From: Luca Barbieri Date: Sun, 21 Feb 2010 01:26:25 +0000 (+0100) Subject: nv30, nv40: non-trivially partially unify nv[34]0_shader.h X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=d9e396ce4a124529fa92ad967f2b3ff72534079b;p=mesa.git nv30, nv40: non-trivially partially unify nv[34]0_shader.h shader.h is similar, except for the following differences: 1. The instruction sets are not exactly the same, but mostly similar 2. Vertex program fields are in different bit positions This patch unifies all parts of nv[34]0_shader.h except the vertex program fields. Vertex opcodes are also changed so that the constant names includes SCA if it is a scalar opcode and VEC if it is a vector opcode. --- diff --git a/src/gallium/drivers/nv30/nv30_fragprog.c b/src/gallium/drivers/nv30/nv30_fragprog.c index ae246ffd647..4ce16b8f0e3 100644 --- a/src/gallium/drivers/nv30/nv30_fragprog.c +++ b/src/gallium/drivers/nv30/nv30_fragprog.c @@ -19,14 +19,14 @@ #define MASK_Z 4 #define MASK_W 8 #define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W) -#define DEF_SCALE NV30_FP_OP_DST_SCALE_1X -#define DEF_CTEST NV30_FP_OP_COND_TR -#include "nv30_shader.h" +#define DEF_SCALE NVFX_FP_OP_DST_SCALE_1X +#define DEF_CTEST NVFX_FP_OP_COND_TR +#include "nvfx_shader.h" -#define swz(s,x,y,z,w) nv30_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w) -#define neg(s) nv30_sr_neg((s)) -#define abs(s) nv30_sr_abs((s)) -#define scale(s,v) nv30_sr_scale((s), NV30_FP_OP_DST_SCALE_##v) +#define 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)) +#define abs(s) nvfx_sr_abs((s)) +#define scale(s,v) nvfx_sr_scale((s), NVFX_FP_OP_DST_SCALE_##v) #define MAX_CONSTS 128 #define MAX_IMM 32 @@ -50,21 +50,21 @@ struct nv30_fpc { } consts[MAX_CONSTS]; int nr_consts; - struct nv30_sreg imm[MAX_IMM]; + struct nvfx_sreg imm[MAX_IMM]; unsigned nr_imm; }; -static INLINE struct nv30_sreg +static INLINE struct nvfx_sreg temp(struct nv30_fpc *fpc) { int idx; idx = fpc->temp_temp_count++; idx += fpc->high_temp + 1; - return nv30_sr(NV30SR_TEMP, idx); + return nvfx_sr(NVFXSR_TEMP, idx); } -static INLINE struct nv30_sreg +static INLINE struct nvfx_sreg constant(struct nv30_fpc *fpc, int pipe, float vals[4]) { int idx; @@ -76,14 +76,14 @@ constant(struct nv30_fpc *fpc, int pipe, float vals[4]) fpc->consts[idx].pipe = pipe; if (pipe == -1) memcpy(fpc->consts[idx].vals, vals, 4 * sizeof(float)); - return nv30_sr(NV30SR_CONST, idx); + return nvfx_sr(NVFXSR_CONST, idx); } #define arith(cc,s,o,d,m,s0,s1,s2) \ - nv30_fp_arith((cc), (s), NV30_FP_OP_OPCODE_##o, \ + nv30_fp_arith((cc), (s), NVFX_FP_OP_OPCODE_##o, \ (d), (m), (s0), (s1), (s2)) #define tex(cc,s,o,u,d,m,s0,s1,s2) \ - nv30_fp_tex((cc), (s), NV30_FP_OP_OPCODE_##o, (u), \ + nv30_fp_tex((cc), (s), NVFX_FP_OP_OPCODE_##o, (u), \ (d), (m), (s0), none, none) static void @@ -96,25 +96,25 @@ grow_insns(struct nv30_fpc *fpc, int size) } static void -emit_src(struct nv30_fpc *fpc, int pos, struct nv30_sreg src) +emit_src(struct nv30_fpc *fpc, int pos, struct nvfx_sreg src) { struct nvfx_fragment_program *fp = fpc->fp; uint32_t *hw = &fp->insn[fpc->inst_offset]; uint32_t sr = 0; switch (src.type) { - case NV30SR_INPUT: - sr |= (NV30_FP_REG_TYPE_INPUT << NV30_FP_REG_TYPE_SHIFT); - hw[0] |= (src.index << NV30_FP_OP_INPUT_SRC_SHIFT); + case NVFXSR_INPUT: + sr |= (NVFX_FP_REG_TYPE_INPUT << NVFX_FP_REG_TYPE_SHIFT); + hw[0] |= (src.index << NVFX_FP_OP_INPUT_SRC_SHIFT); break; - case NV30SR_OUTPUT: - sr |= NV30_FP_REG_SRC_HALF; + case NVFXSR_OUTPUT: + sr |= NVFX_FP_REG_SRC_HALF; /* fall-through */ - case NV30SR_TEMP: - sr |= (NV30_FP_REG_TYPE_TEMP << NV30_FP_REG_TYPE_SHIFT); - sr |= (src.index << NV30_FP_REG_SRC_SHIFT); + case NVFXSR_TEMP: + sr |= (NVFX_FP_REG_TYPE_TEMP << NVFX_FP_REG_TYPE_SHIFT); + sr |= (src.index << NVFX_FP_REG_SRC_SHIFT); break; - case NV30SR_CONST: + case NVFXSR_CONST: grow_insns(fpc, 4); hw = &fp->insn[fpc->inst_offset]; if (fpc->consts[src.index].pipe >= 0) { @@ -132,61 +132,61 @@ emit_src(struct nv30_fpc *fpc, int pos, struct nv30_sreg src) sizeof(uint32_t) * 4); } - sr |= (NV30_FP_REG_TYPE_CONST << NV30_FP_REG_TYPE_SHIFT); + sr |= (NVFX_FP_REG_TYPE_CONST << NVFX_FP_REG_TYPE_SHIFT); break; - case NV30SR_NONE: - sr |= (NV30_FP_REG_TYPE_INPUT << NV30_FP_REG_TYPE_SHIFT); + case NVFXSR_NONE: + sr |= (NVFX_FP_REG_TYPE_INPUT << NVFX_FP_REG_TYPE_SHIFT); break; default: assert(0); } if (src.negate) - sr |= NV30_FP_REG_NEGATE; + sr |= NVFX_FP_REG_NEGATE; if (src.abs) hw[1] |= (1 << (29 + pos)); - sr |= ((src.swz[0] << NV30_FP_REG_SWZ_X_SHIFT) | - (src.swz[1] << NV30_FP_REG_SWZ_Y_SHIFT) | - (src.swz[2] << NV30_FP_REG_SWZ_Z_SHIFT) | - (src.swz[3] << NV30_FP_REG_SWZ_W_SHIFT)); + sr |= ((src.swz[0] << NVFX_FP_REG_SWZ_X_SHIFT) | + (src.swz[1] << NVFX_FP_REG_SWZ_Y_SHIFT) | + (src.swz[2] << NVFX_FP_REG_SWZ_Z_SHIFT) | + (src.swz[3] << NVFX_FP_REG_SWZ_W_SHIFT)); hw[pos + 1] |= sr; } static void -emit_dst(struct nv30_fpc *fpc, struct nv30_sreg dst) +emit_dst(struct nv30_fpc *fpc, struct nvfx_sreg dst) { struct nvfx_fragment_program *fp = fpc->fp; uint32_t *hw = &fp->insn[fpc->inst_offset]; switch (dst.type) { - case NV30SR_TEMP: + case NVFXSR_TEMP: if (fpc->num_regs < (dst.index + 1)) fpc->num_regs = dst.index + 1; break; - case NV30SR_OUTPUT: + case NVFXSR_OUTPUT: if (dst.index == 1) { fp->fp_control |= 0xe; } else { - hw[0] |= NV30_FP_OP_OUT_REG_HALF; + hw[0] |= NVFX_FP_OP_OUT_REG_HALF; } break; - case NV30SR_NONE: + case NVFXSR_NONE: hw[0] |= (1 << 30); break; default: assert(0); } - hw[0] |= (dst.index << NV30_FP_OP_OUT_REG_SHIFT); + hw[0] |= (dst.index << NVFX_FP_OP_OUT_REG_SHIFT); } static void nv30_fp_arith(struct nv30_fpc *fpc, int sat, int op, - struct nv30_sreg dst, int mask, - struct nv30_sreg s0, struct nv30_sreg s1, struct nv30_sreg s2) + struct nvfx_sreg dst, int mask, + struct nvfx_sreg s0, struct nvfx_sreg s1, struct nvfx_sreg s2) { struct nvfx_fragment_program *fp = fpc->fp; uint32_t *hw; @@ -196,22 +196,22 @@ nv30_fp_arith(struct nv30_fpc *fpc, int sat, int op, hw = &fp->insn[fpc->inst_offset]; memset(hw, 0, sizeof(uint32_t) * 4); - if (op == NV30_FP_OP_OPCODE_KIL) + if (op == NVFX_FP_OP_OPCODE_KIL) fp->fp_control |= NV34TCL_FP_CONTROL_USES_KIL; - hw[0] |= (op << NV30_FP_OP_OPCODE_SHIFT); - hw[0] |= (mask << NV30_FP_OP_OUTMASK_SHIFT); - hw[2] |= (dst.dst_scale << NV30_FP_OP_DST_SCALE_SHIFT); + hw[0] |= (op << NVFX_FP_OP_OPCODE_SHIFT); + hw[0] |= (mask << NVFX_FP_OP_OUTMASK_SHIFT); + hw[2] |= (dst.dst_scale << NVFX_FP_OP_DST_SCALE_SHIFT); if (sat) - hw[0] |= NV30_FP_OP_OUT_SAT; + hw[0] |= NVFX_FP_OP_OUT_SAT; if (dst.cc_update) - hw[0] |= NV30_FP_OP_COND_WRITE_ENABLE; - hw[1] |= (dst.cc_test << NV30_FP_OP_COND_SHIFT); - hw[1] |= ((dst.cc_swz[0] << NV30_FP_OP_COND_SWZ_X_SHIFT) | - (dst.cc_swz[1] << NV30_FP_OP_COND_SWZ_Y_SHIFT) | - (dst.cc_swz[2] << NV30_FP_OP_COND_SWZ_Z_SHIFT) | - (dst.cc_swz[3] << NV30_FP_OP_COND_SWZ_W_SHIFT)); + hw[0] |= NVFX_FP_OP_COND_WRITE_ENABLE; + hw[1] |= (dst.cc_test << NVFX_FP_OP_COND_SHIFT); + hw[1] |= ((dst.cc_swz[0] << NVFX_FP_OP_COND_SWZ_X_SHIFT) | + (dst.cc_swz[1] << NVFX_FP_OP_COND_SWZ_Y_SHIFT) | + (dst.cc_swz[2] << NVFX_FP_OP_COND_SWZ_Z_SHIFT) | + (dst.cc_swz[3] << NVFX_FP_OP_COND_SWZ_W_SHIFT)); emit_dst(fpc, dst); emit_src(fpc, 0, s0); @@ -221,25 +221,25 @@ nv30_fp_arith(struct nv30_fpc *fpc, int sat, int op, static void nv30_fp_tex(struct nv30_fpc *fpc, int sat, int op, int unit, - struct nv30_sreg dst, int mask, - struct nv30_sreg s0, struct nv30_sreg s1, struct nv30_sreg s2) + struct nvfx_sreg dst, int mask, + struct nvfx_sreg s0, struct nvfx_sreg s1, struct nvfx_sreg s2) { struct nvfx_fragment_program *fp = fpc->fp; nv30_fp_arith(fpc, sat, op, dst, mask, s0, s1, s2); - fp->insn[fpc->inst_offset] |= (unit << NV30_FP_OP_TEX_UNIT_SHIFT); + fp->insn[fpc->inst_offset] |= (unit << NVFX_FP_OP_TEX_UNIT_SHIFT); fp->samplers |= (1 << unit); } -static INLINE struct nv30_sreg +static INLINE struct nvfx_sreg tgsi_src(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc) { - struct nv30_sreg src; + struct nvfx_sreg src; switch (fsrc->Register.File) { case TGSI_FILE_INPUT: - src = nv30_sr(NV30SR_INPUT, + src = nvfx_sr(NVFXSR_INPUT, fpc->attrib_map[fsrc->Register.Index]); break; case TGSI_FILE_CONSTANT: @@ -250,7 +250,7 @@ tgsi_src(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc) src = fpc->imm[fsrc->Register.Index]; break; case TGSI_FILE_TEMPORARY: - src = nv30_sr(NV30SR_TEMP, fsrc->Register.Index + 1); + src = nvfx_sr(NVFXSR_TEMP, fsrc->Register.Index + 1); if (fpc->high_temp < src.index) fpc->high_temp = src.index; break; @@ -259,9 +259,9 @@ tgsi_src(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc) */ case TGSI_FILE_OUTPUT: if (fsrc->Register.Index == fpc->colour_id) - return nv30_sr(NV30SR_OUTPUT, 0); + return nvfx_sr(NVFXSR_OUTPUT, 0); else - return nv30_sr(NV30SR_OUTPUT, 1); + return nvfx_sr(NVFXSR_OUTPUT, 1); break; default: NOUVEAU_ERR("bad src file\n"); @@ -277,27 +277,27 @@ tgsi_src(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc) return src; } -static INLINE struct nv30_sreg +static INLINE struct nvfx_sreg tgsi_dst(struct nv30_fpc *fpc, const struct tgsi_full_dst_register *fdst) { int idx; switch (fdst->Register.File) { case TGSI_FILE_OUTPUT: if (fdst->Register.Index == fpc->colour_id) - return nv30_sr(NV30SR_OUTPUT, 0); + return nvfx_sr(NVFXSR_OUTPUT, 0); else - return nv30_sr(NV30SR_OUTPUT, 1); + return nvfx_sr(NVFXSR_OUTPUT, 1); break; case TGSI_FILE_TEMPORARY: idx = fdst->Register.Index + 1; if (fpc->high_temp < idx) fpc->high_temp = idx; - return nv30_sr(NV30SR_TEMP, idx); + return nvfx_sr(NVFXSR_TEMP, idx); case TGSI_FILE_NULL: - return nv30_sr(NV30SR_NONE, 0); + return nvfx_sr(NVFXSR_NONE, 0); default: NOUVEAU_ERR("bad dst file %d\n", fdst->Register.File); - return nv30_sr(NV30SR_NONE, 0); + return nvfx_sr(NVFXSR_NONE, 0); } } @@ -315,10 +315,10 @@ tgsi_mask(uint tgsi) static boolean src_native_swz(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc, - struct nv30_sreg *src) + struct nvfx_sreg *src) { - const struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0); - struct nv30_sreg tgsi = tgsi_src(fpc, fsrc); + const struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0); + struct nvfx_sreg tgsi = tgsi_src(fpc, fsrc); uint mask = 0; uint c; @@ -350,8 +350,8 @@ static boolean nv30_fragprog_parse_instruction(struct nv30_fpc *fpc, const struct tgsi_full_instruction *finst) { - const struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0); - struct nv30_sreg src[3], dst, tmp; + const struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0); + struct nvfx_sreg src[3], dst, tmp; int mask, sat, unit = 0; int ai = -1, ci = -1; int i; @@ -435,12 +435,12 @@ nv30_fragprog_parse_instruction(struct nv30_fpc *fpc, arith(fpc, sat, ADD, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_CMP: - tmp = nv30_sr(NV30SR_NONE, 0); + tmp = nvfx_sr(NVFXSR_NONE, 0); tmp.cc_update = 1; arith(fpc, 0, MOV, tmp, 0xf, src[0], none, none); - dst.cc_test = NV30_VP_INST_COND_GE; + dst.cc_test = NVFX_VP_INST_COND_GE; arith(fpc, sat, MOV, dst, mask, src[2], none, none); - dst.cc_test = NV30_VP_INST_COND_LT; + dst.cc_test = NVFX_VP_INST_COND_LT; arith(fpc, sat, MOV, dst, mask, src[1], none, none); break; case TGSI_OPCODE_COS: @@ -474,10 +474,10 @@ nv30_fragprog_parse_instruction(struct nv30_fpc *fpc, arith(fpc, 0, KIL, none, 0, none, none, none); break; case TGSI_OPCODE_KIL: - dst = nv30_sr(NV30SR_NONE, 0); + dst = nvfx_sr(NVFXSR_NONE, 0); dst.cc_update = 1; arith(fpc, 0, MOV, dst, MASK_ALL, src[0], none, none); - dst.cc_update = 0; dst.cc_test = NV30_FP_OP_COND_LT; + dst.cc_update = 0; dst.cc_test = NVFX_FP_OP_COND_LT; arith(fpc, 0, KIL, dst, 0, none, none, none); break; case TGSI_OPCODE_LG2: @@ -485,7 +485,7 @@ nv30_fragprog_parse_instruction(struct nv30_fpc *fpc, break; // case TGSI_OPCODE_LIT: case TGSI_OPCODE_LRP: - arith(fpc, sat, LRP, dst, mask, src[0], src[1], src[2]); + arith(fpc, sat, LRP_NV30, dst, mask, src[0], src[1], src[2]); break; case TGSI_OPCODE_MAD: arith(fpc, sat, MAD, dst, mask, src[0], src[1], src[2]); @@ -503,7 +503,7 @@ nv30_fragprog_parse_instruction(struct nv30_fpc *fpc, arith(fpc, sat, MUL, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_POW: - arith(fpc, sat, POW, dst, mask, src[0], src[1], none); + arith(fpc, sat, POW_NV30, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_RCP: arith(fpc, sat, RCP, dst, mask, src[0], none, none); @@ -512,10 +512,10 @@ nv30_fragprog_parse_instruction(struct nv30_fpc *fpc, assert(0); break; case TGSI_OPCODE_RFL: - arith(fpc, 0, RFL, dst, mask, src[0], src[1], none); + arith(fpc, 0, RFL_NV30, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_RSQ: - arith(fpc, sat, RSQ, dst, mask, abs(swz(src[0], X, X, X, X)), none, none); + arith(fpc, sat, RSQ_NV30, dst, mask, abs(swz(src[0], X, X, X, X)), none, none); break; case TGSI_OPCODE_SCS: /* avoid overwriting the source */ @@ -590,25 +590,25 @@ nv30_fragprog_parse_decl_attrib(struct nv30_fpc *fpc, switch (fdec->Semantic.Name) { case TGSI_SEMANTIC_POSITION: - hw = NV30_FP_OP_INPUT_SRC_POSITION; + hw = NVFX_FP_OP_INPUT_SRC_POSITION; break; case TGSI_SEMANTIC_COLOR: if (fdec->Semantic.Index == 0) { - hw = NV30_FP_OP_INPUT_SRC_COL0; + hw = NVFX_FP_OP_INPUT_SRC_COL0; } else if (fdec->Semantic.Index == 1) { - hw = NV30_FP_OP_INPUT_SRC_COL1; + hw = NVFX_FP_OP_INPUT_SRC_COL1; } else { NOUVEAU_ERR("bad colour semantic index\n"); return FALSE; } break; case TGSI_SEMANTIC_FOG: - hw = NV30_FP_OP_INPUT_SRC_FOGC; + hw = NVFX_FP_OP_INPUT_SRC_FOGC; break; case TGSI_SEMANTIC_GENERIC: if (fdec->Semantic.Index <= 7) { - hw = NV30_FP_OP_INPUT_SRC_TC(fdec->Semantic. + hw = NVFX_FP_OP_INPUT_SRC_TC(fdec->Semantic. Index); } else { NOUVEAU_ERR("bad generic semantic index\n"); @@ -702,7 +702,7 @@ nv30_fragprog_prepare(struct nv30_fpc *fpc) tgsi_parse_free(&p); /*if (++high_temp) { - fpc->r_temp = CALLOC(high_temp, sizeof(struct nv30_sreg)); + fpc->r_temp = CALLOC(high_temp, sizeof(struct nvfx_sreg)); for (i = 0; i < high_temp; i++) fpc->r_temp[i] = temp(fpc); fpc->r_temps_discard = 0; diff --git a/src/gallium/drivers/nv30/nv30_shader.h b/src/gallium/drivers/nv30/nv30_shader.h index dd3a36f78f3..f19efb5aa4d 100644 --- a/src/gallium/drivers/nv30/nv30_shader.h +++ b/src/gallium/drivers/nv30/nv30_shader.h @@ -72,14 +72,6 @@ #define NV30_VP_INST_COND_TEST_ENABLE (1<<14) #define NV30_VP_INST_COND_SHIFT 11 #define NV30_VP_INST_COND_MASK (0x07 << 11) -# define NV30_VP_INST_COND_FL 0 /* guess */ -# define NV30_VP_INST_COND_LT 1 -# define NV30_VP_INST_COND_EQ 2 -# define NV30_VP_INST_COND_LE 3 -# define NV30_VP_INST_COND_GT 4 -# define NV30_VP_INST_COND_NE 5 -# define NV30_VP_INST_COND_GE 6 -# define NV30_VP_INST_COND_TR 7 /* guess */ #define NV30_VP_INST_COND_SWZ_X_SHIFT 9 #define NV30_VP_INST_COND_SWZ_X_MASK (0x03 << 9) #define NV30_VP_INST_COND_SWZ_Y_SHIFT 7 @@ -98,59 +90,12 @@ /* DWORD 1 */ #define NV30_VP_INST_SCA_OPCODEL_SHIFT 28 #define NV30_VP_INST_SCA_OPCODEL_MASK (0x0F << 28) -# define NV30_VP_INST_OP_NOP 0x00 -# define NV30_VP_INST_OP_RCP 0x02 -# define NV30_VP_INST_OP_RCC 0x03 -# define NV30_VP_INST_OP_RSQ 0x04 -# define NV30_VP_INST_OP_EXP 0x05 -# define NV30_VP_INST_OP_LOG 0x06 -# define NV30_VP_INST_OP_LIT 0x07 -# define NV30_VP_INST_OP_BRA 0x09 -# define NV30_VP_INST_OP_CAL 0x0B -# define NV30_VP_INST_OP_RET 0x0C -# define NV30_VP_INST_OP_LG2 0x0D -# define NV30_VP_INST_OP_EX2 0x0E -# define NV30_VP_INST_OP_SIN 0x0F -# define NV30_VP_INST_OP_COS 0x10 #define NV30_VP_INST_VEC_OPCODE_SHIFT 23 #define NV30_VP_INST_VEC_OPCODE_MASK (0x1F << 23) -# define NV30_VP_INST_OP_NOPV 0x00 -# define NV30_VP_INST_OP_MOV 0x01 -# define NV30_VP_INST_OP_MUL 0x02 -# define NV30_VP_INST_OP_ADD 0x03 -# define NV30_VP_INST_OP_MAD 0x04 -# define NV30_VP_INST_OP_DP3 0x05 -# define NV30_VP_INST_OP_DP4 0x07 -# define NV30_VP_INST_OP_DPH 0x06 -# define NV30_VP_INST_OP_DST 0x08 -# define NV30_VP_INST_OP_MIN 0x09 -# define NV30_VP_INST_OP_MAX 0x0A -# define NV30_VP_INST_OP_SLT 0x0B -# define NV30_VP_INST_OP_SGE 0x0C -# define NV30_VP_INST_OP_ARL 0x0D -# define NV30_VP_INST_OP_FRC 0x0E -# define NV30_VP_INST_OP_FLR 0x0F -# define NV30_VP_INST_OP_SEQ 0x10 -# define NV30_VP_INST_OP_SFL 0x11 -# define NV30_VP_INST_OP_SGT 0x12 -# define NV30_VP_INST_OP_SLE 0x13 -# define NV30_VP_INST_OP_SNE 0x14 -# define NV30_VP_INST_OP_STR 0x15 -# define NV30_VP_INST_OP_SSG 0x16 -# define NV30_VP_INST_OP_ARR 0x17 -# define NV30_VP_INST_OP_ARA 0x18 #define NV30_VP_INST_CONST_SRC_SHIFT 14 #define NV30_VP_INST_CONST_SRC_MASK (0xFF << 14) #define NV30_VP_INST_INPUT_SRC_SHIFT 9 /*NV20*/ #define NV30_VP_INST_INPUT_SRC_MASK (0x0F << 9) /*NV20*/ -# define NV30_VP_INST_IN_POS 0 /* These seem to match the bindings specified in */ -# define NV30_VP_INST_IN_WEIGHT 1 /* the ARB_v_p spec (2.14.3.1) */ -# define NV30_VP_INST_IN_NORMAL 2 -# define NV30_VP_INST_IN_COL0 3 /* Should probably confirm them all though */ -# define NV30_VP_INST_IN_COL1 4 -# define NV30_VP_INST_IN_FOGC 5 -# define NV30_VP_INST_IN_TC0 8 -# define NV30_VP_INST_IN_TC(n) (8+n) #define NV30_VP_INST_SRC0H_SHIFT 0 /*NV20*/ #define NV30_VP_INST_SRC0H_MASK (0x1FF << 0) /*NV20*/ @@ -190,8 +135,6 @@ # define NV30_VP_INST_DEST_PSZ 6 # define NV30_VP_INST_DEST_TC(n) (8+n) -#define NV30_VP_INST_LAST (1 << 0) - /* Useful to split the source selection regs into their pieces */ #define NV30_VP_SRC0_HIGH_SHIFT 6 #define NV30_VP_SRC0_HIGH_MASK 0x00007FC0 @@ -221,270 +164,6 @@ #define NV30_VP_SRC_REG_TYPE_INPUT 2 #define NV30_VP_SRC_REG_TYPE_CONST 3 /* guess */ -/* - * Each fragment program opcode appears to be comprised of 4 32-bit values. - * - * 0 - Opcode, output reg/mask, ATTRIB source - * 1 - Source 0 - * 2 - Source 1 - * 3 - Source 2 - * - * There appears to be no special difference between result regs and temp regs. - * result.color == R0.xyzw - * result.depth == R1.z - * When the fragprog contains instructions to write depth, NV30_TCL_PRIMITIVE_3D_UNK1D78=0 - * otherwise it is set to 1. - * - * Constants are inserted directly after the instruction that uses them. - * - * It appears that it's not possible to use two input registers in one - * instruction as the input sourcing is done in the instruction dword - * and not the source selection dwords. As such instructions such as: - * - * ADD result.color, fragment.color, fragment.texcoord[0]; - * - * must be split into two MOV's and then an ADD (nvidia does this) but - * I'm not sure why it's not just one MOV and then source the second input - * in the ADD instruction.. - * - * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary - * negation requires multiplication with a const. - * - * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO/SWIZZLE_ONE - * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as SWIZZLE_ZERO - * is implemented simply by not writing to the relevant components of the destination. - * - * Conditional execution - * TODO - * - * Non-native instructions: - * LIT - * LRP - MAD+MAD - * SUB - ADD, negate second source - * RSQ - LG2 + EX2 - * POW - LG2 + MUL + EX2 - * SCS - COS + SIN - * XPD - */ - -//== Opcode / Destination selection == -#define NV30_FP_OP_PROGRAM_END (1 << 0) -#define NV30_FP_OP_OUT_REG_SHIFT 1 -#define NV30_FP_OP_OUT_REG_MASK (31 << 1) /* uncertain */ -/* Needs to be set when writing outputs to get expected result.. */ -#define NV30_FP_OP_OUT_REG_HALF (1 << 7) -#define NV30_FP_OP_COND_WRITE_ENABLE (1 << 8) -#define NV30_FP_OP_OUTMASK_SHIFT 9 -#define NV30_FP_OP_OUTMASK_MASK (0xF << 9) -# define NV30_FP_OP_OUT_X (1<<9) -# define NV30_FP_OP_OUT_Y (1<<10) -# define NV30_FP_OP_OUT_Z (1<<11) -# define NV30_FP_OP_OUT_W (1<<12) -/* Uncertain about these, especially the input_src values.. it's possible that - * they can be dynamically changed. - */ -#define NV30_FP_OP_INPUT_SRC_SHIFT 13 -#define NV30_FP_OP_INPUT_SRC_MASK (15 << 13) -# define NV30_FP_OP_INPUT_SRC_POSITION 0x0 -# define NV30_FP_OP_INPUT_SRC_COL0 0x1 -# define NV30_FP_OP_INPUT_SRC_COL1 0x2 -# define NV30_FP_OP_INPUT_SRC_FOGC 0x3 -# define NV30_FP_OP_INPUT_SRC_TC0 0x4 -# define NV30_FP_OP_INPUT_SRC_TC(n) (0x4 + n) -#define NV30_FP_OP_TEX_UNIT_SHIFT 17 -#define NV30_FP_OP_TEX_UNIT_MASK (0xF << 17) /* guess */ -#define NV30_FP_OP_PRECISION_SHIFT 22 -#define NV30_FP_OP_PRECISION_MASK (3 << 22) -# define NV30_FP_PRECISION_FP32 0 -# define NV30_FP_PRECISION_FP16 1 -# define NV30_FP_PRECISION_FX12 2 -#define NV30_FP_OP_OPCODE_SHIFT 24 -#define NV30_FP_OP_OPCODE_MASK (0x3F << 24) -# define NV30_FP_OP_OPCODE_NOP 0x00 -# define NV30_FP_OP_OPCODE_MOV 0x01 -# define NV30_FP_OP_OPCODE_MUL 0x02 -# define NV30_FP_OP_OPCODE_ADD 0x03 -# define NV30_FP_OP_OPCODE_MAD 0x04 -# define NV30_FP_OP_OPCODE_DP3 0x05 -# define NV30_FP_OP_OPCODE_DP4 0x06 -# define NV30_FP_OP_OPCODE_DST 0x07 -# define NV30_FP_OP_OPCODE_MIN 0x08 -# define NV30_FP_OP_OPCODE_MAX 0x09 -# define NV30_FP_OP_OPCODE_SLT 0x0A -# define NV30_FP_OP_OPCODE_SGE 0x0B -# define NV30_FP_OP_OPCODE_SLE 0x0C -# define NV30_FP_OP_OPCODE_SGT 0x0D -# define NV30_FP_OP_OPCODE_SNE 0x0E -# define NV30_FP_OP_OPCODE_SEQ 0x0F -# define NV30_FP_OP_OPCODE_FRC 0x10 -# define NV30_FP_OP_OPCODE_FLR 0x11 -# define NV30_FP_OP_OPCODE_KIL 0x12 -# define NV30_FP_OP_OPCODE_PK4B 0x13 -# define NV30_FP_OP_OPCODE_UP4B 0x14 -# define NV30_FP_OP_OPCODE_DDX 0x15 /* can only write XY */ -# define NV30_FP_OP_OPCODE_DDY 0x16 /* can only write XY */ -# define NV30_FP_OP_OPCODE_TEX 0x17 -# define NV30_FP_OP_OPCODE_TXP 0x18 -# define NV30_FP_OP_OPCODE_TXD 0x19 -# define NV30_FP_OP_OPCODE_RCP 0x1A -# define NV30_FP_OP_OPCODE_RSQ 0x1B -# define NV30_FP_OP_OPCODE_EX2 0x1C -# define NV30_FP_OP_OPCODE_LG2 0x1D -# define NV30_FP_OP_OPCODE_LIT 0x1E -# define NV30_FP_OP_OPCODE_LRP 0x1F -# define NV30_FP_OP_OPCODE_STR 0x20 -# define NV30_FP_OP_OPCODE_SFL 0x21 -# define NV30_FP_OP_OPCODE_COS 0x22 -# define NV30_FP_OP_OPCODE_SIN 0x23 -# define NV30_FP_OP_OPCODE_PK2H 0x24 -# define NV30_FP_OP_OPCODE_UP2H 0x25 -# define NV30_FP_OP_OPCODE_POW 0x26 -# define NV30_FP_OP_OPCODE_PK4UB 0x27 -# define NV30_FP_OP_OPCODE_UP4UB 0x28 -# define NV30_FP_OP_OPCODE_PK2US 0x29 -# define NV30_FP_OP_OPCODE_UP2US 0x2A -# define NV30_FP_OP_OPCODE_DP2A 0x2E -# define NV30_FP_OP_OPCODE_TXB 0x31 -# define NV30_FP_OP_OPCODE_RFL 0x36 -# define NV30_FP_OP_OPCODE_DIV 0x3A -#define NV30_FP_OP_OUT_SAT (1 << 31) - -/* high order bits of SRC0 */ -#define NV30_FP_OP_OUT_ABS (1 << 29) -#define NV30_FP_OP_COND_SWZ_W_SHIFT 27 -#define NV30_FP_OP_COND_SWZ_W_MASK (3 << 27) -#define NV30_FP_OP_COND_SWZ_Z_SHIFT 25 -#define NV30_FP_OP_COND_SWZ_Z_MASK (3 << 25) -#define NV30_FP_OP_COND_SWZ_Y_SHIFT 23 -#define NV30_FP_OP_COND_SWZ_Y_MASK (3 << 23) -#define NV30_FP_OP_COND_SWZ_X_SHIFT 21 -#define NV30_FP_OP_COND_SWZ_X_MASK (3 << 21) -#define NV30_FP_OP_COND_SWZ_ALL_SHIFT 21 -#define NV30_FP_OP_COND_SWZ_ALL_MASK (0xFF << 21) -#define NV30_FP_OP_COND_SHIFT 18 -#define NV30_FP_OP_COND_MASK (0x07 << 18) -# define NV30_FP_OP_COND_FL 0 -# define NV30_FP_OP_COND_LT 1 -# define NV30_FP_OP_COND_EQ 2 -# define NV30_FP_OP_COND_LE 3 -# define NV30_FP_OP_COND_GT 4 -# define NV30_FP_OP_COND_NE 5 -# define NV30_FP_OP_COND_GE 6 -# define NV30_FP_OP_COND_TR 7 - -/* high order bits of SRC1 */ -#define NV30_FP_OP_DST_SCALE_SHIFT 28 -#define NV30_FP_OP_DST_SCALE_MASK (3 << 28) -#define NV30_FP_OP_DST_SCALE_1X 0 -#define NV30_FP_OP_DST_SCALE_2X 1 -#define NV30_FP_OP_DST_SCALE_4X 2 -#define NV30_FP_OP_DST_SCALE_8X 3 -#define NV30_FP_OP_DST_SCALE_INV_2X 5 -#define NV30_FP_OP_DST_SCALE_INV_4X 6 -#define NV30_FP_OP_DST_SCALE_INV_8X 7 - - -/* high order bits of SRC2 */ -#define NV30_FP_OP_INDEX_INPUT (1 << 30) - -//== Register selection == -#define NV30_FP_REG_TYPE_SHIFT 0 -#define NV30_FP_REG_TYPE_MASK (3 << 0) -# define NV30_FP_REG_TYPE_TEMP 0 -# define NV30_FP_REG_TYPE_INPUT 1 -# define NV30_FP_REG_TYPE_CONST 2 -#define NV30_FP_REG_SRC_SHIFT 2 /* uncertain */ -#define NV30_FP_REG_SRC_MASK (31 << 2) -#define NV30_FP_REG_SRC_HALF (1 << 8) -#define NV30_FP_REG_SWZ_ALL_SHIFT 9 -#define NV30_FP_REG_SWZ_ALL_MASK (255 << 9) -#define NV30_FP_REG_SWZ_X_SHIFT 9 -#define NV30_FP_REG_SWZ_X_MASK (3 << 9) -#define NV30_FP_REG_SWZ_Y_SHIFT 11 -#define NV30_FP_REG_SWZ_Y_MASK (3 << 11) -#define NV30_FP_REG_SWZ_Z_SHIFT 13 -#define NV30_FP_REG_SWZ_Z_MASK (3 << 13) -#define NV30_FP_REG_SWZ_W_SHIFT 15 -#define NV30_FP_REG_SWZ_W_MASK (3 << 15) -# define NV30_FP_SWIZZLE_X 0 -# define NV30_FP_SWIZZLE_Y 1 -# define NV30_FP_SWIZZLE_Z 2 -# define NV30_FP_SWIZZLE_W 3 -#define NV30_FP_REG_NEGATE (1 << 17) - -#define NV30SR_NONE 0 -#define NV30SR_OUTPUT 1 -#define NV30SR_INPUT 2 -#define NV30SR_TEMP 3 -#define NV30SR_CONST 4 - -struct nv30_sreg { - int type; - int index; - - int dst_scale; - - int negate; - int abs; - int swz[4]; - - int cc_update; - int cc_update_reg; - int cc_test; - int cc_test_reg; - int cc_swz[4]; -}; - -static INLINE struct nv30_sreg -nv30_sr(int type, int index) -{ - struct nv30_sreg temp = { - .type = type, - .index = index, - .dst_scale = DEF_SCALE, - .abs = 0, - .negate = 0, - .swz = { 0, 1, 2, 3 }, - .cc_update = 0, - .cc_update_reg = 0, - .cc_test = DEF_CTEST, - .cc_test_reg = 0, - .cc_swz = { 0, 1, 2, 3 }, - }; - return temp; -} - -static INLINE struct nv30_sreg -nv30_sr_swz(struct nv30_sreg src, int x, int y, int z, int w) -{ - struct nv30_sreg dst = src; - - dst.swz[SWZ_X] = src.swz[x]; - dst.swz[SWZ_Y] = src.swz[y]; - dst.swz[SWZ_Z] = src.swz[z]; - dst.swz[SWZ_W] = src.swz[w]; - return dst; -} - -static INLINE struct nv30_sreg -nv30_sr_neg(struct nv30_sreg src) -{ - src.negate = !src.negate; - return src; -} - -static INLINE struct nv30_sreg -nv30_sr_abs(struct nv30_sreg src) -{ - src.abs = 1; - return src; -} - -static INLINE struct nv30_sreg -nv30_sr_scale(struct nv30_sreg src, int scale) -{ - src.dst_scale = scale; - return src; -} +#include "nvfx_shader.h" #endif diff --git a/src/gallium/drivers/nv30/nv30_vertprog.c b/src/gallium/drivers/nv30/nv30_vertprog.c index cf910e34b11..ec6d63889bc 100644 --- a/src/gallium/drivers/nv30/nv30_vertprog.c +++ b/src/gallium/drivers/nv30/nv30_vertprog.c @@ -35,9 +35,9 @@ #define DEF_CTEST 0 #include "nv30_shader.h" -#define swz(s,x,y,z,w) nv30_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w) -#define neg(s) nv30_sr_neg((s)) -#define abs(s) nv30_sr_abs((s)) +#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)) +#define abs(s) nvfx_sr_abs((s)) struct nv30_vpc { struct nvfx_vertex_program *vp; @@ -49,21 +49,21 @@ struct nv30_vpc { int high_temp; int temp_temp_count; - struct nv30_sreg *imm; + struct nvfx_sreg *imm; unsigned nr_imm; }; -static struct nv30_sreg +static struct nvfx_sreg temp(struct nv30_vpc *vpc) { int idx; idx = vpc->temp_temp_count++; idx += vpc->high_temp + 1; - return nv30_sr(NV30SR_TEMP, idx); + return nvfx_sr(NVFXSR_TEMP, idx); } -static struct nv30_sreg +static struct nvfx_sreg constant(struct nv30_vpc *vpc, int pipe, float x, float y, float z, float w) { struct nvfx_vertex_program *vp = vpc->vp; @@ -73,7 +73,7 @@ constant(struct nv30_vpc *vpc, int pipe, float x, float y, float z, float w) if (pipe >= 0) { for (idx = 0; idx < vp->nr_consts; idx++) { if (vp->consts[idx].index == pipe) - return nv30_sr(NV30SR_CONST, idx); + return nvfx_sr(NVFXSR_CONST, idx); } } @@ -86,37 +86,37 @@ constant(struct nv30_vpc *vpc, int pipe, float x, float y, float z, float w) vpd->value[1] = y; vpd->value[2] = z; vpd->value[3] = w; - return nv30_sr(NV30SR_CONST, idx); + return nvfx_sr(NVFXSR_CONST, idx); } #define arith(cc,s,o,d,m,s0,s1,s2) \ - nv30_vp_arith((cc), (s), NV30_VP_INST_##o, (d), (m), (s0), (s1), (s2)) + nv30_vp_arith((cc), NVFX_VP_INST_SLOT_##s, NVFX_VP_INST_##s##_OP_##o, (d), (m), (s0), (s1), (s2)) static void -emit_src(struct nv30_vpc *vpc, uint32_t *hw, int pos, struct nv30_sreg src) +emit_src(struct nv30_vpc *vpc, uint32_t *hw, int pos, struct nvfx_sreg src) { struct nvfx_vertex_program *vp = vpc->vp; uint32_t sr = 0; switch (src.type) { - case NV30SR_TEMP: + case NVFXSR_TEMP: sr |= (NV30_VP_SRC_REG_TYPE_TEMP << NV30_VP_SRC_REG_TYPE_SHIFT); sr |= (src.index << NV30_VP_SRC_TEMP_SRC_SHIFT); break; - case NV30SR_INPUT: + case NVFXSR_INPUT: sr |= (NV30_VP_SRC_REG_TYPE_INPUT << NV30_VP_SRC_REG_TYPE_SHIFT); vp->ir |= (1 << src.index); hw[1] |= (src.index << NV30_VP_INST_INPUT_SRC_SHIFT); break; - case NV30SR_CONST: + case NVFXSR_CONST: sr |= (NV30_VP_SRC_REG_TYPE_CONST << NV30_VP_SRC_REG_TYPE_SHIFT); assert(vpc->vpi->const_index == -1 || vpc->vpi->const_index == src.index); vpc->vpi->const_index = src.index; break; - case NV30SR_NONE: + case NVFXSR_NONE: sr |= (NV30_VP_SRC_REG_TYPE_INPUT << NV30_VP_SRC_REG_TYPE_SHIFT); break; @@ -164,15 +164,15 @@ emit_src(struct nv30_vpc *vpc, uint32_t *hw, int pos, struct nv30_sreg src) } static void -emit_dst(struct nv30_vpc *vpc, uint32_t *hw, int slot, struct nv30_sreg dst) +emit_dst(struct nv30_vpc *vpc, uint32_t *hw, int slot, struct nvfx_sreg dst) { struct nvfx_vertex_program *vp = vpc->vp; switch (dst.type) { - case NV30SR_TEMP: + case NVFXSR_TEMP: hw[0] |= (dst.index << NV30_VP_INST_DEST_TEMP_ID_SHIFT); break; - case NV30SR_OUTPUT: + case NVFXSR_OUTPUT: switch (dst.index) { case NV30_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break; case NV30_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break; @@ -207,9 +207,9 @@ emit_dst(struct nv30_vpc *vpc, uint32_t *hw, int slot, struct nv30_sreg dst) static void nv30_vp_arith(struct nv30_vpc *vpc, int slot, int op, - struct nv30_sreg dst, int mask, - struct nv30_sreg s0, struct nv30_sreg s1, - struct nv30_sreg s2) + struct nvfx_sreg dst, int mask, + struct nvfx_sreg s0, struct nvfx_sreg s1, + struct nvfx_sreg s2) { struct nvfx_vertex_program *vp = vpc->vp; uint32_t *hw; @@ -221,7 +221,7 @@ nv30_vp_arith(struct nv30_vpc *vpc, int slot, int op, hw = vpc->vpi->data; - hw[0] |= (NV30_VP_INST_COND_TR << NV30_VP_INST_COND_SHIFT); + hw[0] |= (NVFX_VP_INST_COND_TR << NV30_VP_INST_COND_SHIFT); hw[0] |= ((0 << NV30_VP_INST_COND_SWZ_X_SHIFT) | (1 << NV30_VP_INST_COND_SWZ_Y_SHIFT) | (2 << NV30_VP_INST_COND_SWZ_Z_SHIFT) | @@ -231,7 +231,7 @@ nv30_vp_arith(struct nv30_vpc *vpc, int slot, int op, // hw[3] |= NV30_VP_INST_SCA_DEST_TEMP_MASK; // hw[3] |= (mask << NV30_VP_INST_VEC_WRITEMASK_SHIFT); - if (dst.type == NV30SR_OUTPUT) { + if (dst.type == NVFXSR_OUTPUT) { if (slot) hw[3] |= (mask << NV30_VP_INST_SDEST_WRITEMASK_SHIFT); else @@ -249,13 +249,13 @@ nv30_vp_arith(struct nv30_vpc *vpc, int slot, int op, emit_src(vpc, hw, 2, s2); } -static INLINE struct nv30_sreg +static INLINE struct nvfx_sreg tgsi_src(struct nv30_vpc *vpc, const struct tgsi_full_src_register *fsrc) { - struct nv30_sreg src; + struct nvfx_sreg src; switch (fsrc->Register.File) { case TGSI_FILE_INPUT: - src = nv30_sr(NV30SR_INPUT, fsrc->Register.Index); + src = nvfx_sr(NVFXSR_INPUT, fsrc->Register.Index); break; case TGSI_FILE_CONSTANT: src = constant(vpc, fsrc->Register.Index, 0, 0, 0, 0); @@ -266,7 +266,7 @@ tgsi_src(struct nv30_vpc *vpc, const struct tgsi_full_src_register *fsrc) { case TGSI_FILE_TEMPORARY: if (vpc->high_temp < fsrc->Register.Index) vpc->high_temp = fsrc->Register.Index; - src = nv30_sr(NV30SR_TEMP, fsrc->Register.Index); + src = nvfx_sr(NVFXSR_TEMP, fsrc->Register.Index); break; default: NOUVEAU_ERR("bad src file\n"); @@ -282,18 +282,18 @@ tgsi_src(struct nv30_vpc *vpc, const struct tgsi_full_src_register *fsrc) { return src; } -static INLINE struct nv30_sreg +static INLINE struct nvfx_sreg tgsi_dst(struct nv30_vpc *vpc, const struct tgsi_full_dst_register *fdst) { - struct nv30_sreg dst; + struct nvfx_sreg dst; switch (fdst->Register.File) { case TGSI_FILE_OUTPUT: - dst = nv30_sr(NV30SR_OUTPUT, + dst = nvfx_sr(NVFXSR_OUTPUT, vpc->output_map[fdst->Register.Index]); break; case TGSI_FILE_TEMPORARY: - dst = nv30_sr(NV30SR_TEMP, fdst->Register.Index); + dst = nvfx_sr(NVFXSR_TEMP, fdst->Register.Index); if (vpc->high_temp < dst.index) vpc->high_temp = dst.index; break; @@ -321,8 +321,8 @@ static boolean nv30_vertprog_parse_instruction(struct nv30_vpc *vpc, const struct tgsi_full_instruction *finst) { - struct nv30_sreg src[3], dst, tmp; - struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0); + struct nvfx_sreg src[3], dst, tmp; + struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0); int mask; int ai = -1, ci = -1; int i; @@ -351,7 +351,7 @@ nv30_vertprog_parse_instruction(struct nv30_vpc *vpc, src[i] = tgsi_src(vpc, fsrc); } else { src[i] = temp(vpc); - arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + arith(vpc, VEC, MOV, src[i], MASK_ALL, tgsi_src(vpc, fsrc), none, none); } break; @@ -365,7 +365,7 @@ nv30_vertprog_parse_instruction(struct nv30_vpc *vpc, src[i] = tgsi_src(vpc, fsrc); } else { src[i] = temp(vpc); - arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + arith(vpc, VEC, MOV, src[i], MASK_ALL, tgsi_src(vpc, fsrc), none, none); } break; @@ -383,96 +383,96 @@ nv30_vertprog_parse_instruction(struct nv30_vpc *vpc, switch (finst->Instruction.Opcode) { case TGSI_OPCODE_ABS: - arith(vpc, 0, OP_MOV, dst, mask, abs(src[0]), none, none); + arith(vpc, VEC, MOV, dst, mask, abs(src[0]), none, none); break; case TGSI_OPCODE_ADD: - arith(vpc, 0, OP_ADD, dst, mask, src[0], none, src[1]); + arith(vpc, VEC, ADD, dst, mask, src[0], none, src[1]); break; case TGSI_OPCODE_ARL: - arith(vpc, 0, OP_ARL, dst, mask, src[0], none, none); + arith(vpc, VEC, ARL, dst, mask, src[0], none, none); break; case TGSI_OPCODE_DP3: - arith(vpc, 0, OP_DP3, dst, mask, src[0], src[1], none); + arith(vpc, VEC, DP3, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_DP4: - arith(vpc, 0, OP_DP4, dst, mask, src[0], src[1], none); + arith(vpc, VEC, DP4, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_DPH: - arith(vpc, 0, OP_DPH, dst, mask, src[0], src[1], none); + arith(vpc, VEC, DPH, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_DST: - arith(vpc, 0, OP_DST, dst, mask, src[0], src[1], none); + arith(vpc, VEC, DST, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_EX2: - arith(vpc, 1, OP_EX2, dst, mask, none, none, src[0]); + arith(vpc, SCA, EX2, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_EXP: - arith(vpc, 1, OP_EXP, dst, mask, none, none, src[0]); + arith(vpc, SCA, EXP, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_FLR: - arith(vpc, 0, OP_FLR, dst, mask, src[0], none, none); + arith(vpc, VEC, FLR, dst, mask, src[0], none, none); break; case TGSI_OPCODE_FRC: - arith(vpc, 0, OP_FRC, dst, mask, src[0], none, none); + arith(vpc, VEC, FRC, dst, mask, src[0], none, none); break; case TGSI_OPCODE_LG2: - arith(vpc, 1, OP_LG2, dst, mask, none, none, src[0]); + arith(vpc, SCA, LG2, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_LIT: - arith(vpc, 1, OP_LIT, dst, mask, none, none, src[0]); + arith(vpc, SCA, LIT, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_LOG: - arith(vpc, 1, OP_LOG, dst, mask, none, none, src[0]); + arith(vpc, SCA, LOG, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_MAD: - arith(vpc, 0, OP_MAD, dst, mask, src[0], src[1], src[2]); + arith(vpc, VEC, MAD, dst, mask, src[0], src[1], src[2]); break; case TGSI_OPCODE_MAX: - arith(vpc, 0, OP_MAX, dst, mask, src[0], src[1], none); + arith(vpc, VEC, MAX, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_MIN: - arith(vpc, 0, OP_MIN, dst, mask, src[0], src[1], none); + arith(vpc, VEC, MIN, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_MOV: - arith(vpc, 0, OP_MOV, dst, mask, src[0], none, none); + arith(vpc, VEC, MOV, dst, mask, src[0], none, none); break; case TGSI_OPCODE_MUL: - arith(vpc, 0, OP_MUL, dst, mask, src[0], src[1], none); + arith(vpc, VEC, MUL, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_POW: tmp = temp(vpc); - arith(vpc, 1, OP_LG2, tmp, MASK_X, none, none, + arith(vpc, SCA, LG2, tmp, MASK_X, none, none, swz(src[0], X, X, X, X)); - arith(vpc, 0, OP_MUL, tmp, MASK_X, swz(tmp, X, X, X, X), + arith(vpc, VEC, MUL, tmp, MASK_X, swz(tmp, X, X, X, X), swz(src[1], X, X, X, X), none); - arith(vpc, 1, OP_EX2, dst, mask, none, none, + arith(vpc, SCA, EX2, dst, mask, none, none, swz(tmp, X, X, X, X)); break; case TGSI_OPCODE_RCP: - arith(vpc, 1, OP_RCP, dst, mask, none, none, src[0]); + arith(vpc, SCA, RCP, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_RET: break; case TGSI_OPCODE_RSQ: - arith(vpc, 1, OP_RSQ, dst, mask, none, none, src[0]); + arith(vpc, SCA, RSQ, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_SGE: - arith(vpc, 0, OP_SGE, dst, mask, src[0], src[1], none); + arith(vpc, VEC, SGE, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_SGT: - arith(vpc, 0, OP_SGT, dst, mask, src[0], src[1], none); + arith(vpc, VEC, SGT, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_SLT: - arith(vpc, 0, OP_SLT, dst, mask, src[0], src[1], none); + arith(vpc, VEC, SLT, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_SUB: - arith(vpc, 0, OP_ADD, dst, mask, src[0], none, neg(src[1])); + arith(vpc, VEC, ADD, dst, mask, src[0], none, neg(src[1])); break; case TGSI_OPCODE_XPD: tmp = temp(vpc); - arith(vpc, 0, OP_MUL, tmp, mask, + arith(vpc, VEC, MUL, tmp, mask, swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none); - arith(vpc, 0, OP_MAD, dst, (mask & ~MASK_W), + arith(vpc, VEC, MAD, dst, (mask & ~MASK_W), swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y), neg(tmp)); break; @@ -564,7 +564,7 @@ nv30_vertprog_prepare(struct nv30_vpc *vpc) tgsi_parse_free(&p); if (nr_imm) { - vpc->imm = CALLOC(nr_imm, sizeof(struct nv30_sreg)); + vpc->imm = CALLOC(nr_imm, sizeof(struct nvfx_sreg)); assert(vpc->imm); } @@ -639,7 +639,7 @@ nv30_vertprog_translate(struct nvfx_context *nvfx, } } - vp->insns[vp->nr_insns - 1].data[3] |= NV30_VP_INST_LAST; + vp->insns[vp->nr_insns - 1].data[3] |= NVFX_VP_INST_LAST; vp->translated = TRUE; out_err: tgsi_parse_free(&parse); diff --git a/src/gallium/drivers/nv40/nv40_draw.c b/src/gallium/drivers/nv40/nv40_draw.c index 87d2689d54b..4ed87779fd6 100644 --- a/src/gallium/drivers/nv40/nv40_draw.c +++ b/src/gallium/drivers/nv40/nv40_draw.c @@ -8,7 +8,7 @@ #include "draw/draw_pipe.h" #include "nv40_context.h" -#define NV40_SHADER_NO_FUCKEDNESS +#define NVFX_SHADER_NO_FUCKEDNESS #include "nv40_shader.h" /* Simple, but crappy, swtnl path, hopefully we wont need to hit this very diff --git a/src/gallium/drivers/nv40/nv40_fragprog.c b/src/gallium/drivers/nv40/nv40_fragprog.c index 2a0ab0cf310..e044f367a0b 100644 --- a/src/gallium/drivers/nv40/nv40_fragprog.c +++ b/src/gallium/drivers/nv40/nv40_fragprog.c @@ -18,14 +18,14 @@ #define MASK_Z 4 #define MASK_W 8 #define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W) -#define DEF_SCALE NV40_FP_OP_DST_SCALE_1X -#define DEF_CTEST NV40_FP_OP_COND_TR -#include "nv40_shader.h" +#define DEF_SCALE NVFX_FP_OP_DST_SCALE_1X +#define DEF_CTEST NVFX_FP_OP_COND_TR +#include "nvfx_shader.h" -#define swz(s,x,y,z,w) nv40_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w) -#define neg(s) nv40_sr_neg((s)) -#define abs(s) nv40_sr_abs((s)) -#define scale(s,v) nv40_sr_scale((s), NV40_FP_OP_DST_SCALE_##v) +#define 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)) +#define abs(s) nvfx_sr_abs((s)) +#define scale(s,v) nvfx_sr_scale((s), NVFX_FP_OP_DST_SCALE_##v) #define MAX_CONSTS 128 #define MAX_IMM 32 @@ -36,8 +36,8 @@ struct nv40_fpc { unsigned r_temps; unsigned r_temps_discard; - struct nv40_sreg r_result[PIPE_MAX_SHADER_OUTPUTS]; - struct nv40_sreg *r_temp; + struct nvfx_sreg r_result[PIPE_MAX_SHADER_OUTPUTS]; + struct nvfx_sreg *r_temp; int num_regs; @@ -50,11 +50,11 @@ struct nv40_fpc { } consts[MAX_CONSTS]; int nr_consts; - struct nv40_sreg imm[MAX_IMM]; + struct nvfx_sreg imm[MAX_IMM]; unsigned nr_imm; }; -static INLINE struct nv40_sreg +static INLINE struct nvfx_sreg temp(struct nv40_fpc *fpc) { int idx = ffs(~fpc->r_temps) - 1; @@ -62,12 +62,12 @@ temp(struct nv40_fpc *fpc) if (idx < 0) { NOUVEAU_ERR("out of temps!!\n"); assert(0); - return nv40_sr(NV40SR_TEMP, 0); + return nvfx_sr(NVFXSR_TEMP, 0); } fpc->r_temps |= (1 << idx); fpc->r_temps_discard |= (1 << idx); - return nv40_sr(NV40SR_TEMP, idx); + return nvfx_sr(NVFXSR_TEMP, idx); } static INLINE void @@ -77,7 +77,7 @@ release_temps(struct nv40_fpc *fpc) fpc->r_temps_discard = 0; } -static INLINE struct nv40_sreg +static INLINE struct nvfx_sreg constant(struct nv40_fpc *fpc, int pipe, float vals[4]) { int idx; @@ -89,14 +89,14 @@ constant(struct nv40_fpc *fpc, int pipe, float vals[4]) fpc->consts[idx].pipe = pipe; if (pipe == -1) memcpy(fpc->consts[idx].vals, vals, 4 * sizeof(float)); - return nv40_sr(NV40SR_CONST, idx); + return nvfx_sr(NVFXSR_CONST, idx); } #define arith(cc,s,o,d,m,s0,s1,s2) \ - nv40_fp_arith((cc), (s), NV40_FP_OP_OPCODE_##o, \ + nv40_fp_arith((cc), (s), NVFX_FP_OP_OPCODE_##o, \ (d), (m), (s0), (s1), (s2)) #define tex(cc,s,o,u,d,m,s0,s1,s2) \ - nv40_fp_tex((cc), (s), NV40_FP_OP_OPCODE_##o, (u), \ + nv40_fp_tex((cc), (s), NVFX_FP_OP_OPCODE_##o, (u), \ (d), (m), (s0), none, none) static void @@ -109,25 +109,25 @@ grow_insns(struct nv40_fpc *fpc, int size) } static void -emit_src(struct nv40_fpc *fpc, int pos, struct nv40_sreg src) +emit_src(struct nv40_fpc *fpc, int pos, struct nvfx_sreg src) { struct nvfx_fragment_program *fp = fpc->fp; uint32_t *hw = &fp->insn[fpc->inst_offset]; uint32_t sr = 0; switch (src.type) { - case NV40SR_INPUT: - sr |= (NV40_FP_REG_TYPE_INPUT << NV40_FP_REG_TYPE_SHIFT); - hw[0] |= (src.index << NV40_FP_OP_INPUT_SRC_SHIFT); + case NVFXSR_INPUT: + sr |= (NVFX_FP_REG_TYPE_INPUT << NVFX_FP_REG_TYPE_SHIFT); + hw[0] |= (src.index << NVFX_FP_OP_INPUT_SRC_SHIFT); break; - case NV40SR_OUTPUT: - sr |= NV40_FP_REG_SRC_HALF; + case NVFXSR_OUTPUT: + sr |= NVFX_FP_REG_SRC_HALF; /* fall-through */ - case NV40SR_TEMP: - sr |= (NV40_FP_REG_TYPE_TEMP << NV40_FP_REG_TYPE_SHIFT); - sr |= (src.index << NV40_FP_REG_SRC_SHIFT); + case NVFXSR_TEMP: + sr |= (NVFX_FP_REG_TYPE_TEMP << NVFX_FP_REG_TYPE_SHIFT); + sr |= (src.index << NVFX_FP_REG_SRC_SHIFT); break; - case NV40SR_CONST: + case NVFXSR_CONST: if (!fpc->have_const) { grow_insns(fpc, 4); fpc->have_const = 1; @@ -149,61 +149,61 @@ emit_src(struct nv40_fpc *fpc, int pos, struct nv40_sreg src) sizeof(uint32_t) * 4); } - sr |= (NV40_FP_REG_TYPE_CONST << NV40_FP_REG_TYPE_SHIFT); + sr |= (NVFX_FP_REG_TYPE_CONST << NVFX_FP_REG_TYPE_SHIFT); break; - case NV40SR_NONE: - sr |= (NV40_FP_REG_TYPE_INPUT << NV40_FP_REG_TYPE_SHIFT); + case NVFXSR_NONE: + sr |= (NVFX_FP_REG_TYPE_INPUT << NVFX_FP_REG_TYPE_SHIFT); break; default: assert(0); } if (src.negate) - sr |= NV40_FP_REG_NEGATE; + sr |= NVFX_FP_REG_NEGATE; if (src.abs) hw[1] |= (1 << (29 + pos)); - sr |= ((src.swz[0] << NV40_FP_REG_SWZ_X_SHIFT) | - (src.swz[1] << NV40_FP_REG_SWZ_Y_SHIFT) | - (src.swz[2] << NV40_FP_REG_SWZ_Z_SHIFT) | - (src.swz[3] << NV40_FP_REG_SWZ_W_SHIFT)); + sr |= ((src.swz[0] << NVFX_FP_REG_SWZ_X_SHIFT) | + (src.swz[1] << NVFX_FP_REG_SWZ_Y_SHIFT) | + (src.swz[2] << NVFX_FP_REG_SWZ_Z_SHIFT) | + (src.swz[3] << NVFX_FP_REG_SWZ_W_SHIFT)); hw[pos + 1] |= sr; } static void -emit_dst(struct nv40_fpc *fpc, struct nv40_sreg dst) +emit_dst(struct nv40_fpc *fpc, struct nvfx_sreg dst) { struct nvfx_fragment_program *fp = fpc->fp; uint32_t *hw = &fp->insn[fpc->inst_offset]; switch (dst.type) { - case NV40SR_TEMP: + case NVFXSR_TEMP: if (fpc->num_regs < (dst.index + 1)) fpc->num_regs = dst.index + 1; break; - case NV40SR_OUTPUT: + case NVFXSR_OUTPUT: if (dst.index == 1) { fp->fp_control |= 0xe; } else { - hw[0] |= NV40_FP_OP_OUT_REG_HALF; + hw[0] |= NVFX_FP_OP_OUT_REG_HALF; } break; - case NV40SR_NONE: + case NVFXSR_NONE: hw[0] |= (1 << 30); break; default: assert(0); } - hw[0] |= (dst.index << NV40_FP_OP_OUT_REG_SHIFT); + hw[0] |= (dst.index << NVFX_FP_OP_OUT_REG_SHIFT); } static void nv40_fp_arith(struct nv40_fpc *fpc, int sat, int op, - struct nv40_sreg dst, int mask, - struct nv40_sreg s0, struct nv40_sreg s1, struct nv40_sreg s2) + struct nvfx_sreg dst, int mask, + struct nvfx_sreg s0, struct nvfx_sreg s1, struct nvfx_sreg s2) { struct nvfx_fragment_program *fp = fpc->fp; uint32_t *hw; @@ -214,22 +214,22 @@ nv40_fp_arith(struct nv40_fpc *fpc, int sat, int op, hw = &fp->insn[fpc->inst_offset]; memset(hw, 0, sizeof(uint32_t) * 4); - if (op == NV40_FP_OP_OPCODE_KIL) + if (op == NVFX_FP_OP_OPCODE_KIL) fp->fp_control |= NV40TCL_FP_CONTROL_KIL; - hw[0] |= (op << NV40_FP_OP_OPCODE_SHIFT); - hw[0] |= (mask << NV40_FP_OP_OUTMASK_SHIFT); - hw[2] |= (dst.dst_scale << NV40_FP_OP_DST_SCALE_SHIFT); + hw[0] |= (op << NVFX_FP_OP_OPCODE_SHIFT); + hw[0] |= (mask << NVFX_FP_OP_OUTMASK_SHIFT); + hw[2] |= (dst.dst_scale << NVFX_FP_OP_DST_SCALE_SHIFT); if (sat) - hw[0] |= NV40_FP_OP_OUT_SAT; + hw[0] |= NVFX_FP_OP_OUT_SAT; if (dst.cc_update) - hw[0] |= NV40_FP_OP_COND_WRITE_ENABLE; - hw[1] |= (dst.cc_test << NV40_FP_OP_COND_SHIFT); - hw[1] |= ((dst.cc_swz[0] << NV40_FP_OP_COND_SWZ_X_SHIFT) | - (dst.cc_swz[1] << NV40_FP_OP_COND_SWZ_Y_SHIFT) | - (dst.cc_swz[2] << NV40_FP_OP_COND_SWZ_Z_SHIFT) | - (dst.cc_swz[3] << NV40_FP_OP_COND_SWZ_W_SHIFT)); + hw[0] |= NVFX_FP_OP_COND_WRITE_ENABLE; + hw[1] |= (dst.cc_test << NVFX_FP_OP_COND_SHIFT); + hw[1] |= ((dst.cc_swz[0] << NVFX_FP_OP_COND_SWZ_X_SHIFT) | + (dst.cc_swz[1] << NVFX_FP_OP_COND_SWZ_Y_SHIFT) | + (dst.cc_swz[2] << NVFX_FP_OP_COND_SWZ_Z_SHIFT) | + (dst.cc_swz[3] << NVFX_FP_OP_COND_SWZ_W_SHIFT)); emit_dst(fpc, dst); emit_src(fpc, 0, s0); @@ -239,25 +239,25 @@ nv40_fp_arith(struct nv40_fpc *fpc, int sat, int op, static void nv40_fp_tex(struct nv40_fpc *fpc, int sat, int op, int unit, - struct nv40_sreg dst, int mask, - struct nv40_sreg s0, struct nv40_sreg s1, struct nv40_sreg s2) + struct nvfx_sreg dst, int mask, + struct nvfx_sreg s0, struct nvfx_sreg s1, struct nvfx_sreg s2) { struct nvfx_fragment_program *fp = fpc->fp; nv40_fp_arith(fpc, sat, op, dst, mask, s0, s1, s2); - fp->insn[fpc->inst_offset] |= (unit << NV40_FP_OP_TEX_UNIT_SHIFT); + fp->insn[fpc->inst_offset] |= (unit << NVFX_FP_OP_TEX_UNIT_SHIFT); fp->samplers |= (1 << unit); } -static INLINE struct nv40_sreg +static INLINE struct nvfx_sreg tgsi_src(struct nv40_fpc *fpc, const struct tgsi_full_src_register *fsrc) { - struct nv40_sreg src; + struct nvfx_sreg src; switch (fsrc->Register.File) { case TGSI_FILE_INPUT: - src = nv40_sr(NV40SR_INPUT, + src = nvfx_sr(NVFXSR_INPUT, fpc->attrib_map[fsrc->Register.Index]); break; case TGSI_FILE_CONSTANT: @@ -288,7 +288,7 @@ tgsi_src(struct nv40_fpc *fpc, const struct tgsi_full_src_register *fsrc) return src; } -static INLINE struct nv40_sreg +static INLINE struct nvfx_sreg tgsi_dst(struct nv40_fpc *fpc, const struct tgsi_full_dst_register *fdst) { switch (fdst->Register.File) { case TGSI_FILE_OUTPUT: @@ -296,10 +296,10 @@ tgsi_dst(struct nv40_fpc *fpc, const struct tgsi_full_dst_register *fdst) { case TGSI_FILE_TEMPORARY: return fpc->r_temp[fdst->Register.Index]; case TGSI_FILE_NULL: - return nv40_sr(NV40SR_NONE, 0); + return nvfx_sr(NVFXSR_NONE, 0); default: NOUVEAU_ERR("bad dst file %d\n", fdst->Register.File); - return nv40_sr(NV40SR_NONE, 0); + return nvfx_sr(NVFXSR_NONE, 0); } } @@ -317,10 +317,10 @@ tgsi_mask(uint tgsi) static boolean src_native_swz(struct nv40_fpc *fpc, const struct tgsi_full_src_register *fsrc, - struct nv40_sreg *src) + struct nvfx_sreg *src) { - const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0); - struct nv40_sreg tgsi = tgsi_src(fpc, fsrc); + const struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0); + struct nvfx_sreg tgsi = tgsi_src(fpc, fsrc); uint mask = 0; uint c; @@ -352,8 +352,8 @@ static boolean nv40_fragprog_parse_instruction(struct nv40_fpc *fpc, const struct tgsi_full_instruction *finst) { - const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0); - struct nv40_sreg src[3], dst, tmp; + const struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0); + struct nvfx_sreg src[3], dst, tmp; int mask, sat, unit; int ai = -1, ci = -1, ii = -1; int i; @@ -445,12 +445,12 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc, arith(fpc, sat, ADD, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_CMP: - tmp = nv40_sr(NV40SR_NONE, 0); + tmp = nvfx_sr(NVFXSR_NONE, 0); tmp.cc_update = 1; arith(fpc, 0, MOV, tmp, 0xf, src[0], none, none); - dst.cc_test = NV40_VP_INST_COND_GE; + dst.cc_test = NVFX_VP_INST_COND_GE; arith(fpc, sat, MOV, dst, mask, src[2], none, none); - dst.cc_test = NV40_VP_INST_COND_LT; + dst.cc_test = NVFX_VP_INST_COND_LT; arith(fpc, sat, MOV, dst, mask, src[1], none, none); break; case TGSI_OPCODE_COS: @@ -512,10 +512,10 @@ nv40_fragprog_parse_instruction(struct nv40_fpc *fpc, arith(fpc, 0, KIL, none, 0, none, none, none); break; case TGSI_OPCODE_KIL: - dst = nv40_sr(NV40SR_NONE, 0); + dst = nvfx_sr(NVFXSR_NONE, 0); dst.cc_update = 1; arith(fpc, 0, MOV, dst, MASK_ALL, src[0], none, none); - dst.cc_update = 0; dst.cc_test = NV40_FP_OP_COND_LT; + dst.cc_update = 0; dst.cc_test = NVFX_FP_OP_COND_LT; arith(fpc, 0, KIL, dst, 0, none, none, none); break; case TGSI_OPCODE_LG2: @@ -662,25 +662,25 @@ nv40_fragprog_parse_decl_attrib(struct nv40_fpc *fpc, switch (fdec->Semantic.Name) { case TGSI_SEMANTIC_POSITION: - hw = NV40_FP_OP_INPUT_SRC_POSITION; + hw = NVFX_FP_OP_INPUT_SRC_POSITION; break; case TGSI_SEMANTIC_COLOR: if (fdec->Semantic.Index == 0) { - hw = NV40_FP_OP_INPUT_SRC_COL0; + hw = NVFX_FP_OP_INPUT_SRC_COL0; } else if (fdec->Semantic.Index == 1) { - hw = NV40_FP_OP_INPUT_SRC_COL1; + hw = NVFX_FP_OP_INPUT_SRC_COL1; } else { NOUVEAU_ERR("bad colour semantic index\n"); return FALSE; } break; case TGSI_SEMANTIC_FOG: - hw = NV40_FP_OP_INPUT_SRC_FOGC; + hw = NVFX_FP_OP_INPUT_SRC_FOGC; break; case TGSI_SEMANTIC_GENERIC: if (fdec->Semantic.Index <= 7) { - hw = NV40_FP_OP_INPUT_SRC_TC(fdec->Semantic. + hw = NVFX_FP_OP_INPUT_SRC_TC(fdec->Semantic. Index); } else { NOUVEAU_ERR("bad generic semantic index\n"); @@ -723,7 +723,7 @@ nv40_fragprog_parse_decl_output(struct nv40_fpc *fpc, return FALSE; } - fpc->r_result[idx] = nv40_sr(NV40SR_OUTPUT, hw); + fpc->r_result[idx] = nvfx_sr(NVFXSR_OUTPUT, hw); fpc->r_temps |= (1 << hw); return TRUE; } @@ -787,7 +787,7 @@ nv40_fragprog_prepare(struct nv40_fpc *fpc) tgsi_parse_free(&p); if (++high_temp) { - fpc->r_temp = CALLOC(high_temp, sizeof(struct nv40_sreg)); + fpc->r_temp = CALLOC(high_temp, sizeof(struct nvfx_sreg)); for (i = 0; i < high_temp; i++) fpc->r_temp[i] = temp(fpc); fpc->r_temps_discard = 0; diff --git a/src/gallium/drivers/nv40/nv40_shader.h b/src/gallium/drivers/nv40/nv40_shader.h index 854dccf5486..8d28137e9de 100644 --- a/src/gallium/drivers/nv40/nv40_shader.h +++ b/src/gallium/drivers/nv40/nv40_shader.h @@ -48,14 +48,6 @@ #define NV40_VP_INST_COND_TEST_ENABLE (1 << 13) #define NV40_VP_INST_COND_SHIFT 10 #define NV40_VP_INST_COND_MASK (0x7 << 10) -# define NV40_VP_INST_COND_FL 0 -# define NV40_VP_INST_COND_LT 1 -# define NV40_VP_INST_COND_EQ 2 -# define NV40_VP_INST_COND_LE 3 -# define NV40_VP_INST_COND_GT 4 -# define NV40_VP_INST_COND_NE 5 -# define NV40_VP_INST_COND_GE 6 -# define NV40_VP_INST_COND_TR 7 #define NV40_VP_INST_COND_SWZ_X_SHIFT 8 #define NV40_VP_INST_COND_SWZ_X_MASK (3 << 8) #define NV40_VP_INST_COND_SWZ_Y_SHIFT 6 @@ -84,63 +76,12 @@ /* ---- OPCODE BITS 95:64 / data DWORD 1 --- */ #define NV40_VP_INST_VEC_OPCODE_SHIFT 22 #define NV40_VP_INST_VEC_OPCODE_MASK (0x1F << 22) -# define NV40_VP_INST_OP_NOP 0x00 -# define NV40_VP_INST_OP_MOV 0x01 -# define NV40_VP_INST_OP_MUL 0x02 -# define NV40_VP_INST_OP_ADD 0x03 -# define NV40_VP_INST_OP_MAD 0x04 -# define NV40_VP_INST_OP_DP3 0x05 -# define NV40_VP_INST_OP_DPH 0x06 -# define NV40_VP_INST_OP_DP4 0x07 -# define NV40_VP_INST_OP_DST 0x08 -# define NV40_VP_INST_OP_MIN 0x09 -# define NV40_VP_INST_OP_MAX 0x0A -# define NV40_VP_INST_OP_SLT 0x0B -# define NV40_VP_INST_OP_SGE 0x0C -# define NV40_VP_INST_OP_ARL 0x0D -# define NV40_VP_INST_OP_FRC 0x0E -# define NV40_VP_INST_OP_FLR 0x0F -# define NV40_VP_INST_OP_SEQ 0x10 -# define NV40_VP_INST_OP_SFL 0x11 -# define NV40_VP_INST_OP_SGT 0x12 -# define NV40_VP_INST_OP_SLE 0x13 -# define NV40_VP_INST_OP_SNE 0x14 -# define NV40_VP_INST_OP_STR 0x15 -# define NV40_VP_INST_OP_SSG 0x16 -# define NV40_VP_INST_OP_ARR 0x17 -# define NV40_VP_INST_OP_ARA 0x18 -# define NV40_VP_INST_OP_TXL 0x19 #define NV40_VP_INST_SCA_OPCODE_SHIFT 27 #define NV40_VP_INST_SCA_OPCODE_MASK (0x1F << 27) -# define NV40_VP_INST_OP_NOP 0x00 -# define NV40_VP_INST_OP_MOV 0x01 -# define NV40_VP_INST_OP_RCP 0x02 -# define NV40_VP_INST_OP_RCC 0x03 -# define NV40_VP_INST_OP_RSQ 0x04 -# define NV40_VP_INST_OP_EXP 0x05 -# define NV40_VP_INST_OP_LOG 0x06 -# define NV40_VP_INST_OP_LIT 0x07 -# define NV40_VP_INST_OP_BRA 0x09 -# define NV40_VP_INST_OP_CAL 0x0B -# define NV40_VP_INST_OP_RET 0x0C -# define NV40_VP_INST_OP_LG2 0x0D -# define NV40_VP_INST_OP_EX2 0x0E -# define NV40_VP_INST_OP_SIN 0x0F -# define NV40_VP_INST_OP_COS 0x10 -# define NV40_VP_INST_OP_PUSHA 0x13 -# define NV40_VP_INST_OP_POPA 0x14 #define NV40_VP_INST_CONST_SRC_SHIFT 12 #define NV40_VP_INST_CONST_SRC_MASK (0xFF << 12) #define NV40_VP_INST_INPUT_SRC_SHIFT 8 #define NV40_VP_INST_INPUT_SRC_MASK (0x0F << 8) -# define NV40_VP_INST_IN_POS 0 -# define NV40_VP_INST_IN_WEIGHT 1 -# define NV40_VP_INST_IN_NORMAL 2 -# define NV40_VP_INST_IN_COL0 3 -# define NV40_VP_INST_IN_COL1 4 -# define NV40_VP_INST_IN_FOGC 5 -# define NV40_VP_INST_IN_TC0 8 -# define NV40_VP_INST_IN_TC(n) (8+n) #define NV40_VP_INST_SRC0H_SHIFT 0 #define NV40_VP_INST_SRC0H_MASK (0xFF << 0) #define NV40_VP_INST1_KNOWN ( \ @@ -194,7 +135,6 @@ # define NV40_VP_INST_DEST_TC(n) (7+n) # define NV40_VP_INST_DEST_TEMP 0x1F #define NV40_VP_INST_INDEX_CONST (1 << 1) -#define NV40_VP_INST_LAST (1 << 0) #define NV40_VP_INST3_KNOWN ( \ NV40_VP_INST_SRC2L_MASK |\ NV40_VP_INST_SCA_WRITEMASK_MASK |\ @@ -232,325 +172,7 @@ # define NV40_VP_SRC_REG_TYPE_INPUT 2 # define NV40_VP_SRC_REG_TYPE_CONST 3 +#include "nvfx_shader.h" -/* - * Each fragment program opcode appears to be comprised of 4 32-bit values. - * - * 0 - Opcode, output reg/mask, ATTRIB source - * 1 - Source 0 - * 2 - Source 1 - * 3 - Source 2 - * - * There appears to be no special difference between result regs and temp regs. - * result.color == R0.xyzw - * result.depth == R1.z - * When the fragprog contains instructions to write depth, - * NV30_TCL_PRIMITIVE_3D_UNK1D78=0 otherwise it is set to 1. - * - * Constants are inserted directly after the instruction that uses them. - * - * It appears that it's not possible to use two input registers in one - * instruction as the input sourcing is done in the instruction dword - * and not the source selection dwords. As such instructions such as: - * - * ADD result.color, fragment.color, fragment.texcoord[0]; - * - * must be split into two MOV's and then an ADD (nvidia does this) but - * I'm not sure why it's not just one MOV and then source the second input - * in the ADD instruction.. - * - * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary - * negation requires multiplication with a const. - * - * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO and - * SWIZZLE_ONE. - * - * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as - * SWIZZLE_ZERO is implemented simply by not writing to the relevant components - * of the destination. - * - * Looping - * Loops appear to be fairly expensive on NV40 at least, the proprietary - * driver goes to a lot of effort to avoid using the native looping - * instructions. If the total number of *executed* instructions between - * REP/ENDREP or LOOP/ENDLOOP is <=500, the driver will unroll the loop. - * The maximum loop count is 255. - * - * Conditional execution - * TODO - * - * Non-native instructions: - * LIT - * LRP - MAD+MAD - * SUB - ADD, negate second source - * RSQ - LG2 + EX2 - * POW - LG2 + MUL + EX2 - * SCS - COS + SIN - * XPD - * DP2 - MUL + ADD - * NRM - */ - -//== Opcode / Destination selection == -#define NV40_FP_OP_PROGRAM_END (1 << 0) -#define NV40_FP_OP_OUT_REG_SHIFT 1 -#define NV40_FP_OP_OUT_REG_MASK (63 << 1) -/* Needs to be set when writing outputs to get expected result.. */ -#define NV40_FP_OP_OUT_REG_HALF (1 << 7) -#define NV40_FP_OP_COND_WRITE_ENABLE (1 << 8) -#define NV40_FP_OP_OUTMASK_SHIFT 9 -#define NV40_FP_OP_OUTMASK_MASK (0xF << 9) -# define NV40_FP_OP_OUT_X (1 << 9) -# define NV40_FP_OP_OUT_Y (1 <<10) -# define NV40_FP_OP_OUT_Z (1 <<11) -# define NV40_FP_OP_OUT_W (1 <<12) -/* Uncertain about these, especially the input_src values.. it's possible that - * they can be dynamically changed. - */ -#define NV40_FP_OP_INPUT_SRC_SHIFT 13 -#define NV40_FP_OP_INPUT_SRC_MASK (15 << 13) -# define NV40_FP_OP_INPUT_SRC_POSITION 0x0 -# define NV40_FP_OP_INPUT_SRC_COL0 0x1 -# define NV40_FP_OP_INPUT_SRC_COL1 0x2 -# define NV40_FP_OP_INPUT_SRC_FOGC 0x3 -# define NV40_FP_OP_INPUT_SRC_TC0 0x4 -# define NV40_FP_OP_INPUT_SRC_TC(n) (0x4 + n) -# define NV40_FP_OP_INPUT_SRC_FACING 0xE -#define NV40_FP_OP_TEX_UNIT_SHIFT 17 -#define NV40_FP_OP_TEX_UNIT_MASK (0xF << 17) -#define NV40_FP_OP_PRECISION_SHIFT 22 -#define NV40_FP_OP_PRECISION_MASK (3 << 22) -# define NV40_FP_PRECISION_FP32 0 -# define NV40_FP_PRECISION_FP16 1 -# define NV40_FP_PRECISION_FX12 2 -#define NV40_FP_OP_OPCODE_SHIFT 24 -#define NV40_FP_OP_OPCODE_MASK (0x3F << 24) -# define NV40_FP_OP_OPCODE_NOP 0x00 -# define NV40_FP_OP_OPCODE_MOV 0x01 -# define NV40_FP_OP_OPCODE_MUL 0x02 -# define NV40_FP_OP_OPCODE_ADD 0x03 -# define NV40_FP_OP_OPCODE_MAD 0x04 -# define NV40_FP_OP_OPCODE_DP3 0x05 -# define NV40_FP_OP_OPCODE_DP4 0x06 -# define NV40_FP_OP_OPCODE_DST 0x07 -# define NV40_FP_OP_OPCODE_MIN 0x08 -# define NV40_FP_OP_OPCODE_MAX 0x09 -# define NV40_FP_OP_OPCODE_SLT 0x0A -# define NV40_FP_OP_OPCODE_SGE 0x0B -# define NV40_FP_OP_OPCODE_SLE 0x0C -# define NV40_FP_OP_OPCODE_SGT 0x0D -# define NV40_FP_OP_OPCODE_SNE 0x0E -# define NV40_FP_OP_OPCODE_SEQ 0x0F -# define NV40_FP_OP_OPCODE_FRC 0x10 -# define NV40_FP_OP_OPCODE_FLR 0x11 -# define NV40_FP_OP_OPCODE_KIL 0x12 -# define NV40_FP_OP_OPCODE_PK4B 0x13 -# define NV40_FP_OP_OPCODE_UP4B 0x14 -/* DDX/DDY can only write to XY */ -# define NV40_FP_OP_OPCODE_DDX 0x15 -# define NV40_FP_OP_OPCODE_DDY 0x16 -# define NV40_FP_OP_OPCODE_TEX 0x17 -# define NV40_FP_OP_OPCODE_TXP 0x18 -# define NV40_FP_OP_OPCODE_TXD 0x19 -# define NV40_FP_OP_OPCODE_RCP 0x1A -# define NV40_FP_OP_OPCODE_EX2 0x1C -# define NV40_FP_OP_OPCODE_LG2 0x1D -# define NV40_FP_OP_OPCODE_STR 0x20 -# define NV40_FP_OP_OPCODE_SFL 0x21 -# define NV40_FP_OP_OPCODE_COS 0x22 -# define NV40_FP_OP_OPCODE_SIN 0x23 -# define NV40_FP_OP_OPCODE_PK2H 0x24 -# define NV40_FP_OP_OPCODE_UP2H 0x25 -# define NV40_FP_OP_OPCODE_PK4UB 0x27 -# define NV40_FP_OP_OPCODE_UP4UB 0x28 -# define NV40_FP_OP_OPCODE_PK2US 0x29 -# define NV40_FP_OP_OPCODE_UP2US 0x2A -# define NV40_FP_OP_OPCODE_DP2A 0x2E -# define NV40_FP_OP_OPCODE_TXL 0x2F -# define NV40_FP_OP_OPCODE_TXB 0x31 -# define NV40_FP_OP_OPCODE_DIV 0x3A -# define NV40_FP_OP_OPCODE_UNK_LIT 0x3C -/* The use of these instructions appears to be indicated by bit 31 of DWORD 2.*/ -# define NV40_FP_OP_BRA_OPCODE_BRK 0x0 -# define NV40_FP_OP_BRA_OPCODE_CAL 0x1 -# define NV40_FP_OP_BRA_OPCODE_IF 0x2 -# define NV40_FP_OP_BRA_OPCODE_LOOP 0x3 -# define NV40_FP_OP_BRA_OPCODE_REP 0x4 -# define NV40_FP_OP_BRA_OPCODE_RET 0x5 -#define NV40_FP_OP_OUT_SAT (1 << 31) - -/* high order bits of SRC0 */ -#define NV40_FP_OP_OUT_ABS (1 << 29) -#define NV40_FP_OP_COND_SWZ_W_SHIFT 27 -#define NV40_FP_OP_COND_SWZ_W_MASK (3 << 27) -#define NV40_FP_OP_COND_SWZ_Z_SHIFT 25 -#define NV40_FP_OP_COND_SWZ_Z_MASK (3 << 25) -#define NV40_FP_OP_COND_SWZ_Y_SHIFT 23 -#define NV40_FP_OP_COND_SWZ_Y_MASK (3 << 23) -#define NV40_FP_OP_COND_SWZ_X_SHIFT 21 -#define NV40_FP_OP_COND_SWZ_X_MASK (3 << 21) -#define NV40_FP_OP_COND_SWZ_ALL_SHIFT 21 -#define NV40_FP_OP_COND_SWZ_ALL_MASK (0xFF << 21) -#define NV40_FP_OP_COND_SHIFT 18 -#define NV40_FP_OP_COND_MASK (0x07 << 18) -# define NV40_FP_OP_COND_FL 0 -# define NV40_FP_OP_COND_LT 1 -# define NV40_FP_OP_COND_EQ 2 -# define NV40_FP_OP_COND_LE 3 -# define NV40_FP_OP_COND_GT 4 -# define NV40_FP_OP_COND_NE 5 -# define NV40_FP_OP_COND_GE 6 -# define NV40_FP_OP_COND_TR 7 - -/* high order bits of SRC1 */ -#define NV40_FP_OP_OPCODE_IS_BRANCH (1<<31) -#define NV40_FP_OP_DST_SCALE_SHIFT 28 -#define NV40_FP_OP_DST_SCALE_MASK (3 << 28) -#define NV40_FP_OP_DST_SCALE_1X 0 -#define NV40_FP_OP_DST_SCALE_2X 1 -#define NV40_FP_OP_DST_SCALE_4X 2 -#define NV40_FP_OP_DST_SCALE_8X 3 -#define NV40_FP_OP_DST_SCALE_INV_2X 5 -#define NV40_FP_OP_DST_SCALE_INV_4X 6 -#define NV40_FP_OP_DST_SCALE_INV_8X 7 - -/* SRC1 LOOP */ -#define NV40_FP_OP_LOOP_INCR_SHIFT 19 -#define NV40_FP_OP_LOOP_INCR_MASK (0xFF << 19) -#define NV40_FP_OP_LOOP_INDEX_SHIFT 10 -#define NV40_FP_OP_LOOP_INDEX_MASK (0xFF << 10) -#define NV40_FP_OP_LOOP_COUNT_SHIFT 2 -#define NV40_FP_OP_LOOP_COUNT_MASK (0xFF << 2) - -/* SRC1 IF */ -#define NV40_FP_OP_ELSE_ID_SHIFT 2 -#define NV40_FP_OP_ELSE_ID_MASK (0xFF << 2) - -/* SRC1 CAL */ -#define NV40_FP_OP_IADDR_SHIFT 2 -#define NV40_FP_OP_IADDR_MASK (0xFF << 2) - -/* SRC1 REP - * I have no idea why there are 3 count values here.. but they - * have always been filled with the same value in my tests so - * far.. - */ -#define NV40_FP_OP_REP_COUNT1_SHIFT 2 -#define NV40_FP_OP_REP_COUNT1_MASK (0xFF << 2) -#define NV40_FP_OP_REP_COUNT2_SHIFT 10 -#define NV40_FP_OP_REP_COUNT2_MASK (0xFF << 10) -#define NV40_FP_OP_REP_COUNT3_SHIFT 19 -#define NV40_FP_OP_REP_COUNT3_MASK (0xFF << 19) - -/* SRC2 REP/IF */ -#define NV40_FP_OP_END_ID_SHIFT 2 -#define NV40_FP_OP_END_ID_MASK (0xFF << 2) - -// SRC2 high-order -#define NV40_FP_OP_INDEX_INPUT (1 << 30) -#define NV40_FP_OP_ADDR_INDEX_SHIFT 19 -#define NV40_FP_OP_ADDR_INDEX_MASK (0xF << 19) - -//== Register selection == -#define NV40_FP_REG_TYPE_SHIFT 0 -#define NV40_FP_REG_TYPE_MASK (3 << 0) -# define NV40_FP_REG_TYPE_TEMP 0 -# define NV40_FP_REG_TYPE_INPUT 1 -# define NV40_FP_REG_TYPE_CONST 2 -#define NV40_FP_REG_SRC_SHIFT 2 -#define NV40_FP_REG_SRC_MASK (63 << 2) -#define NV40_FP_REG_SRC_HALF (1 << 8) -#define NV40_FP_REG_SWZ_ALL_SHIFT 9 -#define NV40_FP_REG_SWZ_ALL_MASK (255 << 9) -#define NV40_FP_REG_SWZ_X_SHIFT 9 -#define NV40_FP_REG_SWZ_X_MASK (3 << 9) -#define NV40_FP_REG_SWZ_Y_SHIFT 11 -#define NV40_FP_REG_SWZ_Y_MASK (3 << 11) -#define NV40_FP_REG_SWZ_Z_SHIFT 13 -#define NV40_FP_REG_SWZ_Z_MASK (3 << 13) -#define NV40_FP_REG_SWZ_W_SHIFT 15 -#define NV40_FP_REG_SWZ_W_MASK (3 << 15) -# define NV40_FP_SWIZZLE_X 0 -# define NV40_FP_SWIZZLE_Y 1 -# define NV40_FP_SWIZZLE_Z 2 -# define NV40_FP_SWIZZLE_W 3 -#define NV40_FP_REG_NEGATE (1 << 17) - -#ifndef NV40_SHADER_NO_FUCKEDNESS -#define NV40SR_NONE 0 -#define NV40SR_OUTPUT 1 -#define NV40SR_INPUT 2 -#define NV40SR_TEMP 3 -#define NV40SR_CONST 4 - -struct nv40_sreg { - int type; - int index; - - int dst_scale; - - int negate; - int abs; - int swz[4]; - - int cc_update; - int cc_update_reg; - int cc_test; - int cc_test_reg; - int cc_swz[4]; -}; - -static INLINE struct nv40_sreg -nv40_sr(int type, int index) -{ - struct nv40_sreg temp = { - .type = type, - .index = index, - .dst_scale = DEF_SCALE, - .abs = 0, - .negate = 0, - .swz = { 0, 1, 2, 3 }, - .cc_update = 0, - .cc_update_reg = 0, - .cc_test = DEF_CTEST, - .cc_test_reg = 0, - .cc_swz = { 0, 1, 2, 3 }, - }; - return temp; -} - -static INLINE struct nv40_sreg -nv40_sr_swz(struct nv40_sreg src, int x, int y, int z, int w) -{ - struct nv40_sreg dst = src; - - dst.swz[SWZ_X] = src.swz[x]; - dst.swz[SWZ_Y] = src.swz[y]; - dst.swz[SWZ_Z] = src.swz[z]; - dst.swz[SWZ_W] = src.swz[w]; - return dst; -} - -static INLINE struct nv40_sreg -nv40_sr_neg(struct nv40_sreg src) -{ - src.negate = !src.negate; - return src; -} - -static INLINE struct nv40_sreg -nv40_sr_abs(struct nv40_sreg src) -{ - src.abs = 1; - return src; -} - -static INLINE struct nv40_sreg -nv40_sr_scale(struct nv40_sreg src, int scale) -{ - src.dst_scale = scale; - return src; -} #endif -#endif diff --git a/src/gallium/drivers/nv40/nv40_vertprog.c b/src/gallium/drivers/nv40/nv40_vertprog.c index a199f0766e4..752cd0d1b3d 100644 --- a/src/gallium/drivers/nv40/nv40_vertprog.c +++ b/src/gallium/drivers/nv40/nv40_vertprog.c @@ -34,9 +34,9 @@ #define DEF_CTEST 0 #include "nv40_shader.h" -#define swz(s,x,y,z,w) nv40_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w) -#define neg(s) nv40_sr_neg((s)) -#define abs(s) nv40_sr_abs((s)) +#define 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)) +#define abs(s) nvfx_sr_abs((s)) #define NV40_VP_INST_DEST_CLIP(n) ((~0 - 6) + (n)) @@ -47,17 +47,17 @@ struct nv40_vpc { unsigned r_temps; unsigned r_temps_discard; - struct nv40_sreg r_result[PIPE_MAX_SHADER_OUTPUTS]; - struct nv40_sreg *r_address; - struct nv40_sreg *r_temp; + struct nvfx_sreg r_result[PIPE_MAX_SHADER_OUTPUTS]; + struct nvfx_sreg *r_address; + struct nvfx_sreg *r_temp; - struct nv40_sreg *imm; + struct nvfx_sreg *imm; unsigned nr_imm; unsigned hpos_idx; }; -static struct nv40_sreg +static struct nvfx_sreg temp(struct nv40_vpc *vpc) { int idx = ffs(~vpc->r_temps) - 1; @@ -65,12 +65,12 @@ temp(struct nv40_vpc *vpc) if (idx < 0) { NOUVEAU_ERR("out of temps!!\n"); assert(0); - return nv40_sr(NV40SR_TEMP, 0); + return nvfx_sr(NVFXSR_TEMP, 0); } vpc->r_temps |= (1 << idx); vpc->r_temps_discard |= (1 << idx); - return nv40_sr(NV40SR_TEMP, idx); + return nvfx_sr(NVFXSR_TEMP, idx); } static INLINE void @@ -80,7 +80,7 @@ release_temps(struct nv40_vpc *vpc) vpc->r_temps_discard = 0; } -static struct nv40_sreg +static struct nvfx_sreg constant(struct nv40_vpc *vpc, int pipe, float x, float y, float z, float w) { struct nvfx_vertex_program *vp = vpc->vp; @@ -90,7 +90,7 @@ constant(struct nv40_vpc *vpc, int pipe, float x, float y, float z, float w) if (pipe >= 0) { for (idx = 0; idx < vp->nr_consts; idx++) { if (vp->consts[idx].index == pipe) - return nv40_sr(NV40SR_CONST, idx); + return nvfx_sr(NVFXSR_CONST, idx); } } @@ -103,37 +103,37 @@ constant(struct nv40_vpc *vpc, int pipe, float x, float y, float z, float w) vpd->value[1] = y; vpd->value[2] = z; vpd->value[3] = w; - return nv40_sr(NV40SR_CONST, idx); + return nvfx_sr(NVFXSR_CONST, idx); } #define arith(cc,s,o,d,m,s0,s1,s2) \ - nv40_vp_arith((cc), (s), NV40_VP_INST_##o, (d), (m), (s0), (s1), (s2)) + nv40_vp_arith((cc), NVFX_VP_INST_SLOT_##s, NVFX_VP_INST_##s##_OP_##o, (d), (m), (s0), (s1), (s2)) static void -emit_src(struct nv40_vpc *vpc, uint32_t *hw, int pos, struct nv40_sreg src) +emit_src(struct nv40_vpc *vpc, uint32_t *hw, int pos, struct nvfx_sreg src) { struct nvfx_vertex_program *vp = vpc->vp; uint32_t sr = 0; switch (src.type) { - case NV40SR_TEMP: + case NVFXSR_TEMP: sr |= (NV40_VP_SRC_REG_TYPE_TEMP << NV40_VP_SRC_REG_TYPE_SHIFT); sr |= (src.index << NV40_VP_SRC_TEMP_SRC_SHIFT); break; - case NV40SR_INPUT: + case NVFXSR_INPUT: sr |= (NV40_VP_SRC_REG_TYPE_INPUT << NV40_VP_SRC_REG_TYPE_SHIFT); vp->ir |= (1 << src.index); hw[1] |= (src.index << NV40_VP_INST_INPUT_SRC_SHIFT); break; - case NV40SR_CONST: + case NVFXSR_CONST: sr |= (NV40_VP_SRC_REG_TYPE_CONST << NV40_VP_SRC_REG_TYPE_SHIFT); assert(vpc->vpi->const_index == -1 || vpc->vpi->const_index == src.index); vpc->vpi->const_index = src.index; break; - case NV40SR_NONE: + case NVFXSR_NONE: sr |= (NV40_VP_SRC_REG_TYPE_INPUT << NV40_VP_SRC_REG_TYPE_SHIFT); break; @@ -174,12 +174,12 @@ emit_src(struct nv40_vpc *vpc, uint32_t *hw, int pos, struct nv40_sreg src) } static void -emit_dst(struct nv40_vpc *vpc, uint32_t *hw, int slot, struct nv40_sreg dst) +emit_dst(struct nv40_vpc *vpc, uint32_t *hw, int slot, struct nvfx_sreg dst) { struct nvfx_vertex_program *vp = vpc->vp; switch (dst.type) { - case NV40SR_TEMP: + case NVFXSR_TEMP: hw[3] |= NV40_VP_INST_DEST_MASK; if (slot == 0) { hw[0] |= (dst.index << @@ -189,7 +189,7 @@ emit_dst(struct nv40_vpc *vpc, uint32_t *hw, int slot, struct nv40_sreg dst) NV40_VP_INST_SCA_DEST_TEMP_SHIFT); } break; - case NV40SR_OUTPUT: + case NVFXSR_OUTPUT: switch (dst.index) { case NV40_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break; case NV40_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break; @@ -255,9 +255,9 @@ emit_dst(struct nv40_vpc *vpc, uint32_t *hw, int slot, struct nv40_sreg dst) static void nv40_vp_arith(struct nv40_vpc *vpc, int slot, int op, - struct nv40_sreg dst, int mask, - struct nv40_sreg s0, struct nv40_sreg s1, - struct nv40_sreg s2) + struct nvfx_sreg dst, int mask, + struct nvfx_sreg s0, struct nvfx_sreg s1, + struct nvfx_sreg s2) { struct nvfx_vertex_program *vp = vpc->vp; uint32_t *hw; @@ -269,7 +269,7 @@ nv40_vp_arith(struct nv40_vpc *vpc, int slot, int op, hw = vpc->vpi->data; - hw[0] |= (NV40_VP_INST_COND_TR << NV40_VP_INST_COND_SHIFT); + hw[0] |= (NVFX_VP_INST_COND_TR << NV40_VP_INST_COND_SHIFT); hw[0] |= ((0 << NV40_VP_INST_COND_SWZ_X_SHIFT) | (1 << NV40_VP_INST_COND_SWZ_Y_SHIFT) | (2 << NV40_VP_INST_COND_SWZ_Z_SHIFT) | @@ -291,13 +291,13 @@ nv40_vp_arith(struct nv40_vpc *vpc, int slot, int op, emit_src(vpc, hw, 2, s2); } -static INLINE struct nv40_sreg +static INLINE struct nvfx_sreg tgsi_src(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc) { - struct nv40_sreg src; + struct nvfx_sreg src; switch (fsrc->Register.File) { case TGSI_FILE_INPUT: - src = nv40_sr(NV40SR_INPUT, fsrc->Register.Index); + src = nvfx_sr(NVFXSR_INPUT, fsrc->Register.Index); break; case TGSI_FILE_CONSTANT: src = constant(vpc, fsrc->Register.Index, 0, 0, 0, 0); @@ -322,9 +322,9 @@ tgsi_src(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc) { return src; } -static INLINE struct nv40_sreg +static INLINE struct nvfx_sreg tgsi_dst(struct nv40_vpc *vpc, const struct tgsi_full_dst_register *fdst) { - struct nv40_sreg dst; + struct nvfx_sreg dst; switch (fdst->Register.File) { case TGSI_FILE_OUTPUT: @@ -358,10 +358,10 @@ tgsi_mask(uint tgsi) static boolean src_native_swz(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc, - struct nv40_sreg *src) + struct nvfx_sreg *src) { - const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0); - struct nv40_sreg tgsi = tgsi_src(vpc, fsrc); + const struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0); + struct nvfx_sreg tgsi = tgsi_src(vpc, fsrc); uint mask = 0; uint c; @@ -384,7 +384,7 @@ src_native_swz(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc, *src = temp(vpc); if (mask) - arith(vpc, 0, OP_MOV, *src, mask, tgsi, none, none); + arith(vpc, VEC, MOV, *src, mask, tgsi, none, none); return FALSE; } @@ -393,8 +393,8 @@ static boolean nv40_vertprog_parse_instruction(struct nv40_vpc *vpc, const struct tgsi_full_instruction *finst) { - struct nv40_sreg src[3], dst, tmp; - struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0); + struct nvfx_sreg src[3], dst, tmp; + struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0); int mask; int ai = -1, ci = -1, ii = -1; int i; @@ -434,7 +434,7 @@ nv40_vertprog_parse_instruction(struct nv40_vpc *vpc, src[i] = tgsi_src(vpc, fsrc); } else { src[i] = temp(vpc); - arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + arith(vpc, VEC, MOV, src[i], MASK_ALL, tgsi_src(vpc, fsrc), none, none); } break; @@ -445,7 +445,7 @@ nv40_vertprog_parse_instruction(struct nv40_vpc *vpc, src[i] = tgsi_src(vpc, fsrc); } else { src[i] = temp(vpc); - arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + arith(vpc, VEC, MOV, src[i], MASK_ALL, tgsi_src(vpc, fsrc), none, none); } break; @@ -456,7 +456,7 @@ nv40_vertprog_parse_instruction(struct nv40_vpc *vpc, src[i] = tgsi_src(vpc, fsrc); } else { src[i] = temp(vpc); - arith(vpc, 0, OP_MOV, src[i], MASK_ALL, + arith(vpc, VEC, MOV, src[i], MASK_ALL, tgsi_src(vpc, fsrc), none, none); } break; @@ -474,93 +474,93 @@ nv40_vertprog_parse_instruction(struct nv40_vpc *vpc, switch (finst->Instruction.Opcode) { case TGSI_OPCODE_ABS: - arith(vpc, 0, OP_MOV, dst, mask, abs(src[0]), none, none); + arith(vpc, VEC, MOV, dst, mask, abs(src[0]), none, none); break; case TGSI_OPCODE_ADD: - arith(vpc, 0, OP_ADD, dst, mask, src[0], none, src[1]); + arith(vpc, VEC, ADD, dst, mask, src[0], none, src[1]); break; case TGSI_OPCODE_ARL: - arith(vpc, 0, OP_ARL, dst, mask, src[0], none, none); + arith(vpc, VEC, ARL, dst, mask, src[0], none, none); break; case TGSI_OPCODE_DP3: - arith(vpc, 0, OP_DP3, dst, mask, src[0], src[1], none); + arith(vpc, VEC, DP3, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_DP4: - arith(vpc, 0, OP_DP4, dst, mask, src[0], src[1], none); + arith(vpc, VEC, DP4, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_DPH: - arith(vpc, 0, OP_DPH, dst, mask, src[0], src[1], none); + arith(vpc, VEC, DPH, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_DST: - arith(vpc, 0, OP_DST, dst, mask, src[0], src[1], none); + arith(vpc, VEC, DST, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_EX2: - arith(vpc, 1, OP_EX2, dst, mask, none, none, src[0]); + arith(vpc, SCA, EX2, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_EXP: - arith(vpc, 1, OP_EXP, dst, mask, none, none, src[0]); + arith(vpc, SCA, EXP, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_FLR: - arith(vpc, 0, OP_FLR, dst, mask, src[0], none, none); + arith(vpc, VEC, FLR, dst, mask, src[0], none, none); break; case TGSI_OPCODE_FRC: - arith(vpc, 0, OP_FRC, dst, mask, src[0], none, none); + arith(vpc, VEC, FRC, dst, mask, src[0], none, none); break; case TGSI_OPCODE_LG2: - arith(vpc, 1, OP_LG2, dst, mask, none, none, src[0]); + arith(vpc, SCA, LG2, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_LIT: - arith(vpc, 1, OP_LIT, dst, mask, none, none, src[0]); + arith(vpc, SCA, LIT, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_LOG: - arith(vpc, 1, OP_LOG, dst, mask, none, none, src[0]); + arith(vpc, SCA, LOG, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_MAD: - arith(vpc, 0, OP_MAD, dst, mask, src[0], src[1], src[2]); + arith(vpc, VEC, MAD, dst, mask, src[0], src[1], src[2]); break; case TGSI_OPCODE_MAX: - arith(vpc, 0, OP_MAX, dst, mask, src[0], src[1], none); + arith(vpc, VEC, MAX, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_MIN: - arith(vpc, 0, OP_MIN, dst, mask, src[0], src[1], none); + arith(vpc, VEC, MIN, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_MOV: - arith(vpc, 0, OP_MOV, dst, mask, src[0], none, none); + arith(vpc, VEC, MOV, dst, mask, src[0], none, none); break; case TGSI_OPCODE_MUL: - arith(vpc, 0, OP_MUL, dst, mask, src[0], src[1], none); + arith(vpc, VEC, MUL, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_POW: tmp = temp(vpc); - arith(vpc, 1, OP_LG2, tmp, MASK_X, none, none, + arith(vpc, SCA, LG2, tmp, MASK_X, none, none, swz(src[0], X, X, X, X)); - arith(vpc, 0, OP_MUL, tmp, MASK_X, swz(tmp, X, X, X, X), + arith(vpc, VEC, MUL, tmp, MASK_X, swz(tmp, X, X, X, X), swz(src[1], X, X, X, X), none); - arith(vpc, 1, OP_EX2, dst, mask, none, none, + arith(vpc, SCA, EX2, dst, mask, none, none, swz(tmp, X, X, X, X)); break; case TGSI_OPCODE_RCP: - arith(vpc, 1, OP_RCP, dst, mask, none, none, src[0]); + arith(vpc, SCA, RCP, dst, mask, none, none, src[0]); break; case TGSI_OPCODE_RET: break; case TGSI_OPCODE_RSQ: - arith(vpc, 1, OP_RSQ, dst, mask, none, none, abs(src[0])); + arith(vpc, SCA, RSQ, dst, mask, none, none, abs(src[0])); break; case TGSI_OPCODE_SGE: - arith(vpc, 0, OP_SGE, dst, mask, src[0], src[1], none); + arith(vpc, VEC, SGE, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_SLT: - arith(vpc, 0, OP_SLT, dst, mask, src[0], src[1], none); + arith(vpc, VEC, SLT, dst, mask, src[0], src[1], none); break; case TGSI_OPCODE_SUB: - arith(vpc, 0, OP_ADD, dst, mask, src[0], none, neg(src[1])); + arith(vpc, VEC, ADD, dst, mask, src[0], none, neg(src[1])); break; case TGSI_OPCODE_XPD: tmp = temp(vpc); - arith(vpc, 0, OP_MUL, tmp, mask, + arith(vpc, VEC, MUL, tmp, mask, swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none); - arith(vpc, 0, OP_MAD, dst, (mask & ~MASK_W), + arith(vpc, VEC, MAD, dst, (mask & ~MASK_W), swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y), neg(tmp)); break; @@ -630,7 +630,7 @@ nv40_vertprog_parse_decl_output(struct nv40_vpc *vpc, return FALSE; } - vpc->r_result[idx] = nv40_sr(NV40SR_OUTPUT, hw); + vpc->r_result[idx] = nvfx_sr(NVFXSR_OUTPUT, hw); return TRUE; } @@ -702,18 +702,18 @@ nv40_vertprog_prepare(struct nv40_vpc *vpc) tgsi_parse_free(&p); if (nr_imm) { - vpc->imm = CALLOC(nr_imm, sizeof(struct nv40_sreg)); + vpc->imm = CALLOC(nr_imm, sizeof(struct nvfx_sreg)); assert(vpc->imm); } if (++high_temp) { - vpc->r_temp = CALLOC(high_temp, sizeof(struct nv40_sreg)); + vpc->r_temp = CALLOC(high_temp, sizeof(struct nvfx_sreg)); for (i = 0; i < high_temp; i++) vpc->r_temp[i] = temp(vpc); } if (++high_addr) { - vpc->r_address = CALLOC(high_addr, sizeof(struct nv40_sreg)); + vpc->r_address = CALLOC(high_addr, sizeof(struct nvfx_sreg)); for (i = 0; i < high_addr; i++) vpc->r_address[i] = temp(vpc); } @@ -728,7 +728,7 @@ nv40_vertprog_translate(struct nvfx_context *nvfx, { struct tgsi_parse_context parse; struct nv40_vpc *vpc = NULL; - struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0); + struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0); int i; vpc = CALLOC(1, sizeof(struct nv40_vpc)); @@ -785,24 +785,24 @@ nv40_vertprog_translate(struct nvfx_context *nvfx, } /* Write out HPOS if it was redirected to a temp earlier */ - if (vpc->r_result[vpc->hpos_idx].type != NV40SR_OUTPUT) { - struct nv40_sreg hpos = nv40_sr(NV40SR_OUTPUT, + if (vpc->r_result[vpc->hpos_idx].type != NVFXSR_OUTPUT) { + struct nvfx_sreg hpos = nvfx_sr(NVFXSR_OUTPUT, NV40_VP_INST_DEST_POS); - struct nv40_sreg htmp = vpc->r_result[vpc->hpos_idx]; + struct nvfx_sreg htmp = vpc->r_result[vpc->hpos_idx]; - arith(vpc, 0, OP_MOV, hpos, MASK_ALL, htmp, none, none); + arith(vpc, VEC, MOV, hpos, MASK_ALL, htmp, none, none); } /* Insert code to handle user clip planes */ for (i = 0; i < vp->ucp.nr; i++) { - struct nv40_sreg cdst = nv40_sr(NV40SR_OUTPUT, + struct nvfx_sreg cdst = nvfx_sr(NVFXSR_OUTPUT, NV40_VP_INST_DEST_CLIP(i)); - struct nv40_sreg ceqn = constant(vpc, -1, + struct nvfx_sreg ceqn = constant(vpc, -1, nvfx->clip.ucp[i][0], nvfx->clip.ucp[i][1], nvfx->clip.ucp[i][2], nvfx->clip.ucp[i][3]); - struct nv40_sreg htmp = vpc->r_result[vpc->hpos_idx]; + struct nvfx_sreg htmp = vpc->r_result[vpc->hpos_idx]; unsigned mask; switch (i) { @@ -814,10 +814,10 @@ nv40_vertprog_translate(struct nvfx_context *nvfx, goto out_err; } - arith(vpc, 0, OP_DP4, cdst, mask, htmp, ceqn, none); + arith(vpc, VEC, DP4, cdst, mask, htmp, ceqn, none); } - vp->insns[vp->nr_insns - 1].data[3] |= NV40_VP_INST_LAST; + vp->insns[vp->nr_insns - 1].data[3] |= NVFX_VP_INST_LAST; vp->translated = TRUE; out_err: tgsi_parse_free(&parse); diff --git a/src/gallium/drivers/nvfx/nvfx_shader.h b/src/gallium/drivers/nvfx/nvfx_shader.h new file mode 100644 index 00000000000..191131a40a1 --- /dev/null +++ b/src/gallium/drivers/nvfx/nvfx_shader.h @@ -0,0 +1,407 @@ +#ifndef __NVFX_SHADER_H__ +#define __NVFX_SHADER_H__ + +/* 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_COND_FL 0 /* guess */ +#define NVFX_VP_INST_COND_LT 1 +#define NVFX_VP_INST_COND_EQ 2 +#define NVFX_VP_INST_COND_LE 3 +#define NVFX_VP_INST_COND_GT 4 +#define NVFX_VP_INST_COND_NE 5 +#define NVFX_VP_INST_COND_GE 6 +#define NVFX_VP_INST_COND_TR 7 /* guess */ + +#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 - Opcode, output reg/mask, ATTRIB source + * 1 - Source 0 + * 2 - Source 1 + * 3 - Source 2 + * + * There appears to be no special difference between result regs and temp regs. + * result.color == R0.xyzw + * result.depth == R1.z + * When the fragprog contains instructions to write depth, NV30_TCL_PRIMITIVE_3D_UNK1D78=0 + * otherwise it is set to 1. + * + * Constants are inserted directly after the instruction that uses them. + * + * It appears that it's not possible to use two input registers in one + * instruction as the input sourcing is done in the instruction dword + * and not the source selection dwords. As such instructions such as: + * + * ADD result.color, fragment.color, fragment.texcoord[0]; + * + * must be split into two MOV's and then an ADD (nvidia does this) but + * I'm not sure why it's not just one MOV and then source the second input + * in the ADD instruction.. + * + * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary + * negation requires multiplication with a const. + * + * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO/SWIZZLE_ONE + * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as SWIZZLE_ZERO + * is implemented simply by not writing to the relevant components of the destination. + * + * Conditional execution + * TODO + * + * Non-native instructions: + * LIT + * LRP - MAD+MAD + * SUB - ADD, negate second source + * RSQ - LG2 + EX2 + * POW - LG2 + MUL + EX2 + * SCS - COS + SIN + * XPD + * + * 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 0x31 +/* 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 NVFX_FP_OP_OUT_SAT (1 << 31) + +/* high order bits of SRC0 */ +#define NVFX_FP_OP_OUT_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 + +/* SRC1 LOOP */ +#define NV40_FP_OP_LOOP_INCR_SHIFT 19 +#define NV40_FP_OP_LOOP_INCR_MASK (0xFF << 19) +#define NV40_FP_OP_LOOP_INDEX_SHIFT 10 +#define NV40_FP_OP_LOOP_INDEX_MASK (0xFF << 10) +#define NV40_FP_OP_LOOP_COUNT_SHIFT 2 +#define NV40_FP_OP_LOOP_COUNT_MASK (0xFF << 2) + +/* SRC1 IF */ +#define NV40_FP_OP_ELSE_ID_SHIFT 2 +#define NV40_FP_OP_ELSE_ID_MASK (0xFF << 2) + +/* SRC1 CAL */ +#define NV40_FP_OP_IADDR_SHIFT 2 +#define NV40_FP_OP_IADDR_MASK (0xFF << 2) + +/* SRC1 REP + * I have no idea why there are 3 count values here.. but they + * have always been filled with the same value in my tests so + * far.. + */ +#define NV40_FP_OP_REP_COUNT1_SHIFT 2 +#define NV40_FP_OP_REP_COUNT1_MASK (0xFF << 2) +#define NV40_FP_OP_REP_COUNT2_SHIFT 10 +#define NV40_FP_OP_REP_COUNT2_MASK (0xFF << 10) +#define NV40_FP_OP_REP_COUNT3_SHIFT 19 +#define NV40_FP_OP_REP_COUNT3_MASK (0xFF << 19) + +/* SRC2 REP/IF */ +#define NV40_FP_OP_END_ID_SHIFT 2 +#define NV40_FP_OP_END_ID_MASK (0xFF << 2) + +/* 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) + +#ifndef NVFX_SHADER_NO_FUCKEDNESS +#define NVFXSR_NONE 0 +#define NVFXSR_OUTPUT 1 +#define NVFXSR_INPUT 2 +#define NVFXSR_TEMP 3 +#define NVFXSR_CONST 4 + +struct nvfx_sreg { + int type; + int index; + + int dst_scale; + + int negate; + int abs; + int swz[4]; + + int cc_update; + int cc_update_reg; + int cc_test; + int cc_test_reg; + int cc_swz[4]; +}; + +static INLINE struct nvfx_sreg +nvfx_sr(int type, int index) +{ + struct nvfx_sreg temp = { + .type = type, + .index = index, + .dst_scale = DEF_SCALE, + .abs = 0, + .negate = 0, + .swz = { 0, 1, 2, 3 }, + .cc_update = 0, + .cc_update_reg = 0, + .cc_test = DEF_CTEST, + .cc_test_reg = 0, + .cc_swz = { 0, 1, 2, 3 }, + }; + return temp; +} + +static INLINE struct nvfx_sreg +nvfx_sr_swz(struct nvfx_sreg src, int x, int y, int z, int w) +{ + struct nvfx_sreg dst = src; + + dst.swz[SWZ_X] = src.swz[x]; + dst.swz[SWZ_Y] = src.swz[y]; + dst.swz[SWZ_Z] = src.swz[z]; + dst.swz[SWZ_W] = src.swz[w]; + return dst; +} + +static INLINE struct nvfx_sreg +nvfx_sr_neg(struct nvfx_sreg src) +{ + src.negate = !src.negate; + return src; +} + +static INLINE struct nvfx_sreg +nvfx_sr_abs(struct nvfx_sreg 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; +} +#endif + +#endif