From f9d09a2e7859a2cf025d71b7c3cb189edb6688c4 Mon Sep 17 00:00:00 2001 From: Luca Barbieri Date: Sun, 21 Feb 2010 15:13:12 +0100 Subject: nv30, nv40: move last files to nvfx/ and rm -rf nv30 nv40 This is the last nvfx unification patch. nv[34]0_fragtex.c are moved to the common directory nv[34]0_shader.h are renamed to nv[34]0_vertprog.h and moved to the common directory The separate nv30 and nv40 directories are removed from the build system --- src/gallium/drivers/nvfx/Makefile | 2 + src/gallium/drivers/nvfx/nv30_fragtex.c | 147 +++++++++++++++++++++++++ src/gallium/drivers/nvfx/nv30_vertprog.h | 169 +++++++++++++++++++++++++++++ src/gallium/drivers/nvfx/nv40_fragtex.c | 174 ++++++++++++++++++++++++++++++ src/gallium/drivers/nvfx/nv40_vertprog.h | 177 +++++++++++++++++++++++++++++++ src/gallium/drivers/nvfx/nvfx_draw.c | 4 +- src/gallium/drivers/nvfx/nvfx_vertprog.c | 4 +- 7 files changed, 673 insertions(+), 4 deletions(-) create mode 100644 src/gallium/drivers/nvfx/nv30_fragtex.c create mode 100644 src/gallium/drivers/nvfx/nv30_vertprog.h create mode 100644 src/gallium/drivers/nvfx/nv40_fragtex.c create mode 100644 src/gallium/drivers/nvfx/nv40_vertprog.h (limited to 'src/gallium/drivers/nvfx') diff --git a/src/gallium/drivers/nvfx/Makefile b/src/gallium/drivers/nvfx/Makefile index 51fa34cfad..e912177b21 100644 --- a/src/gallium/drivers/nvfx/Makefile +++ b/src/gallium/drivers/nvfx/Makefile @@ -9,6 +9,8 @@ C_SOURCES = \ nvfx_draw.c \ nvfx_fragprog.c \ nvfx_fragtex.c \ + nv30_fragtex.c \ + nv40_fragtex.c \ nvfx_miptree.c \ nvfx_query.c \ nvfx_screen.c \ diff --git a/src/gallium/drivers/nvfx/nv30_fragtex.c b/src/gallium/drivers/nvfx/nv30_fragtex.c new file mode 100644 index 0000000000..2b56f45492 --- /dev/null +++ b/src/gallium/drivers/nvfx/nv30_fragtex.c @@ -0,0 +1,147 @@ +#include "util/u_format.h" + +#include "nvfx_context.h" +#include "nouveau/nouveau_util.h" +#include "nvfx_tex.h" + +void +nv30_sampler_state_init(struct pipe_context *pipe, + struct nvfx_sampler_state *ps, + const struct pipe_sampler_state *cso) +{ + if (cso->max_anisotropy >= 8) { + ps->en |= NV34TCL_TX_ENABLE_ANISO_8X; + } else + if (cso->max_anisotropy >= 4) { + ps->en |= NV34TCL_TX_ENABLE_ANISO_4X; + } else + if (cso->max_anisotropy >= 2) { + ps->en |= NV34TCL_TX_ENABLE_ANISO_2X; + } + + { + float limit; + + limit = CLAMP(cso->lod_bias, -16.0, 15.0); + ps->filt |= (int)(cso->lod_bias * 256.0) & 0x1fff; + + limit = CLAMP(cso->max_lod, 0.0, 15.0); + ps->en |= (int)(limit) << 14 /*NV34TCL_TX_ENABLE_MIPMAP_MAX_LOD_SHIFT*/; + + limit = CLAMP(cso->min_lod, 0.0, 15.0); + ps->en |= (int)(limit) << 26 /*NV34TCL_TX_ENABLE_MIPMAP_MIN_LOD_SHIFT*/; + } +} + +#define _(m,tf,ts0x,ts0y,ts0z,ts0w,ts1x,ts1y,ts1z,ts1w) \ +{ \ + TRUE, \ + PIPE_FORMAT_##m, \ + NV34TCL_TX_FORMAT_FORMAT_##tf, \ + (NV34TCL_TX_SWIZZLE_S0_X_##ts0x | NV34TCL_TX_SWIZZLE_S0_Y_##ts0y | \ + NV34TCL_TX_SWIZZLE_S0_Z_##ts0z | NV34TCL_TX_SWIZZLE_S0_W_##ts0w | \ + NV34TCL_TX_SWIZZLE_S1_X_##ts1x | NV34TCL_TX_SWIZZLE_S1_Y_##ts1y | \ + NV34TCL_TX_SWIZZLE_S1_Z_##ts1z | NV34TCL_TX_SWIZZLE_S1_W_##ts1w) \ +} + +struct nv30_texture_format { + boolean defined; + uint pipe; + int format; + int swizzle; +}; + +static struct nv30_texture_format +nv30_texture_formats[] = { + _(B8G8R8X8_UNORM, A8R8G8B8, S1, S1, S1, ONE, X, Y, Z, W), + _(B8G8R8A8_UNORM, A8R8G8B8, S1, S1, S1, S1, X, Y, Z, W), + _(B5G5R5A1_UNORM, A1R5G5B5, S1, S1, S1, S1, X, Y, Z, W), + _(B4G4R4A4_UNORM, A4R4G4B4, S1, S1, S1, S1, X, Y, Z, W), + _(B5G6R5_UNORM , R5G6B5 , S1, S1, S1, ONE, X, Y, Z, W), + _(L8_UNORM , L8 , S1, S1, S1, ONE, X, X, X, X), + _(A8_UNORM , L8 , ZERO, ZERO, ZERO, S1, X, X, X, X), + _(I8_UNORM , L8 , S1, S1, S1, S1, X, X, X, X), + _(L8A8_UNORM , A8L8 , S1, S1, S1, S1, X, X, X, Y), + _(Z16_UNORM , R5G6B5 , S1, S1, S1, ONE, X, X, X, X), + _(S8Z24_UNORM , A8R8G8B8, S1, S1, S1, ONE, X, X, X, X), + _(DXT1_RGB , DXT1 , S1, S1, S1, ONE, X, Y, Z, W), + _(DXT1_RGBA , DXT1 , S1, S1, S1, S1, X, Y, Z, W), + _(DXT3_RGBA , DXT3 , S1, S1, S1, S1, X, Y, Z, W), + _(DXT5_RGBA , DXT5 , S1, S1, S1, S1, X, Y, Z, W), + {}, +}; + +static struct nv30_texture_format * +nv30_fragtex_format(uint pipe_format) +{ + struct nv30_texture_format *tf = nv30_texture_formats; + + while (tf->defined) { + if (tf->pipe == pipe_format) + return tf; + tf++; + } + + NOUVEAU_ERR("unknown texture format %s\n", util_format_name(pipe_format)); + return NULL; +} + + +struct nouveau_stateobj * +nv30_fragtex_build(struct nvfx_context *nvfx, int unit) +{ + struct nvfx_sampler_state *ps = nvfx->tex_sampler[unit]; + struct nvfx_miptree *nv30mt = nvfx->tex_miptree[unit]; + struct pipe_texture *pt = &nv30mt->base; + struct nouveau_bo *bo = nouveau_bo(nv30mt->buffer); + struct nv30_texture_format *tf; + struct nouveau_stateobj *so; + uint32_t txf, txs; + unsigned tex_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD; + + tf = nv30_fragtex_format(pt->format); + if (!tf) + return NULL; + + txf = tf->format; + txf |= ((pt->last_level>0) ? NV34TCL_TX_FORMAT_MIPMAP : 0); + txf |= log2i(pt->width0) << NV34TCL_TX_FORMAT_BASE_SIZE_U_SHIFT; + txf |= log2i(pt->height0) << NV34TCL_TX_FORMAT_BASE_SIZE_V_SHIFT; + txf |= log2i(pt->depth0) << NV34TCL_TX_FORMAT_BASE_SIZE_W_SHIFT; + txf |= NV34TCL_TX_FORMAT_NO_BORDER | 0x10000; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE: + txf |= NV34TCL_TX_FORMAT_CUBIC; + /* fall-through */ + case PIPE_TEXTURE_2D: + txf |= NV34TCL_TX_FORMAT_DIMS_2D; + break; + case PIPE_TEXTURE_3D: + txf |= NV34TCL_TX_FORMAT_DIMS_3D; + break; + case PIPE_TEXTURE_1D: + txf |= NV34TCL_TX_FORMAT_DIMS_1D; + break; + default: + NOUVEAU_ERR("Unknown target %d\n", pt->target); + return NULL; + } + + txs = tf->swizzle; + + so = so_new(1, 8, 2); + so_method(so, nvfx->screen->eng3d, NV34TCL_TX_OFFSET(unit), 8); + so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR, + NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1); + so_data (so, ps->wrap); + so_data (so, NV34TCL_TX_ENABLE_ENABLE | ps->en); + so_data (so, txs); + so_data (so, ps->filt | 0x2000 /*voodoo*/); + so_data (so, (pt->width0 << NV34TCL_TX_NPOT_SIZE_W_SHIFT) | + pt->height0); + so_data (so, ps->bcol); + + return so; +} diff --git a/src/gallium/drivers/nvfx/nv30_vertprog.h b/src/gallium/drivers/nvfx/nv30_vertprog.h new file mode 100644 index 0000000000..ec0444c07f --- /dev/null +++ b/src/gallium/drivers/nvfx/nv30_vertprog.h @@ -0,0 +1,169 @@ +#ifndef __NV30_SHADER_H__ +#define __NV30_SHADER_H__ + +/* Vertex programs instruction set + * + * 128bit opcodes, split into 4 32-bit ones for ease of use. + * + * Non-native instructions + * ABS - MOV + NV40_VP_INST0_DEST_ABS + * POW - EX2 + MUL + LG2 + * SUB - ADD, second source negated + * SWZ - MOV + * XPD - + * + * Register access + * - Only one INPUT can be accessed per-instruction (move extras into TEMPs) + * - Only one CONST can be accessed per-instruction (move extras into TEMPs) + * + * Relative Addressing + * According to the value returned for + * MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB + * + * there are only two address registers available. The destination in the + * ARL instruction is set to TEMP (The temp isn't actually written). + * + * When using vanilla ARB_v_p, the proprietary driver will squish both the + * available ADDRESS regs into the first hardware reg in the X and Y + * components. + * + * To use an address reg as an index into consts, the CONST_SRC is set to + * (const_base + offset) and INDEX_CONST is set. + * + * To access the second address reg use ADDR_REG_SELECT_1. A particular + * component of the address regs is selected with ADDR_SWZ. + * + * Only one address register can be accessed per instruction. + * + * Conditional execution (see NV_vertex_program{2,3} for details) Conditional + * execution of an instruction is enabled by setting COND_TEST_ENABLE, and + * selecting the condition which will allow the test to pass with + * COND_{FL,LT,...}. It is possible to swizzle the values in the condition + * register, which allows for testing against an individual component. + * + * Branching: + * + * The BRA/CAL instructions seem to follow a slightly different opcode + * layout. The destination instruction ID (IADDR) overlaps a source field. + * Instruction ID's seem to be numbered based on the UPLOAD_FROM_ID FIFO + * command, and is incremented automatically on each UPLOAD_INST FIFO + * command. + * + * Conditional branching is achieved by using the condition tests described + * above. There doesn't appear to be dedicated looping instructions, but + * this can be done using a temp reg + conditional branching. + * + * Subroutines may be uploaded before the main program itself, but the first + * executed instruction is determined by the PROGRAM_START_ID FIFO command. + * + */ + +/* DWORD 0 */ + +#define NV30_VP_INST_ADDR_REG_SELECT_1 (1 << 24) +#define NV30_VP_INST_SRC2_ABS (1 << 23) /* guess */ +#define NV30_VP_INST_SRC1_ABS (1 << 22) /* guess */ +#define NV30_VP_INST_SRC0_ABS (1 << 21) /* guess */ +#define NV30_VP_INST_VEC_RESULT (1 << 20) +#define NV30_VP_INST_DEST_TEMP_ID_SHIFT 16 +#define NV30_VP_INST_DEST_TEMP_ID_MASK (0x0F << 16) +#define NV30_VP_INST_COND_UPDATE_ENABLE (1<<15) +#define NV30_VP_INST_VEC_DEST_TEMP_MASK (0xF << 16) +#define NV30_VP_INST_COND_TEST_ENABLE (1<<14) +#define NV30_VP_INST_COND_SHIFT 11 +#define NV30_VP_INST_COND_MASK (0x07 << 11) +#define NV30_VP_INST_COND_SWZ_X_SHIFT 9 +#define NV30_VP_INST_COND_SWZ_X_MASK (0x03 << 9) +#define NV30_VP_INST_COND_SWZ_Y_SHIFT 7 +#define NV30_VP_INST_COND_SWZ_Y_MASK (0x03 << 7) +#define NV30_VP_INST_COND_SWZ_Z_SHIFT 5 +#define NV30_VP_INST_COND_SWZ_Z_MASK (0x03 << 5) +#define NV30_VP_INST_COND_SWZ_W_SHIFT 3 +#define NV30_VP_INST_COND_SWZ_W_MASK (0x03 << 3) +#define NV30_VP_INST_COND_SWZ_ALL_SHIFT 3 +#define NV30_VP_INST_COND_SWZ_ALL_MASK (0xFF << 3) +#define NV30_VP_INST_ADDR_SWZ_SHIFT 1 +#define NV30_VP_INST_ADDR_SWZ_MASK (0x03 << 1) +#define NV30_VP_INST_SCA_OPCODEH_SHIFT 0 +#define NV30_VP_INST_SCA_OPCODEH_MASK (0x01 << 0) + +/* DWORD 1 */ +#define NV30_VP_INST_SCA_OPCODEL_SHIFT 28 +#define NV30_VP_INST_SCA_OPCODEL_MASK (0x0F << 28) +#define NV30_VP_INST_VEC_OPCODE_SHIFT 23 +#define NV30_VP_INST_VEC_OPCODE_MASK (0x1F << 23) +#define NV30_VP_INST_CONST_SRC_SHIFT 14 +#define NV30_VP_INST_CONST_SRC_MASK (0xFF << 14) +#define NV30_VP_INST_INPUT_SRC_SHIFT 9 /*NV20*/ +#define NV30_VP_INST_INPUT_SRC_MASK (0x0F << 9) /*NV20*/ +#define NV30_VP_INST_SRC0H_SHIFT 0 /*NV20*/ +#define NV30_VP_INST_SRC0H_MASK (0x1FF << 0) /*NV20*/ + +/* Please note: the IADDR fields overlap other fields because they are used + * only for branch instructions. See Branching: label above + * + * DWORD 2 + */ +#define NV30_VP_INST_SRC0L_SHIFT 26 /*NV20*/ +#define NV30_VP_INST_SRC0L_MASK (0x3F <<26) /* NV30_VP_SRC0_LOW_MASK << 26 */ +#define NV30_VP_INST_SRC1_SHIFT 11 /*NV20*/ +#define NV30_VP_INST_SRC1_MASK (0x7FFF<<11) /*NV20*/ +#define NV30_VP_INST_SRC2H_SHIFT 0 /*NV20*/ +#define NV30_VP_INST_SRC2H_MASK (0x7FF << 0) /* NV30_VP_SRC2_HIGH_MASK >> 4*/ +#define NV30_VP_INST_IADDR_SHIFT 2 +#define NV30_VP_INST_IADDR_MASK (0xF << 28) /* NV30_VP_SRC2_LOW_MASK << 28 */ + +/* DWORD 3 */ +#define NV30_VP_INST_SRC2L_SHIFT 28 /*NV20*/ +#define NV30_VP_INST_SRC2L_MASK (0x0F <<28) /*NV20*/ +#define NV30_VP_INST_STEMP_WRITEMASK_SHIFT 24 +#define NV30_VP_INST_STEMP_WRITEMASK_MASK (0x0F << 24) +#define NV30_VP_INST_VTEMP_WRITEMASK_SHIFT 20 +#define NV30_VP_INST_VTEMP_WRITEMASK_MASK (0x0F << 20) +#define NV30_VP_INST_SDEST_WRITEMASK_SHIFT 16 +#define NV30_VP_INST_SDEST_WRITEMASK_MASK (0x0F << 16) +#define NV30_VP_INST_VDEST_WRITEMASK_SHIFT 12 /*NV20*/ +#define NV30_VP_INST_VDEST_WRITEMASK_MASK (0x0F << 12) /*NV20*/ +#define NV30_VP_INST_DEST_SHIFT 2 +#define NV30_VP_INST_DEST_MASK (0x0F << 2) +# define NV30_VP_INST_DEST_POS 0 +# define NV30_VP_INST_DEST_BFC0 1 +# define NV30_VP_INST_DEST_BFC1 2 +# define NV30_VP_INST_DEST_COL0 3 +# define NV30_VP_INST_DEST_COL1 4 +# define NV30_VP_INST_DEST_FOGC 5 +# define NV30_VP_INST_DEST_PSZ 6 +# define NV30_VP_INST_DEST_TC(n) (8+n) + +/* Useful to split the source selection regs into their pieces */ +#define NV30_VP_SRC0_HIGH_SHIFT 6 +#define NV30_VP_SRC0_HIGH_MASK 0x00007FC0 +#define NV30_VP_SRC0_LOW_MASK 0x0000003F +#define NV30_VP_SRC2_HIGH_SHIFT 4 +#define NV30_VP_SRC2_HIGH_MASK 0x00007FF0 +#define NV30_VP_SRC2_LOW_MASK 0x0000000F + + +/* Source-register definition - matches NV20 exactly */ +#define NV30_VP_SRC_NEGATE (1<<14) +#define NV30_VP_SRC_SWZ_X_SHIFT 12 +#define NV30_VP_SRC_REG_SWZ_X_MASK (0x03 <<12) +#define NV30_VP_SRC_SWZ_Y_SHIFT 10 +#define NV30_VP_SRC_REG_SWZ_Y_MASK (0x03 <<10) +#define NV30_VP_SRC_SWZ_Z_SHIFT 8 +#define NV30_VP_SRC_REG_SWZ_Z_MASK (0x03 << 8) +#define NV30_VP_SRC_SWZ_W_SHIFT 6 +#define NV30_VP_SRC_REG_SWZ_W_MASK (0x03 << 6) +#define NV30_VP_SRC_REG_SWZ_ALL_SHIFT 6 +#define NV30_VP_SRC_REG_SWZ_ALL_MASK (0xFF << 6) +#define NV30_VP_SRC_TEMP_SRC_SHIFT 2 +#define NV30_VP_SRC_REG_TEMP_ID_MASK (0x0F << 0) +#define NV30_VP_SRC_REG_TYPE_SHIFT 0 +#define NV30_VP_SRC_REG_TYPE_MASK (0x03 << 0) +#define NV30_VP_SRC_REG_TYPE_TEMP 1 +#define NV30_VP_SRC_REG_TYPE_INPUT 2 +#define NV30_VP_SRC_REG_TYPE_CONST 3 /* guess */ + +#include "nvfx_shader.h" + +#endif diff --git a/src/gallium/drivers/nvfx/nv40_fragtex.c b/src/gallium/drivers/nvfx/nv40_fragtex.c new file mode 100644 index 0000000000..5889b5e40d --- /dev/null +++ b/src/gallium/drivers/nvfx/nv40_fragtex.c @@ -0,0 +1,174 @@ +#include "util/u_format.h" +#include "nvfx_context.h" +#include "nvfx_tex.h" + +void +nv40_sampler_state_init(struct pipe_context *pipe, + struct nvfx_sampler_state *ps, + const struct pipe_sampler_state *cso) +{ + if (cso->max_anisotropy >= 2) { + /* no idea, binary driver sets it, works without it.. meh.. */ + ps->wrap |= (1 << 5); + + if (cso->max_anisotropy >= 16) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_16X; + } else + if (cso->max_anisotropy >= 12) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_12X; + } else + if (cso->max_anisotropy >= 10) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_10X; + } else + if (cso->max_anisotropy >= 8) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_8X; + } else + if (cso->max_anisotropy >= 6) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_6X; + } else + if (cso->max_anisotropy >= 4) { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_4X; + } else { + ps->en |= NV40TCL_TEX_ENABLE_ANISO_2X; + } + } + + { + float limit; + + limit = CLAMP(cso->lod_bias, -16.0, 15.0); + ps->filt |= (int)(cso->lod_bias * 256.0) & 0x1fff; + + limit = CLAMP(cso->max_lod, 0.0, 15.0); + ps->en |= (int)(limit * 256.0) << 7; + + limit = CLAMP(cso->min_lod, 0.0, 15.0); + ps->en |= (int)(limit * 256.0) << 19; + } +} + +#define _(m,tf,ts0x,ts0y,ts0z,ts0w,ts1x,ts1y,ts1z,ts1w,sx,sy,sz,sw) \ +{ \ + TRUE, \ + PIPE_FORMAT_##m, \ + NV40TCL_TEX_FORMAT_FORMAT_##tf, \ + (NV34TCL_TX_SWIZZLE_S0_X_##ts0x | NV34TCL_TX_SWIZZLE_S0_Y_##ts0y | \ + NV34TCL_TX_SWIZZLE_S0_Z_##ts0z | NV34TCL_TX_SWIZZLE_S0_W_##ts0w | \ + NV34TCL_TX_SWIZZLE_S1_X_##ts1x | NV34TCL_TX_SWIZZLE_S1_Y_##ts1y | \ + NV34TCL_TX_SWIZZLE_S1_Z_##ts1z | NV34TCL_TX_SWIZZLE_S1_W_##ts1w), \ + ((NV34TCL_TX_FILTER_SIGNED_RED*sx) | (NV34TCL_TX_FILTER_SIGNED_GREEN*sy) | \ + (NV34TCL_TX_FILTER_SIGNED_BLUE*sz) | (NV34TCL_TX_FILTER_SIGNED_ALPHA*sw)) \ +} + +struct nv40_texture_format { + boolean defined; + uint pipe; + int format; + int swizzle; + int sign; +}; + +static struct nv40_texture_format +nv40_texture_formats[] = { + _(B8G8R8X8_UNORM, A8R8G8B8, S1, S1, S1, ONE, X, Y, Z, W, 0, 0, 0, 0), + _(B8G8R8A8_UNORM, A8R8G8B8, S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + _(B5G5R5A1_UNORM, A1R5G5B5, S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + _(B4G4R4A4_UNORM, A4R4G4B4, S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + _(B5G6R5_UNORM , R5G6B5 , S1, S1, S1, ONE, X, Y, Z, W, 0, 0, 0, 0), + _(L8_UNORM , L8 , S1, S1, S1, ONE, X, X, X, X, 0, 0, 0, 0), + _(A8_UNORM , L8 , ZERO, ZERO, ZERO, S1, X, X, X, X, 0, 0, 0, 0), + _(R16_SNORM , A16 , ZERO, ZERO, S1, ONE, X, X, X, Y, 1, 1, 1, 1), + _(I8_UNORM , L8 , S1, S1, S1, S1, X, X, X, X, 0, 0, 0, 0), + _(L8A8_UNORM , A8L8 , S1, S1, S1, S1, X, X, X, Y, 0, 0, 0, 0), + _(Z16_UNORM , Z16 , S1, S1, S1, ONE, X, X, X, X, 0, 0, 0, 0), + _(S8Z24_UNORM , Z24 , S1, S1, S1, ONE, X, X, X, X, 0, 0, 0, 0), + _(DXT1_RGB , DXT1 , S1, S1, S1, ONE, X, Y, Z, W, 0, 0, 0, 0), + _(DXT1_RGBA , DXT1 , S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + _(DXT3_RGBA , DXT3 , S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + _(DXT5_RGBA , DXT5 , S1, S1, S1, S1, X, Y, Z, W, 0, 0, 0, 0), + {}, +}; + +static struct nv40_texture_format * +nv40_fragtex_format(uint pipe_format) +{ + struct nv40_texture_format *tf = nv40_texture_formats; + + while (tf->defined) { + if (tf->pipe == pipe_format) + return tf; + tf++; + } + + NOUVEAU_ERR("unknown texture format %s\n", util_format_name(pipe_format)); + return NULL; +} + + +struct nouveau_stateobj * +nv40_fragtex_build(struct nvfx_context *nvfx, int unit) +{ + struct nvfx_sampler_state *ps = nvfx->tex_sampler[unit]; + struct nvfx_miptree *nv40mt = nvfx->tex_miptree[unit]; + struct nouveau_bo *bo = nouveau_bo(nv40mt->buffer); + struct pipe_texture *pt = &nv40mt->base; + struct nv40_texture_format *tf; + struct nouveau_stateobj *so; + uint32_t txf, txs, txp; + unsigned tex_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD; + + tf = nv40_fragtex_format(pt->format); + if (!tf) + assert(0); + + txf = ps->fmt; + txf |= tf->format | 0x8000; + txf |= ((pt->last_level + 1) << NV40TCL_TEX_FORMAT_MIPMAP_COUNT_SHIFT); + + if (1) /* XXX */ + txf |= NV34TCL_TX_FORMAT_NO_BORDER; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE: + txf |= NV34TCL_TX_FORMAT_CUBIC; + /* fall-through */ + case PIPE_TEXTURE_2D: + txf |= NV34TCL_TX_FORMAT_DIMS_2D; + break; + case PIPE_TEXTURE_3D: + txf |= NV34TCL_TX_FORMAT_DIMS_3D; + break; + case PIPE_TEXTURE_1D: + txf |= NV34TCL_TX_FORMAT_DIMS_1D; + break; + default: + NOUVEAU_ERR("Unknown target %d\n", pt->target); + return NULL; + } + + if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) { + txp = 0; + } else { + txp = nv40mt->level[0].pitch; + txf |= NV40TCL_TEX_FORMAT_LINEAR; + } + + txs = tf->swizzle; + + so = so_new(2, 9, 2); + so_method(so, nvfx->screen->eng3d, NV34TCL_TX_OFFSET(unit), 8); + so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR, + NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1); + so_data (so, ps->wrap); + so_data (so, NV40TCL_TEX_ENABLE_ENABLE | ps->en); + so_data (so, txs); + so_data (so, ps->filt | tf->sign | 0x2000 /*voodoo*/); + so_data (so, (pt->width0 << NV34TCL_TX_NPOT_SIZE_W_SHIFT) | + pt->height0); + so_data (so, ps->bcol); + so_method(so, nvfx->screen->eng3d, NV40TCL_TEX_SIZE1(unit), 1); + so_data (so, (pt->depth0 << NV40TCL_TEX_SIZE1_DEPTH_SHIFT) | txp); + + return so; +} diff --git a/src/gallium/drivers/nvfx/nv40_vertprog.h b/src/gallium/drivers/nvfx/nv40_vertprog.h new file mode 100644 index 0000000000..7337293bab --- /dev/null +++ b/src/gallium/drivers/nvfx/nv40_vertprog.h @@ -0,0 +1,177 @@ +#ifndef __NV40_SHADER_H__ +#define __NV40_SHADER_H__ + +/* Vertex programs instruction set + * + * The NV40 instruction set is very similar to NV30. Most fields are in + * a slightly different position in the instruction however. + * + * Merged instructions + * In some cases it is possible to put two instructions into one opcode + * slot. The rules for when this is OK is not entirely clear to me yet. + * + * There are separate writemasks and dest temp register fields for each + * grouping of instructions. There is however only one field with the + * ID of a result register. Writing to temp/result regs is selected by + * setting VEC_RESULT/SCA_RESULT. + * + * Temporary registers + * The source/dest temp register fields have been extended by 1 bit, to + * give a total of 32 temporary registers. + * + * Relative Addressing + * NV40 can use an address register to index into vertex attribute regs. + * This is done by putting the offset value into INPUT_SRC and setting + * the INDEX_INPUT flag. + * + * Conditional execution (see NV_vertex_program{2,3} for details) + * There is a second condition code register on NV40, it's use is enabled + * by setting the COND_REG_SELECT_1 flag. + * + * Texture lookup + * TODO + */ + +/* ---- OPCODE BITS 127:96 / data DWORD 0 --- */ +#define NV40_VP_INST_VEC_RESULT (1 << 30) +/* uncertain.. */ +#define NV40_VP_INST_COND_UPDATE_ENABLE ((1 << 14)|1<<29) +/* use address reg as index into attribs */ +#define NV40_VP_INST_INDEX_INPUT (1 << 27) +#define NV40_VP_INST_COND_REG_SELECT_1 (1 << 25) +#define NV40_VP_INST_ADDR_REG_SELECT_1 (1 << 24) +#define NV40_VP_INST_SRC2_ABS (1 << 23) +#define NV40_VP_INST_SRC1_ABS (1 << 22) +#define NV40_VP_INST_SRC0_ABS (1 << 21) +#define NV40_VP_INST_VEC_DEST_TEMP_SHIFT 15 +#define NV40_VP_INST_VEC_DEST_TEMP_MASK (0x1F << 15) +#define NV40_VP_INST_COND_TEST_ENABLE (1 << 13) +#define NV40_VP_INST_COND_SHIFT 10 +#define NV40_VP_INST_COND_MASK (0x7 << 10) +#define NV40_VP_INST_COND_SWZ_X_SHIFT 8 +#define NV40_VP_INST_COND_SWZ_X_MASK (3 << 8) +#define NV40_VP_INST_COND_SWZ_Y_SHIFT 6 +#define NV40_VP_INST_COND_SWZ_Y_MASK (3 << 6) +#define NV40_VP_INST_COND_SWZ_Z_SHIFT 4 +#define NV40_VP_INST_COND_SWZ_Z_MASK (3 << 4) +#define NV40_VP_INST_COND_SWZ_W_SHIFT 2 +#define NV40_VP_INST_COND_SWZ_W_MASK (3 << 2) +#define NV40_VP_INST_COND_SWZ_ALL_SHIFT 2 +#define NV40_VP_INST_COND_SWZ_ALL_MASK (0xFF << 2) +#define NV40_VP_INST_ADDR_SWZ_SHIFT 0 +#define NV40_VP_INST_ADDR_SWZ_MASK (0x03 << 0) +#define NV40_VP_INST0_KNOWN ( \ + NV40_VP_INST_INDEX_INPUT | \ + NV40_VP_INST_COND_REG_SELECT_1 | \ + NV40_VP_INST_ADDR_REG_SELECT_1 | \ + NV40_VP_INST_SRC2_ABS | \ + NV40_VP_INST_SRC1_ABS | \ + NV40_VP_INST_SRC0_ABS | \ + NV40_VP_INST_VEC_DEST_TEMP_MASK | \ + NV40_VP_INST_COND_TEST_ENABLE | \ + NV40_VP_INST_COND_MASK | \ + NV40_VP_INST_COND_SWZ_ALL_MASK | \ + NV40_VP_INST_ADDR_SWZ_MASK) + +/* ---- OPCODE BITS 95:64 / data DWORD 1 --- */ +#define NV40_VP_INST_VEC_OPCODE_SHIFT 22 +#define NV40_VP_INST_VEC_OPCODE_MASK (0x1F << 22) +#define NV40_VP_INST_SCA_OPCODE_SHIFT 27 +#define NV40_VP_INST_SCA_OPCODE_MASK (0x1F << 27) +#define NV40_VP_INST_CONST_SRC_SHIFT 12 +#define NV40_VP_INST_CONST_SRC_MASK (0xFF << 12) +#define NV40_VP_INST_INPUT_SRC_SHIFT 8 +#define NV40_VP_INST_INPUT_SRC_MASK (0x0F << 8) +#define NV40_VP_INST_SRC0H_SHIFT 0 +#define NV40_VP_INST_SRC0H_MASK (0xFF << 0) +#define NV40_VP_INST1_KNOWN ( \ + NV40_VP_INST_VEC_OPCODE_MASK | \ + NV40_VP_INST_SCA_OPCODE_MASK | \ + NV40_VP_INST_CONST_SRC_MASK | \ + NV40_VP_INST_INPUT_SRC_MASK | \ + NV40_VP_INST_SRC0H_MASK \ + ) + +/* ---- OPCODE BITS 63:32 / data DWORD 2 --- */ +#define NV40_VP_INST_SRC0L_SHIFT 23 +#define NV40_VP_INST_SRC0L_MASK (0x1FF << 23) +#define NV40_VP_INST_SRC1_SHIFT 6 +#define NV40_VP_INST_SRC1_MASK (0x1FFFF << 6) +#define NV40_VP_INST_SRC2H_SHIFT 0 +#define NV40_VP_INST_SRC2H_MASK (0x3F << 0) +#define NV40_VP_INST_IADDRH_SHIFT 0 +#define NV40_VP_INST_IADDRH_MASK (0x1F << 0) + +/* ---- OPCODE BITS 31:0 / data DWORD 3 --- */ +#define NV40_VP_INST_IADDRL_SHIFT 29 +#define NV40_VP_INST_IADDRL_MASK (7 << 29) +#define NV40_VP_INST_SRC2L_SHIFT 21 +#define NV40_VP_INST_SRC2L_MASK (0x7FF << 21) +#define NV40_VP_INST_SCA_WRITEMASK_SHIFT 17 +#define NV40_VP_INST_SCA_WRITEMASK_MASK (0xF << 17) +# define NV40_VP_INST_SCA_WRITEMASK_X (1 << 20) +# define NV40_VP_INST_SCA_WRITEMASK_Y (1 << 19) +# define NV40_VP_INST_SCA_WRITEMASK_Z (1 << 18) +# define NV40_VP_INST_SCA_WRITEMASK_W (1 << 17) +#define NV40_VP_INST_VEC_WRITEMASK_SHIFT 13 +#define NV40_VP_INST_VEC_WRITEMASK_MASK (0xF << 13) +# define NV40_VP_INST_VEC_WRITEMASK_X (1 << 16) +# define NV40_VP_INST_VEC_WRITEMASK_Y (1 << 15) +# define NV40_VP_INST_VEC_WRITEMASK_Z (1 << 14) +# define NV40_VP_INST_VEC_WRITEMASK_W (1 << 13) +#define NV40_VP_INST_SCA_RESULT (1 << 12) +#define NV40_VP_INST_SCA_DEST_TEMP_SHIFT 7 +#define NV40_VP_INST_SCA_DEST_TEMP_MASK (0x1F << 7) +#define NV40_VP_INST_DEST_SHIFT 2 +#define NV40_VP_INST_DEST_MASK (31 << 2) +# define NV40_VP_INST_DEST_POS 0 +# define NV40_VP_INST_DEST_COL0 1 +# define NV40_VP_INST_DEST_COL1 2 +# define NV40_VP_INST_DEST_BFC0 3 +# define NV40_VP_INST_DEST_BFC1 4 +# define NV40_VP_INST_DEST_FOGC 5 +# define NV40_VP_INST_DEST_PSZ 6 +# define NV40_VP_INST_DEST_TC0 7 +# define NV40_VP_INST_DEST_TC(n) (7+n) +# define NV40_VP_INST_DEST_TEMP 0x1F +#define NV40_VP_INST_INDEX_CONST (1 << 1) +#define NV40_VP_INST3_KNOWN ( \ + NV40_VP_INST_SRC2L_MASK |\ + NV40_VP_INST_SCA_WRITEMASK_MASK |\ + NV40_VP_INST_VEC_WRITEMASK_MASK |\ + NV40_VP_INST_SCA_DEST_TEMP_MASK |\ + NV40_VP_INST_DEST_MASK |\ + NV40_VP_INST_INDEX_CONST) + +/* Useful to split the source selection regs into their pieces */ +#define NV40_VP_SRC0_HIGH_SHIFT 9 +#define NV40_VP_SRC0_HIGH_MASK 0x0001FE00 +#define NV40_VP_SRC0_LOW_MASK 0x000001FF +#define NV40_VP_SRC2_HIGH_SHIFT 11 +#define NV40_VP_SRC2_HIGH_MASK 0x0001F800 +#define NV40_VP_SRC2_LOW_MASK 0x000007FF + +/* Source selection - these are the bits you fill NV40_VP_INST_SRCn with */ +#define NV40_VP_SRC_NEGATE (1 << 16) +#define NV40_VP_SRC_SWZ_X_SHIFT 14 +#define NV40_VP_SRC_SWZ_X_MASK (3 << 14) +#define NV40_VP_SRC_SWZ_Y_SHIFT 12 +#define NV40_VP_SRC_SWZ_Y_MASK (3 << 12) +#define NV40_VP_SRC_SWZ_Z_SHIFT 10 +#define NV40_VP_SRC_SWZ_Z_MASK (3 << 10) +#define NV40_VP_SRC_SWZ_W_SHIFT 8 +#define NV40_VP_SRC_SWZ_W_MASK (3 << 8) +#define NV40_VP_SRC_SWZ_ALL_SHIFT 8 +#define NV40_VP_SRC_SWZ_ALL_MASK (0xFF << 8) +#define NV40_VP_SRC_TEMP_SRC_SHIFT 2 +#define NV40_VP_SRC_TEMP_SRC_MASK (0x1F << 2) +#define NV40_VP_SRC_REG_TYPE_SHIFT 0 +#define NV40_VP_SRC_REG_TYPE_MASK (3 << 0) +# define NV40_VP_SRC_REG_TYPE_UNK0 0 +# define NV40_VP_SRC_REG_TYPE_TEMP 1 +# define NV40_VP_SRC_REG_TYPE_INPUT 2 +# define NV40_VP_SRC_REG_TYPE_CONST 3 + +#include "nvfx_shader.h" + +#endif diff --git a/src/gallium/drivers/nvfx/nvfx_draw.c b/src/gallium/drivers/nvfx/nvfx_draw.c index 8700e14329..7308f0667c 100644 --- a/src/gallium/drivers/nvfx/nvfx_draw.c +++ b/src/gallium/drivers/nvfx/nvfx_draw.c @@ -9,8 +9,8 @@ #include "nvfx_context.h" #define NVFX_SHADER_NO_FUCKEDNESS -#include "nv30/nv30_shader.h" -#include "nv40/nv40_shader.h" +#include "nv30_vertprog.h" +#include "nv40_vertprog.h" /* Simple, but crappy, swtnl path, hopefully we wont need to hit this very * often at all. Uses "quadro style" vertex submission + a fixed vertex diff --git a/src/gallium/drivers/nvfx/nvfx_vertprog.c b/src/gallium/drivers/nvfx/nvfx_vertprog.c index 730361a982..3d0e8c23a1 100644 --- a/src/gallium/drivers/nvfx/nvfx_vertprog.c +++ b/src/gallium/drivers/nvfx/nvfx_vertprog.c @@ -33,8 +33,8 @@ #define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W) #define DEF_SCALE 0 #define DEF_CTEST 0 -#include "nv30/nv30_shader.h" -#include "nv40/nv40_shader.h" +#include "nv30_vertprog.h" +#include "nv40_vertprog.h" #define swz(s,x,y,z,w) nvfx_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w) #define neg(s) nvfx_sr_neg((s)) -- cgit v1.2.3