[PATCH v4 0/2] MR497: vkd3d-shader/dxil: Implement DX intrinsic Unary.
-- v4: vkd3d-shader/dxil: Apply metadata attachments to instructions. https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/497
From: Conor McCarthy <cmccarthy(a)codeweavers.com> --- libs/vkd3d-shader/dxil.c | 153 +++++++++++++++++++++++++++++-- tests/hlsl/ceil.shader_test | 6 +- tests/hlsl/ddxddy.shader_test | 8 +- tests/hlsl/distance.shader_test | 2 +- tests/hlsl/exp.shader_test | 4 +- tests/hlsl/floor.shader_test | 6 +- tests/hlsl/frac.shader_test | 2 +- tests/hlsl/ldexp.shader_test | 4 +- tests/hlsl/length.shader_test | 6 +- tests/hlsl/lit.shader_test | 8 +- tests/hlsl/log.shader_test | 6 +- tests/hlsl/normalize.shader_test | 4 +- tests/hlsl/pow.shader_test | 2 +- tests/hlsl/round.shader_test | 6 +- tests/hlsl/sqrt.shader_test | 4 +- tests/hlsl/trunc.shader_test | 8 +- 16 files changed, 185 insertions(+), 44 deletions(-) diff --git a/libs/vkd3d-shader/dxil.c b/libs/vkd3d-shader/dxil.c index 1709212fa..202cee49e 100644 --- a/libs/vkd3d-shader/dxil.c +++ b/libs/vkd3d-shader/dxil.c @@ -296,8 +296,28 @@ enum dx_intrinsic_opcode { DX_LOAD_INPUT = 4, DX_STORE_OUTPUT = 5, + DX_EXP = 21, + DX_FRC = 22, + DX_LOG = 23, + DX_SQRT = 24, + DX_RSQRT = 25, + DX_ROUND_NE = 26, + DX_ROUND_NI = 27, + DX_ROUND_PI = 28, + DX_ROUND_Z = 29, + DX_BFREV = 30, + DX_COUNT_BITS = 31, + DX_FIRST_BIT_LO = 32, + DX_FIRST_BIT_HI = 33, + DX_FIRST_BIT_SHI = 34, DX_CREATE_HANDLE = 57, DX_CBUFFER_LOAD_LEGACY = 59, + DX_DERIV_COARSEX = 83, + DX_DERIV_COARSEY = 84, + DX_DERIV_FINEX = 85, + DX_DERIV_FINEY = 86, + DX_LEGACY_F32TOF16 = 130, + DX_LEGACY_F16TOF32 = 131, }; enum dxil_cast_code @@ -1572,6 +1592,11 @@ static bool sm6_type_is_bool_i16_i32_i64(const struct sm6_type *type) return type->class == TYPE_CLASS_INTEGER && (type->u.width == 1 || type->u.width >= 16); } +static bool sm6_type_is_i16_i32_i64(const struct sm6_type *type) +{ + return type->class == TYPE_CLASS_INTEGER && type->u.width >= 16; +} + static bool sm6_type_is_bool(const struct sm6_type *type) { return type->class == TYPE_CLASS_INTEGER && type->u.width == 1; @@ -1587,6 +1612,16 @@ static inline bool sm6_type_is_i32(const struct sm6_type *type) return type->class == TYPE_CLASS_INTEGER && type->u.width == 32; } +static bool sm6_type_is_float(const struct sm6_type *type) +{ + return type->class == TYPE_CLASS_FLOAT && type->u.width == 32; +} + +static bool sm6_type_is_f16_f32(const struct sm6_type *type) +{ + return type->class == TYPE_CLASS_FLOAT && (type->u.width == 16 || type->u.width == 32); +} + static inline bool sm6_type_is_floating_point(const struct sm6_type *type) { return type->class == TYPE_CLASS_FLOAT; @@ -3170,6 +3205,67 @@ static void sm6_parser_emit_binop(struct sm6_parser *sm6, const struct dxil_reco } } +static enum vkd3d_shader_opcode map_dx_unary_op(enum dx_intrinsic_opcode op) +{ + switch (op) + { + case DX_EXP: + return VKD3DSIH_EXP; + case DX_FRC: + return VKD3DSIH_FRC; + case DX_LOG: + return VKD3DSIH_LOG; + case DX_SQRT: + return VKD3DSIH_SQRT; + case DX_RSQRT: + return VKD3DSIH_RSQ; + case DX_ROUND_NE: + return VKD3DSIH_ROUND_NE; + case DX_ROUND_NI: + return VKD3DSIH_ROUND_NI; + case DX_ROUND_PI: + return VKD3DSIH_ROUND_PI; + case DX_ROUND_Z: + return VKD3DSIH_ROUND_Z; + case DX_BFREV: + return VKD3DSIH_BFREV; + case DX_COUNT_BITS: + return VKD3DSIH_COUNTBITS; + case DX_FIRST_BIT_LO: + return VKD3DSIH_FIRSTBIT_LO; + case DX_FIRST_BIT_HI: + return VKD3DSIH_FIRSTBIT_HI; + case DX_FIRST_BIT_SHI: + return VKD3DSIH_FIRSTBIT_SHI; + case DX_DERIV_COARSEX: + return VKD3DSIH_DSX_COARSE; + case DX_DERIV_COARSEY: + return VKD3DSIH_DSY_COARSE; + case DX_DERIV_FINEX: + return VKD3DSIH_DSX_FINE; + case DX_DERIV_FINEY: + return VKD3DSIH_DSY_FINE; + case DX_LEGACY_F32TOF16: + return VKD3DSIH_F32TOF16; + case DX_LEGACY_F16TOF32: + return VKD3DSIH_F16TOF32; + default: + vkd3d_unreachable(); + } +} + +static void sm6_parser_emit_dx_unary(struct sm6_parser *sm6, struct sm6_block *code_block, + enum dx_intrinsic_opcode op, const struct sm6_value **operands, struct vkd3d_shader_instruction *ins) +{ + struct vkd3d_shader_src_param *src_param; + + vsir_instruction_init(ins, &sm6->p.location, map_dx_unary_op(op)); + src_param = instruction_src_params_alloc(ins, 1, sm6); + src_param_init_from_value(src_param, operands[0]); + + instruction_dst_param_init_ssa_scalar(ins, sm6); +} + static void sm6_parser_emit_dx_cbuffer_load(struct sm6_parser *sm6, struct sm6_block *code_block, enum dx_intrinsic_opcode op, const struct sm6_value **operands, struct vkd3d_shader_instruction *ins) { @@ -3343,7 +3439,7 @@ static void sm6_parser_emit_dx_store_output(struct sm6_parser *sm6, struct sm6_b struct sm6_dx_opcode_info { - const char ret_type; + const char *ret_type; const char *operand_info; void (*handler)(struct sm6_parser *, struct sm6_block *, enum dx_intrinsic_opcode, const struct sm6_value **, struct vkd3d_shader_instruction *); @@ -3354,16 +3450,40 @@ struct sm6_dx_opcode_info b -> constant int1 c -> constant int8/16/32 i -> int32 + m -> int16/32/64 + f -> float + e -> half/float + g -> half/float/double H -> handle v -> void o -> overloaded */ static const struct sm6_dx_opcode_info sm6_dx_op_table[] = { - [DX_CBUFFER_LOAD_LEGACY ] = {'o', "Hi", sm6_parser_emit_dx_cbuffer_load}, - [DX_CREATE_HANDLE ] = {'H', "ccib", sm6_parser_emit_dx_create_handle}, - [DX_LOAD_INPUT ] = {'o', "ii8i", sm6_parser_emit_dx_load_input}, - [DX_STORE_OUTPUT ] = {'v', "ii8o", sm6_parser_emit_dx_store_output}, + [DX_BFREV ] = {"m0", "m", sm6_parser_emit_dx_unary}, + [DX_CBUFFER_LOAD_LEGACY ] = {"o", "Hi", sm6_parser_emit_dx_cbuffer_load}, + [DX_COUNT_BITS ] = {"i", "m", sm6_parser_emit_dx_unary}, + [DX_CREATE_HANDLE ] = {"H", "ccib", sm6_parser_emit_dx_create_handle}, + [DX_DERIV_COARSEX ] = {"e0", "e", sm6_parser_emit_dx_unary}, + [DX_DERIV_COARSEY ] = {"e0", "e", sm6_parser_emit_dx_unary}, + [DX_DERIV_FINEX ] = {"e0", "e", sm6_parser_emit_dx_unary}, + [DX_DERIV_FINEY ] = {"e0", "e", sm6_parser_emit_dx_unary}, + [DX_EXP ] = {"g0", "g", sm6_parser_emit_dx_unary}, + [DX_FIRST_BIT_HI ] = {"i", "m", sm6_parser_emit_dx_unary}, + [DX_FIRST_BIT_LO ] = {"i", "m", sm6_parser_emit_dx_unary}, + [DX_FIRST_BIT_SHI ] = {"i", "m", sm6_parser_emit_dx_unary}, + [DX_FRC ] = {"g0", "g", sm6_parser_emit_dx_unary}, + [DX_LEGACY_F16TOF32 ] = {"f", "i", sm6_parser_emit_dx_unary}, + [DX_LEGACY_F32TOF16 ] = {"i", "f", sm6_parser_emit_dx_unary}, + [DX_LOAD_INPUT ] = {"o", "ii8i", sm6_parser_emit_dx_load_input}, + [DX_LOG ] = {"g0", "g", sm6_parser_emit_dx_unary}, + [DX_ROUND_NE ] = {"g0", "g", sm6_parser_emit_dx_unary}, + [DX_ROUND_NI ] = {"g0", "g", sm6_parser_emit_dx_unary}, + [DX_ROUND_PI ] = {"g0", "g", sm6_parser_emit_dx_unary}, + [DX_ROUND_Z ] = {"g0", "g", sm6_parser_emit_dx_unary}, + [DX_RSQRT ] = {"g0", "g", sm6_parser_emit_dx_unary}, + [DX_SQRT ] = {"g0", "g", sm6_parser_emit_dx_unary}, + [DX_STORE_OUTPUT ] = {"v", "ii8o", sm6_parser_emit_dx_store_output}, }; static bool sm6_parser_validate_operand_type(struct sm6_parser *sm6, const struct sm6_value *value, char info_type, @@ -3388,6 +3508,14 @@ static bool sm6_parser_validate_operand_type(struct sm6_parser *sm6, const struc && type->u.width <= 32; case 'i': return sm6_type_is_i32(type); + case 'm': + return sm6_type_is_i16_i32_i64(type); + case 'f': + return sm6_type_is_float(type); + case 'e': + return sm6_type_is_f16_f32(type); + case 'g': + return sm6_type_is_floating_point(type); case 'H': return (is_return || sm6_value_is_handle(value)) && type == sm6->handle_type; case 'v': @@ -3401,6 +3529,17 @@ static bool sm6_parser_validate_operand_type(struct sm6_parser *sm6, const struc } } +static bool operand_types_match(struct sm6_value *dst, const struct sm6_value **operands, unsigned int operand_count, + char index_char) +{ + unsigned int i = index_char - '0'; + + assert(i < 10); + if (i >= operand_count) + return false; + return dst->type == operands[i]->type; +} + static bool sm6_parser_validate_dx_op(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, const char *name, const struct sm6_value **operands, unsigned int operand_count, struct sm6_value *dst) { @@ -3409,7 +3548,9 @@ static bool sm6_parser_validate_dx_op(struct sm6_parser *sm6, enum dx_intrinsic_ info = &sm6_dx_op_table[op]; - if (!sm6_parser_validate_operand_type(sm6, dst, info->ret_type, true)) + assert(info->ret_type[0]); + if (!sm6_parser_validate_operand_type(sm6, dst, info->ret_type[0], true) + || (info->ret_type[1] && !operand_types_match(dst, operands, operand_count, info->ret_type[1]))) { WARN("Failed to validate return type for dx intrinsic id %u, '%s'.\n", op, name); /* Return type validation failure is not so critical. We only need to set diff --git a/tests/hlsl/ceil.shader_test b/tests/hlsl/ceil.shader_test index 0ff9938b9..ef26cc8ef 100644 --- a/tests/hlsl/ceil.shader_test +++ b/tests/hlsl/ceil.shader_test @@ -18,7 +18,7 @@ float4 main() : sv_target [test] uniform 0 float4 -0.5 6.5 7.5 3.4 -todo(sm>=6) draw quad +draw quad probe all rgba (0.0, 7.0, 8.0, 4.0) 4 [pixel shader] @@ -34,7 +34,7 @@ float4 main() : sv_target [test] uniform 0 float4 -0.5 6.5 7.5 3.4 -todo(sm>=6) draw quad +draw quad probe all rgba (7.0, 8.0, 0.0, 4.0) 4 [require] @@ -53,5 +53,5 @@ float4 main() : sv_target [test] uniform 0 int4 -1 6 7 3 -todo(sm>=6) draw quad +draw quad probe all rgba (6.0, 7.0, -1.0, 3.0) 4 diff --git a/tests/hlsl/ddxddy.shader_test b/tests/hlsl/ddxddy.shader_test index ba7215ca0..4986c233f 100644 --- a/tests/hlsl/ddxddy.shader_test +++ b/tests/hlsl/ddxddy.shader_test @@ -8,7 +8,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo(sm>=6) draw quad +draw quad probe all rgba (1.0, 1.0, 0.0, 0.0) @@ -29,7 +29,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo(sm>=6) draw quad +draw quad probe (10, 10) rgba (-16.0, -5.0, 3.0, 0.0) probe (11, 10) rgba (-21.0, -5.0, 3.0, 0.0) probe (10, 11) rgba (-13.0, -5.0, 3.0, 0.0) @@ -55,7 +55,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo(sm>=6) draw quad +draw quad probe (10, 10) rgba (-16.0, -5.0, 3.0, 0.0) probe (11, 10) rgba (-21.0, -5.0, 3.0, 0.0) probe (10, 11) rgba (-13.0, -5.0, 3.0, 0.0) @@ -74,7 +74,7 @@ float4 main(float4 pos : sv_position) : sv_target } [test] -todo(sm>=6) draw quad +draw quad probe (10, 10) rgba (-0.524999976, -0.164999843, 0.104999900, 0.0) 16 probe (11, 10) rgba (-0.689999819, -0.164999843, 0.114999890, 0.0) 32 probe (10, 11) rgba (-0.420000076, -0.154999852, 0.104999900, 0.0) 32 diff --git a/tests/hlsl/distance.shader_test b/tests/hlsl/distance.shader_test index 6527c218a..3f5446451 100644 --- a/tests/hlsl/distance.shader_test +++ b/tests/hlsl/distance.shader_test @@ -10,7 +10,7 @@ float4 main() : sv_target [test] uniform 0 float4 -2.0 3.0 4.0 0.1 uniform 4 float4 2.0 -1.0 4.0 5.0 -todo(sm>=6) draw quad +draw quad probe all rgba (7.483983, 7.483983, 7.483983, 7.483983) 1 [pixel shader] diff --git a/tests/hlsl/exp.shader_test b/tests/hlsl/exp.shader_test index 38f8750fd..1d1889977 100644 --- a/tests/hlsl/exp.shader_test +++ b/tests/hlsl/exp.shader_test @@ -8,7 +8,7 @@ float4 main() : sv_target [test] uniform 0 float4 -1.0 0.0 1.0 2.0 -todo(sm>=6) draw quad +draw quad probe all rgba (0.5, 1.0, 2.0, 4.0) 2 [pixel shader] @@ -21,5 +21,5 @@ float4 main() : sv_target [test] uniform 0 float4 -1.0 0.0 1.0 2.0 -todo(sm>=6) draw quad +draw quad probe all rgba (0.36787948, 1.0, 2.7182815, 7.38905573) 2 diff --git a/tests/hlsl/floor.shader_test b/tests/hlsl/floor.shader_test index e6562c4aa..19f9437a9 100644 --- a/tests/hlsl/floor.shader_test +++ b/tests/hlsl/floor.shader_test @@ -18,7 +18,7 @@ float4 main() : sv_target [test] uniform 0 float4 -0.5 6.5 7.5 3.4 -todo(sm>=6) draw quad +draw quad probe all rgba (-1.0, 6.0, 7.0, 3.0) 4 [pixel shader] @@ -34,7 +34,7 @@ float4 main() : sv_target [test] uniform 0 float4 -0.5 6.5 7.5 3.4 -todo(sm>=6) draw quad +draw quad probe all rgba (6.0, 7.0, -1.0, 3.0) 4 [require] @@ -53,5 +53,5 @@ float4 main() : sv_target [test] uniform 0 int4 -1 6 7 3 -todo(sm>=6) draw quad +draw quad probe all rgba (6.0, 7.0, -1.0, 3.0) 4 diff --git a/tests/hlsl/frac.shader_test b/tests/hlsl/frac.shader_test index 2ac9b530d..f54f3fe88 100644 --- a/tests/hlsl/frac.shader_test +++ b/tests/hlsl/frac.shader_test @@ -8,5 +8,5 @@ float4 main() : sv_target [test] uniform 0 float4 -1.1 1.6 1.3 0.5 -todo(sm>=6) draw quad +draw quad probe all rgba (0.9, 0.6, 0.3, 0.5) 2 diff --git a/tests/hlsl/ldexp.shader_test b/tests/hlsl/ldexp.shader_test index 2db624067..f8ad40d8e 100644 --- a/tests/hlsl/ldexp.shader_test +++ b/tests/hlsl/ldexp.shader_test @@ -10,7 +10,7 @@ float4 main() : SV_TARGET [test] uniform 0 float4 2.0 3.0 4.0 5.0 uniform 4 float4 0.0 -10.0 10.0 100.0 -todo(sm>=6) draw quad +draw quad probe all rgba (2.0, 0.00292968750, 4096.0, 6.33825300e+030) 2 [require] @@ -28,7 +28,7 @@ float4 main() : SV_TARGET [test] uniform 0 int4 2 3 4 5 uniform 4 int4 0 -10 10 100 -todo(sm>=6) draw quad +draw quad probe all rgba (2.0, 0.00292968750, 4096.0, 6.33825300e+030) 2 diff --git a/tests/hlsl/length.shader_test b/tests/hlsl/length.shader_test index 8653942e0..44b4e5254 100644 --- a/tests/hlsl/length.shader_test +++ b/tests/hlsl/length.shader_test @@ -8,7 +8,7 @@ float4 main() : SV_TARGET [test] uniform 0 float4 2.0 3.0 4.0 5.0 -todo(sm>=6) draw quad +draw quad probe all rgba (7.34846926, 7.34846926, 7.34846926, 7.34846926) 1 [pixel shader] @@ -21,7 +21,7 @@ float4 main() : SV_TARGET [test] uniform 0 float4 2.0 3.0 4.0 0.0 -todo(sm>=6) draw quad +draw quad probe all rgba (5.38516474, 5.38516474, 5.38516474, 5.38516474) 1 [pixel shader] @@ -34,7 +34,7 @@ float4 main() : SV_TARGET [test] uniform 0 float4 2.0 3.0 0.0 0.0 -todo(sm>=6) draw quad +draw quad probe all rgba (3.60555124, 3.60555124, 3.60555124, 3.60555124) 1 [pixel shader] diff --git a/tests/hlsl/lit.shader_test b/tests/hlsl/lit.shader_test index 5014c1ed0..fcf37d78b 100644 --- a/tests/hlsl/lit.shader_test +++ b/tests/hlsl/lit.shader_test @@ -8,17 +8,17 @@ float4 main() : sv_target [test] uniform 0 float4 -0.1 10.0 0.0 0.0 -todo(sm>=6) draw quad +draw quad probe all rgba (1.0, 0.0, 0.0, 1.0) [test] uniform 0 float4 1.2 -0.1 0.0 0.0 -todo(sm>=6) draw quad +draw quad probe all rgba (1.0, 1.2, 0.0, 1.0) [test] uniform 0 float4 1.2 2.0 3.0 0.0 -todo(sm>=6) draw quad +draw quad probe all rgba (1.0, 1.2, 8.0, 1.0) [pixel shader] @@ -31,7 +31,7 @@ float4 main() : sv_target [test] uniform 0 float4 1.2 2.0 3.0 0.0 -todo(sm>=6) draw quad +draw quad probe all rgba (2.0, 2.4, 16.0, 2.0) [pixel shader fail] diff --git a/tests/hlsl/log.shader_test b/tests/hlsl/log.shader_test index 2f7d5c7c0..b0f405d11 100644 --- a/tests/hlsl/log.shader_test +++ b/tests/hlsl/log.shader_test @@ -8,7 +8,7 @@ float4 main() : sv_target [test] uniform 0 float4 2.0 4.0 5.0 0.4 -todo(sm>=6) draw quad +draw quad probe all rgba (1.0, 2.0, 2.32192802, -1.32192802) 1 [pixel shader] @@ -21,7 +21,7 @@ float4 main() : sv_target [test] uniform 0 float4 10.0 100.0 1.0 0.1 -todo(sm>=6) draw quad +draw quad probe all rgba (1.0, 2.0, 0.0, -1.0) 1 [pixel shader] @@ -34,5 +34,5 @@ float4 main() : sv_target [test] uniform 0 float4 3.0 10.0 1.0 0.1 -todo(sm>=6) draw quad +draw quad probe all rgba (1.0986123, 2.302585, 0.0, -2.302585) 2 diff --git a/tests/hlsl/normalize.shader_test b/tests/hlsl/normalize.shader_test index 0fe0dea3c..8b181dfa5 100644 --- a/tests/hlsl/normalize.shader_test +++ b/tests/hlsl/normalize.shader_test @@ -47,7 +47,7 @@ float4 main() : SV_TARGET [test] uniform 0 float4 2.0 0.0 0.0 0.0 -todo(sm>=6) draw quad +draw quad probe all rgba (1.0, 1.0, 1.0, 1.0) [pixel shader] @@ -60,7 +60,7 @@ float4 main() : SV_TARGET [test] uniform 0 float4 2.0 0.0 0.0 0.0 -todo(sm>=6) draw quad +draw quad probe all rgba (1.0, 1.0, 1.0, 1.0) [pixel shader fail] diff --git a/tests/hlsl/pow.shader_test b/tests/hlsl/pow.shader_test index d8f5eec9f..1c9394f06 100644 --- a/tests/hlsl/pow.shader_test +++ b/tests/hlsl/pow.shader_test @@ -8,7 +8,7 @@ float4 main() : sv_target [test] uniform 0 float4 0.4 0.8 2.5 2.0 -todo(sm>=6) draw quad +draw quad probe all rgba (0.512, 0.101192884, 0.64, 0.25) 4 diff --git a/tests/hlsl/round.shader_test b/tests/hlsl/round.shader_test index 2a32b3ec2..b9234b010 100644 --- a/tests/hlsl/round.shader_test +++ b/tests/hlsl/round.shader_test @@ -8,7 +8,7 @@ float4 main() : sv_target [test] uniform 0 float4 -0.4 -6.6 7.6 3.4 -todo(sm>=6) draw quad +draw quad probe all rgba (0.0, -7.0, 8.0, 3.0) 4 @@ -26,7 +26,7 @@ float4 main() : sv_target [test] uniform 0 float4 -0.4 -6.6 7.6 3.4 -todo(sm>=6) draw quad +draw quad probe all rgba (-7.0, 8.0, 0.0, 3.0) 4 @@ -42,5 +42,5 @@ float4 main() : sv_target [test] uniform 0 float4 -1 0 2 10 -todo(sm>=6) draw quad +draw quad probe all rgba (-1.0, 0.0, 2.0, 10.0) 4 diff --git a/tests/hlsl/sqrt.shader_test b/tests/hlsl/sqrt.shader_test index 73ecba899..78d89d38f 100644 --- a/tests/hlsl/sqrt.shader_test +++ b/tests/hlsl/sqrt.shader_test @@ -8,7 +8,7 @@ float4 main() : sv_target [test] uniform 0 float4 1.0 9.0 32.3 46.5 -todo(sm>=6) draw quad +draw quad probe all rgba (1.0, 3.0, 5.683309, 6.819091) 1 [pixel shader] @@ -21,5 +21,5 @@ float4 main() : sv_target [test] uniform 0 float4 1.0 9.0 4.0 16.0 -todo(sm>=6) draw quad +draw quad probe all rgba (1.0, 0.33333333, 0.5, 0.25) 1 diff --git a/tests/hlsl/trunc.shader_test b/tests/hlsl/trunc.shader_test index a359cd03a..76ffdbbad 100644 --- a/tests/hlsl/trunc.shader_test +++ b/tests/hlsl/trunc.shader_test @@ -8,10 +8,10 @@ float4 main() : sv_target [test] uniform 0 float4 -0.5 6.5 7.5 3.4 -todo(sm>=6) draw quad +draw quad probe all rgba (0.0, 6.0, 7.0, 3.0) uniform 0 float4 -1.5 6.5 7.5 3.4 -todo(sm>=6) draw quad +draw quad probe all rgba (-1.0, 6.0, 7.0, 3.0) [pixel shader] @@ -27,7 +27,7 @@ float4 main() : sv_target [test] uniform 0 float4 -0.5 6.5 7.5 3.4 -todo(sm>=6) draw quad +draw quad probe all rgba (6.0, 7.0, 0.0, 3.0) [require] @@ -46,5 +46,5 @@ float4 main() : sv_target [test] uniform 0 int4 -1 6 7 3 -todo(sm>=6) draw quad +draw quad probe all rgba (6.0, 7.0, -1.0, 3.0) -- GitLab https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/497
From: Conor McCarthy <cmccarthy(a)codeweavers.com> These are apperently only used for 'dx.op' intrinsics, because the instructions based on native LLVM ones have their own way to apply attributes. --- libs/vkd3d-shader/dxil.c | 215 +++++++++++++++++++++++ libs/vkd3d-shader/vkd3d_shader_private.h | 1 + 2 files changed, 216 insertions(+) diff --git a/libs/vkd3d-shader/dxil.c b/libs/vkd3d-shader/dxil.c index 202cee49e..e1df27b67 100644 --- a/libs/vkd3d-shader/dxil.c +++ b/libs/vkd3d-shader/dxil.c @@ -459,6 +459,7 @@ struct dxil_record { unsigned int code; unsigned int operand_count; + const struct dxil_record *attachment; uint64_t operands[]; }; @@ -816,6 +817,7 @@ static enum vkd3d_result sm6_parser_read_unabbrev_record(struct sm6_parser *sm6) record->code = code; record->operand_count = count; + record->attachment = NULL; for (i = 0; i < count; ++i) record->operands[i] = sm6_parser_read_vbr(sm6, 6); @@ -1032,6 +1034,7 @@ static enum vkd3d_result sm6_parser_read_abbrev_record(struct sm6_parser *sm6, u if (!abbrev->operands[i + 1].read_operand(sm6, abbrev->operands[i + 1].context, &record->operands[i])) goto fail; record->operand_count = count; + record->attachment = NULL; /* An array can occur only as the last operand. */ if (abbrev->is_array) @@ -2623,6 +2626,13 @@ static enum vkd3d_result sm6_parser_constants_init(struct sm6_parser *sm6, const break; } + if (record->attachment) + { + WARN("Ignoring metadata attachment.\n"); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "Ignoring a metadata attachment for a constant."); + } + ++sm6->value_count; } @@ -4293,6 +4303,206 @@ static bool sm6_metadata_get_uint64_value(const struct sm6_parser *sm6, return true; } +static void sm6_parser_metadata_attachment_block_init(struct sm6_parser *sm6, const struct dxil_block *target_block, + const struct dxil_block *block) +{ + struct dxil_record *target_record; + const struct dxil_record *record; + unsigned int i; + uint64_t index; + + for (i = 0; i < block->record_count; ++i) + { + record = block->records[i]; + if (record->code != METADATA_ATTACHMENT) + { + WARN("Ignoring record with code %u.\n", record->code); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "Ignoring a metadata attachment record with code %u.", record->code); + continue; + } + if (!(record->operand_count & 1)) + { + WARN("Ignoring function attachment.\n"); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "Ignoring a metadata function attachment."); + continue; + } + + index = record->operands[0]; + if (!target_block->record_count || index >= target_block->record_count - 1) + { + WARN("Invalid record index %"PRIu64".\n", index); + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + "Invalid record index %"PRIu64" for a metadata attachment.", index); + continue; + } + /* 'index' is an instruction index, but records[0] is DECLAREBLOCKS, not an instruction. */ + target_record = target_block->records[index + 1]; + if (target_record->attachment) + { + WARN("Overwriting record attachment.\n"); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "The target record for a metadata attachment already has an attachment."); + } + target_record->attachment = record; + } +} + +static void sm6_parser_metadata_attachments_init(struct sm6_parser *sm6, const struct dxil_block *block) +{ + unsigned int i; + + for (i = 0; i < block->child_block_count; ++i) + { + if (block->child_blocks[i]->id == METADATA_ATTACHMENT_BLOCK) + sm6_parser_metadata_attachment_block_init(sm6, block, block->child_blocks[i]); + } +} + +static const struct sm6_metadata_value *sm6_parser_find_metadata_kind(const struct sm6_parser *sm6, uint64_t id) +{ + unsigned int i, j; + + for (i = 0; i < ARRAY_SIZE(sm6->metadata_tables); ++i) + { + for (j = 0; j < sm6->metadata_tables[i].count; ++j) + { + if (sm6->metadata_tables[i].values[j].type == VKD3D_METADATA_KIND + && sm6->metadata_tables[i].values[j].u.kind.id == id) + return &sm6->metadata_tables[i].values[j]; + } + } + + return NULL; +} + +static const struct sm6_metadata_value *sm6_parser_metadata_get_value(const struct sm6_parser *sm6, uint64_t index) +{ + unsigned int i; + + for (i = 0; i < ARRAY_SIZE(sm6->metadata_tables); ++i) + { + if (sm6->metadata_tables[i].count > index) + break; + index -= sm6->metadata_tables[i].count; + } + + return (index < sm6->metadata_tables[i].count) ? &sm6->metadata_tables[i].values[index] : NULL; +} + +static bool metadata_node_get_unary_uint(const struct sm6_metadata_node *node, unsigned int *operand, + struct sm6_parser *sm6) +{ + if (node->operand_count != 1) + { + FIXME("Ignoring node with %u operands.\n", node->operand_count); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "Ignoring metadata attachment node with %u operands; expected unary.", node->operand_count); + return false; + } + if (!sm6_metadata_value_is_value(node->operands[0]) + || !sm6_metadata_get_uint_value(sm6, node->operands[0], operand)) + { + WARN("Failed to get operand value.\n"); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "Failed to get a metadata attachment operand value; ignoring the attachment."); + return false; + } + + return true; +} + +static void metadata_attachment_record_apply(const struct dxil_record *record, + struct vkd3d_shader_instruction *ins, struct sm6_value *dst, struct sm6_parser *sm6) +{ + static const char *ignored_names[] = + { + "alias.scope", + "dx.controlflow.hints", + "llvm.loop", + "noalias", + "tbaa", + "range", + }; + const struct sm6_metadata_node *node; + const struct sm6_metadata_value *m; + unsigned int i, j, operand; + const char *name; + + if (record->attachment) + { + WARN("Ignoring nested metadata attachment.\n"); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "Ignoring a nested metadata attachment."); + } + + if (record->code != FUNC_CODE_INST_CALL) + WARN("Metadata attachment target is not a function call.\n"); + + assert(record->operand_count & 1); + for (i = 1; i < record->operand_count; i += 2) + { + if (!(m = sm6_parser_find_metadata_kind(sm6, record->operands[i]))) + { + WARN("Failed to find metadata kind %"PRIx64".\n", record->operands[i]); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "Failed to find metadata kind %"PRIx64" for an attachment.", record->operands[i]); + continue; + } + name = m->u.kind.name; + + m = sm6_parser_metadata_get_value(sm6, record->operands[i + 1]); + if (!m || !sm6_metadata_value_is_node(m)) + { + WARN("Failed to retrieve metadata attachment node.\n"); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "Failed to retrieve a metadata attachment node."); + continue; + } + node = m->u.node; + + if (!strcmp(name, "dx.precise")) + { + if (!sm6_value_is_register(dst)) + { + WARN("Precise value is not a register.\n"); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "A value marked as precise is not a register."); + } + else if (metadata_node_get_unary_uint(node, &operand, sm6) && operand) + { + ins->flags |= sm6_type_is_scalar(dst->type) ? VKD3DSI_PRECISE_X : VKD3DSI_PRECISE_XYZW; + } + } + else if (!strcmp(name, "dx.nonuniform")) + { + if (!sm6_value_is_register(dst)) + { + WARN("Non-uniform value is not a register.\n"); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "A value marked as non-uniform is not a register."); + } + else if (metadata_node_get_unary_uint(node, &operand, sm6)) + { + dst->u.reg.non_uniform = !!operand; + } + } + else + { + for (j = 0; j < ARRAY_SIZE(ignored_names); ++j) + if (!strcmp(name, ignored_names[j])) + break; + if (j == ARRAY_SIZE(ignored_names)) + { + WARN("Ignoring metadata attachment '%s'.\n", name); + vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT, + "Ignoring a metadata attachment named '%s'.", name); + } + } + } +} + static enum vkd3d_result sm6_parser_function_init(struct sm6_parser *sm6, const struct dxil_block *block, struct sm6_function *function) { @@ -4418,6 +4628,9 @@ static enum vkd3d_result sm6_parser_function_init(struct sm6_parser *sm6, const return VKD3D_ERROR; assert(ins->handler_idx != VKD3DSIH_INVALID); + if (record->attachment) + metadata_attachment_record_apply(record->attachment, ins, dst, sm6); + if (is_terminator) { ++block_idx; @@ -4471,6 +4684,8 @@ static enum vkd3d_result sm6_parser_module_init(struct sm6_parser *sm6, const st sm6->p.location.line = block->id; sm6->p.location.column = 0; + sm6_parser_metadata_attachments_init(sm6, block); + switch (block->id) { case CONSTANTS_BLOCK: diff --git a/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d-shader/vkd3d_shader_private.h index d3989672b..08669520c 100644 --- a/libs/vkd3d-shader/vkd3d_shader_private.h +++ b/libs/vkd3d-shader/vkd3d_shader_private.h @@ -196,6 +196,7 @@ enum vkd3d_shader_error VKD3D_SHADER_WARNING_DXIL_ENTRY_POINT_MISMATCH = 8306, VKD3D_SHADER_WARNING_DXIL_INVALID_MASK = 8307, VKD3D_SHADER_WARNING_DXIL_INVALID_OPERATION = 8308, + VKD3D_SHADER_WARNING_DXIL_IGNORING_ATTACHMENT = 8309, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED = 9000, VKD3D_SHADER_ERROR_VSIR_INVALID_HANDLER = 9001, -- GitLab https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/497
On Tue Nov 28 11:48:14 2023 +0000, Giovanni Mascellani wrote:
That means that even for dxcompiler F16 values are typed as integers, rather than `half`? For these legacy instructions, yes they store two f16 in a u32 as TPF does.
-- https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/497#note_54106
participants (2)
-
Conor McCarthy -
Conor McCarthy (＠cmccarthy)