diff --git a/src/freedreno/ir2/disasm-a2xx.c b/src/freedreno/ir2/disasm-a2xx.c index 36567093663..61a20c49b02 100644 --- a/src/freedreno/ir2/disasm-a2xx.c +++ b/src/freedreno/ir2/disasm-a2xx.c @@ -21,34 +21,34 @@ * SOFTWARE. */ +#include +#include #include #include -#include -#include -#include -#include -#include #include +#include +#include +#include #include "disasm.h" #include "instr-a2xx.h" static const char *levels[] = { - "\t", - "\t\t", - "\t\t\t", - "\t\t\t\t", - "\t\t\t\t\t", - "\t\t\t\t\t\t", - "\t\t\t\t\t\t\t", - "\t\t\t\t\t\t\t\t", - "\t\t\t\t\t\t\t\t\t", - "x", - "x", - "x", - "x", - "x", - "x", + "\t", + "\t\t", + "\t\t\t", + "\t\t\t\t", + "\t\t\t\t\t", + "\t\t\t\t\t\t", + "\t\t\t\t\t\t\t", + "\t\t\t\t\t\t\t\t", + "\t\t\t\t\t\t\t\t\t", + "x", + "x", + "x", + "x", + "x", + "x", }; static enum debug_t debug; @@ -58,535 +58,556 @@ static enum debug_t debug; */ static const char chan_names[] = { - 'x', 'y', 'z', 'w', - /* these only apply to FETCH dst's: */ - '0', '1', '?', '_', + 'x', + 'y', + 'z', + 'w', + /* these only apply to FETCH dst's: */ + '0', + '1', + '?', + '_', }; -static void print_srcreg(uint32_t num, uint32_t type, - uint32_t swiz, uint32_t negate, uint32_t abs) +static void +print_srcreg(uint32_t num, uint32_t type, uint32_t swiz, uint32_t negate, + uint32_t abs) { - if (negate) - printf("-"); - if (abs) - printf("|"); - printf("%c%u", type ? 'R' : 'C', num); - if (swiz) { - int i; - printf("."); - for (i = 0; i < 4; i++) { - printf("%c", chan_names[(swiz + i) & 0x3]); - swiz >>= 2; - } - } - if (abs) - printf("|"); + if (negate) + printf("-"); + if (abs) + printf("|"); + printf("%c%u", type ? 'R' : 'C', num); + if (swiz) { + int i; + printf("."); + for (i = 0; i < 4; i++) { + printf("%c", chan_names[(swiz + i) & 0x3]); + swiz >>= 2; + } + } + if (abs) + printf("|"); } -static void print_dstreg(uint32_t num, uint32_t mask, uint32_t dst_exp) +static void +print_dstreg(uint32_t num, uint32_t mask, uint32_t dst_exp) { - printf("%s%u", dst_exp ? "export" : "R", num); - if (mask != 0xf) { - int i; - printf("."); - for (i = 0; i < 4; i++) { - printf("%c", (mask & 0x1) ? chan_names[i] : '_'); - mask >>= 1; - } - } + printf("%s%u", dst_exp ? "export" : "R", num); + if (mask != 0xf) { + int i; + printf("."); + for (i = 0; i < 4; i++) { + printf("%c", (mask & 0x1) ? chan_names[i] : '_'); + mask >>= 1; + } + } } -static void print_export_comment(uint32_t num, gl_shader_stage type) +static void +print_export_comment(uint32_t num, gl_shader_stage type) { - const char *name = NULL; - switch (type) { - case MESA_SHADER_VERTEX: - switch (num) { - case 62: name = "gl_Position"; break; - case 63: name = "gl_PointSize"; break; - } - break; - case MESA_SHADER_FRAGMENT: - switch (num) { - case 0: name = "gl_FragColor"; break; - } - break; - default: - assert(!"not reached"); - } - /* if we had a symbol table here, we could look - * up the name of the varying.. - */ - if (name) { - printf("\t; %s", name); - } + const char *name = NULL; + switch (type) { + case MESA_SHADER_VERTEX: + switch (num) { + case 62: + name = "gl_Position"; + break; + case 63: + name = "gl_PointSize"; + break; + } + break; + case MESA_SHADER_FRAGMENT: + switch (num) { + case 0: + name = "gl_FragColor"; + break; + } + break; + default: + assert(!"not reached"); + } + /* if we had a symbol table here, we could look + * up the name of the varying.. + */ + if (name) { + printf("\t; %s", name); + } } struct { - uint32_t num_srcs; - const char *name; + uint32_t num_srcs; + const char *name; } vector_instructions[0x20] = { -#define INSTR(opc, num_srcs) [opc] = { num_srcs, #opc } - INSTR(ADDv, 2), - INSTR(MULv, 2), - INSTR(MAXv, 2), - INSTR(MINv, 2), - INSTR(SETEv, 2), - INSTR(SETGTv, 2), - INSTR(SETGTEv, 2), - INSTR(SETNEv, 2), - INSTR(FRACv, 1), - INSTR(TRUNCv, 1), - INSTR(FLOORv, 1), - INSTR(MULADDv, 3), - INSTR(CNDEv, 3), - INSTR(CNDGTEv, 3), - INSTR(CNDGTv, 3), - INSTR(DOT4v, 2), - INSTR(DOT3v, 2), - INSTR(DOT2ADDv, 3), // ??? - INSTR(CUBEv, 2), - INSTR(MAX4v, 1), - INSTR(PRED_SETE_PUSHv, 2), - INSTR(PRED_SETNE_PUSHv, 2), - INSTR(PRED_SETGT_PUSHv, 2), - INSTR(PRED_SETGTE_PUSHv, 2), - INSTR(KILLEv, 2), - INSTR(KILLGTv, 2), - INSTR(KILLGTEv, 2), - INSTR(KILLNEv, 2), - INSTR(DSTv, 2), - INSTR(MOVAv, 1), +#define INSTR(opc, num_srcs) [opc] = {num_srcs, #opc} + INSTR(ADDv, 2), + INSTR(MULv, 2), + INSTR(MAXv, 2), + INSTR(MINv, 2), + INSTR(SETEv, 2), + INSTR(SETGTv, 2), + INSTR(SETGTEv, 2), + INSTR(SETNEv, 2), + INSTR(FRACv, 1), + INSTR(TRUNCv, 1), + INSTR(FLOORv, 1), + INSTR(MULADDv, 3), + INSTR(CNDEv, 3), + INSTR(CNDGTEv, 3), + INSTR(CNDGTv, 3), + INSTR(DOT4v, 2), + INSTR(DOT3v, 2), + INSTR(DOT2ADDv, 3), // ??? + INSTR(CUBEv, 2), + INSTR(MAX4v, 1), + INSTR(PRED_SETE_PUSHv, 2), + INSTR(PRED_SETNE_PUSHv, 2), + INSTR(PRED_SETGT_PUSHv, 2), + INSTR(PRED_SETGTE_PUSHv, 2), + INSTR(KILLEv, 2), + INSTR(KILLGTv, 2), + INSTR(KILLGTEv, 2), + INSTR(KILLNEv, 2), + INSTR(DSTv, 2), + INSTR(MOVAv, 1), }, scalar_instructions[0x40] = { - INSTR(ADDs, 1), - INSTR(ADD_PREVs, 1), - INSTR(MULs, 1), - INSTR(MUL_PREVs, 1), - INSTR(MUL_PREV2s, 1), - INSTR(MAXs, 1), - INSTR(MINs, 1), - INSTR(SETEs, 1), - INSTR(SETGTs, 1), - INSTR(SETGTEs, 1), - INSTR(SETNEs, 1), - INSTR(FRACs, 1), - INSTR(TRUNCs, 1), - INSTR(FLOORs, 1), - INSTR(EXP_IEEE, 1), - INSTR(LOG_CLAMP, 1), - INSTR(LOG_IEEE, 1), - INSTR(RECIP_CLAMP, 1), - INSTR(RECIP_FF, 1), - INSTR(RECIP_IEEE, 1), - INSTR(RECIPSQ_CLAMP, 1), - INSTR(RECIPSQ_FF, 1), - INSTR(RECIPSQ_IEEE, 1), - INSTR(MOVAs, 1), - INSTR(MOVA_FLOORs, 1), - INSTR(SUBs, 1), - INSTR(SUB_PREVs, 1), - INSTR(PRED_SETEs, 1), - INSTR(PRED_SETNEs, 1), - INSTR(PRED_SETGTs, 1), - INSTR(PRED_SETGTEs, 1), - INSTR(PRED_SET_INVs, 1), - INSTR(PRED_SET_POPs, 1), - INSTR(PRED_SET_CLRs, 1), - INSTR(PRED_SET_RESTOREs, 1), - INSTR(KILLEs, 1), - INSTR(KILLGTs, 1), - INSTR(KILLGTEs, 1), - INSTR(KILLNEs, 1), - INSTR(KILLONEs, 1), - INSTR(SQRT_IEEE, 1), - INSTR(MUL_CONST_0, 1), - INSTR(MUL_CONST_1, 1), - INSTR(ADD_CONST_0, 1), - INSTR(ADD_CONST_1, 1), - INSTR(SUB_CONST_0, 1), - INSTR(SUB_CONST_1, 1), - INSTR(SIN, 1), - INSTR(COS, 1), - INSTR(RETAIN_PREV, 1), + INSTR(ADDs, 1), + INSTR(ADD_PREVs, 1), + INSTR(MULs, 1), + INSTR(MUL_PREVs, 1), + INSTR(MUL_PREV2s, 1), + INSTR(MAXs, 1), + INSTR(MINs, 1), + INSTR(SETEs, 1), + INSTR(SETGTs, 1), + INSTR(SETGTEs, 1), + INSTR(SETNEs, 1), + INSTR(FRACs, 1), + INSTR(TRUNCs, 1), + INSTR(FLOORs, 1), + INSTR(EXP_IEEE, 1), + INSTR(LOG_CLAMP, 1), + INSTR(LOG_IEEE, 1), + INSTR(RECIP_CLAMP, 1), + INSTR(RECIP_FF, 1), + INSTR(RECIP_IEEE, 1), + INSTR(RECIPSQ_CLAMP, 1), + INSTR(RECIPSQ_FF, 1), + INSTR(RECIPSQ_IEEE, 1), + INSTR(MOVAs, 1), + INSTR(MOVA_FLOORs, 1), + INSTR(SUBs, 1), + INSTR(SUB_PREVs, 1), + INSTR(PRED_SETEs, 1), + INSTR(PRED_SETNEs, 1), + INSTR(PRED_SETGTs, 1), + INSTR(PRED_SETGTEs, 1), + INSTR(PRED_SET_INVs, 1), + INSTR(PRED_SET_POPs, 1), + INSTR(PRED_SET_CLRs, 1), + INSTR(PRED_SET_RESTOREs, 1), + INSTR(KILLEs, 1), + INSTR(KILLGTs, 1), + INSTR(KILLGTEs, 1), + INSTR(KILLNEs, 1), + INSTR(KILLONEs, 1), + INSTR(SQRT_IEEE, 1), + INSTR(MUL_CONST_0, 1), + INSTR(MUL_CONST_1, 1), + INSTR(ADD_CONST_0, 1), + INSTR(ADD_CONST_1, 1), + INSTR(SUB_CONST_0, 1), + INSTR(SUB_CONST_1, 1), + INSTR(SIN, 1), + INSTR(COS, 1), + INSTR(RETAIN_PREV, 1), #undef INSTR }; -static int disasm_alu(uint32_t *dwords, uint32_t alu_off, - int level, int sync, gl_shader_stage type) +static int +disasm_alu(uint32_t *dwords, uint32_t alu_off, int level, int sync, + gl_shader_stage type) { - instr_alu_t *alu = (instr_alu_t *)dwords; + instr_alu_t *alu = (instr_alu_t *)dwords; - printf("%s", levels[level]); - if (debug & PRINT_RAW) { - printf("%02x: %08x %08x %08x\t", alu_off, - dwords[0], dwords[1], dwords[2]); - } + printf("%s", levels[level]); + if (debug & PRINT_RAW) { + printf("%02x: %08x %08x %08x\t", alu_off, dwords[0], dwords[1], + dwords[2]); + } - printf(" %sALU:\t", sync ? "(S)" : " "); + printf(" %sALU:\t", sync ? "(S)" : " "); - printf("%s", vector_instructions[alu->vector_opc].name); + printf("%s", vector_instructions[alu->vector_opc].name); - if (alu->pred_select & 0x2) { - /* seems to work similar to conditional execution in ARM instruction - * set, so let's use a similar syntax for now: - */ - printf((alu->pred_select & 0x1) ? "EQ" : "NE"); - } + if (alu->pred_select & 0x2) { + /* seems to work similar to conditional execution in ARM instruction + * set, so let's use a similar syntax for now: + */ + printf((alu->pred_select & 0x1) ? "EQ" : "NE"); + } - printf("\t"); + printf("\t"); - print_dstreg(alu->vector_dest, alu->vector_write_mask, alu->export_data); - printf(" = "); - if (vector_instructions[alu->vector_opc].num_srcs == 3) { - print_srcreg(alu->src3_reg, alu->src3_sel, alu->src3_swiz, - alu->src3_reg_negate, alu->src3_reg_abs); - printf(", "); - } - print_srcreg(alu->src1_reg, alu->src1_sel, alu->src1_swiz, - alu->src1_reg_negate, alu->src1_reg_abs); - if (vector_instructions[alu->vector_opc].num_srcs > 1) { - printf(", "); - print_srcreg(alu->src2_reg, alu->src2_sel, alu->src2_swiz, - alu->src2_reg_negate, alu->src2_reg_abs); - } + print_dstreg(alu->vector_dest, alu->vector_write_mask, alu->export_data); + printf(" = "); + if (vector_instructions[alu->vector_opc].num_srcs == 3) { + print_srcreg(alu->src3_reg, alu->src3_sel, alu->src3_swiz, + alu->src3_reg_negate, alu->src3_reg_abs); + printf(", "); + } + print_srcreg(alu->src1_reg, alu->src1_sel, alu->src1_swiz, + alu->src1_reg_negate, alu->src1_reg_abs); + if (vector_instructions[alu->vector_opc].num_srcs > 1) { + printf(", "); + print_srcreg(alu->src2_reg, alu->src2_sel, alu->src2_swiz, + alu->src2_reg_negate, alu->src2_reg_abs); + } - if (alu->vector_clamp) - printf(" CLAMP"); + if (alu->vector_clamp) + printf(" CLAMP"); - if (alu->export_data) - print_export_comment(alu->vector_dest, type); + if (alu->export_data) + print_export_comment(alu->vector_dest, type); - printf("\n"); + printf("\n"); - if (alu->scalar_write_mask || !alu->vector_write_mask) { - /* 2nd optional scalar op: */ + if (alu->scalar_write_mask || !alu->vector_write_mask) { + /* 2nd optional scalar op: */ - printf("%s", levels[level]); - if (debug & PRINT_RAW) - printf(" \t"); + printf("%s", levels[level]); + if (debug & PRINT_RAW) + printf(" \t"); - if (scalar_instructions[alu->scalar_opc].name) { - printf("\t \t%s\t", scalar_instructions[alu->scalar_opc].name); - } else { - printf("\t \tOP(%u)\t", alu->scalar_opc); - } + if (scalar_instructions[alu->scalar_opc].name) { + printf("\t \t%s\t", scalar_instructions[alu->scalar_opc].name); + } else { + printf("\t \tOP(%u)\t", alu->scalar_opc); + } - print_dstreg(alu->scalar_dest, alu->scalar_write_mask, alu->export_data); - printf(" = "); - print_srcreg(alu->src3_reg, alu->src3_sel, alu->src3_swiz, - alu->src3_reg_negate, alu->src3_reg_abs); - // TODO ADD/MUL must have another src?!? - if (alu->scalar_clamp) - printf(" CLAMP"); - if (alu->export_data) - print_export_comment(alu->scalar_dest, type); - printf("\n"); - } + print_dstreg(alu->scalar_dest, alu->scalar_write_mask, alu->export_data); + printf(" = "); + print_srcreg(alu->src3_reg, alu->src3_sel, alu->src3_swiz, + alu->src3_reg_negate, alu->src3_reg_abs); + // TODO ADD/MUL must have another src?!? + if (alu->scalar_clamp) + printf(" CLAMP"); + if (alu->export_data) + print_export_comment(alu->scalar_dest, type); + printf("\n"); + } - return 0; + return 0; } - /* * FETCH instructions: */ struct { - const char *name; + const char *name; } fetch_types[0xff] = { -#define TYPE(id) [id] = { #id } - TYPE(FMT_1_REVERSE), - TYPE(FMT_32_FLOAT), - TYPE(FMT_32_32_FLOAT), - TYPE(FMT_32_32_32_FLOAT), - TYPE(FMT_32_32_32_32_FLOAT), - TYPE(FMT_16), - TYPE(FMT_16_16), - TYPE(FMT_16_16_16_16), - TYPE(FMT_8), - TYPE(FMT_8_8), - TYPE(FMT_8_8_8_8), - TYPE(FMT_32), - TYPE(FMT_32_32), - TYPE(FMT_32_32_32_32), +#define TYPE(id) [id] = {#id} + TYPE(FMT_1_REVERSE), + TYPE(FMT_32_FLOAT), + TYPE(FMT_32_32_FLOAT), + TYPE(FMT_32_32_32_FLOAT), + TYPE(FMT_32_32_32_32_FLOAT), + TYPE(FMT_16), + TYPE(FMT_16_16), + TYPE(FMT_16_16_16_16), + TYPE(FMT_8), + TYPE(FMT_8_8), + TYPE(FMT_8_8_8_8), + TYPE(FMT_32), + TYPE(FMT_32_32), + TYPE(FMT_32_32_32_32), #undef TYPE }; -static void print_fetch_dst(uint32_t dst_reg, uint32_t dst_swiz) +static void +print_fetch_dst(uint32_t dst_reg, uint32_t dst_swiz) { - int i; - printf("\tR%u.", dst_reg); - for (i = 0; i < 4; i++) { - printf("%c", chan_names[dst_swiz & 0x7]); - dst_swiz >>= 3; - } + int i; + printf("\tR%u.", dst_reg); + for (i = 0; i < 4; i++) { + printf("%c", chan_names[dst_swiz & 0x7]); + dst_swiz >>= 3; + } } -static void print_fetch_vtx(instr_fetch_t *fetch) +static void +print_fetch_vtx(instr_fetch_t *fetch) { - instr_fetch_vtx_t *vtx = &fetch->vtx; + instr_fetch_vtx_t *vtx = &fetch->vtx; - if (vtx->pred_select) { - /* seems to work similar to conditional execution in ARM instruction - * set, so let's use a similar syntax for now: - */ - printf(vtx->pred_condition ? "EQ" : "NE"); - } + if (vtx->pred_select) { + /* seems to work similar to conditional execution in ARM instruction + * set, so let's use a similar syntax for now: + */ + printf(vtx->pred_condition ? "EQ" : "NE"); + } - print_fetch_dst(vtx->dst_reg, vtx->dst_swiz); - printf(" = R%u.", vtx->src_reg); - printf("%c", chan_names[vtx->src_swiz & 0x3]); - if (fetch_types[vtx->format].name) { - printf(" %s", fetch_types[vtx->format].name); - } else { - printf(" TYPE(0x%x)", vtx->format); - } - printf(" %s", vtx->format_comp_all ? "SIGNED" : "UNSIGNED"); - if (!vtx->num_format_all) - printf(" NORMALIZED"); - printf(" STRIDE(%u)", vtx->stride); - if (vtx->offset) - printf(" OFFSET(%u)", vtx->offset); - printf(" CONST(%u, %u)", vtx->const_index, vtx->const_index_sel); - if (0) { - // XXX - printf(" src_reg_am=%u", vtx->src_reg_am); - printf(" dst_reg_am=%u", vtx->dst_reg_am); - printf(" num_format_all=%u", vtx->num_format_all); - printf(" signed_rf_mode_all=%u", vtx->signed_rf_mode_all); - printf(" exp_adjust_all=%u", vtx->exp_adjust_all); - } + print_fetch_dst(vtx->dst_reg, vtx->dst_swiz); + printf(" = R%u.", vtx->src_reg); + printf("%c", chan_names[vtx->src_swiz & 0x3]); + if (fetch_types[vtx->format].name) { + printf(" %s", fetch_types[vtx->format].name); + } else { + printf(" TYPE(0x%x)", vtx->format); + } + printf(" %s", vtx->format_comp_all ? "SIGNED" : "UNSIGNED"); + if (!vtx->num_format_all) + printf(" NORMALIZED"); + printf(" STRIDE(%u)", vtx->stride); + if (vtx->offset) + printf(" OFFSET(%u)", vtx->offset); + printf(" CONST(%u, %u)", vtx->const_index, vtx->const_index_sel); + if (0) { + // XXX + printf(" src_reg_am=%u", vtx->src_reg_am); + printf(" dst_reg_am=%u", vtx->dst_reg_am); + printf(" num_format_all=%u", vtx->num_format_all); + printf(" signed_rf_mode_all=%u", vtx->signed_rf_mode_all); + printf(" exp_adjust_all=%u", vtx->exp_adjust_all); + } } -static void print_fetch_tex(instr_fetch_t *fetch) +static void +print_fetch_tex(instr_fetch_t *fetch) { - static const char *filter[] = { - [TEX_FILTER_POINT] = "POINT", - [TEX_FILTER_LINEAR] = "LINEAR", - [TEX_FILTER_BASEMAP] = "BASEMAP", - }; - static const char *aniso_filter[] = { - [ANISO_FILTER_DISABLED] = "DISABLED", - [ANISO_FILTER_MAX_1_1] = "MAX_1_1", - [ANISO_FILTER_MAX_2_1] = "MAX_2_1", - [ANISO_FILTER_MAX_4_1] = "MAX_4_1", - [ANISO_FILTER_MAX_8_1] = "MAX_8_1", - [ANISO_FILTER_MAX_16_1] = "MAX_16_1", - }; - static const char *arbitrary_filter[] = { - [ARBITRARY_FILTER_2X4_SYM] = "2x4_SYM", - [ARBITRARY_FILTER_2X4_ASYM] = "2x4_ASYM", - [ARBITRARY_FILTER_4X2_SYM] = "4x2_SYM", - [ARBITRARY_FILTER_4X2_ASYM] = "4x2_ASYM", - [ARBITRARY_FILTER_4X4_SYM] = "4x4_SYM", - [ARBITRARY_FILTER_4X4_ASYM] = "4x4_ASYM", - }; - static const char *sample_loc[] = { - [SAMPLE_CENTROID] = "CENTROID", - [SAMPLE_CENTER] = "CENTER", - }; - instr_fetch_tex_t *tex = &fetch->tex; - uint32_t src_swiz = tex->src_swiz; - int i; + static const char *filter[] = { + [TEX_FILTER_POINT] = "POINT", + [TEX_FILTER_LINEAR] = "LINEAR", + [TEX_FILTER_BASEMAP] = "BASEMAP", + }; + static const char *aniso_filter[] = { + [ANISO_FILTER_DISABLED] = "DISABLED", + [ANISO_FILTER_MAX_1_1] = "MAX_1_1", + [ANISO_FILTER_MAX_2_1] = "MAX_2_1", + [ANISO_FILTER_MAX_4_1] = "MAX_4_1", + [ANISO_FILTER_MAX_8_1] = "MAX_8_1", + [ANISO_FILTER_MAX_16_1] = "MAX_16_1", + }; + static const char *arbitrary_filter[] = { + [ARBITRARY_FILTER_2X4_SYM] = "2x4_SYM", + [ARBITRARY_FILTER_2X4_ASYM] = "2x4_ASYM", + [ARBITRARY_FILTER_4X2_SYM] = "4x2_SYM", + [ARBITRARY_FILTER_4X2_ASYM] = "4x2_ASYM", + [ARBITRARY_FILTER_4X4_SYM] = "4x4_SYM", + [ARBITRARY_FILTER_4X4_ASYM] = "4x4_ASYM", + }; + static const char *sample_loc[] = { + [SAMPLE_CENTROID] = "CENTROID", + [SAMPLE_CENTER] = "CENTER", + }; + instr_fetch_tex_t *tex = &fetch->tex; + uint32_t src_swiz = tex->src_swiz; + int i; - if (tex->pred_select) { - /* seems to work similar to conditional execution in ARM instruction - * set, so let's use a similar syntax for now: - */ - printf(tex->pred_condition ? "EQ" : "NE"); - } + if (tex->pred_select) { + /* seems to work similar to conditional execution in ARM instruction + * set, so let's use a similar syntax for now: + */ + printf(tex->pred_condition ? "EQ" : "NE"); + } - print_fetch_dst(tex->dst_reg, tex->dst_swiz); - printf(" = R%u.", tex->src_reg); - for (i = 0; i < 3; i++) { - printf("%c", chan_names[src_swiz & 0x3]); - src_swiz >>= 2; - } - printf(" CONST(%u)", tex->const_idx); - if (tex->fetch_valid_only) - printf(" VALID_ONLY"); - if (tex->tx_coord_denorm) - printf(" DENORM"); - if (tex->mag_filter != TEX_FILTER_USE_FETCH_CONST) - printf(" MAG(%s)", filter[tex->mag_filter]); - if (tex->min_filter != TEX_FILTER_USE_FETCH_CONST) - printf(" MIN(%s)", filter[tex->min_filter]); - if (tex->mip_filter != TEX_FILTER_USE_FETCH_CONST) - printf(" MIP(%s)", filter[tex->mip_filter]); - if (tex->aniso_filter != ANISO_FILTER_USE_FETCH_CONST) - printf(" ANISO(%s)", aniso_filter[tex->aniso_filter]); - if (tex->arbitrary_filter != ARBITRARY_FILTER_USE_FETCH_CONST) - printf(" ARBITRARY(%s)", arbitrary_filter[tex->arbitrary_filter]); - if (tex->vol_mag_filter != TEX_FILTER_USE_FETCH_CONST) - printf(" VOL_MAG(%s)", filter[tex->vol_mag_filter]); - if (tex->vol_min_filter != TEX_FILTER_USE_FETCH_CONST) - printf(" VOL_MIN(%s)", filter[tex->vol_min_filter]); - if (!tex->use_comp_lod) { - printf(" LOD(%u)", tex->use_comp_lod); - printf(" LOD_BIAS(%u)", tex->lod_bias); - } - if (tex->use_reg_lod) { - printf(" REG_LOD(%u)", tex->use_reg_lod); - } - if (tex->use_reg_gradients) - printf(" USE_REG_GRADIENTS"); - printf(" LOCATION(%s)", sample_loc[tex->sample_location]); - if (tex->offset_x || tex->offset_y || tex->offset_z) - printf(" OFFSET(%u,%u,%u)", tex->offset_x, tex->offset_y, tex->offset_z); + print_fetch_dst(tex->dst_reg, tex->dst_swiz); + printf(" = R%u.", tex->src_reg); + for (i = 0; i < 3; i++) { + printf("%c", chan_names[src_swiz & 0x3]); + src_swiz >>= 2; + } + printf(" CONST(%u)", tex->const_idx); + if (tex->fetch_valid_only) + printf(" VALID_ONLY"); + if (tex->tx_coord_denorm) + printf(" DENORM"); + if (tex->mag_filter != TEX_FILTER_USE_FETCH_CONST) + printf(" MAG(%s)", filter[tex->mag_filter]); + if (tex->min_filter != TEX_FILTER_USE_FETCH_CONST) + printf(" MIN(%s)", filter[tex->min_filter]); + if (tex->mip_filter != TEX_FILTER_USE_FETCH_CONST) + printf(" MIP(%s)", filter[tex->mip_filter]); + if (tex->aniso_filter != ANISO_FILTER_USE_FETCH_CONST) + printf(" ANISO(%s)", aniso_filter[tex->aniso_filter]); + if (tex->arbitrary_filter != ARBITRARY_FILTER_USE_FETCH_CONST) + printf(" ARBITRARY(%s)", arbitrary_filter[tex->arbitrary_filter]); + if (tex->vol_mag_filter != TEX_FILTER_USE_FETCH_CONST) + printf(" VOL_MAG(%s)", filter[tex->vol_mag_filter]); + if (tex->vol_min_filter != TEX_FILTER_USE_FETCH_CONST) + printf(" VOL_MIN(%s)", filter[tex->vol_min_filter]); + if (!tex->use_comp_lod) { + printf(" LOD(%u)", tex->use_comp_lod); + printf(" LOD_BIAS(%u)", tex->lod_bias); + } + if (tex->use_reg_lod) { + printf(" REG_LOD(%u)", tex->use_reg_lod); + } + if (tex->use_reg_gradients) + printf(" USE_REG_GRADIENTS"); + printf(" LOCATION(%s)", sample_loc[tex->sample_location]); + if (tex->offset_x || tex->offset_y || tex->offset_z) + printf(" OFFSET(%u,%u,%u)", tex->offset_x, tex->offset_y, tex->offset_z); } struct { - const char *name; - void (*fxn)(instr_fetch_t *cf); + const char *name; + void (*fxn)(instr_fetch_t *cf); } fetch_instructions[] = { -#define INSTR(opc, name, fxn) [opc] = { name, fxn } - INSTR(VTX_FETCH, "VERTEX", print_fetch_vtx), - INSTR(TEX_FETCH, "SAMPLE", print_fetch_tex), - INSTR(TEX_GET_BORDER_COLOR_FRAC, "?", print_fetch_tex), - INSTR(TEX_GET_COMP_TEX_LOD, "?", print_fetch_tex), - INSTR(TEX_GET_GRADIENTS, "?", print_fetch_tex), - INSTR(TEX_GET_WEIGHTS, "?", print_fetch_tex), - INSTR(TEX_SET_TEX_LOD, "SET_TEX_LOD", print_fetch_tex), - INSTR(TEX_SET_GRADIENTS_H, "?", print_fetch_tex), - INSTR(TEX_SET_GRADIENTS_V, "?", print_fetch_tex), - INSTR(TEX_RESERVED_4, "?", print_fetch_tex), +#define INSTR(opc, name, fxn) [opc] = {name, fxn} + INSTR(VTX_FETCH, "VERTEX", print_fetch_vtx), + INSTR(TEX_FETCH, "SAMPLE", print_fetch_tex), + INSTR(TEX_GET_BORDER_COLOR_FRAC, "?", print_fetch_tex), + INSTR(TEX_GET_COMP_TEX_LOD, "?", print_fetch_tex), + INSTR(TEX_GET_GRADIENTS, "?", print_fetch_tex), + INSTR(TEX_GET_WEIGHTS, "?", print_fetch_tex), + INSTR(TEX_SET_TEX_LOD, "SET_TEX_LOD", print_fetch_tex), + INSTR(TEX_SET_GRADIENTS_H, "?", print_fetch_tex), + INSTR(TEX_SET_GRADIENTS_V, "?", print_fetch_tex), + INSTR(TEX_RESERVED_4, "?", print_fetch_tex), #undef INSTR }; -static int disasm_fetch(uint32_t *dwords, uint32_t alu_off, int level, int sync) +static int +disasm_fetch(uint32_t *dwords, uint32_t alu_off, int level, int sync) { - instr_fetch_t *fetch = (instr_fetch_t *)dwords; + instr_fetch_t *fetch = (instr_fetch_t *)dwords; - printf("%s", levels[level]); - if (debug & PRINT_RAW) { - printf("%02x: %08x %08x %08x\t", alu_off, - dwords[0], dwords[1], dwords[2]); - } + printf("%s", levels[level]); + if (debug & PRINT_RAW) { + printf("%02x: %08x %08x %08x\t", alu_off, dwords[0], dwords[1], + dwords[2]); + } - printf(" %sFETCH:\t", sync ? "(S)" : " "); - printf("%s", fetch_instructions[fetch->opc].name); - fetch_instructions[fetch->opc].fxn(fetch); - printf("\n"); + printf(" %sFETCH:\t", sync ? "(S)" : " "); + printf("%s", fetch_instructions[fetch->opc].name); + fetch_instructions[fetch->opc].fxn(fetch); + printf("\n"); - return 0; + return 0; } /* * CF instructions: */ -static int cf_exec(instr_cf_t *cf) +static int +cf_exec(instr_cf_t *cf) { - return (cf->opc == EXEC) || - (cf->opc == EXEC_END) || - (cf->opc == COND_EXEC) || - (cf->opc == COND_EXEC_END) || - (cf->opc == COND_PRED_EXEC) || - (cf->opc == COND_PRED_EXEC_END) || - (cf->opc == COND_EXEC_PRED_CLEAN) || - (cf->opc == COND_EXEC_PRED_CLEAN_END); + return (cf->opc == EXEC) || (cf->opc == EXEC_END) || + (cf->opc == COND_EXEC) || (cf->opc == COND_EXEC_END) || + (cf->opc == COND_PRED_EXEC) || (cf->opc == COND_PRED_EXEC_END) || + (cf->opc == COND_EXEC_PRED_CLEAN) || + (cf->opc == COND_EXEC_PRED_CLEAN_END); } -static int cf_cond_exec(instr_cf_t *cf) +static int +cf_cond_exec(instr_cf_t *cf) { - return (cf->opc == COND_EXEC) || - (cf->opc == COND_EXEC_END) || - (cf->opc == COND_PRED_EXEC) || - (cf->opc == COND_PRED_EXEC_END) || - (cf->opc == COND_EXEC_PRED_CLEAN) || - (cf->opc == COND_EXEC_PRED_CLEAN_END); + return (cf->opc == COND_EXEC) || (cf->opc == COND_EXEC_END) || + (cf->opc == COND_PRED_EXEC) || (cf->opc == COND_PRED_EXEC_END) || + (cf->opc == COND_EXEC_PRED_CLEAN) || + (cf->opc == COND_EXEC_PRED_CLEAN_END); } -static void print_cf_nop(instr_cf_t *cf) +static void +print_cf_nop(instr_cf_t *cf) { } -static void print_cf_exec(instr_cf_t *cf) +static void +print_cf_exec(instr_cf_t *cf) { - printf(" ADDR(0x%x) CNT(0x%x)", cf->exec.address, cf->exec.count); - if (cf->exec.yeild) - printf(" YIELD"); - if (cf->exec.vc) - printf(" VC(0x%x)", cf->exec.vc); - if (cf->exec.bool_addr) - printf(" BOOL_ADDR(0x%x)", cf->exec.bool_addr); - if (cf->exec.address_mode == ABSOLUTE_ADDR) - printf(" ABSOLUTE_ADDR"); - if (cf_cond_exec(cf)) - printf(" COND(%d)", cf->exec.condition); + printf(" ADDR(0x%x) CNT(0x%x)", cf->exec.address, cf->exec.count); + if (cf->exec.yeild) + printf(" YIELD"); + if (cf->exec.vc) + printf(" VC(0x%x)", cf->exec.vc); + if (cf->exec.bool_addr) + printf(" BOOL_ADDR(0x%x)", cf->exec.bool_addr); + if (cf->exec.address_mode == ABSOLUTE_ADDR) + printf(" ABSOLUTE_ADDR"); + if (cf_cond_exec(cf)) + printf(" COND(%d)", cf->exec.condition); } -static void print_cf_loop(instr_cf_t *cf) +static void +print_cf_loop(instr_cf_t *cf) { - printf(" ADDR(0x%x) LOOP_ID(%d)", cf->loop.address, cf->loop.loop_id); - if (cf->loop.address_mode == ABSOLUTE_ADDR) - printf(" ABSOLUTE_ADDR"); + printf(" ADDR(0x%x) LOOP_ID(%d)", cf->loop.address, cf->loop.loop_id); + if (cf->loop.address_mode == ABSOLUTE_ADDR) + printf(" ABSOLUTE_ADDR"); } -static void print_cf_jmp_call(instr_cf_t *cf) +static void +print_cf_jmp_call(instr_cf_t *cf) { - printf(" ADDR(0x%x) DIR(%d)", cf->jmp_call.address, cf->jmp_call.direction); - if (cf->jmp_call.force_call) - printf(" FORCE_CALL"); - if (cf->jmp_call.predicated_jmp) - printf(" COND(%d)", cf->jmp_call.condition); - if (cf->jmp_call.bool_addr) - printf(" BOOL_ADDR(0x%x)", cf->jmp_call.bool_addr); - if (cf->jmp_call.address_mode == ABSOLUTE_ADDR) - printf(" ABSOLUTE_ADDR"); + printf(" ADDR(0x%x) DIR(%d)", cf->jmp_call.address, cf->jmp_call.direction); + if (cf->jmp_call.force_call) + printf(" FORCE_CALL"); + if (cf->jmp_call.predicated_jmp) + printf(" COND(%d)", cf->jmp_call.condition); + if (cf->jmp_call.bool_addr) + printf(" BOOL_ADDR(0x%x)", cf->jmp_call.bool_addr); + if (cf->jmp_call.address_mode == ABSOLUTE_ADDR) + printf(" ABSOLUTE_ADDR"); } -static void print_cf_alloc(instr_cf_t *cf) +static void +print_cf_alloc(instr_cf_t *cf) { - static const char *bufname[] = { - [SQ_NO_ALLOC] = "NO ALLOC", - [SQ_POSITION] = "POSITION", - [SQ_PARAMETER_PIXEL] = "PARAM/PIXEL", - [SQ_MEMORY] = "MEMORY", - }; - printf(" %s SIZE(0x%x)", bufname[cf->alloc.buffer_select], cf->alloc.size); - if (cf->alloc.no_serial) - printf(" NO_SERIAL"); - if (cf->alloc.alloc_mode) // ??? - printf(" ALLOC_MODE"); + static const char *bufname[] = { + [SQ_NO_ALLOC] = "NO ALLOC", + [SQ_POSITION] = "POSITION", + [SQ_PARAMETER_PIXEL] = "PARAM/PIXEL", + [SQ_MEMORY] = "MEMORY", + }; + printf(" %s SIZE(0x%x)", bufname[cf->alloc.buffer_select], cf->alloc.size); + if (cf->alloc.no_serial) + printf(" NO_SERIAL"); + if (cf->alloc.alloc_mode) // ??? + printf(" ALLOC_MODE"); } struct { - const char *name; - void (*fxn)(instr_cf_t *cf); + const char *name; + void (*fxn)(instr_cf_t *cf); } cf_instructions[] = { -#define INSTR(opc, fxn) [opc] = { #opc, fxn } - INSTR(NOP, print_cf_nop), - INSTR(EXEC, print_cf_exec), - INSTR(EXEC_END, print_cf_exec), - INSTR(COND_EXEC, print_cf_exec), - INSTR(COND_EXEC_END, print_cf_exec), - INSTR(COND_PRED_EXEC, print_cf_exec), - INSTR(COND_PRED_EXEC_END, print_cf_exec), - INSTR(LOOP_START, print_cf_loop), - INSTR(LOOP_END, print_cf_loop), - INSTR(COND_CALL, print_cf_jmp_call), - INSTR(RETURN, print_cf_jmp_call), - INSTR(COND_JMP, print_cf_jmp_call), - INSTR(ALLOC, print_cf_alloc), - INSTR(COND_EXEC_PRED_CLEAN, print_cf_exec), - INSTR(COND_EXEC_PRED_CLEAN_END, print_cf_exec), - INSTR(MARK_VS_FETCH_DONE, print_cf_nop), // ?? +#define INSTR(opc, fxn) [opc] = {#opc, fxn} + INSTR(NOP, print_cf_nop), + INSTR(EXEC, print_cf_exec), + INSTR(EXEC_END, print_cf_exec), + INSTR(COND_EXEC, print_cf_exec), + INSTR(COND_EXEC_END, print_cf_exec), + INSTR(COND_PRED_EXEC, print_cf_exec), + INSTR(COND_PRED_EXEC_END, print_cf_exec), + INSTR(LOOP_START, print_cf_loop), + INSTR(LOOP_END, print_cf_loop), + INSTR(COND_CALL, print_cf_jmp_call), + INSTR(RETURN, print_cf_jmp_call), + INSTR(COND_JMP, print_cf_jmp_call), + INSTR(ALLOC, print_cf_alloc), + INSTR(COND_EXEC_PRED_CLEAN, print_cf_exec), + INSTR(COND_EXEC_PRED_CLEAN_END, print_cf_exec), + INSTR(MARK_VS_FETCH_DONE, print_cf_nop), // ?? #undef INSTR }; -static void print_cf(instr_cf_t *cf, int level) +static void +print_cf(instr_cf_t *cf, int level) { - printf("%s", levels[level]); - if (debug & PRINT_RAW) { - uint16_t words[3]; - memcpy(&words, cf, sizeof(words)); - printf(" %04x %04x %04x \t", - words[0], words[1], words[2]); - } - printf("%s", cf_instructions[cf->opc].name); - cf_instructions[cf->opc].fxn(cf); - printf("\n"); + printf("%s", levels[level]); + if (debug & PRINT_RAW) { + uint16_t words[3]; + memcpy(&words, cf, sizeof(words)); + printf(" %04x %04x %04x \t", words[0], words[1], words[2]); + } + printf("%s", cf_instructions[cf->opc].name); + cf_instructions[cf->opc].fxn(cf); + printf("\n"); } /* @@ -596,43 +617,47 @@ static void print_cf(instr_cf_t *cf, int level) * 2) ALU and FETCH instructions */ -int disasm_a2xx(uint32_t *dwords, int sizedwords, int level, gl_shader_stage type) +int +disasm_a2xx(uint32_t *dwords, int sizedwords, int level, gl_shader_stage type) { - instr_cf_t *cfs = (instr_cf_t *)dwords; - int idx, max_idx; + instr_cf_t *cfs = (instr_cf_t *)dwords; + int idx, max_idx; - for (idx = 0; ; idx++) { - instr_cf_t *cf = &cfs[idx]; - if (cf_exec(cf)) { - max_idx = 2 * cf->exec.address; - break; - } - } + for (idx = 0;; idx++) { + instr_cf_t *cf = &cfs[idx]; + if (cf_exec(cf)) { + max_idx = 2 * cf->exec.address; + break; + } + } - for (idx = 0; idx < max_idx; idx++) { - instr_cf_t *cf = &cfs[idx]; + for (idx = 0; idx < max_idx; idx++) { + instr_cf_t *cf = &cfs[idx]; - print_cf(cf, level); + print_cf(cf, level); - if (cf_exec(cf)) { - uint32_t sequence = cf->exec.serialize; - uint32_t i; - for (i = 0; i < cf->exec.count; i++) { - uint32_t alu_off = (cf->exec.address + i); - if (sequence & 0x1) { - disasm_fetch(dwords + alu_off * 3, alu_off, level, sequence & 0x2); - } else { - disasm_alu(dwords + alu_off * 3, alu_off, level, sequence & 0x2, type); - } - sequence >>= 2; - } - } - } + if (cf_exec(cf)) { + uint32_t sequence = cf->exec.serialize; + uint32_t i; + for (i = 0; i < cf->exec.count; i++) { + uint32_t alu_off = (cf->exec.address + i); + if (sequence & 0x1) { + disasm_fetch(dwords + alu_off * 3, alu_off, level, + sequence & 0x2); + } else { + disasm_alu(dwords + alu_off * 3, alu_off, level, sequence & 0x2, + type); + } + sequence >>= 2; + } + } + } - return 0; + return 0; } -void disasm_a2xx_set_debug(enum debug_t d) +void +disasm_a2xx_set_debug(enum debug_t d) { - debug = d; + debug = d; } diff --git a/src/freedreno/ir2/instr-a2xx.h b/src/freedreno/ir2/instr-a2xx.h index 5049c553870..ead9a2f550d 100644 --- a/src/freedreno/ir2/instr-a2xx.h +++ b/src/freedreno/ir2/instr-a2xx.h @@ -31,377 +31,372 @@ #include "adreno_pm4.xml.h" #include "a2xx.xml.h" - /* * ALU instructions: */ typedef enum { - ADDs = 0, - ADD_PREVs = 1, - MULs = 2, - MUL_PREVs = 3, - MUL_PREV2s = 4, - MAXs = 5, - MINs = 6, - SETEs = 7, - SETGTs = 8, - SETGTEs = 9, - SETNEs = 10, - FRACs = 11, - TRUNCs = 12, - FLOORs = 13, - EXP_IEEE = 14, - LOG_CLAMP = 15, - LOG_IEEE = 16, - RECIP_CLAMP = 17, - RECIP_FF = 18, - RECIP_IEEE = 19, - RECIPSQ_CLAMP = 20, - RECIPSQ_FF = 21, - RECIPSQ_IEEE = 22, - MOVAs = 23, - MOVA_FLOORs = 24, - SUBs = 25, - SUB_PREVs = 26, - PRED_SETEs = 27, - PRED_SETNEs = 28, - PRED_SETGTs = 29, - PRED_SETGTEs = 30, - PRED_SET_INVs = 31, - PRED_SET_POPs = 32, - PRED_SET_CLRs = 33, - PRED_SET_RESTOREs = 34, - KILLEs = 35, - KILLGTs = 36, - KILLGTEs = 37, - KILLNEs = 38, - KILLONEs = 39, - SQRT_IEEE = 40, - MUL_CONST_0 = 42, - MUL_CONST_1 = 43, - ADD_CONST_0 = 44, - ADD_CONST_1 = 45, - SUB_CONST_0 = 46, - SUB_CONST_1 = 47, - SIN = 48, - COS = 49, - RETAIN_PREV = 50, - SCALAR_NONE = 63, + ADDs = 0, + ADD_PREVs = 1, + MULs = 2, + MUL_PREVs = 3, + MUL_PREV2s = 4, + MAXs = 5, + MINs = 6, + SETEs = 7, + SETGTs = 8, + SETGTEs = 9, + SETNEs = 10, + FRACs = 11, + TRUNCs = 12, + FLOORs = 13, + EXP_IEEE = 14, + LOG_CLAMP = 15, + LOG_IEEE = 16, + RECIP_CLAMP = 17, + RECIP_FF = 18, + RECIP_IEEE = 19, + RECIPSQ_CLAMP = 20, + RECIPSQ_FF = 21, + RECIPSQ_IEEE = 22, + MOVAs = 23, + MOVA_FLOORs = 24, + SUBs = 25, + SUB_PREVs = 26, + PRED_SETEs = 27, + PRED_SETNEs = 28, + PRED_SETGTs = 29, + PRED_SETGTEs = 30, + PRED_SET_INVs = 31, + PRED_SET_POPs = 32, + PRED_SET_CLRs = 33, + PRED_SET_RESTOREs = 34, + KILLEs = 35, + KILLGTs = 36, + KILLGTEs = 37, + KILLNEs = 38, + KILLONEs = 39, + SQRT_IEEE = 40, + MUL_CONST_0 = 42, + MUL_CONST_1 = 43, + ADD_CONST_0 = 44, + ADD_CONST_1 = 45, + SUB_CONST_0 = 46, + SUB_CONST_1 = 47, + SIN = 48, + COS = 49, + RETAIN_PREV = 50, + SCALAR_NONE = 63, } instr_scalar_opc_t; typedef enum { - ADDv = 0, - MULv = 1, - MAXv = 2, - MINv = 3, - SETEv = 4, - SETGTv = 5, - SETGTEv = 6, - SETNEv = 7, - FRACv = 8, - TRUNCv = 9, - FLOORv = 10, - MULADDv = 11, - CNDEv = 12, - CNDGTEv = 13, - CNDGTv = 14, - DOT4v = 15, - DOT3v = 16, - DOT2ADDv = 17, - CUBEv = 18, - MAX4v = 19, - PRED_SETE_PUSHv = 20, - PRED_SETNE_PUSHv = 21, - PRED_SETGT_PUSHv = 22, - PRED_SETGTE_PUSHv = 23, - KILLEv = 24, - KILLGTv = 25, - KILLGTEv = 26, - KILLNEv = 27, - DSTv = 28, - MOVAv = 29, - VECTOR_NONE = 31, + ADDv = 0, + MULv = 1, + MAXv = 2, + MINv = 3, + SETEv = 4, + SETGTv = 5, + SETGTEv = 6, + SETNEv = 7, + FRACv = 8, + TRUNCv = 9, + FLOORv = 10, + MULADDv = 11, + CNDEv = 12, + CNDGTEv = 13, + CNDGTv = 14, + DOT4v = 15, + DOT3v = 16, + DOT2ADDv = 17, + CUBEv = 18, + MAX4v = 19, + PRED_SETE_PUSHv = 20, + PRED_SETNE_PUSHv = 21, + PRED_SETGT_PUSHv = 22, + PRED_SETGTE_PUSHv = 23, + KILLEv = 24, + KILLGTv = 25, + KILLGTEv = 26, + KILLNEv = 27, + DSTv = 28, + MOVAv = 29, + VECTOR_NONE = 31, } instr_vector_opc_t; typedef struct PACKED { - /* dword0: */ - uint8_t vector_dest : 6; - uint8_t vector_dest_rel : 1; - uint8_t low_precision_16b_fp : 1; - uint8_t scalar_dest : 6; - uint8_t scalar_dest_rel : 1; - uint8_t export_data : 1; - uint8_t vector_write_mask : 4; - uint8_t scalar_write_mask : 4; - uint8_t vector_clamp : 1; - uint8_t scalar_clamp : 1; - instr_scalar_opc_t scalar_opc : 6; - /* dword1: */ - uint8_t src3_swiz : 8; - uint8_t src2_swiz : 8; - uint8_t src1_swiz : 8; - uint8_t src3_reg_negate : 1; - uint8_t src2_reg_negate : 1; - uint8_t src1_reg_negate : 1; - uint8_t pred_select : 2; - uint8_t relative_addr : 1; - uint8_t const_1_rel_abs : 1; - uint8_t const_0_rel_abs : 1; - /* dword2: */ - union { - struct { - uint8_t src3_reg : 6; - uint8_t src3_reg_select : 1; - uint8_t src3_reg_abs : 1; - uint8_t src2_reg : 6; - uint8_t src2_reg_select : 1; - uint8_t src2_reg_abs : 1; - uint8_t src1_reg : 6; - uint8_t src1_reg_select : 1; - uint8_t src1_reg_abs : 1; - }; - /* constants have full 8-bit index */ - struct { - uint8_t src3_reg_byte : 8; - uint8_t src2_reg_byte : 8; - uint8_t src1_reg_byte : 8; - }; - }; - instr_vector_opc_t vector_opc : 5; - uint8_t src3_sel : 1; - uint8_t src2_sel : 1; - uint8_t src1_sel : 1; + /* dword0: */ + uint8_t vector_dest : 6; + uint8_t vector_dest_rel : 1; + uint8_t low_precision_16b_fp : 1; + uint8_t scalar_dest : 6; + uint8_t scalar_dest_rel : 1; + uint8_t export_data : 1; + uint8_t vector_write_mask : 4; + uint8_t scalar_write_mask : 4; + uint8_t vector_clamp : 1; + uint8_t scalar_clamp : 1; + instr_scalar_opc_t scalar_opc : 6; + /* dword1: */ + uint8_t src3_swiz : 8; + uint8_t src2_swiz : 8; + uint8_t src1_swiz : 8; + uint8_t src3_reg_negate : 1; + uint8_t src2_reg_negate : 1; + uint8_t src1_reg_negate : 1; + uint8_t pred_select : 2; + uint8_t relative_addr : 1; + uint8_t const_1_rel_abs : 1; + uint8_t const_0_rel_abs : 1; + /* dword2: */ + union { + struct { + uint8_t src3_reg : 6; + uint8_t src3_reg_select : 1; + uint8_t src3_reg_abs : 1; + uint8_t src2_reg : 6; + uint8_t src2_reg_select : 1; + uint8_t src2_reg_abs : 1; + uint8_t src1_reg : 6; + uint8_t src1_reg_select : 1; + uint8_t src1_reg_abs : 1; + }; + /* constants have full 8-bit index */ + struct { + uint8_t src3_reg_byte : 8; + uint8_t src2_reg_byte : 8; + uint8_t src1_reg_byte : 8; + }; + }; + instr_vector_opc_t vector_opc : 5; + uint8_t src3_sel : 1; + uint8_t src2_sel : 1; + uint8_t src1_sel : 1; } instr_alu_t; - - /* * CF instructions: */ typedef enum { - NOP = 0, - EXEC = 1, - EXEC_END = 2, - COND_EXEC = 3, - COND_EXEC_END = 4, - COND_PRED_EXEC = 5, - COND_PRED_EXEC_END = 6, - LOOP_START = 7, - LOOP_END = 8, - COND_CALL = 9, - RETURN = 10, - COND_JMP = 11, - ALLOC = 12, - COND_EXEC_PRED_CLEAN = 13, - COND_EXEC_PRED_CLEAN_END = 14, - MARK_VS_FETCH_DONE = 15, + NOP = 0, + EXEC = 1, + EXEC_END = 2, + COND_EXEC = 3, + COND_EXEC_END = 4, + COND_PRED_EXEC = 5, + COND_PRED_EXEC_END = 6, + LOOP_START = 7, + LOOP_END = 8, + COND_CALL = 9, + RETURN = 10, + COND_JMP = 11, + ALLOC = 12, + COND_EXEC_PRED_CLEAN = 13, + COND_EXEC_PRED_CLEAN_END = 14, + MARK_VS_FETCH_DONE = 15, } instr_cf_opc_t; typedef enum { - RELATIVE_ADDR = 0, - ABSOLUTE_ADDR = 1, + RELATIVE_ADDR = 0, + ABSOLUTE_ADDR = 1, } instr_addr_mode_t; typedef enum { - SQ_NO_ALLOC = 0, - SQ_POSITION = 1, - SQ_PARAMETER_PIXEL = 2, - SQ_MEMORY = 3, + SQ_NO_ALLOC = 0, + SQ_POSITION = 1, + SQ_PARAMETER_PIXEL = 2, + SQ_MEMORY = 3, } instr_alloc_type_t; typedef struct PACKED { - uint16_t address : 9; - uint8_t reserved0 : 3; - uint8_t count : 3; - uint8_t yeild : 1; - uint16_t serialize : 12; - uint32_t vc : 6; /* vertex cache? */ - uint32_t bool_addr : 8; - uint8_t condition : 1; - instr_addr_mode_t address_mode : 1; - instr_cf_opc_t opc : 4; + uint16_t address : 9; + uint8_t reserved0 : 3; + uint8_t count : 3; + uint8_t yeild : 1; + uint16_t serialize : 12; + uint32_t vc : 6; /* vertex cache? */ + uint32_t bool_addr : 8; + uint8_t condition : 1; + instr_addr_mode_t address_mode : 1; + instr_cf_opc_t opc : 4; } instr_cf_exec_t; typedef struct PACKED { - uint16_t address : 10; - uint8_t reserved0 : 6; - uint8_t loop_id : 5; - uint32_t reserved1 : 22; - instr_addr_mode_t address_mode : 1; - instr_cf_opc_t opc : 4; + uint16_t address : 10; + uint8_t reserved0 : 6; + uint8_t loop_id : 5; + uint32_t reserved1 : 22; + instr_addr_mode_t address_mode : 1; + instr_cf_opc_t opc : 4; } instr_cf_loop_t; typedef struct PACKED { - uint16_t address : 10; - uint8_t reserved0 : 3; - uint8_t force_call : 1; - uint8_t predicated_jmp : 1; - uint32_t reserved1 : 18; - uint8_t direction : 1; - uint32_t bool_addr : 8; - uint8_t condition : 1; - instr_addr_mode_t address_mode : 1; - instr_cf_opc_t opc : 4; + uint16_t address : 10; + uint8_t reserved0 : 3; + uint8_t force_call : 1; + uint8_t predicated_jmp : 1; + uint32_t reserved1 : 18; + uint8_t direction : 1; + uint32_t bool_addr : 8; + uint8_t condition : 1; + instr_addr_mode_t address_mode : 1; + instr_cf_opc_t opc : 4; } instr_cf_jmp_call_t; typedef struct PACKED { - uint8_t size : 4; - uint64_t reserved0 : 36; - uint8_t no_serial : 1; - instr_alloc_type_t buffer_select : 2; - uint8_t alloc_mode : 1; - instr_cf_opc_t opc : 4; + uint8_t size : 4; + uint64_t reserved0 : 36; + uint8_t no_serial : 1; + instr_alloc_type_t buffer_select : 2; + uint8_t alloc_mode : 1; + instr_cf_opc_t opc : 4; } instr_cf_alloc_t; typedef union PACKED { - instr_cf_exec_t exec; - instr_cf_loop_t loop; - instr_cf_jmp_call_t jmp_call; - instr_cf_alloc_t alloc; - struct PACKED { - uint64_t dummy : 44; - instr_cf_opc_t opc : 4; - }; + instr_cf_exec_t exec; + instr_cf_loop_t loop; + instr_cf_jmp_call_t jmp_call; + instr_cf_alloc_t alloc; + struct PACKED { + uint64_t dummy : 44; + instr_cf_opc_t opc : 4; + }; } instr_cf_t; - - /* * FETCH instructions: */ typedef enum { - VTX_FETCH = 0, - TEX_FETCH = 1, - TEX_GET_BORDER_COLOR_FRAC = 16, - TEX_GET_COMP_TEX_LOD = 17, - TEX_GET_GRADIENTS = 18, - TEX_GET_WEIGHTS = 19, - TEX_SET_TEX_LOD = 24, - TEX_SET_GRADIENTS_H = 25, - TEX_SET_GRADIENTS_V = 26, - TEX_RESERVED_4 = 27, + VTX_FETCH = 0, + TEX_FETCH = 1, + TEX_GET_BORDER_COLOR_FRAC = 16, + TEX_GET_COMP_TEX_LOD = 17, + TEX_GET_GRADIENTS = 18, + TEX_GET_WEIGHTS = 19, + TEX_SET_TEX_LOD = 24, + TEX_SET_GRADIENTS_H = 25, + TEX_SET_GRADIENTS_V = 26, + TEX_RESERVED_4 = 27, } instr_fetch_opc_t; typedef enum { - TEX_FILTER_POINT = 0, - TEX_FILTER_LINEAR = 1, - TEX_FILTER_BASEMAP = 2, /* only applicable for mip-filter */ - TEX_FILTER_USE_FETCH_CONST = 3, + TEX_FILTER_POINT = 0, + TEX_FILTER_LINEAR = 1, + TEX_FILTER_BASEMAP = 2, /* only applicable for mip-filter */ + TEX_FILTER_USE_FETCH_CONST = 3, } instr_tex_filter_t; typedef enum { - ANISO_FILTER_DISABLED = 0, - ANISO_FILTER_MAX_1_1 = 1, - ANISO_FILTER_MAX_2_1 = 2, - ANISO_FILTER_MAX_4_1 = 3, - ANISO_FILTER_MAX_8_1 = 4, - ANISO_FILTER_MAX_16_1 = 5, - ANISO_FILTER_USE_FETCH_CONST = 7, + ANISO_FILTER_DISABLED = 0, + ANISO_FILTER_MAX_1_1 = 1, + ANISO_FILTER_MAX_2_1 = 2, + ANISO_FILTER_MAX_4_1 = 3, + ANISO_FILTER_MAX_8_1 = 4, + ANISO_FILTER_MAX_16_1 = 5, + ANISO_FILTER_USE_FETCH_CONST = 7, } instr_aniso_filter_t; typedef enum { - ARBITRARY_FILTER_2X4_SYM = 0, - ARBITRARY_FILTER_2X4_ASYM = 1, - ARBITRARY_FILTER_4X2_SYM = 2, - ARBITRARY_FILTER_4X2_ASYM = 3, - ARBITRARY_FILTER_4X4_SYM = 4, - ARBITRARY_FILTER_4X4_ASYM = 5, - ARBITRARY_FILTER_USE_FETCH_CONST = 7, + ARBITRARY_FILTER_2X4_SYM = 0, + ARBITRARY_FILTER_2X4_ASYM = 1, + ARBITRARY_FILTER_4X2_SYM = 2, + ARBITRARY_FILTER_4X2_ASYM = 3, + ARBITRARY_FILTER_4X4_SYM = 4, + ARBITRARY_FILTER_4X4_ASYM = 5, + ARBITRARY_FILTER_USE_FETCH_CONST = 7, } instr_arbitrary_filter_t; typedef enum { - SAMPLE_CENTROID = 0, - SAMPLE_CENTER = 1, + SAMPLE_CENTROID = 0, + SAMPLE_CENTER = 1, } instr_sample_loc_t; typedef enum a2xx_sq_surfaceformat instr_surf_fmt_t; typedef struct PACKED { - /* dword0: */ - instr_fetch_opc_t opc : 5; - uint32_t src_reg : 6; - uint8_t src_reg_am : 1; - uint32_t dst_reg : 6; - uint8_t dst_reg_am : 1; - uint8_t fetch_valid_only : 1; - uint32_t const_idx : 5; - uint8_t tx_coord_denorm : 1; - uint8_t src_swiz : 6; - /* dword1: */ - uint16_t dst_swiz : 12; - instr_tex_filter_t mag_filter : 2; - instr_tex_filter_t min_filter : 2; - instr_tex_filter_t mip_filter : 2; - instr_aniso_filter_t aniso_filter : 3; - instr_arbitrary_filter_t arbitrary_filter : 3; - instr_tex_filter_t vol_mag_filter : 2; - instr_tex_filter_t vol_min_filter : 2; - uint8_t use_comp_lod : 1; - uint8_t use_reg_lod : 2; /* 0 for cube, 1 for 2d */ - uint8_t pred_select : 1; - /* dword2: */ - uint8_t use_reg_gradients : 1; - instr_sample_loc_t sample_location : 1; - uint32_t lod_bias : 7; - uint8_t unused : 7; - uint8_t offset_x : 5; - uint32_t offset_y : 5; - uint8_t offset_z : 5; - uint8_t pred_condition : 1; + /* dword0: */ + instr_fetch_opc_t opc : 5; + uint32_t src_reg : 6; + uint8_t src_reg_am : 1; + uint32_t dst_reg : 6; + uint8_t dst_reg_am : 1; + uint8_t fetch_valid_only : 1; + uint32_t const_idx : 5; + uint8_t tx_coord_denorm : 1; + uint8_t src_swiz : 6; + /* dword1: */ + uint16_t dst_swiz : 12; + instr_tex_filter_t mag_filter : 2; + instr_tex_filter_t min_filter : 2; + instr_tex_filter_t mip_filter : 2; + instr_aniso_filter_t aniso_filter : 3; + instr_arbitrary_filter_t arbitrary_filter : 3; + instr_tex_filter_t vol_mag_filter : 2; + instr_tex_filter_t vol_min_filter : 2; + uint8_t use_comp_lod : 1; + uint8_t use_reg_lod : 2; /* 0 for cube, 1 for 2d */ + uint8_t pred_select : 1; + /* dword2: */ + uint8_t use_reg_gradients : 1; + instr_sample_loc_t sample_location : 1; + uint32_t lod_bias : 7; + uint8_t unused : 7; + uint8_t offset_x : 5; + uint32_t offset_y : 5; + uint8_t offset_z : 5; + uint8_t pred_condition : 1; } instr_fetch_tex_t; typedef struct PACKED { - /* dword0: */ - instr_fetch_opc_t opc : 5; - uint32_t src_reg : 6; - uint8_t src_reg_am : 1; - uint32_t dst_reg : 6; - uint8_t dst_reg_am : 1; - uint8_t must_be_one : 1; - uint32_t const_index : 5; - uint8_t const_index_sel : 2; - uint8_t reserved0 : 3; - uint8_t src_swiz : 2; - /* dword1: */ - uint16_t dst_swiz : 12; - uint8_t format_comp_all : 1; /* '1' for signed, '0' for unsigned? */ - uint8_t num_format_all : 1; /* '0' for normalized, '1' for unnormalized */ - uint8_t signed_rf_mode_all : 1; - uint8_t reserved1 : 1; - instr_surf_fmt_t format : 6; - uint8_t reserved2 : 2; - uint8_t exp_adjust_all : 6; - uint8_t reserved3 : 1; - uint8_t pred_select : 1; - /* dword2: */ - uint8_t stride : 8; - uint32_t offset : 22; - uint8_t reserved4 : 1; - uint8_t pred_condition : 1; + /* dword0: */ + instr_fetch_opc_t opc : 5; + uint32_t src_reg : 6; + uint8_t src_reg_am : 1; + uint32_t dst_reg : 6; + uint8_t dst_reg_am : 1; + uint8_t must_be_one : 1; + uint32_t const_index : 5; + uint8_t const_index_sel : 2; + uint8_t reserved0 : 3; + uint8_t src_swiz : 2; + /* dword1: */ + uint16_t dst_swiz : 12; + uint8_t format_comp_all : 1; /* '1' for signed, '0' for unsigned? */ + uint8_t num_format_all : 1; /* '0' for normalized, '1' for unnormalized */ + uint8_t signed_rf_mode_all : 1; + uint8_t reserved1 : 1; + instr_surf_fmt_t format : 6; + uint8_t reserved2 : 2; + uint8_t exp_adjust_all : 6; + uint8_t reserved3 : 1; + uint8_t pred_select : 1; + /* dword2: */ + uint8_t stride : 8; + uint32_t offset : 22; + uint8_t reserved4 : 1; + uint8_t pred_condition : 1; } instr_fetch_vtx_t; typedef union PACKED { - instr_fetch_tex_t tex; - instr_fetch_vtx_t vtx; - struct PACKED { - /* dword0: */ - instr_fetch_opc_t opc : 5; - uint32_t dummy0 : 27; - /* dword1: */ - uint32_t dummy1 : 31; - uint8_t pred_select : 1; - /* dword2: */ - uint32_t dummy2 : 31; - uint8_t pred_condition : 1; - }; + instr_fetch_tex_t tex; + instr_fetch_vtx_t vtx; + struct PACKED { + /* dword0: */ + instr_fetch_opc_t opc : 5; + uint32_t dummy0 : 27; + /* dword1: */ + uint32_t dummy1 : 31; + uint8_t pred_select : 1; + /* dword2: */ + uint32_t dummy2 : 31; + uint8_t pred_condition : 1; + }; } instr_fetch_t; typedef union PACKED { - instr_alu_t alu; - instr_fetch_t fetch; + instr_alu_t alu; + instr_fetch_t fetch; } instr_t; #endif /* INSTR_H_ */