diff options
Diffstat (limited to 'src/gallium/drivers/nvfx/nvfx_shader.h')
| -rw-r--r-- | src/gallium/drivers/nvfx/nvfx_shader.h | 194 |
1 files changed, 143 insertions, 51 deletions
diff --git a/src/gallium/drivers/nvfx/nvfx_shader.h b/src/gallium/drivers/nvfx/nvfx_shader.h index 50830b39164..35006eec3d4 100644 --- a/src/gallium/drivers/nvfx/nvfx_shader.h +++ b/src/gallium/drivers/nvfx/nvfx_shader.h @@ -1,6 +1,12 @@ #ifndef __NVFX_SHADER_H__ #define __NVFX_SHADER_H__ +#include <stdint.h> + +#include "pipe/p_compiler.h" + +#define NVFX_SWZ_IDENTITY ((3 << 6) | (2 << 4) | (1 << 2) | (0 << 0)) + /* this will resolve to either the NV30 or the NV40 version * depending on the current hardware */ /* unusual, but very fast and compact method */ @@ -71,11 +77,58 @@ /* * Each fragment program opcode appears to be comprised of 4 32-bit values. * - * 0 - Opcode, output reg/mask, ATTRIB source - * 1 - Source 0 - * 2 - Source 1 - * 3 - Source 2 + * 0: OPDEST + * 0: program end + * 1-6: destination register + * 7: destination register is fp16?? (use for outputs) + * 8: set condition code + * 9: writemask x + * 10: writemask y + * 11: writemask z + * 12: writemask w + * 13-16: source attribute register number (e.g. COL0) + * 17-20: texture unit number + * 21: expand value on texture operation (x -> 2x - 1) + * 22-23: precision 0 = fp32, 1 = fp16, 2 = s1.10 fixed, 3 = s0.8 fixed (nv40-only)) + * 24-29: opcode + * 30: no destination + * 31: saturate + * 1 - SRC0 + * 0-17: see common source fields + * 18: execute if condition code less + * 19: execute if condition code equal + * 20: execute if condition code greater + * 21-22: condition code swizzle x source component + * 23-24: condition code swizzle y source component + * 25-26: condition code swizzle z source component + * 27-28: condition code swizzle w source component + * 29: source 0 absolute + * 30: always 0 in renouveau tests + * 31: always 0 in renouveau tests + * 2 - SRC1 + * 0-17: see common source fields + * 18: source 1 absolute + * 19-20: input precision 0 = fp32, 1 = fp16, 2 = s1.10 fixed, 3 = ??? + * 21-27: always 0 in renouveau tests + * 28-30: scale (0 = 1x, 1 = 2x, 2 = 4x, 3 = 8x, 4 = ???, 5, = 1/2, 6 = 1/4, 7 = 1/8) + * 31: opcode is branch + * 3 - SRC2 + * 0-17: see common source fields + * 18: source 2 absolute + * 19-29: address register displacement + * 30: use index register + * 31: disable perspective-correct interpolation? * +* Common fields of 0, 1, 2 - SRC + * 0-1: source register type (0 = temp, 1 = input, 2 = immediate, 3 = ???) + * 2-7: source temp register index + * 8: source register is fp16?? + * 9-10: source swizzle x source component + * 11-12: source swizzle y source component + * 13-14: source swizzle z source component + * 15-16: source swizzle w source component + * 17: negate + * There appears to be no special difference between result regs and temp regs. * result.color == R0.xyzw * result.depth == R1.z @@ -210,6 +263,7 @@ /* NV40 only fragment program opcodes */ #define NVFX_FP_OP_OPCODE_TXL_NV40 0x2F + /* The use of these instructions appears to be indicated by bit 31 of DWORD 2.*/ #define NV40_FP_OP_BRA_OPCODE_BRK 0x0 #define NV40_FP_OP_BRA_OPCODE_CAL 0x1 @@ -218,10 +272,11 @@ #define NV40_FP_OP_BRA_OPCODE_REP 0x4 #define NV40_FP_OP_BRA_OPCODE_RET 0x5 +#define NV40_FP_OP_OUT_NONE (1 << 30) #define NVFX_FP_OP_OUT_SAT (1 << 31) /* high order bits of SRC0 */ -#define NVFX_FP_OP_OUT_ABS (1 << 29) +#define NVFX_FP_OP_SRC0_ABS (1 << 29) #define NVFX_FP_OP_COND_SWZ_W_SHIFT 27 #define NVFX_FP_OP_COND_SWZ_W_MASK (3 << 27) #define NVFX_FP_OP_COND_SWZ_Z_SHIFT 25 @@ -254,6 +309,7 @@ #define NVFX_FP_OP_DST_SCALE_INV_2X 5 #define NVFX_FP_OP_DST_SCALE_INV_4X 6 #define NVFX_FP_OP_DST_SCALE_INV_8X 7 +#define NVFX_FP_OP_SRC1_ABS (1 << 18) /* SRC1 LOOP */ #define NV40_FP_OP_LOOP_INCR_SHIFT 19 @@ -263,13 +319,13 @@ #define NV40_FP_OP_LOOP_COUNT_SHIFT 2 #define NV40_FP_OP_LOOP_COUNT_MASK (0xFF << 2) -/* SRC1 IF */ -#define NV40_FP_OP_ELSE_ID_SHIFT 2 -#define NV40_FP_OP_ELSE_ID_MASK (0xFF << 2) +/* SRC1 IF: absolute offset in dwords */ +#define NV40_FP_OP_ELSE_OFFSET_SHIFT 0 +#define NV40_FP_OP_ELSE_OFFSET_MASK (0x7FFFFFFF << 0) /* SRC1 CAL */ -#define NV40_FP_OP_IADDR_SHIFT 2 -#define NV40_FP_OP_IADDR_MASK (0xFF << 2) +#define NV40_FP_OP_SUB_OFFSET_SHIFT 0 +#define NV40_FP_OP_SUB_OFFSET_MASK (0x7FFFFFFF << 0) /* SRC1 REP * I have no idea why there are 3 count values here.. but they @@ -283,9 +339,9 @@ #define NV40_FP_OP_REP_COUNT3_SHIFT 19 #define NV40_FP_OP_REP_COUNT3_MASK (0xFF << 19) -/* SRC2 REP/IF */ -#define NV40_FP_OP_END_ID_SHIFT 2 -#define NV40_FP_OP_END_ID_MASK (0xFF << 2) +/* SRC2 REP/IF: absolute offset in dwords */ +#define NV40_FP_OP_END_OFFSET_SHIFT 0 +#define NV40_FP_OP_END_OFFSET_MASK (0x7FFFFFFF << 0) /* high order bits of SRC2 */ #define NVFX_FP_OP_INDEX_INPUT (1 << 30) @@ -323,6 +379,7 @@ #define NVFXSR_INPUT 2 #define NVFXSR_TEMP 3 #define NVFXSR_CONST 4 +#define NVFXSR_RELOCATED 5 #define NVFX_COND_FL 0 #define NVFX_COND_LT 1 @@ -352,51 +409,88 @@ #define NVFX_SWZ_Z 2 #define NVFX_SWZ_W 3 -#define swz(s,x,y,z,w) nvfx_sr_swz((s), NVFX_SWZ_##x, NVFX_SWZ_##y, NVFX_SWZ_##z, NVFX_SWZ_##w) -#define neg(s) nvfx_sr_neg((s)) -#define abs(s) nvfx_sr_abs((s)) -#define scale(s,v) nvfx_sr_scale((s), NVFX_FP_OP_DST_SCALE_##v) +#define swz(s,x,y,z,w) nvfx_src_swz((s), NVFX_SWZ_##x, NVFX_SWZ_##y, NVFX_SWZ_##z, NVFX_SWZ_##w) +#define neg(s) nvfx_src_neg((s)) +#define abs(s) nvfx_src_abs((s)) -struct nvfx_sreg { - int type; - int index; +struct nvfx_reg { + uint8_t type; + uint32_t index; +}; - int dst_scale; +struct nvfx_src { + struct nvfx_reg reg; - int negate; - int abs; - int swz[4]; + /* src only */ + uint8_t negate : 1; + uint8_t abs : 1; + uint8_t swz[4]; +}; - int cc_update; - int cc_update_reg; - int cc_test; - int cc_test_reg; - int cc_swz[4]; +struct nvfx_insn +{ + uint8_t op; + char scale; + int8_t unit; + uint8_t mask; + uint8_t cc_swz[4]; + + uint8_t sat : 1; + uint8_t cc_update : 1; + uint8_t cc_update_reg : 1; + uint8_t cc_test : 3; + uint8_t cc_test_reg : 1; + + struct nvfx_reg dst; + struct nvfx_src src[3]; }; -static INLINE struct nvfx_sreg -nvfx_sr(int type, int index) +static INLINE struct nvfx_insn +nvfx_insn(boolean sat, unsigned op, int unit, struct nvfx_reg dst, unsigned mask, struct nvfx_src s0, struct nvfx_src s1, struct nvfx_src s2) { - struct nvfx_sreg temp = { - .type = type, - .index = index, - .dst_scale = 0, - .abs = 0, - .negate = 0, - .swz = { 0, 1, 2, 3 }, + struct nvfx_insn insn = { + .op = op, + .scale = 0, + .unit = unit, + .sat = sat, + .mask = mask, .cc_update = 0, .cc_update_reg = 0, .cc_test = NVFX_COND_TR, .cc_test_reg = 0, .cc_swz = { 0, 1, 2, 3 }, + .dst = dst, + .src = {s0, s1, s2} + }; + return insn; +} + +static INLINE struct nvfx_reg +nvfx_reg(int type, int index) +{ + struct nvfx_reg temp = { + .type = type, + .index = index, }; return temp; } -static INLINE struct nvfx_sreg -nvfx_sr_swz(struct nvfx_sreg src, int x, int y, int z, int w) +static INLINE struct nvfx_src +nvfx_src(struct nvfx_reg reg) { - struct nvfx_sreg dst = src; + struct nvfx_src temp = { + .reg = reg, + .abs = 0, + .negate = 0, + .swz = { 0, 1, 2, 3 }, + }; + return temp; +} + +static INLINE struct nvfx_src +nvfx_src_swz(struct nvfx_src src, int x, int y, int z, int w) +{ + struct nvfx_src dst = src; dst.swz[NVFX_SWZ_X] = src.swz[x]; dst.swz[NVFX_SWZ_Y] = src.swz[y]; @@ -405,25 +499,23 @@ nvfx_sr_swz(struct nvfx_sreg src, int x, int y, int z, int w) return dst; } -static INLINE struct nvfx_sreg -nvfx_sr_neg(struct nvfx_sreg src) +static INLINE struct nvfx_src +nvfx_src_neg(struct nvfx_src src) { src.negate = !src.negate; return src; } -static INLINE struct nvfx_sreg -nvfx_sr_abs(struct nvfx_sreg src) +static INLINE struct nvfx_src +nvfx_src_abs(struct nvfx_src src) { src.abs = 1; return src; } -static INLINE struct nvfx_sreg -nvfx_sr_scale(struct nvfx_sreg src, int scale) -{ - src.dst_scale = scale; - return src; -} +struct nvfx_relocation { + unsigned location; + unsigned target; +}; #endif |