diff options
Diffstat (limited to 'src/gallium/drivers/softpipe')
38 files changed, 1226 insertions, 874 deletions
diff --git a/src/gallium/drivers/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile index 120bdfd9dd..516e3992fd 100644 --- a/src/gallium/drivers/softpipe/Makefile +++ b/src/gallium/drivers/softpipe/Makefile @@ -14,7 +14,7 @@ C_SOURCES = \ sp_draw_arrays.c \ sp_prim_setup.c \ sp_prim_vbuf.c \ - sp_quad.c \ + sp_quad_pipe.c \ sp_quad_alpha_test.c \ sp_quad_blend.c \ sp_quad_colormask.c \ @@ -42,6 +42,3 @@ C_SOURCES = \ sp_surface.c include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/drivers/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript index c1f7daa8ab..f8720638a7 100644 --- a/src/gallium/drivers/softpipe/SConscript +++ b/src/gallium/drivers/softpipe/SConscript @@ -17,7 +17,7 @@ softpipe = env.ConvenienceLibrary( 'sp_setup.c', 'sp_quad_alpha_test.c', 'sp_quad_blend.c', - 'sp_quad.c', + 'sp_quad_pipe.c', 'sp_quad_colormask.c', 'sp_quad_coverage.c', 'sp_quad_depth_test.c', diff --git a/src/gallium/drivers/softpipe/sp_clear.c b/src/gallium/drivers/softpipe/sp_clear.c index dfa46c9fb7..ad108ec446 100644 --- a/src/gallium/drivers/softpipe/sp_clear.c +++ b/src/gallium/drivers/softpipe/sp_clear.c @@ -85,7 +85,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, #endif } - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + for (i = 0; i < softpipe->framebuffer.nr_cbufs; i++) { if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[i])) { unsigned cv; if (ps->format != PIPE_FORMAT_A8R8G8B8_UNORM) { diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index cd1e6663d8..ff5d1b54a4 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 @@ -52,15 +53,15 @@ * Map any drawing surfaces which aren't already mapped */ void -softpipe_map_surfaces(struct softpipe_context *sp) +softpipe_map_transfers(struct softpipe_context *sp) { unsigned i; - for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - sp_tile_cache_map_surfaces(sp->cbuf_cache[i]); + for (i = 0; i < sp->framebuffer.nr_cbufs; i++) { + sp_tile_cache_map_transfers(sp->cbuf_cache[i]); } - sp_tile_cache_map_surfaces(sp->zsbuf_cache); + sp_tile_cache_map_transfers(sp->zsbuf_cache); } @@ -68,25 +69,25 @@ softpipe_map_surfaces(struct softpipe_context *sp) * Unmap any mapped drawing surfaces */ void -softpipe_unmap_surfaces(struct softpipe_context *sp) +softpipe_unmap_transfers(struct softpipe_context *sp) { uint i; - for (i = 0; i < sp->framebuffer.num_cbufs; i++) + for (i = 0; i < sp->framebuffer.nr_cbufs; i++) sp_flush_tile_cache(sp, sp->cbuf_cache[i]); sp_flush_tile_cache(sp, sp->zsbuf_cache); - for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]); + for (i = 0; i < sp->framebuffer.nr_cbufs; i++) { + sp_tile_cache_unmap_transfers(sp->cbuf_cache[i]); } - sp_tile_cache_unmap_surfaces(sp->zsbuf_cache); + sp_tile_cache_unmap_transfers(sp->zsbuf_cache); } static void softpipe_destroy( struct pipe_context *pipe ) { struct softpipe_context *softpipe = softpipe_context( pipe ); - struct pipe_winsys *ws = pipe->winsys; + struct pipe_screen *screen = pipe->screen; uint i; if (softpipe->draw) @@ -115,7 +116,7 @@ static void softpipe_destroy( struct pipe_context *pipe ) for (i = 0; i < Elements(softpipe->constants); i++) { if (softpipe->constants[i].buffer) { - winsys_buffer_reference(ws, &softpipe->constants[i].buffer, NULL); + pipe_buffer_reference(screen, &softpipe->constants[i].buffer, NULL); } } @@ -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..59d6df8f2d 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -32,11 +32,11 @@ #define SP_CONTEXT_H #include "pipe/p_context.h" -#include "pipe/p_defines.h" #include "draw/draw_vertex.h" -#include "sp_quad.h" +#include "sp_quad_pipe.h" +#include "sp_tex_sample.h" /** @@ -50,7 +50,6 @@ */ #define SP_NUM_QUAD_THREADS 1 -struct softpipe_winsys; struct softpipe_vbuf_render; struct draw_context; struct draw_stage; @@ -62,15 +61,15 @@ struct sp_vertex_shader; struct softpipe_context { struct pipe_context pipe; /**< base class */ - /* The most recent drawing state as set by the driver: - */ - const struct pipe_blend_state *blend; + /** Constant state objects */ + const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct pipe_depth_stencil_alpha_state *depth_stencil; + const struct pipe_depth_stencil_alpha_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; const struct sp_fragment_shader *fs; const struct sp_vertex_shader *vs; + /** Other rendering state */ struct pipe_blend_color blend_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; @@ -81,23 +80,20 @@ struct softpipe_context { struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; - unsigned dirty; unsigned num_samplers; unsigned num_textures; unsigned num_vertex_elements; unsigned num_vertex_buffers; - boolean no_rast; + unsigned dirty; /**< Mask of SP_NEW_x flags */ /* Counter for occlusion queries. Note this supports overlapping * queries. */ - uint64 occlusion_count; + uint64_t occlusion_count; - /* - * Mapped vertex buffers - */ + /** Mapped vertex buffers */ ubyte *mapped_vbuffer[PIPE_MAX_ATTRIBS]; /** Mapped constant buffers */ @@ -107,16 +103,11 @@ struct softpipe_context { struct vertex_info vertex_info; struct vertex_info vertex_info_vbuf; + /** Which vertex shader output slot contains point size */ int psize_slot; unsigned reduced_api_prim; /**< PIPE_PRIM_POINTS, _LINES or _TRIANGLES */ -#if 0 - /* Stipple derived state: - */ - ubyte stipple_masks[16][16]; -#endif - /** Derived from scissor and surface bounds: */ struct pipe_scissor_state cliprect; @@ -139,6 +130,14 @@ 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; @@ -150,8 +149,9 @@ struct softpipe_context { struct softpipe_tile_cache *tex_cache[PIPE_MAX_SAMPLERS]; - int use_sse : 1; - int dump_fs : 1; + unsigned use_sse : 1; + unsigned dump_fs : 1; + unsigned no_rast : 1; }; diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 424bd56846..f117096bf7 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -33,7 +33,7 @@ #include "pipe/p_defines.h" #include "pipe/p_context.h" -#include "pipe/p_winsys.h" +#include "pipe/internal/p_winsys_screen.h" #include "pipe/p_inlines.h" #include "sp_context.h" @@ -47,16 +47,22 @@ static void softpipe_map_constant_buffers(struct softpipe_context *sp) { struct pipe_winsys *ws = sp->pipe.winsys; - uint i; + uint i, size; + for (i = 0; i < PIPE_SHADER_TYPES; i++) { - if (sp->constants[i].size) + if (sp->constants[i].buffer && sp->constants[i].buffer->size) sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, PIPE_BUFFER_USAGE_CPU_READ); } + if (sp->constants[PIPE_SHADER_VERTEX].buffer) + size = sp->constants[PIPE_SHADER_VERTEX].buffer->size; + else + size = 0; + draw_set_mapped_constant_buffer(sp->draw, sp->mapped_constants[PIPE_SHADER_VERTEX], - sp->constants[PIPE_SHADER_VERTEX].size); + size); } static void @@ -73,7 +79,7 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp) draw_set_mapped_constant_buffer(sp->draw, NULL, 0); for (i = 0; i < 2; i++) { - if (sp->constants[i].size) + if (sp->constants[i].buffer && sp->constants[i].buffer->size) ws->buffer_unmap(ws, sp->constants[i].buffer); sp->mapped_constants[i] = NULL; } @@ -128,7 +134,7 @@ softpipe_draw_range_elements(struct pipe_context *pipe, if (sp->dirty) softpipe_update_derived( sp ); - softpipe_map_surfaces(sp); + softpipe_map_transfers(sp); softpipe_map_constant_buffers(sp); /* diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c index 401764bb43..035f4b963e 100644 --- a/src/gallium/drivers/softpipe/sp_flush.c +++ b/src/gallium/drivers/softpipe/sp_flush.c @@ -57,7 +57,7 @@ softpipe_flush( struct pipe_context *pipe, } if (flags & PIPE_FLUSH_RENDER_CACHE) { - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) + for (i = 0; i < softpipe->framebuffer.nr_cbufs; i++) if (softpipe->cbuf_cache[i]) sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]); @@ -70,7 +70,7 @@ softpipe_flush( struct pipe_context *pipe, * that's called before swapbuffers because we don't always want * to unmap surfaces when flushing. */ - softpipe_unmap_surfaces(softpipe); + softpipe_unmap_transfers(softpipe); } /* Enable to dump BMPs of the color/depth buffers each frame */ diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c index 701ee4c72f..3c7ba565d6 100644 --- a/src/gallium/drivers/softpipe/sp_fs_exec.c +++ b/src/gallium/drivers/softpipe/sp_fs_exec.c @@ -29,7 +29,7 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_fs.h" -#include "sp_headers.h" +#include "sp_quad.h" #include "pipe/p_state.h" @@ -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..7e22081132 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -29,7 +29,7 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_fs.h" -#include "sp_headers.h" +#include "sp_quad.h" #include "pipe/p_state.h" @@ -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_headers.h b/src/gallium/drivers/softpipe/sp_headers.h deleted file mode 100644 index 4a42cb3c19..0000000000 --- a/src/gallium/drivers/softpipe/sp_headers.h +++ /dev/null @@ -1,95 +0,0 @@ -/************************************************************************** - * - * 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. - * - **************************************************************************/ - -/* Authors: Keith Whitwell <keith@tungstengraphics.com> - */ - -#ifndef SP_HEADERS_H -#define SP_HEADERS_H - -#include "pipe/p_state.h" -#include "tgsi/tgsi_exec.h" - -#define PRIM_POINT 1 -#define PRIM_LINE 2 -#define PRIM_TRI 3 - - -/* The rasterizer generates 2x2 quads of fragment and feeds them to - * the current fp_machine (see below). - * Remember that Y=0=top with Y increasing down the window. - */ -#define QUAD_TOP_LEFT 0 -#define QUAD_TOP_RIGHT 1 -#define QUAD_BOTTOM_LEFT 2 -#define QUAD_BOTTOM_RIGHT 3 - -#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) -#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) -#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) -#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) -#define MASK_ALL 0xf - - -/** - * Encodes everything we need to know about a 2x2 pixel block. Uses - * "Channel-Serial" or "SoA" layout. - */ -struct quad_header_input -{ - int x0; - int y0; - float coverage[QUAD_SIZE]; /** fragment coverage for antialiasing */ - unsigned facing:1; /**< Front (0) or back (1) facing? */ - unsigned prim:2; /**< PRIM_POINT, LINE, TRI */ -}; - -struct quad_header_inout -{ - unsigned mask:4; -}; - -struct quad_header_output -{ - /** colors in SOA format (rrrr, gggg, bbbb, aaaa) */ - float color[PIPE_MAX_COLOR_BUFS][NUM_CHANNELS][QUAD_SIZE]; - float depth[QUAD_SIZE]; -}; - -struct quad_header { - struct quad_header_input input; - struct quad_header_inout inout; - struct quad_header_output output; - - const struct tgsi_interp_coef *coef; - const struct tgsi_interp_coef *posCoef; - - unsigned nr_attrs; -}; - -#endif /* SP_HEADERS_H */ - diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 425e13cd28..d56eed80a4 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 @@ -60,6 +60,7 @@ struct softpipe_vbuf_render struct softpipe_context *softpipe; uint prim; uint vertex_size; + uint vertex_buffer_size; void *vertex_buffer; }; @@ -80,26 +81,44 @@ sp_vbuf_get_vertex_info(struct vbuf_render *vbr) } -static void * +static boolean sp_vbuf_allocate_vertices(struct vbuf_render *vbr, - ushort vertex_size, ushort nr_vertices) + ushort vertex_size, ushort nr_vertices) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - assert(!cvbr->vertex_buffer); - cvbr->vertex_buffer = align_malloc(vertex_size * nr_vertices, 16); + unsigned size = vertex_size * nr_vertices; + + if (cvbr->vertex_buffer_size < size) { + align_free(cvbr->vertex_buffer); + cvbr->vertex_buffer = align_malloc(size, 16); + cvbr->vertex_buffer_size = size; + } + cvbr->vertex_size = vertex_size; - return cvbr->vertex_buffer; + return cvbr->vertex_buffer != NULL; } - static void -sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, - unsigned vertex_size, unsigned vertices_used) +sp_vbuf_release_vertices(struct vbuf_render *vbr) +{ + /* keep the old allocation for next time */ +} + +static void * +sp_vbuf_map_vertices(struct vbuf_render *vbr) +{ + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + return cvbr->vertex_buffer; +} + +static void +sp_vbuf_unmap_vertices(struct vbuf_render *vbr, + ushort min_index, + ushort max_index ) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - align_free(vertices); - assert(vertices == cvbr->vertex_buffer); - cvbr->vertex_buffer = NULL; + assert( cvbr->vertex_buffer_size >= (max_index+1) * cvbr->vertex_size ); + /* do nothing */ } @@ -115,8 +134,6 @@ sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) setup_prepare( setup_ctx ); - - cvbr->prim = prim; return TRUE; @@ -131,21 +148,23 @@ 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 +277,16 @@ 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: @@ -389,6 +411,8 @@ sp_init_vbuf(struct softpipe_context *sp) sp->vbuf_render->base.get_vertex_info = sp_vbuf_get_vertex_info; sp->vbuf_render->base.allocate_vertices = sp_vbuf_allocate_vertices; + sp->vbuf_render->base.map_vertices = sp_vbuf_map_vertices; + sp->vbuf_render->base.unmap_vertices = sp_vbuf_unmap_vertices; sp->vbuf_render->base.set_primitive = sp_vbuf_set_primitive; sp->vbuf_render->base.draw = sp_vbuf_draw; sp->vbuf_render->base.draw_arrays = sp_vbuf_draw_arrays; diff --git a/src/gallium/drivers/softpipe/sp_quad.h b/src/gallium/drivers/softpipe/sp_quad.h index 08513cb95f..bd6c6cb912 100644 --- a/src/gallium/drivers/softpipe/sp_quad.h +++ b/src/gallium/drivers/softpipe/sp_quad.h @@ -31,39 +31,76 @@ #ifndef SP_QUAD_H #define SP_QUAD_H +#include "pipe/p_state.h" +#include "tgsi/tgsi_exec.h" -struct softpipe_context; -struct quad_header; +#define QUAD_PRIM_POINT 1 +#define QUAD_PRIM_LINE 2 +#define QUAD_PRIM_TRI 3 -struct quad_stage { - struct softpipe_context *softpipe; - struct quad_stage *next; +/* The rasterizer generates 2x2 quads of fragment and feeds them to + * the current fp_machine (see below). + * Remember that Y=0=top with Y increasing down the window. + */ +#define QUAD_TOP_LEFT 0 +#define QUAD_TOP_RIGHT 1 +#define QUAD_BOTTOM_LEFT 2 +#define QUAD_BOTTOM_RIGHT 3 + +#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) +#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) +#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) +#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) +#define MASK_ALL 0xf + + +/** + * Quad stage inputs (pos, coverage, front/back face, etc) + */ +struct quad_header_input +{ + int x0, y0; /**< quad window pos, always even */ + float coverage[QUAD_SIZE]; /**< fragment coverage for antialiasing */ + unsigned facing:1; /**< Front (0) or back (1) facing? */ + unsigned prim:2; /**< QUAD_PRIM_POINT, LINE, TRI */ +}; - void (*begin)(struct quad_stage *qs); - /** the stage action */ - void (*run)(struct quad_stage *qs, struct quad_header *quad); +/** + * Quad stage inputs/outputs. + */ +struct quad_header_inout +{ + unsigned mask:4; +}; + - void (*destroy)(struct quad_stage *qs); +/** + * Quad stage outputs (color & depth). + */ +struct quad_header_output +{ + /** colors in SOA format (rrrr, gggg, bbbb, aaaa) */ + float color[PIPE_MAX_COLOR_BUFS][NUM_CHANNELS][QUAD_SIZE]; + float depth[QUAD_SIZE]; }; -struct quad_stage *sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_earlyz_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); +/** + * Encodes everything we need to know about a 2x2 pixel block. Uses + * "Channel-Serial" or "SoA" layout. + */ +struct quad_header { + struct quad_header_input input; + struct quad_header_inout inout; + struct quad_header_output output; -void sp_build_quad_pipeline(struct softpipe_context *sp); + const struct tgsi_interp_coef *coef; + const struct tgsi_interp_coef *posCoef; -void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad); + unsigned nr_attrs; +}; #endif /* SP_QUAD_H */ diff --git a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c index 5bebd141e9..0845bae0e6 100644 --- a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c @@ -4,8 +4,8 @@ */ #include "sp_context.h" -#include "sp_headers.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" #include "pipe/p_defines.h" #include "util/u_memory.h" @@ -14,7 +14,7 @@ static void alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - const float ref = softpipe->depth_stencil->alpha.ref; + const float ref = softpipe->depth_stencil->alpha.ref_value; unsigned passMask = 0x0, j; const uint cbuf = 0; /* only output[0].alpha is tested */ const float *aaaa = quad->output.color[cbuf][3]; diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index 6f64c6e584..e134e44337 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -34,10 +34,10 @@ #include "util/u_math.h" #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" +#include "sp_quad.h" #include "sp_surface.h" #include "sp_tile_cache.h" -#include "sp_quad.h" +#include "sp_quad_pipe.h" #define VEC4_COPY(DST, SRC) \ @@ -105,7 +105,7 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) uint cbuf; /* loop over colorbuffer outputs */ - for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + for (cbuf = 0; cbuf < softpipe->framebuffer.nr_cbufs; cbuf++) { float dest[4][QUAD_SIZE]; ubyte src[4][4], dst[4][4], res[4][4]; uint *src4 = (uint *) src; @@ -239,7 +239,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) } /* loop over colorbuffer outputs */ - for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + for (cbuf = 0; cbuf < softpipe->framebuffer.nr_cbufs; cbuf++) { float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; struct softpipe_cached_tile *tile = sp_get_cached_tile(softpipe, diff --git a/src/gallium/drivers/softpipe/sp_quad_bufloop.c b/src/gallium/drivers/softpipe/sp_quad_bufloop.c index 92e9af09c1..953d8516b9 100644 --- a/src/gallium/drivers/softpipe/sp_quad_bufloop.c +++ b/src/gallium/drivers/softpipe/sp_quad_bufloop.c @@ -1,9 +1,9 @@ #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" +#include "sp_surface.h" +#include "sp_quad_pipe.h" /** @@ -17,7 +17,7 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) unsigned i; assert(sizeof(quad->outputs.color) == sizeof(tmp)); - assert(softpipe->framebuffer.num_cbufs <= PIPE_MAX_COLOR_BUFS); + assert(softpipe->framebuffer.nr_cbufs <= PIPE_MAX_COLOR_BUFS); /* make copy of original colors since they can get modified * by blending and masking. @@ -28,7 +28,7 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) */ memcpy(tmp, quad->outputs.color, sizeof(tmp)); - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + for (i = 0; i < softpipe->framebuffer.nr_cbufs; i++) { /* set current cbuffer */ #if 0 /* obsolete & going away */ softpipe->current_cbuf = i; diff --git a/src/gallium/drivers/softpipe/sp_quad_colormask.c b/src/gallium/drivers/softpipe/sp_quad_colormask.c index f32bdfab78..dc90e5d5e9 100644 --- a/src/gallium/drivers/softpipe/sp_quad_colormask.c +++ b/src/gallium/drivers/softpipe/sp_quad_colormask.c @@ -34,9 +34,9 @@ #include "util/u_math.h" #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" +#include "sp_surface.h" +#include "sp_quad_pipe.h" #include "sp_tile_cache.h" @@ -51,7 +51,7 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad) uint cbuf; /* loop over colorbuffer outputs */ - for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + for (cbuf = 0; cbuf < softpipe->framebuffer.nr_cbufs; cbuf++) { float dest[4][QUAD_SIZE]; struct softpipe_cached_tile *tile = sp_get_cached_tile(softpipe, diff --git a/src/gallium/drivers/softpipe/sp_quad_coverage.c b/src/gallium/drivers/softpipe/sp_quad_coverage.c index ee29aa7dfe..4aeee85870 100644 --- a/src/gallium/drivers/softpipe/sp_quad_coverage.c +++ b/src/gallium/drivers/softpipe/sp_quad_coverage.c @@ -35,8 +35,8 @@ #include "pipe/p_defines.h" #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" /** @@ -46,14 +46,15 @@ static void coverage_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; + const uint prim = quad->input.prim; - if ((softpipe->rasterizer->poly_smooth && quad->input.prim == PRIM_TRI) || - (softpipe->rasterizer->line_smooth && quad->input.prim == PRIM_LINE) || - (softpipe->rasterizer->point_smooth && quad->input.prim == PRIM_POINT)) { + if ((softpipe->rasterizer->poly_smooth && prim == QUAD_PRIM_TRI) || + (softpipe->rasterizer->line_smooth && prim == QUAD_PRIM_LINE) || + (softpipe->rasterizer->point_smooth && prim == QUAD_PRIM_POINT)) { uint cbuf; /* loop over colorbuffer outputs */ - for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + for (cbuf = 0; cbuf < softpipe->framebuffer.nr_cbufs; cbuf++) { float (*quadColor)[4] = quad->output.color[cbuf]; unsigned j; for (j = 0; j < QUAD_SIZE; j++) { diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c index 523bd3e080..d463930bae 100644 --- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c @@ -32,9 +32,9 @@ #include "pipe/p_defines.h" #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" +#include "sp_surface.h" +#include "sp_quad_pipe.h" #include "sp_tile_cache.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_earlyz.c b/src/gallium/drivers/softpipe/sp_quad_earlyz.c index 6e2dde304e..496fd39ed1 100644 --- a/src/gallium/drivers/softpipe/sp_quad_earlyz.c +++ b/src/gallium/drivers/softpipe/sp_quad_earlyz.c @@ -31,8 +31,8 @@ #include "pipe/p_defines.h" #include "util/u_memory.h" -#include "sp_headers.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" /** diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 1f0cb3e035..adca5df73d 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 @@ -42,16 +43,15 @@ #include "sp_context.h" #include "sp_state.h" -#include "sp_headers.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" #include "sp_texture.h" #include "sp_tex_sample.h" 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); } @@ -178,7 +171,6 @@ static void shade_destroy(struct quad_stage *qs) struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) { struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); - uint i; /* allocate storage for program inputs/outputs, aligned to 16 bytes */ qss->inputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->inputs) + 16); @@ -191,14 +183,6 @@ 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_quad_occlusion.c b/src/gallium/drivers/softpipe/sp_quad_occlusion.c index 169bd82876..dfa7ff3b1d 100644 --- a/src/gallium/drivers/softpipe/sp_quad_occlusion.c +++ b/src/gallium/drivers/softpipe/sp_quad_occlusion.c @@ -35,9 +35,9 @@ #include "pipe/p_defines.h" #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" +#include "sp_surface.h" +#include "sp_quad_pipe.h" static unsigned count_bits( unsigned val ) { diff --git a/src/gallium/drivers/softpipe/sp_quad_output.c b/src/gallium/drivers/softpipe/sp_quad_output.c index b7aac7f84a..92d5f9f3c1 100644 --- a/src/gallium/drivers/softpipe/sp_quad_output.c +++ b/src/gallium/drivers/softpipe/sp_quad_output.c @@ -27,9 +27,9 @@ #include "util/u_memory.h" #include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" +#include "sp_surface.h" +#include "sp_quad_pipe.h" #include "sp_tile_cache.h" @@ -48,7 +48,7 @@ output_quad(struct quad_stage *qs, struct quad_header *quad) uint cbuf; /* loop over colorbuffer outputs */ - for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + for (cbuf = 0; cbuf < softpipe->framebuffer.nr_cbufs; cbuf++) { struct softpipe_cached_tile *tile = sp_get_cached_tile(softpipe, softpipe->cbuf_cache[cbuf], diff --git a/src/gallium/drivers/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad_pipe.c index 892ef87ee9..892ef87ee9 100644 --- a/src/gallium/drivers/softpipe/sp_quad.c +++ b/src/gallium/drivers/softpipe/sp_quad_pipe.c diff --git a/src/gallium/drivers/softpipe/sp_quad_pipe.h b/src/gallium/drivers/softpipe/sp_quad_pipe.h new file mode 100644 index 0000000000..0e40586ffc --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_pipe.h @@ -0,0 +1,74 @@ +/************************************************************************** + * + * 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. + * + **************************************************************************/ + +/* Authors: Keith Whitwell <keith@tungstengraphics.com> + */ + +#ifndef SP_QUAD_PIPE_H +#define SP_QUAD_PIPE_H + + +struct softpipe_context; +struct quad_header; + + +/** + * Fragment processing is performed on 2x2 blocks of pixels called "quads". + * Quad processing is performed with a pipeline of stages represented by + * this type. + */ +struct quad_stage { + struct softpipe_context *softpipe; + + struct quad_stage *next; + + void (*begin)(struct quad_stage *qs); + + /** the stage action */ + void (*run)(struct quad_stage *qs, struct quad_header *quad); + + void (*destroy)(struct quad_stage *qs); +}; + + +struct quad_stage *sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_earlyz_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); + +void sp_build_quad_pipeline(struct softpipe_context *sp); + +void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad); + +#endif /* SP_QUAD_PIPE_H */ diff --git a/src/gallium/drivers/softpipe/sp_quad_stencil.c b/src/gallium/drivers/softpipe/sp_quad_stencil.c index abb5487748..5e9d447737 100644 --- a/src/gallium/drivers/softpipe/sp_quad_stencil.c +++ b/src/gallium/drivers/softpipe/sp_quad_stencil.c @@ -5,10 +5,10 @@ #include "sp_context.h" -#include "sp_headers.h" +#include "sp_quad.h" #include "sp_surface.h" #include "sp_tile_cache.h" -#include "sp_quad.h" +#include "sp_quad_pipe.h" #include "pipe/p_defines.h" #include "util/u_memory.h" @@ -222,8 +222,8 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) zFailOp = softpipe->depth_stencil->stencil[face].zfail_op; zPassOp = softpipe->depth_stencil->stencil[face].zpass_op; ref = softpipe->depth_stencil->stencil[face].ref_value; - wrtMask = softpipe->depth_stencil->stencil[face].write_mask; - valMask = softpipe->depth_stencil->stencil[face].value_mask; + wrtMask = softpipe->depth_stencil->stencil[face].writemask; + valMask = softpipe->depth_stencil->stencil[face].valuemask; assert(ps); /* shouldn't get here if there's no stencil buffer */ diff --git a/src/gallium/drivers/softpipe/sp_quad_stipple.c b/src/gallium/drivers/softpipe/sp_quad_stipple.c index ccf37f6be5..05e862f097 100644 --- a/src/gallium/drivers/softpipe/sp_quad_stipple.c +++ b/src/gallium/drivers/softpipe/sp_quad_stipple.c @@ -4,8 +4,8 @@ */ #include "sp_context.h" -#include "sp_headers.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" #include "pipe/p_defines.h" #include "util/u_memory.h" @@ -19,11 +19,13 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) static const uint bit31 = 1 << 31; static const uint bit30 = 1 << 30; - if (quad->input.prim == PRIM_TRI) { + if (quad->input.prim == QUAD_PRIM_TRI) { struct softpipe_context *softpipe = qs->softpipe; /* need to invert Y to index into OpenGL's stipple pattern */ int y0, y1; uint stipple0, stipple1; + const int col0 = quad->input.x0 % 32; + if (softpipe->rasterizer->origin_lower_left) { y0 = softpipe->framebuffer.height - 1 - quad->input.y0; y1 = y0 - 1; @@ -32,12 +34,11 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) y0 = quad->input.y0; y1 = y0 + 1; } + stipple0 = softpipe->poly_stipple.stipple[y0 % 32]; stipple1 = softpipe->poly_stipple.stipple[y1 % 32]; -#if 1 - { - const int col0 = quad->input.x0 % 32; + /* turn off quad mask bits that fail the stipple test */ if ((stipple0 & (bit31 >> col0)) == 0) quad->inout.mask &= ~MASK_TOP_LEFT; @@ -49,19 +50,11 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) if ((stipple1 & (bit30 >> col0)) == 0) quad->inout.mask &= ~MASK_BOTTOM_RIGHT; - } -#else - /* We'd like to use this code, but we'd need to redefine - * MASK_TOP_LEFT to be (1 << 1) and MASK_TOP_RIGHT to be (1 << 0), - * and similarly for the BOTTOM bits. But that may have undesirable - * side effects elsewhere. - */ - const int col0 = 30 - (quad->input.x0 % 32); - quad->inout.mask &= (((stipple0 >> col0) & 0x3) | - (((stipple1 >> col0) & 0x3) << 2)); -#endif - if (!quad->inout.mask) + + if (!quad->inout.mask) { + /* all fragments failed stipple test, end of quad pipeline */ return; + } } qs->next->run(qs->next, quad); diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c index 2106ee1d23..b0d8e01426 100644 --- a/src/gallium/drivers/softpipe/sp_query.c +++ b/src/gallium/drivers/softpipe/sp_query.c @@ -37,8 +37,8 @@ #include "sp_query.h" struct softpipe_query { - uint64 start; - uint64 end; + uint64_t start; + uint64_t end; }; @@ -87,7 +87,7 @@ static boolean softpipe_get_query_result(struct pipe_context *pipe, struct pipe_query *q, boolean wait, - uint64 *result ) + uint64_t *result ) { struct softpipe_query *sq = softpipe_query(q); *result = sq->end - sq->start; diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index 9644dbd168..7380a6ae2b 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -27,7 +27,8 @@ #include "util/u_memory.h" -#include "pipe/p_winsys.h" +#include "util/u_simple_screen.h" +#include "pipe/internal/p_winsys_screen.h" #include "pipe/p_defines.h" #include "pipe/p_screen.h" @@ -55,7 +56,9 @@ softpipe_get_param(struct pipe_screen *screen, int param) { switch (param) { case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return 8; + return PIPE_MAX_SAMPLERS; + case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: + return PIPE_MAX_SAMPLERS; case PIPE_CAP_NPOT_TEXTURES: return 1; case PIPE_CAP_TWO_SIDED_STENCIL: @@ -172,6 +175,7 @@ softpipe_create_screen(struct pipe_winsys *winsys) screen->base.is_format_supported = softpipe_is_format_supported; softpipe_init_screen_texture_funcs(&screen->base); + u_simple_screen_init(&screen->base); return &screen->base; } diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 13d8017393..0925653b5d 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -32,13 +32,12 @@ * \author Brian Paul */ -#include "sp_setup.h" - #include "sp_context.h" -#include "sp_headers.h" +#include "sp_prim_setup.h" #include "sp_quad.h" +#include "sp_quad_pipe.h" +#include "sp_setup.h" #include "sp_state.h" -#include "sp_prim_setup.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vertex.h" @@ -265,17 +264,20 @@ is_inf_or_nan(float x) } -static boolean cull_tri( struct setup_context *setup, - float det ) +/** + * Do triangle cull test using tri determinant (sign indicates orientation) + * \return true if triangle is to be culled. + */ +static INLINE boolean +cull_tri(const struct setup_context *setup, float det) { - if (det != 0) - { + if (det != 0) { /* if (det < 0 then Z points toward camera and triangle is * counter-clockwise winding. */ unsigned winding = (det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; - - if ((winding & setup->winding) == 0) + + if ((winding & setup->winding) == 0) return FALSE; } @@ -968,7 +970,7 @@ void setup_tri( struct setup_context *setup, setup_tri_coefficients( setup ); setup_tri_edges( setup ); - setup->quad.input.prim = PRIM_TRI; + setup->quad.input.prim = QUAD_PRIM_TRI; setup->span.y = 0; setup->span.y_flags = 0; @@ -1009,7 +1011,7 @@ void setup_tri( struct setup_context *setup, * for a line. */ static void -line_linear_coeff(struct setup_context *setup, +line_linear_coeff(const struct setup_context *setup, struct tgsi_interp_coef *coef, uint vertSlot, uint i) { @@ -1029,9 +1031,9 @@ line_linear_coeff(struct setup_context *setup, * for a line. */ static void -line_persp_coeff(struct setup_context *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) +line_persp_coeff(const struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) { /* XXX double-check/verify this arithmetic */ const float a0 = setup->vmin[vertSlot][i] * setup->vmin[0][3]; @@ -1206,7 +1208,7 @@ setup_line(struct setup_context *setup, setup->quad.input.x0 = setup->quad.input.y0 = -1; setup->quad.inout.mask = 0x0; - setup->quad.input.prim = PRIM_LINE; + setup->quad.input.prim = QUAD_PRIM_LINE; /* XXX temporary: set coverage to 1.0 so the line appears * if AA mode happens to be enabled. */ @@ -1266,7 +1268,7 @@ setup_line(struct setup_context *setup, static void -point_persp_coeff(struct setup_context *setup, +point_persp_coeff(const struct setup_context *setup, const float (*vert)[4], struct tgsi_interp_coef *coef, uint vertSlot, uint i) @@ -1361,7 +1363,7 @@ setup_point( struct setup_context *setup, } } - setup->quad.input.prim = PRIM_POINT; + setup->quad.input.prim = QUAD_PRIM_POINT; if (halfSize <= 0.5 && !round) { /* special case for 1-pixel points */ @@ -1497,7 +1499,7 @@ void setup_prepare( struct setup_context *setup ) } /* Mark surfaces as defined now */ - for (i = 0; i < sp->framebuffer.num_cbufs; i++){ + for (i = 0; i < sp->framebuffer.nr_cbufs; i++){ if (sp->framebuffer.cbufs[i]) { sp->framebuffer.cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED; } diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 476ef3dc8f..6f558e6da5 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: @@ -184,10 +184,10 @@ softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags); void -softpipe_map_surfaces(struct softpipe_context *sp); +softpipe_map_transfers(struct softpipe_context *sp); void -softpipe_unmap_surfaces(struct softpipe_context *sp); +softpipe_unmap_transfers(struct softpipe_context *sp); void softpipe_map_texture_surfaces(struct softpipe_context *sp); diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index e5b609cf6c..4d01a9dbe1 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -32,7 +32,7 @@ #include "pipe/p_defines.h" #include "util/u_memory.h" #include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" +#include "pipe/internal/p_winsys_screen.h" #include "pipe/p_shader_tokens.h" #include "draw/draw_context.h" #include "tgsi/tgsi_dump.h" @@ -146,16 +146,15 @@ softpipe_set_constant_buffer(struct pipe_context *pipe, const struct pipe_constant_buffer *buf) { struct softpipe_context *softpipe = softpipe_context(pipe); - struct pipe_winsys *ws = pipe->winsys; + struct pipe_screen *screen = pipe->screen; assert(shader < PIPE_SHADER_TYPES); assert(index == 0); /* note: reference counting */ - winsys_buffer_reference(ws, + pipe_buffer_reference(screen, &softpipe->constants[shader].buffer, buf ? buf->buffer : NULL); - softpipe->constants[shader].size = buf ? buf->size : 0; softpipe->dirty |= SP_NEW_CONSTANTS; } diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c index b5376e522d..1493c65884 100644 --- a/src/gallium/drivers/softpipe/sp_state_surface.c +++ b/src/gallium/drivers/softpipe/sp_state_surface.c @@ -64,7 +64,7 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, } } - sp->framebuffer.num_cbufs = fb->num_cbufs; + sp->framebuffer.nr_cbufs = fb->nr_cbufs; /* zbuf changing? */ if (sp->framebuffer.zsbuf != fb->zsbuf) { diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 49250ec084..adbd0cb7f0 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 @@ -33,17 +34,18 @@ */ #include "sp_context.h" -#include "sp_headers.h" +#include "sp_quad.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_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index a64dc89f43..28a9784b16 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -33,7 +33,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" +#include "pipe/internal/p_winsys_screen.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -94,49 +94,23 @@ softpipe_texture_layout(struct pipe_screen *screen, return spt->buffer != NULL; } -/* Hack it up to use the old winsys->surface_alloc_storage() - * method for now: - */ static boolean softpipe_displaytarget_layout(struct pipe_screen *screen, struct softpipe_texture * spt) { struct pipe_winsys *ws = screen->winsys; - struct pipe_surface surf; - unsigned flags = (PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_GPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE); - int ret; - - - memset(&surf, 0, sizeof(surf)); - - ret =ws->surface_alloc_storage( ws, - &surf, - spt->base.width[0], - spt->base.height[0], - spt->base.format, - flags, - spt->base.tex_usage); - if(ret != 0) - return FALSE; - - if (!surf.buffer) { - /* allocation failed */ - return FALSE; - } + unsigned usage = (PIPE_BUFFER_USAGE_CPU_READ_WRITE | + PIPE_BUFFER_USAGE_GPU_READ_WRITE); - /* Now extract the goodies: - */ spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]); spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]); - spt->stride[0] = surf.stride; - /* Transfer the reference: - */ - spt->buffer = surf.buffer; - surf.buffer = NULL; + spt->buffer = ws->surface_buffer_create( ws, + spt->base.width[0], + spt->base.height[0], + spt->base.format, + usage, + &spt->stride[0]); return spt->buffer != NULL; } @@ -231,28 +205,21 @@ softpipe_get_tex_surface(struct pipe_screen *screen, unsigned face, unsigned level, unsigned zslice, unsigned usage) { - struct pipe_winsys *ws = screen->winsys; struct softpipe_texture *spt = softpipe_texture(pt); struct pipe_surface *ps; assert(level <= pt->last_level); ps = CALLOC_STRUCT(pipe_surface); - ps->refcount = 1; if (ps) { - assert(ps->refcount); + ps->refcount = 1; pipe_texture_reference(&ps->texture, pt); - pipe_buffer_reference(screen, &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; - + /* Because we are softpipe, anything that the state tracker * thought was going to be done with the GPU will actually get * done with the CPU. Let's adjust the flags to take that into @@ -278,8 +245,7 @@ softpipe_get_tex_surface(struct pipe_screen *screen, if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - ps->nblocksy * - ps->stride; + pt->nblocksy[level] * spt->stride[level]; } else { assert(face == 0); @@ -299,37 +265,108 @@ softpipe_tex_surface_release(struct pipe_screen *screen, * needed post-processing to put them into hardware layout, this is * where it would happen. For softpipe, nothing to do. */ - assert ((*s)->texture); + assert(surf->texture); if (--surf->refcount == 0) { - pipe_texture_reference(&surf->texture, NULL); - pipe_buffer_reference(screen, &surf->buffer, NULL); + pipe_texture_reference(&surf->texture, NULL); FREE(surf); } *s = NULL; } +static struct pipe_transfer * +softpipe_get_tex_transfer(struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, unsigned y, unsigned w, unsigned h) +{ + struct softpipe_texture *sptex = softpipe_texture(texture); + struct softpipe_transfer *spt; + struct pipe_transfer *pt; + + assert(texture); + assert(level <= texture->last_level); + + spt = CALLOC_STRUCT(softpipe_transfer); + pt = &spt->base; + if (spt) { + pt->refcount = 1; + pipe_texture_reference(&pt->texture, texture); + pt->format = texture->format; + pt->block = texture->block; + pt->x = x; + pt->y = y; + pt->width = w; + pt->height = h; + pt->nblocksx = texture->nblocksx[level]; + pt->nblocksy = texture->nblocksy[level]; + pt->stride = sptex->stride[level]; + spt->offset = sptex->level_offset[level]; + pt->usage = usage; + pt->face = face; + pt->level = level; + pt->zslice = zslice; + + if (texture->target == PIPE_TEXTURE_CUBE || + texture->target == PIPE_TEXTURE_3D) { + spt->offset += ((texture->target == PIPE_TEXTURE_CUBE) ? face : + zslice) * pt->nblocksy * pt->stride; + } + else { + assert(face == 0); + assert(zslice == 0); + } + } + return pt; +} + + +static void +softpipe_tex_transfer_release(struct pipe_screen *screen, + struct pipe_transfer **t) +{ + struct softpipe_transfer *transfer = softpipe_transfer(*t); + /* 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 (transfer->base.texture); + if (--transfer->base.refcount == 0) { + pipe_texture_reference(&transfer->base.texture, NULL); + FREE(transfer); + } + *t = NULL; +} + + static void * -softpipe_surface_map( struct pipe_screen *screen, - struct pipe_surface *surface, - unsigned flags ) +softpipe_transfer_map( struct pipe_screen *screen, + struct pipe_transfer *transfer ) { ubyte *map; + struct softpipe_texture *spt; + unsigned flags = 0; - if (flags & ~surface->usage) { - assert(0); - return NULL; + assert(transfer->texture); + spt = softpipe_texture(transfer->texture); + + if (transfer->usage != PIPE_TRANSFER_READ) { + flags |= PIPE_BUFFER_USAGE_CPU_WRITE; + } + + if (transfer->usage != PIPE_TRANSFER_WRITE) { + flags |= PIPE_BUFFER_USAGE_CPU_READ; } - map = pipe_buffer_map( screen, surface->buffer, flags ); + map = pipe_buffer_map(screen, spt->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)) + if (transfer->texture && transfer->usage != PIPE_TRANSFER_READ) { /* Do something to notify sharing contexts of a texture change. * In softpipe, that would mean flushing the texture cache. @@ -337,15 +374,22 @@ softpipe_surface_map( struct pipe_screen *screen, softpipe_screen(screen)->timestamp++; } - return map + surface->offset; + return map + softpipe_transfer(transfer)->offset + + transfer->y / transfer->block.height * transfer->stride + + transfer->x / transfer->block.width * transfer->block.size; } static void -softpipe_surface_unmap(struct pipe_screen *screen, - struct pipe_surface *surface) +softpipe_transfer_unmap(struct pipe_screen *screen, + struct pipe_transfer *transfer) { - pipe_buffer_unmap( screen, surface->buffer ); + struct softpipe_texture *spt; + + assert(transfer->texture); + spt = softpipe_texture(transfer->texture); + + pipe_buffer_unmap( screen, spt->buffer ); } @@ -365,6 +409,8 @@ softpipe_init_screen_texture_funcs(struct pipe_screen *screen) screen->get_tex_surface = softpipe_get_tex_surface; screen->tex_surface_release = softpipe_tex_surface_release; - screen->surface_map = softpipe_surface_map; - screen->surface_unmap = softpipe_surface_unmap; + screen->get_tex_transfer = softpipe_get_tex_transfer; + screen->tex_transfer_release = softpipe_tex_transfer_release; + screen->transfer_map = softpipe_transfer_map; + screen->transfer_unmap = softpipe_transfer_unmap; } diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h index bf437a7c61..893aa7d11d 100644 --- a/src/gallium/drivers/softpipe/sp_texture.h +++ b/src/gallium/drivers/softpipe/sp_texture.h @@ -42,7 +42,7 @@ struct softpipe_texture struct pipe_texture base; unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; - unsigned long stride[PIPE_MAX_TEXTURE_LEVELS]; + unsigned stride[PIPE_MAX_TEXTURE_LEVELS]; /* The data is held here: */ @@ -51,14 +51,27 @@ struct softpipe_texture boolean modified; }; +struct softpipe_transfer +{ + struct pipe_transfer base; + + unsigned long offset; +}; + -/** cast wrapper */ +/** cast wrappers */ static INLINE struct softpipe_texture * softpipe_texture(struct pipe_texture *pt) { return (struct softpipe_texture *) pt; } +static INLINE struct softpipe_transfer * +softpipe_transfer(struct pipe_transfer *pt) +{ + return (struct softpipe_transfer *) pt; +} + extern void softpipe_init_texture_funcs( struct softpipe_context *softpipe ); diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index b50c984513..593360aab0 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -26,7 +26,7 @@ **************************************************************************/ /** - * Framebuffer/surface tile caching. + * Texture tile caching. * * Author: * Brian Paul @@ -40,7 +40,7 @@ #include "sp_texture.h" #include "sp_tile_cache.h" -#define NUM_ENTRIES 32 +#define NUM_ENTRIES 50 /** XXX move these */ @@ -52,7 +52,8 @@ struct softpipe_tile_cache { struct pipe_screen *screen; struct pipe_surface *surface; /**< the surface we're caching */ - void *surface_map; + struct pipe_transfer *transfer; + void *transfer_map; struct pipe_texture *texture; /**< if caching a texture */ struct softpipe_cached_tile entries[NUM_ENTRIES]; uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; @@ -60,8 +61,8 @@ struct softpipe_tile_cache uint clear_val; boolean depth_stencil; /** Is the surface a depth/stencil format? */ - struct pipe_surface *tex_surf; - void *tex_surf_map; + struct pipe_transfer *tex_trans; + void *tex_trans_map; int tex_face, tex_level, tex_z; struct softpipe_cached_tile tile; /**< scratch tile for clears */ @@ -131,16 +132,19 @@ sp_create_tile_cache( struct pipe_screen *screen ) void sp_destroy_tile_cache(struct softpipe_tile_cache *tc) { + struct pipe_screen *screen; uint pos; for (pos = 0; pos < NUM_ENTRIES; pos++) { /*assert(tc->entries[pos].x < 0);*/ } - if (tc->surface) { - pipe_surface_reference(&tc->surface, NULL); + if (tc->transfer) { + screen = tc->transfer->texture->screen; + screen->tex_transfer_release(screen, &tc->transfer); } - if (tc->tex_surf) { - pipe_surface_reference(&tc->tex_surf, NULL); + if (tc->tex_trans) { + screen = tc->tex_trans->texture->screen; + screen->tex_transfer_release(screen, &tc->tex_trans); } FREE( tc ); @@ -156,18 +160,29 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, { assert(!tc->texture); - if (tc->surface_map) { - tc->screen->surface_unmap(tc->screen, tc->surface); - tc->surface_map = NULL; + if (tc->transfer) { + struct pipe_screen *screen = tc->transfer->texture->screen; + + if (ps == tc->surface) + return; + + if (tc->transfer_map) { + tc->screen->transfer_unmap(tc->screen, tc->transfer); + tc->transfer_map = NULL; + } + + screen->tex_transfer_release(screen, &tc->transfer); } - pipe_surface_reference(&tc->surface, ps); + tc->surface = ps; + + if (ps) { + struct pipe_screen *screen = ps->texture->screen; - if (tc->surface) { - if (tc->surface_map) /* XXX: this is always NULL!? */ - tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface, - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE); + tc->transfer = screen->get_tex_transfer(screen, ps->texture, ps->face, + ps->level, ps->zslice, + PIPE_TRANSFER_READ_WRITE, + 0, 0, ps->width, ps->height); tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || ps->format == PIPE_FORMAT_X8Z24_UNORM || @@ -181,7 +196,7 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, /** - * Return the surface being cached. + * Return the transfer being cached. */ struct pipe_surface * sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) @@ -191,30 +206,27 @@ sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) void -sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc) +sp_tile_cache_map_transfers(struct softpipe_tile_cache *tc) { - if (tc->surface && !tc->surface_map) - tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface, - PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_CPU_READ); - - if (tc->tex_surf && !tc->tex_surf_map) - tc->tex_surf_map = tc->screen->surface_map(tc->screen, tc->tex_surf, - PIPE_BUFFER_USAGE_CPU_READ); + if (tc->transfer && !tc->transfer_map) + tc->transfer_map = tc->screen->transfer_map(tc->screen, tc->transfer); + + if (tc->tex_trans && !tc->tex_trans_map) + tc->tex_trans_map = tc->screen->transfer_map(tc->screen, tc->tex_trans); } void -sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc) +sp_tile_cache_unmap_transfers(struct softpipe_tile_cache *tc) { - if (tc->surface_map) { - tc->screen->surface_unmap(tc->screen, tc->surface); - tc->surface_map = NULL; + if (tc->transfer_map) { + tc->screen->transfer_unmap(tc->screen, tc->transfer); + tc->transfer_map = NULL; } - if (tc->tex_surf_map) { - tc->screen->surface_unmap(tc->screen, tc->tex_surf); - tc->tex_surf_map = NULL; + if (tc->tex_trans_map) { + tc->screen->transfer_unmap(tc->screen, tc->tex_trans); + tc->tex_trans_map = NULL; } } @@ -229,15 +241,20 @@ sp_tile_cache_set_texture(struct pipe_context *pipe, { uint i; - assert(!tc->surface); + assert(!tc->transfer); pipe_texture_reference(&tc->texture, texture); - if (tc->tex_surf_map) { - tc->screen->surface_unmap(tc->screen, tc->tex_surf); - tc->tex_surf_map = NULL; + if (tc->transfer) { + struct pipe_screen *screen = tc->transfer->texture->screen; + + if (tc->tex_trans_map) { + tc->screen->transfer_unmap(tc->screen, tc->tex_trans); + tc->tex_trans_map = NULL; + } + + screen->tex_transfer_release(screen, &tc->tex_trans); } - pipe_surface_reference(&tc->tex_surf, NULL); /* mark as entries as invalid/empty */ /* XXX we should try to avoid this when the teximage hasn't changed */ @@ -328,20 +345,20 @@ static void sp_tile_cache_flush_clear(struct pipe_context *pipe, struct softpipe_tile_cache *tc) { - struct pipe_surface *ps = tc->surface; - const uint w = tc->surface->width; - const uint h = tc->surface->height; + struct pipe_transfer *pt = tc->transfer; + const uint w = tc->transfer->width; + const uint h = tc->transfer->height; uint x, y; uint numCleared = 0; /* clear the scratch tile to the clear value */ - clear_tile(&tc->tile, ps->format, tc->clear_val); + clear_tile(&tc->tile, pt->format, tc->clear_val); /* push the tile to all positions marked as clear */ for (y = 0; y < h; y += TILE_SIZE) { for (x = 0; x < w; x += TILE_SIZE) { if (is_clear_flag_set(tc->clear_flags, x, y)) { - pipe_put_tile_raw(ps, + pipe_put_tile_raw(pt, x, y, TILE_SIZE, TILE_SIZE, tc->tile.data.color32, 0/*STRIDE*/); @@ -359,28 +376,28 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, /** - * Flush the tile cache: write all dirty tiles back to the surface. + * Flush the tile cache: write all dirty tiles back to the transfer. * any tiles "flagged" as cleared will be "really" cleared. */ void sp_flush_tile_cache(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc) { - struct pipe_surface *ps = tc->surface; + struct pipe_transfer *pt = tc->transfer; int inuse = 0, pos; - if (ps && ps->buffer) { - /* caching a drawing surface */ + if (pt) { + /* caching a drawing transfer */ for (pos = 0; pos < NUM_ENTRIES; pos++) { struct softpipe_cached_tile *tile = tc->entries + pos; if (tile->x >= 0) { if (tc->depth_stencil) { - pipe_put_tile_raw(ps, + pipe_put_tile_raw(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { - pipe_put_tile_rgba(ps, + pipe_put_tile_rgba(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); } @@ -415,7 +432,7 @@ struct softpipe_cached_tile * sp_get_cached_tile(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc, int x, int y) { - struct pipe_surface *ps = tc->surface; + struct pipe_transfer *pt = tc->transfer; /* tile pos in framebuffer: */ const int tile_x = x & ~(TILE_SIZE - 1); @@ -431,12 +448,12 @@ sp_get_cached_tile(struct softpipe_context *softpipe, if (tile->x != -1) { /* put dirty tile back in framebuffer */ if (tc->depth_stencil) { - pipe_put_tile_raw(ps, + pipe_put_tile_raw(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { - pipe_put_tile_rgba(ps, + pipe_put_tile_rgba(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); } @@ -448,22 +465,22 @@ sp_get_cached_tile(struct softpipe_context *softpipe, if (is_clear_flag_set(tc->clear_flags, x, y)) { /* don't get tile from framebuffer, just clear it */ if (tc->depth_stencil) { - clear_tile(tile, ps->format, tc->clear_val); + clear_tile(tile, pt->format, tc->clear_val); } else { - clear_tile_rgba(tile, ps->format, tc->clear_color); + clear_tile_rgba(tile, pt->format, tc->clear_color); } clear_clear_flag(tc->clear_flags, x, y); } else { - /* get new tile data from surface */ + /* get new tile data from transfer */ if (tc->depth_stencil) { - pipe_get_tile_raw(ps, + pipe_get_tile_raw(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { - pipe_get_tile_rgba(ps, + pipe_get_tile_rgba(pt, tile->x, tile->y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); } @@ -484,7 +501,7 @@ sp_get_cached_tile(struct softpipe_context *softpipe, static INLINE uint tex_cache_pos(int x, int y, int z, int face, int level) { - uint entry = x + y * 5 + z * 4 + face + level; + uint entry = x + y * 9 + z * 3 + face + level * 7; return entry % NUM_ENTRIES; } @@ -494,11 +511,11 @@ 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); @@ -510,8 +527,12 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, if (tc->texture) { struct softpipe_texture *spt = softpipe_texture(tc->texture); if (spt->modified) { - /* texture was modified, force a cache reload */ - tile->x = -1; + /* texture was modified, invalidate all cached tiles */ + uint p; + for (p = 0; p < NUM_ENTRIES; p++) { + tile = tc->entries + p; + tile->x = -1; + } spt->modified = FALSE; } } @@ -523,28 +544,37 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, level != tile->level) { /* cache miss */ - /* check if we need to get a new surface */ - if (!tc->tex_surf || +#if 0 + printf("miss at %u x=%d y=%d z=%d face=%d level=%d\n", pos, + x/TILE_SIZE, y/TILE_SIZE, z, face, level); +#endif + /* check if we need to get a new transfer */ + if (!tc->tex_trans || tc->tex_face != face || tc->tex_level != level || tc->tex_z != z) { - /* get new surface (view into texture) */ + /* get new transfer (view into texture) */ + + if (tc->transfer) { + if (tc->tex_trans_map) + tc->screen->transfer_unmap(tc->screen, tc->tex_trans); - if (tc->tex_surf_map) - tc->screen->surface_unmap(tc->screen, tc->tex_surf); + screen->tex_transfer_release(screen, &tc->tex_trans); + } - tc->tex_surf = screen->get_tex_surface(screen, tc->texture, face, level, z, - PIPE_BUFFER_USAGE_CPU_READ); - tc->tex_surf_map = screen->surface_map(screen, tc->tex_surf, - PIPE_BUFFER_USAGE_CPU_READ); + tc->tex_trans = screen->get_tex_transfer(screen, tc->texture, face, level, z, + PIPE_TRANSFER_READ, 0, 0, + tc->texture->width[level], + tc->texture->height[level]); + tc->tex_trans_map = screen->transfer_map(screen, tc->tex_trans); tc->tex_face = face; tc->tex_level = level; tc->tex_z = z; } - /* get tile from the surface (view into texture) */ - pipe_get_tile_rgba(tc->tex_surf, + /* get tile from the transfer (view into texture) */ + pipe_get_tile_rgba(tc->tex_trans, tile_x, tile_y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); tile->x = tile_x; @@ -571,7 +601,7 @@ sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue) tc->clear_val = clearValue; - switch (tc->surface->format) { + switch (tc->transfer->format) { case PIPE_FORMAT_R8G8B8A8_UNORM: r = (clearValue >> 24) & 0xff; g = (clearValue >> 16) & 0xff; diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.h b/src/gallium/drivers/softpipe/sp_tile_cache.h index bc96c941f6..9ac3fdda94 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.h +++ b/src/gallium/drivers/softpipe/sp_tile_cache.h @@ -74,10 +74,10 @@ extern struct pipe_surface * sp_tile_cache_get_surface(struct softpipe_tile_cache *tc); extern void -sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc); +sp_tile_cache_map_transfers(struct softpipe_tile_cache *tc); extern void -sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc); +sp_tile_cache_unmap_transfers(struct softpipe_tile_cache *tc); extern void sp_tile_cache_set_texture(struct pipe_context *pipe, @@ -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); |