[j], tx1, 0); + get_texel(tgsi_sampler, face, level1, x1[j], y0[j], z1[j], tx1, 1); + get_texel(tgsi_sampler, face, level1, x0[j], y1[j], z1[j], tx1, 2); + get_texel(tgsi_sampler, face, level1, x1[j], y1[j], z1[j], tx1, 3); + + /* interpolate R, G, B, A */ + for (c = 0; c < 4; c++) { + rgba2[c][j] = lerp_3d(xw[j], yw[j], zw[j], + tx0[c][0], tx0[c][1], + tx0[c][2], tx0[c][3], + tx1[c][0], tx1[c][1], + tx1[c][2], tx1[c][3]); + } + + /* blend mipmap levels */ + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]); + } } } } @@ -939,10 +1041,11 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, static void -sp_get_samples_cube(struct tgsi_sampler *sampler, +sp_get_samples_cube(const struct tgsi_sampler *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + boolean computeLambda, float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { @@ -951,67 +1054,80 @@ sp_get_samples_cube(struct tgsi_sampler *sampler, for (j = 0; j < QUAD_SIZE; j++) { faces[j] = choose_cube_face(s[j], t[j], p[j], ssss + j, tttt + j); } - sp_get_samples_2d_common(sampler, ssss, tttt, NULL, lodbias, rgba, faces); + sp_get_samples_2d_common(sampler, ssss, tttt, NULL, + computeLambda, lodbias, rgba, faces); } static void -sp_get_samples_rect(struct tgsi_sampler *sampler, +sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + boolean computeLambda, float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { - //sp_get_samples_2d_common(sampler, s, t, p, lodbias, rgba, faces); - static const uint face = 0; - const uint compare_func = sampler->state->compare_func; + const struct sp_shader_sampler *samp = sp_shader_sampler(tgsi_sampler); + const struct softpipe_context *sp = samp->sp; + const uint unit = samp->unit; + const struct pipe_texture *texture = sp->texture[unit]; + const struct pipe_sampler_state *sampler = sp->sampler[unit]; + const uint face = 0; + const uint compare_func = sampler->compare_func; unsigned level0, level1, j, imgFilter; int width, height; float levelBlend; - choose_mipmap_levels(sampler, s, t, p, lodbias, + choose_mipmap_levels(texture, sampler, s, t, p, computeLambda, lodbias, &level0, &level1, &levelBlend, &imgFilter); /* texture RECTS cannot be mipmapped */ assert(level0 == level1); - width = sampler->texture->width[level0]; - height = sampler->texture->height[level0]; + width = texture->width[level0]; + height = texture->height[level0]; assert(width > 0); switch (imgFilter) { case PIPE_TEX_FILTER_NEAREST: - for (j = 0; j < QUAD_SIZE; j++) { - int x = nearest_texcoord_unnorm(sampler->state->wrap_s, s[j], width); - int y = nearest_texcoord_unnorm(sampler->state->wrap_t, t[j], height); - get_texel(sampler, face, level0, x, y, 0, rgba, j); - if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - shadow_compare(compare_func, rgba, p, j); + { + int x[4], y[4]; + nearest_texcoord_unnorm_4(sampler->wrap_s, s, width, x); + nearest_texcoord_unnorm_4(sampler->wrap_t, t, height, y); + for (j = 0; j < QUAD_SIZE; j++) { + get_texel(tgsi_sampler, face, level0, x[j], y[j], 0, rgba, j); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(compare_func, rgba, p, j); + } } } break; case PIPE_TEX_FILTER_LINEAR: case PIPE_TEX_FILTER_ANISO: - for (j = 0; j < QUAD_SIZE; j++) { - float tx[4][4], a, b; - int x0, y0, x1, y1, c; - linear_texcoord_unnorm(sampler->state->wrap_s, s[j], width, &x0, &x1, &a); - linear_texcoord_unnorm(sampler->state->wrap_t, t[j], height, &y0, &y1, &b); - get_texel(sampler, face, level0, x0, y0, 0, tx, 0); - get_texel(sampler, face, level0, x1, y0, 0, tx, 1); - get_texel(sampler, face, level0, x0, y1, 0, tx, 2); - get_texel(sampler, face, level0, x1, y1, 0, tx, 3); - if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - shadow_compare(compare_func, tx, p, 0); - shadow_compare(compare_func, tx, p, 1); - shadow_compare(compare_func, tx, p, 2); - shadow_compare(compare_func, tx, p, 3); - } - - for (c = 0; c < 4; c++) { - rgba[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + { + int x0[4], y0[4], x1[4], y1[4]; + float xw[4], yw[4]; /* weights */ + linear_texcoord_unnorm_4(sampler->wrap_s, s, width, x0, x1, xw); + linear_texcoord_unnorm_4(sampler->wrap_t, t, height, y0, y1, yw); + for (j = 0; j < QUAD_SIZE; j++) { + float tx[4][4]; /* texels */ + int c; + get_texel(tgsi_sampler, face, level0, x0[j], y0[j], 0, tx, 0); + get_texel(tgsi_sampler, face, level0, x1[j], y0[j], 0, tx, 1); + get_texel(tgsi_sampler, face, level0, x0[j], y1[j], 0, tx, 2); + get_texel(tgsi_sampler, face, level0, x1[j], y1[j], 0, tx, 3); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(compare_func, tx, p, 0); + shadow_compare(compare_func, tx, p, 1); + shadow_compare(compare_func, tx, p, 2); + shadow_compare(compare_func, tx, p, 3); + } + for (c = 0; c < 4; c++) { + rgba[c][j] = lerp_2d(xw[j], yw[j], + tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + } } } break; @@ -1021,49 +1137,45 @@ sp_get_samples_rect(struct tgsi_sampler *sampler, } - - /** - * Called via tgsi_sampler::get_samples() - * Use the sampler's state setting to get a filtered RGBA value - * from the sampler's texture. - * - * XXX we can implement many versions of this function, each - * tightly coded for a specific combination of sampler state - * (nearest + repeat), (bilinear mipmap + clamp), etc. - * - * The update_samplers() function in st_atom_sampler.c could create - * a new tgsi_sampler object for each state combo it finds.... + * Common code for vertex/fragment program texture sampling. */ -void -sp_get_samples(struct tgsi_sampler *sampler, +static INLINE void +sp_get_samples(struct tgsi_sampler *tgsi_sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + boolean computeLambda, float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { - if (!sampler->texture) + const struct sp_shader_sampler *samp = sp_shader_sampler(tgsi_sampler); + const struct softpipe_context *sp = samp->sp; + const uint unit = samp->unit; + const struct pipe_texture *texture = sp->texture[unit]; + const struct pipe_sampler_state *sampler = sp->sampler[unit]; + + if (!texture) return; - switch (sampler->texture->target) { + switch (texture->target) { case PIPE_TEXTURE_1D: - assert(sampler->state->normalized_coords); - sp_get_samples_1d(sampler, s, t, p, lodbias, rgba); + assert(sampler->normalized_coords); + sp_get_samples_1d(tgsi_sampler, s, t, p, computeLambda, lodbias, rgba); break; case PIPE_TEXTURE_2D: - if (sampler->state->normalized_coords) - sp_get_samples_2d(sampler, s, t, p, lodbias, rgba); + if (sampler->normalized_coords) + sp_get_samples_2d(tgsi_sampler, s, t, p, computeLambda, lodbias, rgba); else - sp_get_samples_rect(sampler, s, t, p, lodbias, rgba); + sp_get_samples_rect(tgsi_sampler, s, t, p, computeLambda, lodbias, rgba); break; case PIPE_TEXTURE_3D: - assert(sampler->state->normalized_coords); - sp_get_samples_3d(sampler, s, t, p, lodbias, rgba); + assert(sampler->normalized_coords); + sp_get_samples_3d(tgsi_sampler, s, t, p, computeLambda, lodbias, rgba); break; case PIPE_TEXTURE_CUBE: - assert(sampler->state->normalized_coords); - sp_get_samples_cube(sampler, s, t, p, lodbias, rgba); + assert(sampler->normalized_coords); + sp_get_samples_cube(tgsi_sampler, s, t, p, computeLambda, lodbias, rgba); break; default: assert(0); @@ -1084,3 +1196,34 @@ sp_get_samples(struct tgsi_sampler *sampler, #endif } + +/** + * Called via tgsi_sampler::get_samples() when running a fragment shader. + * Get four filtered RGBA values from the sampler's texture. + */ +void +sp_get_samples_fragment(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + sp_get_samples(tgsi_sampler, s, t, p, TRUE, lodbias, rgba); +} + + +/** + * Called via tgsi_sampler::get_samples() when running a vertex shader. + * Get four filtered RGBA values from the sampler's texture. + */ +void +sp_get_samples_vertex(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + sp_get_samples(tgsi_sampler, s, t, p, FALSE, lodbias, rgba); +} diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.h b/src/gallium/drivers/softpipe/sp_tex_sample.h index 404bfd0c36..40d8eb2c2a 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.h +++ b/src/gallium/drivers/softpipe/sp_tex_sample.h @@ -1,17 +1,73 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #ifndef SP_TEX_SAMPLE_H #define SP_TEX_SAMPLE_H -struct tgsi_sampler; +#include "tgsi/tgsi_exec.h" + + +/** + * Subclass of tgsi_sampler + */ +struct sp_shader_sampler +{ + struct tgsi_sampler base; /**< base class */ + + uint unit; + struct softpipe_context *sp; + struct softpipe_tile_cache *cache; +}; + +static INLINE const struct sp_shader_sampler * +sp_shader_sampler(const struct tgsi_sampler *sampler) +{ + return (const struct sp_shader_sampler *) sampler; +} + + +extern void +sp_get_samples_fragment(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); + extern void -sp_get_samples(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]); +sp_get_samples_vertex(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); #endif /* SP_TEX_SAMPLE_H */ diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index b50c984513..78b0efa46d 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -494,11 +494,11 @@ tex_cache_pos(int x, int y, int z, int face, int level) * Tiles are read-only and indexed with more params. */ const struct softpipe_cached_tile * -sp_get_cached_tile_tex(struct pipe_context *pipe, +sp_get_cached_tile_tex(struct softpipe_context *sp, struct softpipe_tile_cache *tc, int x, int y, int z, int face, int level) { - struct pipe_screen *screen = pipe->screen; + struct pipe_screen *screen = sp->pipe.screen; /* tile pos in framebuffer: */ const int tile_x = x & ~(TILE_SIZE - 1); const int tile_y = y & ~(TILE_SIZE - 1); diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.h b/src/gallium/drivers/softpipe/sp_tile_cache.h index bc96c941f6..a66bb50bcc 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.h +++ b/src/gallium/drivers/softpipe/sp_tile_cache.h @@ -96,7 +96,7 @@ sp_get_cached_tile(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc, int x, int y); extern const struct softpipe_cached_tile * -sp_get_cached_tile_tex(struct pipe_context *pipe, +sp_get_cached_tile_tex(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc, int x, int y, int z, int face, int level); |