#ifndef __NVFX_SHADER_H__ #define __NVFX_SHADER_H__ #include #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 */ #define NVFX_VP(c) ((NV30_VP_##c) + (nvfx->is_nv4x & ((NV40_VP_##c) - (NV30_VP_##c)))) #define NVFX_VP_INST_SLOT_VEC 0 #define NVFX_VP_INST_SLOT_SCA 1 #define NVFX_VP_INST_IN_POS 0 /* These seem to match the bindings specified in */ #define NVFX_VP_INST_IN_WEIGHT 1 /* the ARB_v_p spec (2.14.3.1) */ #define NVFX_VP_INST_IN_NORMAL 2 #define NVFX_VP_INST_IN_COL0 3 /* Should probably confirm them all though */ #define NVFX_VP_INST_IN_COL1 4 #define NVFX_VP_INST_IN_FOGC 5 #define NVFX_VP_INST_IN_TC0 8 #define NVFX_VP_INST_IN_TC(n) (8+n) #define NVFX_VP_INST_SCA_OP_NOP 0x00 #define NVFX_VP_INST_SCA_OP_MOV 0x01 #define NVFX_VP_INST_SCA_OP_RCP 0x02 #define NVFX_VP_INST_SCA_OP_RCC 0x03 #define NVFX_VP_INST_SCA_OP_RSQ 0x04 #define NVFX_VP_INST_SCA_OP_EXP 0x05 #define NVFX_VP_INST_SCA_OP_LOG 0x06 #define NVFX_VP_INST_SCA_OP_LIT 0x07 #define NVFX_VP_INST_SCA_OP_BRA 0x09 #define NVFX_VP_INST_SCA_OP_CAL 0x0B #define NVFX_VP_INST_SCA_OP_RET 0x0C #define NVFX_VP_INST_SCA_OP_LG2 0x0D #define NVFX_VP_INST_SCA_OP_EX2 0x0E #define NVFX_VP_INST_SCA_OP_SIN 0x0F #define NVFX_VP_INST_SCA_OP_COS 0x10 #define NV40_VP_INST_SCA_OP_PUSHA 0x13 #define NV40_VP_INST_SCA_OP_POPA 0x14 #define NVFX_VP_INST_VEC_OP_NOP 0x00 #define NVFX_VP_INST_VEC_OP_MOV 0x01 #define NVFX_VP_INST_VEC_OP_MUL 0x02 #define NVFX_VP_INST_VEC_OP_ADD 0x03 #define NVFX_VP_INST_VEC_OP_MAD 0x04 #define NVFX_VP_INST_VEC_OP_DP3 0x05 #define NVFX_VP_INST_VEC_OP_DPH 0x06 #define NVFX_VP_INST_VEC_OP_DP4 0x07 #define NVFX_VP_INST_VEC_OP_DST 0x08 #define NVFX_VP_INST_VEC_OP_MIN 0x09 #define NVFX_VP_INST_VEC_OP_MAX 0x0A #define NVFX_VP_INST_VEC_OP_SLT 0x0B #define NVFX_VP_INST_VEC_OP_SGE 0x0C #define NVFX_VP_INST_VEC_OP_ARL 0x0D #define NVFX_VP_INST_VEC_OP_FRC 0x0E #define NVFX_VP_INST_VEC_OP_FLR 0x0F #define NVFX_VP_INST_VEC_OP_SEQ 0x10 #define NVFX_VP_INST_VEC_OP_SFL 0x11 #define NVFX_VP_INST_VEC_OP_SGT 0x12 #define NVFX_VP_INST_VEC_OP_SLE 0x13 #define NVFX_VP_INST_VEC_OP_SNE 0x14 #define NVFX_VP_INST_VEC_OP_STR 0x15 #define NVFX_VP_INST_VEC_OP_SSG 0x16 #define NVFX_VP_INST_VEC_OP_ARR 0x17 #define NVFX_VP_INST_VEC_OP_ARA 0x18 #define NV40_VP_INST_VEC_OP_TXL 0x19 /* DWORD 3 */ #define NVFX_VP_INST_LAST (1 << 0) /* * Each fragment program opcode appears to be comprised of 4 32-bit values. * * 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 * When the fragprog contains instructions to write depth, NV30_TCL_PRIMITIVE_3D_UNK1D78=0 * otherwise it is set to 1. * * Constants are inserted directly after the instruction that uses them. * * It appears that it's not possible to use two input registers in one * instruction as the input sourcing is done in the instruction dword * and not the source selection dwords. As such instructions such as: * * ADD result.color, fragment.color, fragment.texcoord[0]; * * must be split into two MOV's and then an ADD (nvidia does this) but * I'm not sure why it's not just one MOV and then source the second input * in the ADD instruction.. * * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary * negation requires multiplication with a const. * * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO/SWIZZLE_ONE * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as SWIZZLE_ZERO * is implemented simply by not writing to the relevant components of the destination. * * Conditional execution * TODO * * Non-native instructions: * LIT * LRP - MAD+MAD * SUB - ADD, negate second source * RSQ - LG2 + EX2 * POW - LG2 + MUL + EX2 * SCS - COS + SIN * XPD * * NV40 Looping * Loops appear to be fairly expensive on NV40 at least, the proprietary * driver goes to a lot of effort to avoid using the native looping * instructions. If the total number of *executed* instructions between * REP/ENDREP or LOOP/ENDLOOP is <=500, the driver will unroll the loop. * The maximum loop count is 255. * */ //== Opcode / Destination selection == #define NVFX_FP_OP_PROGRAM_END (1 << 0) #define NVFX_FP_OP_OUT_REG_SHIFT 1 #define NV30_FP_OP_OUT_REG_MASK (31 << 1) /* uncertain */ #define NV40_FP_OP_OUT_REG_MASK (63 << 1) /* Needs to be set when writing outputs to get expected result.. */ #define NVFX_FP_OP_OUT_REG_HALF (1 << 7) #define NVFX_FP_OP_COND_WRITE_ENABLE (1 << 8) #define NVFX_FP_OP_OUTMASK_SHIFT 9 #define NVFX_FP_OP_OUTMASK_MASK (0xF << 9) # define NVFX_FP_OP_OUT_X (1<<9) # define NVFX_FP_OP_OUT_Y (1<<10) # define NVFX_FP_OP_OUT_Z (1<<11) # define NVFX_FP_OP_OUT_W (1<<12) /* Uncertain about these, especially the input_src values.. it's possible that * they can be dynamically changed. */ #define NVFX_FP_OP_INPUT_SRC_SHIFT 13 #define NVFX_FP_OP_INPUT_SRC_MASK (15 << 13) # define NVFX_FP_OP_INPUT_SRC_POSITION 0x0 # define NVFX_FP_OP_INPUT_SRC_COL0 0x1 # define NVFX_FP_OP_INPUT_SRC_COL1 0x2 # define NVFX_FP_OP_INPUT_SRC_FOGC 0x3 # define NVFX_FP_OP_INPUT_SRC_TC0 0x4 # define NVFX_FP_OP_INPUT_SRC_TC(n) (0x4 + n) # define NV40_FP_OP_INPUT_SRC_FACING 0xE #define NVFX_FP_OP_TEX_UNIT_SHIFT 17 #define NVFX_FP_OP_TEX_UNIT_MASK (0xF << 17) /* guess */ #define NVFX_FP_OP_PRECISION_SHIFT 22 #define NVFX_FP_OP_PRECISION_MASK (3 << 22) # define NVFX_FP_PRECISION_FP32 0 # define NVFX_FP_PRECISION_FP16 1 # define NVFX_FP_PRECISION_FX12 2 #define NVFX_FP_OP_OPCODE_SHIFT 24 #define NVFX_FP_OP_OPCODE_MASK (0x3F << 24) /* NV30/NV40 fragment program opcodes */ #define NVFX_FP_OP_OPCODE_NOP 0x00 #define NVFX_FP_OP_OPCODE_MOV 0x01 #define NVFX_FP_OP_OPCODE_MUL 0x02 #define NVFX_FP_OP_OPCODE_ADD 0x03 #define NVFX_FP_OP_OPCODE_MAD 0x04 #define NVFX_FP_OP_OPCODE_DP3 0x05 #define NVFX_FP_OP_OPCODE_DP4 0x06 #define NVFX_FP_OP_OPCODE_DST 0x07 #define NVFX_FP_OP_OPCODE_MIN 0x08 #define NVFX_FP_OP_OPCODE_MAX 0x09 #define NVFX_FP_OP_OPCODE_SLT 0x0A #define NVFX_FP_OP_OPCODE_SGE 0x0B #define NVFX_FP_OP_OPCODE_SLE 0x0C #define NVFX_FP_OP_OPCODE_SGT 0x0D #define NVFX_FP_OP_OPCODE_SNE 0x0E #define NVFX_FP_OP_OPCODE_SEQ 0x0F #define NVFX_FP_OP_OPCODE_FRC 0x10 #define NVFX_FP_OP_OPCODE_FLR 0x11 #define NVFX_FP_OP_OPCODE_KIL 0x12 #define NVFX_FP_OP_OPCODE_PK4B 0x13 #define NVFX_FP_OP_OPCODE_UP4B 0x14 #define NVFX_FP_OP_OPCODE_DDX 0x15 /* can only write XY */ #define NVFX_FP_OP_OPCODE_DDY 0x16 /* can only write XY */ #define NVFX_FP_OP_OPCODE_TEX 0x17 #define NVFX_FP_OP_OPCODE_TXP 0x18 #define NVFX_FP_OP_OPCODE_TXD 0x19 #define NVFX_FP_OP_OPCODE_RCP 0x1A #define NVFX_FP_OP_OPCODE_EX2 0x1C #define NVFX_FP_OP_OPCODE_LG2 0x1D #define NVFX_FP_OP_OPCODE_STR 0x20 #define NVFX_FP_OP_OPCODE_SFL 0x21 #define NVFX_FP_OP_OPCODE_COS 0x22 #define NVFX_FP_OP_OPCODE_SIN 0x23 #define NVFX_FP_OP_OPCODE_PK2H 0x24 #define NVFX_FP_OP_OPCODE_UP2H 0x25 #define NVFX_FP_OP_OPCODE_PK4UB 0x27 #define NVFX_FP_OP_OPCODE_UP4UB 0x28 #define NVFX_FP_OP_OPCODE_PK2US 0x29 #define NVFX_FP_OP_OPCODE_UP2US 0x2A #define NVFX_FP_OP_OPCODE_DP2A 0x2E #define NVFX_FP_OP_OPCODE_TXB 0x31 #define NVFX_FP_OP_OPCODE_DIV 0x3A /* NV30 only fragment program opcodes */ #define NVFX_FP_OP_OPCODE_RSQ_NV30 0x1B #define NVFX_FP_OP_OPCODE_LIT_NV30 0x1E #define NVFX_FP_OP_OPCODE_LRP_NV30 0x1F #define NVFX_FP_OP_OPCODE_POW_NV30 0x26 #define NVFX_FP_OP_OPCODE_RFL_NV30 0x36 /* NV40 only fragment program opcodes */ #define NVFX_FP_OP_OPCODE_TXL_NV40 0x2F #define NVFX_FP_OP_OPCODE_LITEX2_NV40 0x3C /* The use of these instructions appears to be indicated by bit 31 of DWORD 2.*/ #define NV40_FP_OP_BRA_OPCODE_BRK 0x0 #define NV40_FP_OP_BRA_OPCODE_CAL 0x1 #define NV40_FP_OP_BRA_OPCODE_IF 0x2 #define NV40_FP_OP_BRA_OPCODE_LOOP 0x3 #define NV40_FP_OP_BRA_OPCODE_REP 0x4 #define NV40_FP_OP_BRA_OPCODE_RET 0x5 #define NV40_FP_OP_OUT_NONE (1 << 30) #define NVFX_FP_OP_OUT_SAT (1 << 31) /* high order bits of SRC0 */ #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 #define NVFX_FP_OP_COND_SWZ_Z_MASK (3 << 25) #define NVFX_FP_OP_COND_SWZ_Y_SHIFT 23 #define NVFX_FP_OP_COND_SWZ_Y_MASK (3 << 23) #define NVFX_FP_OP_COND_SWZ_X_SHIFT 21 #define NVFX_FP_OP_COND_SWZ_X_MASK (3 << 21) #define NVFX_FP_OP_COND_SWZ_ALL_SHIFT 21 #define NVFX_FP_OP_COND_SWZ_ALL_MASK (0xFF << 21) #define NVFX_FP_OP_COND_SHIFT 18 #define NVFX_FP_OP_COND_MASK (0x07 << 18) # define NVFX_FP_OP_COND_FL 0 # define NVFX_FP_OP_COND_LT 1 # define NVFX_FP_OP_COND_EQ 2 # define NVFX_FP_OP_COND_LE 3 # define NVFX_FP_OP_COND_GT 4 # define NVFX_FP_OP_COND_NE 5 # define NVFX_FP_OP_COND_GE 6 # define NVFX_FP_OP_COND_TR 7 /* high order bits of SRC1 */ #define NV40_FP_OP_OPCODE_IS_BRANCH (1<<31) #define NVFX_FP_OP_DST_SCALE_SHIFT 28 #define NVFX_FP_OP_DST_SCALE_MASK (3 << 28) #define NVFX_FP_OP_DST_SCALE_1X 0 #define NVFX_FP_OP_DST_SCALE_2X 1 #define NVFX_FP_OP_DST_SCALE_4X 2 #define NVFX_FP_OP_DST_SCALE_8X 3 #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 #define NV40_FP_OP_LOOP_INCR_MASK (0xFF << 19) #define NV40_FP_OP_LOOP_INDEX_SHIFT 10 #define NV40_FP_OP_LOOP_INDEX_MASK (0xFF << 10) #define NV40_FP_OP_LOOP_COUNT_SHIFT 2 #define NV40_FP_OP_LOOP_COUNT_MASK (0xFF << 2) /* SRC1 IF: 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_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 * have always been filled with the same value in my tests so * far.. */ #define NV40_FP_OP_REP_COUNT1_SHIFT 2 #define NV40_FP_OP_REP_COUNT1_MASK (0xFF << 2) #define NV40_FP_OP_REP_COUNT2_SHIFT 10 #define NV40_FP_OP_REP_COUNT2_MASK (0xFF << 10) #define NV40_FP_OP_REP_COUNT3_SHIFT 19 #define NV40_FP_OP_REP_COUNT3_MASK (0xFF << 19) /* SRC2 REP/IF: 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) #define NV40_FP_OP_ADDR_INDEX_SHIFT 19 #define NV40_FP_OP_ADDR_INDEX_MASK (0xF << 19) //== Register selection == #define NVFX_FP_REG_TYPE_SHIFT 0 #define NVFX_FP_REG_TYPE_MASK (3 << 0) # define NVFX_FP_REG_TYPE_TEMP 0 # define NVFX_FP_REG_TYPE_INPUT 1 # define NVFX_FP_REG_TYPE_CONST 2 #define NVFX_FP_REG_SRC_SHIFT 2 #define NV30_FP_REG_SRC_MASK (31 << 2) #define NV40_FP_REG_SRC_MASK (63 << 2) #define NVFX_FP_REG_SRC_HALF (1 << 8) #define NVFX_FP_REG_SWZ_ALL_SHIFT 9 #define NVFX_FP_REG_SWZ_ALL_MASK (255 << 9) #define NVFX_FP_REG_SWZ_X_SHIFT 9 #define NVFX_FP_REG_SWZ_X_MASK (3 << 9) #define NVFX_FP_REG_SWZ_Y_SHIFT 11 #define NVFX_FP_REG_SWZ_Y_MASK (3 << 11) #define NVFX_FP_REG_SWZ_Z_SHIFT 13 #define NVFX_FP_REG_SWZ_Z_MASK (3 << 13) #define NVFX_FP_REG_SWZ_W_SHIFT 15 #define NVFX_FP_REG_SWZ_W_MASK (3 << 15) # define NVFX_FP_SWIZZLE_X 0 # define NVFX_FP_SWIZZLE_Y 1 # define NVFX_FP_SWIZZLE_Z 2 # define NVFX_FP_SWIZZLE_W 3 #define NVFX_FP_REG_NEGATE (1 << 17) #define NVFXSR_NONE 0 #define NVFXSR_OUTPUT 1 #define NVFXSR_INPUT 2 #define NVFXSR_TEMP 3 #define NVFXSR_CONST 5 #define NVFXSR_IMM 6 #define NVFXSR_RELOCATED 7 #define NVFX_COND_FL 0 #define NVFX_COND_LT 1 #define NVFX_COND_EQ 2 #define NVFX_COND_LE 3 #define NVFX_COND_GT 4 #define NVFX_COND_NE 5 #define NVFX_COND_GE 6 #define NVFX_COND_TR 7 /* Yes, this are ordered differently... */ #define NVFX_VP_MASK_X 8 #define NVFX_VP_MASK_Y 4 #define NVFX_VP_MASK_Z 2 #define NVFX_VP_MASK_W 1 #define NVFX_VP_MASK_ALL 0xf #define NVFX_FP_MASK_X 1 #define NVFX_FP_MASK_Y 2 #define NVFX_FP_MASK_Z 4 #define NVFX_FP_MASK_W 8 #define NVFX_FP_MASK_ALL 0xf #define NVFX_SWZ_X 0 #define NVFX_SWZ_Y 1 #define NVFX_SWZ_Z 2 #define NVFX_SWZ_W 3 #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_reg { int8_t type; uint32_t index; }; struct nvfx_src { struct nvfx_reg reg; uint8_t indirect : 1; uint8_t indirect_reg : 1; uint8_t indirect_swz : 2; uint8_t negate : 1; uint8_t abs : 1; uint8_t 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_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_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_src nvfx_src(struct nvfx_reg reg) { struct nvfx_src temp = { .reg = reg, .abs = 0, .negate = 0, .swz = { 0, 1, 2, 3 }, .indirect = 0, }; 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]; dst.swz[NVFX_SWZ_Z] = src.swz[z]; dst.swz[NVFX_SWZ_W] = src.swz[w]; return dst; } static INLINE struct nvfx_src nvfx_src_neg(struct nvfx_src src) { src.negate = !src.negate; return src; } static INLINE struct nvfx_src nvfx_src_abs(struct nvfx_src src) { src.abs = 1; return src; } struct nvfx_relocation { unsigned location; unsigned target; }; #endif