diff options
Diffstat (limited to '')
21 files changed, 128 insertions, 127 deletions
diff --git a/src/shader_recompiler/backend/glasm/emit_glasm.cpp b/src/shader_recompiler/backend/glasm/emit_glasm.cpp index 0cb1e193e..fd4a61a4d 100644 --- a/src/shader_recompiler/backend/glasm/emit_glasm.cpp +++ b/src/shader_recompiler/backend/glasm/emit_glasm.cpp @@ -279,6 +279,8 @@ void SetupOptions(const IR::Program& program, const Profile& profile, header += "OPTION NV_internal;" "OPTION NV_shader_storage_buffer;" "OPTION NV_gpu_program_fp64;"; + // TODO: Enable only when MS is used + header += "OPTION NV_texture_multisample;"; if (info.uses_int64_bit_atomics) { header += "OPTION NV_shader_atomic_int64;"; } diff --git a/src/shader_recompiler/backend/glasm/emit_glasm_bitwise_conversion.cpp b/src/shader_recompiler/backend/glasm/emit_glasm_bitwise_conversion.cpp index 5bfdecc09..2fc2a0ac6 100644 --- a/src/shader_recompiler/backend/glasm/emit_glasm_bitwise_conversion.cpp +++ b/src/shader_recompiler/backend/glasm/emit_glasm_bitwise_conversion.cpp @@ -43,10 +43,6 @@ void EmitBitCastU64F64(EmitContext&, IR::Inst& inst, const IR::Value& value) { Alias(inst, value); } -void EmitBitCastS32F32(EmitContext&, IR::Inst& inst, const IR::Value& value) { - Alias(inst, value); -} - void EmitBitCastF16U16(EmitContext&, IR::Inst& inst, const IR::Value& value) { Alias(inst, value); } diff --git a/src/shader_recompiler/backend/glasm/emit_glasm_image.cpp b/src/shader_recompiler/backend/glasm/emit_glasm_image.cpp index e67e80fac..b7bc11416 100644 --- a/src/shader_recompiler/backend/glasm/emit_glasm_image.cpp +++ b/src/shader_recompiler/backend/glasm/emit_glasm_image.cpp @@ -59,7 +59,7 @@ std::string Image(EmitContext& ctx, IR::TextureInstInfo info, } } -std::string_view TextureType(IR::TextureInstInfo info) { +std::string_view TextureType(IR::TextureInstInfo info, bool is_ms = false) { if (info.is_depth) { switch (info.type) { case TextureType::Color1D: @@ -88,9 +88,9 @@ std::string_view TextureType(IR::TextureInstInfo info) { return "ARRAY1D"; case TextureType::Color2D: case TextureType::Color2DRect: - return "2D"; + return is_ms ? "2DMS" : "2D"; case TextureType::ColorArray2D: - return "ARRAY2D"; + return is_ms ? "ARRAY2DMS" : "ARRAY2D"; case TextureType::Color3D: return "3D"; case TextureType::ColorCube: @@ -510,15 +510,16 @@ void EmitImageFetch(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, const IR::Value& coord, const IR::Value& offset, ScalarS32 lod, ScalarS32 ms) { const auto info{inst.Flags<IR::TextureInstInfo>()}; const auto sparse_inst{PrepareSparse(inst)}; + const bool is_multisample{ms.type != Type::Void}; const std::string_view sparse_mod{sparse_inst ? ".SPARSE" : ""}; - const std::string_view type{TextureType(info)}; + const std::string_view type{TextureType(info, is_multisample)}; const std::string texture{Texture(ctx, info, index)}; const std::string offset_vec{Offset(ctx, offset)}; const auto [coord_vec, coord_alloc]{Coord(ctx, coord)}; const Register ret{ctx.reg_alloc.Define(inst)}; if (info.type == TextureType::Buffer) { ctx.Add("TXF.F{} {},{},{},{}{};", sparse_mod, ret, coord_vec, texture, type, offset_vec); - } else if (ms.type != Type::Void) { + } else if (is_multisample) { ctx.Add("MOV.S {}.w,{};" "TXFMS.F{} {},{},{},{}{};", coord_vec, ms, sparse_mod, ret, coord_vec, texture, type, offset_vec); @@ -531,7 +532,7 @@ void EmitImageFetch(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, } void EmitImageQueryDimensions(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, - ScalarS32 lod) { + ScalarS32 lod, [[maybe_unused]] const IR::Value& skip_mips) { const auto info{inst.Flags<IR::TextureInstInfo>()}; const std::string texture{Texture(ctx, info, index)}; const std::string_view type{TextureType(info)}; diff --git a/src/shader_recompiler/backend/glasm/emit_glasm_instructions.h b/src/shader_recompiler/backend/glasm/emit_glasm_instructions.h index eaaf9ba39..1a1ea61d5 100644 --- a/src/shader_recompiler/backend/glasm/emit_glasm_instructions.h +++ b/src/shader_recompiler/backend/glasm/emit_glasm_instructions.h @@ -197,7 +197,6 @@ void EmitSelectF64(EmitContext& ctx, ScalarS32 cond, Register true_value, Regist void EmitBitCastU16F16(EmitContext& ctx, IR::Inst& inst, const IR::Value& value); void EmitBitCastU32F32(EmitContext& ctx, IR::Inst& inst, const IR::Value& value); void EmitBitCastU64F64(EmitContext& ctx, IR::Inst& inst, const IR::Value& value); -void EmitBitCastS32F32(EmitContext& ctx, IR::Inst& inst, const IR::Value& value); void EmitBitCastF16U16(EmitContext& ctx, IR::Inst& inst, const IR::Value& value); void EmitBitCastF32U32(EmitContext& ctx, IR::Inst& inst, const IR::Value& value); void EmitBitCastF64U64(EmitContext& ctx, IR::Inst& inst, const IR::Value& value); @@ -582,7 +581,7 @@ void EmitImageGatherDref(EmitContext& ctx, IR::Inst& inst, const IR::Value& inde void EmitImageFetch(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, const IR::Value& coord, const IR::Value& offset, ScalarS32 lod, ScalarS32 ms); void EmitImageQueryDimensions(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, - ScalarS32 lod); + ScalarS32 lod, const IR::Value& skip_mips); void EmitImageQueryLod(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, Register coord); void EmitImageGradient(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, const IR::Value& coord, const IR::Value& derivatives, diff --git a/src/shader_recompiler/backend/glsl/emit_glsl_bitwise_conversion.cpp b/src/shader_recompiler/backend/glsl/emit_glsl_bitwise_conversion.cpp index 8e5e6cf1f..1be4a0f59 100644 --- a/src/shader_recompiler/backend/glsl/emit_glsl_bitwise_conversion.cpp +++ b/src/shader_recompiler/backend/glsl/emit_glsl_bitwise_conversion.cpp @@ -48,10 +48,6 @@ void EmitBitCastU64F64(EmitContext& ctx, IR::Inst& inst, std::string_view value) ctx.AddU64("{}=doubleBitsToUint64({});", inst, value); } -void EmitBitCastS32F32(EmitContext& ctx, IR::Inst& inst, std::string_view value) { - ctx.AddF32("{}=ftoi({});", inst, value); -} - void EmitBitCastF16U16([[maybe_unused]] EmitContext& ctx, [[maybe_unused]] IR::Inst& inst) { NotImplemented(); } diff --git a/src/shader_recompiler/backend/glsl/emit_glsl_image.cpp b/src/shader_recompiler/backend/glsl/emit_glsl_image.cpp index cecdbb9d6..4be2c25ec 100644 --- a/src/shader_recompiler/backend/glsl/emit_glsl_image.cpp +++ b/src/shader_recompiler/backend/glsl/emit_glsl_image.cpp @@ -414,7 +414,7 @@ void EmitImageGatherDref(EmitContext& ctx, IR::Inst& inst, const IR::Value& inde void EmitImageFetch(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, std::string_view coords, std::string_view offset, std::string_view lod, - [[maybe_unused]] std::string_view ms) { + std::string_view ms) { const auto info{inst.Flags<IR::TextureInstInfo>()}; if (info.has_bias) { throw NotImplementedException("EmitImageFetch Bias texture samples"); @@ -431,19 +431,24 @@ void EmitImageFetch(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, ctx.AddU1("{}=true;", *sparse_inst); } if (!sparse_inst || !supports_sparse) { - if (!offset.empty()) { - ctx.Add("{}=texelFetchOffset({},{},int({}),{});", texel, texture, - CoordsCastToInt(coords, info), lod, CoordsCastToInt(offset, info)); + const auto int_coords{CoordsCastToInt(coords, info)}; + if (!ms.empty()) { + ctx.Add("{}=texelFetch({},{},int({}));", texel, texture, int_coords, ms); + } else if (!offset.empty()) { + ctx.Add("{}=texelFetchOffset({},{},int({}),{});", texel, texture, int_coords, lod, + CoordsCastToInt(offset, info)); } else { if (info.type == TextureType::Buffer) { ctx.Add("{}=texelFetch({},int({}));", texel, texture, coords); } else { - ctx.Add("{}=texelFetch({},{},int({}));", texel, texture, - CoordsCastToInt(coords, info), lod); + ctx.Add("{}=texelFetch({},{},int({}));", texel, texture, int_coords, lod); } } return; } + if (!ms.empty()) { + throw NotImplementedException("EmitImageFetch Sparse MSAA samples"); + } if (!offset.empty()) { ctx.AddU1("{}=sparseTexelsResidentARB(sparseTexelFetchOffsetARB({},{},int({}),{},{}));", *sparse_inst, texture, CastToIntVec(coords, info), lod, @@ -455,27 +460,27 @@ void EmitImageFetch(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, } void EmitImageQueryDimensions(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, - std::string_view lod) { + std::string_view lod, const IR::Value& skip_mips_val) { const auto info{inst.Flags<IR::TextureInstInfo>()}; const auto texture{Texture(ctx, info, index)}; + const bool skip_mips{skip_mips_val.U1()}; + const auto mips{ + [&] { return skip_mips ? "0u" : fmt::format("uint(textureQueryLevels({}))", texture); }}; switch (info.type) { case TextureType::Color1D: - return ctx.AddU32x4( - "{}=uvec4(uint(textureSize({},int({}))),0u,0u,uint(textureQueryLevels({})));", inst, - texture, lod, texture); + return ctx.AddU32x4("{}=uvec4(uint(textureSize({},int({}))),0u,0u,{});", inst, texture, lod, + mips()); case TextureType::ColorArray1D: case TextureType::Color2D: case TextureType::ColorCube: case TextureType::Color2DRect: - return ctx.AddU32x4( - "{}=uvec4(uvec2(textureSize({},int({}))),0u,uint(textureQueryLevels({})));", inst, - texture, lod, texture); + return ctx.AddU32x4("{}=uvec4(uvec2(textureSize({},int({}))),0u,{});", inst, texture, lod, + mips()); case TextureType::ColorArray2D: case TextureType::Color3D: case TextureType::ColorArrayCube: - return ctx.AddU32x4( - "{}=uvec4(uvec3(textureSize({},int({}))),uint(textureQueryLevels({})));", inst, texture, - lod, texture); + return ctx.AddU32x4("{}=uvec4(uvec3(textureSize({},int({}))),{});", inst, texture, lod, + mips()); case TextureType::Buffer: throw NotImplementedException("EmitImageQueryDimensions Texture buffers"); } diff --git a/src/shader_recompiler/backend/glsl/emit_glsl_instructions.h b/src/shader_recompiler/backend/glsl/emit_glsl_instructions.h index 4151c89de..8d0a65047 100644 --- a/src/shader_recompiler/backend/glsl/emit_glsl_instructions.h +++ b/src/shader_recompiler/backend/glsl/emit_glsl_instructions.h @@ -231,7 +231,6 @@ void EmitSelectF64(EmitContext& ctx, IR::Inst& inst, std::string_view cond, void EmitBitCastU16F16(EmitContext& ctx, IR::Inst& inst); void EmitBitCastU32F32(EmitContext& ctx, IR::Inst& inst, std::string_view value); void EmitBitCastU64F64(EmitContext& ctx, IR::Inst& inst, std::string_view value); -void EmitBitCastS32F32(EmitContext& ctx, IR::Inst& inst, std::string_view value); void EmitBitCastF16U16(EmitContext& ctx, IR::Inst& inst); void EmitBitCastF32U32(EmitContext& ctx, IR::Inst& inst, std::string_view value); void EmitBitCastF64U64(EmitContext& ctx, IR::Inst& inst, std::string_view value); @@ -655,7 +654,7 @@ void EmitImageFetch(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, std::string_view coords, std::string_view offset, std::string_view lod, std::string_view ms); void EmitImageQueryDimensions(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, - std::string_view lod); + std::string_view lod, const IR::Value& skip_mips); void EmitImageQueryLod(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, std::string_view coords); void EmitImageGradient(EmitContext& ctx, IR::Inst& inst, const IR::Value& index, diff --git a/src/shader_recompiler/backend/glsl/glsl_emit_context.cpp b/src/shader_recompiler/backend/glsl/glsl_emit_context.cpp index 5d01ec0cd..1b006e811 100644 --- a/src/shader_recompiler/backend/glsl/glsl_emit_context.cpp +++ b/src/shader_recompiler/backend/glsl/glsl_emit_context.cpp @@ -61,24 +61,28 @@ std::string OutputDecorator(Stage stage, u32 size) { } } -std::string_view SamplerType(TextureType type, bool is_depth) { - if (is_depth) { - switch (type) { - case TextureType::Color1D: - return "sampler1DShadow"; - case TextureType::ColorArray1D: - return "sampler1DArrayShadow"; - case TextureType::Color2D: - return "sampler2DShadow"; - case TextureType::ColorArray2D: - return "sampler2DArrayShadow"; - case TextureType::ColorCube: - return "samplerCubeShadow"; - case TextureType::ColorArrayCube: - return "samplerCubeArrayShadow"; - default: - throw NotImplementedException("Texture type: {}", type); - } +std::string_view DepthSamplerType(TextureType type) { + switch (type) { + case TextureType::Color1D: + return "sampler1DShadow"; + case TextureType::ColorArray1D: + return "sampler1DArrayShadow"; + case TextureType::Color2D: + return "sampler2DShadow"; + case TextureType::ColorArray2D: + return "sampler2DArrayShadow"; + case TextureType::ColorCube: + return "samplerCubeShadow"; + case TextureType::ColorArrayCube: + return "samplerCubeArrayShadow"; + default: + throw NotImplementedException("Texture type: {}", type); + } +} + +std::string_view ColorSamplerType(TextureType type, bool is_multisample = false) { + if (is_multisample) { + ASSERT(type == TextureType::Color2D || type == TextureType::ColorArray2D); } switch (type) { case TextureType::Color1D: @@ -87,9 +91,9 @@ std::string_view SamplerType(TextureType type, bool is_depth) { return "sampler1DArray"; case TextureType::Color2D: case TextureType::Color2DRect: - return "sampler2D"; + return is_multisample ? "sampler2DMS" : "sampler2D"; case TextureType::ColorArray2D: - return "sampler2DArray"; + return is_multisample ? "sampler2DMSArray" : "sampler2DArray"; case TextureType::Color3D: return "sampler3D"; case TextureType::ColorCube: @@ -677,7 +681,7 @@ void EmitContext::SetupTextures(Bindings& bindings) { texture_buffers.reserve(info.texture_buffer_descriptors.size()); for (const auto& desc : info.texture_buffer_descriptors) { texture_buffers.push_back({bindings.texture, desc.count}); - const auto sampler_type{SamplerType(TextureType::Buffer, false)}; + const auto sampler_type{ColorSamplerType(TextureType::Buffer)}; const auto array_decorator{desc.count > 1 ? fmt::format("[{}]", desc.count) : ""}; header += fmt::format("layout(binding={}) uniform {} tex{}{};", bindings.texture, sampler_type, bindings.texture, array_decorator); @@ -686,7 +690,8 @@ void EmitContext::SetupTextures(Bindings& bindings) { textures.reserve(info.texture_descriptors.size()); for (const auto& desc : info.texture_descriptors) { textures.push_back({bindings.texture, desc.count}); - const auto sampler_type{SamplerType(desc.type, desc.is_depth)}; + const auto sampler_type{desc.is_depth ? DepthSamplerType(desc.type) + : ColorSamplerType(desc.type, desc.is_multisample)}; const auto array_decorator{desc.count > 1 ? fmt::format("[{}]", desc.count) : ""}; header += fmt::format("layout(binding={}) uniform {} tex{}{};", bindings.texture, sampler_type, bindings.texture, array_decorator); diff --git a/src/shader_recompiler/backend/spirv/emit_spirv_bitwise_conversion.cpp b/src/shader_recompiler/backend/spirv/emit_spirv_bitwise_conversion.cpp index 50daacd95..c4ca28d11 100644 --- a/src/shader_recompiler/backend/spirv/emit_spirv_bitwise_conversion.cpp +++ b/src/shader_recompiler/backend/spirv/emit_spirv_bitwise_conversion.cpp @@ -18,10 +18,6 @@ void EmitBitCastU64F64(EmitContext&) { throw NotImplementedException("SPIR-V Instruction"); } -void EmitBitCastS32F32(EmitContext&) { - throw NotImplementedException("SPIR-V Instruction"); -} - void EmitBitCastF16U16(EmitContext&) { throw NotImplementedException("SPIR-V Instruction"); } diff --git a/src/shader_recompiler/backend/spirv/emit_spirv_image.cpp b/src/shader_recompiler/backend/spirv/emit_spirv_image.cpp index c898ce12f..3b969d915 100644 --- a/src/shader_recompiler/backend/spirv/emit_spirv_image.cpp +++ b/src/shader_recompiler/backend/spirv/emit_spirv_image.cpp @@ -445,11 +445,13 @@ Id EmitImageFetch(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id c TextureImage(ctx, info, index), coords, operands.MaskOptional(), operands.Span()); } -Id EmitImageQueryDimensions(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id lod) { +Id EmitImageQueryDimensions(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id lod, + const IR::Value& skip_mips_val) { const auto info{inst->Flags<IR::TextureInstInfo>()}; const Id image{TextureImage(ctx, info, index)}; const Id zero{ctx.u32_zero_value}; - const auto mips{[&] { return ctx.OpImageQueryLevels(ctx.U32[1], image); }}; + const bool skip_mips{skip_mips_val.U1()}; + const auto mips{[&] { return skip_mips ? zero : ctx.OpImageQueryLevels(ctx.U32[1], image); }}; switch (info.type) { case TextureType::Color1D: return ctx.OpCompositeConstruct(ctx.U32[4], ctx.OpImageQuerySizeLod(ctx.U32[1], image, lod), diff --git a/src/shader_recompiler/backend/spirv/emit_spirv_instructions.h b/src/shader_recompiler/backend/spirv/emit_spirv_instructions.h index e31cdc5e8..a440b557d 100644 --- a/src/shader_recompiler/backend/spirv/emit_spirv_instructions.h +++ b/src/shader_recompiler/backend/spirv/emit_spirv_instructions.h @@ -179,7 +179,6 @@ Id EmitSelectF64(EmitContext& ctx, Id cond, Id true_value, Id false_value); void EmitBitCastU16F16(EmitContext& ctx); Id EmitBitCastU32F32(EmitContext& ctx, Id value); void EmitBitCastU64F64(EmitContext& ctx); -void EmitBitCastS32F32(EmitContext& ctx); void EmitBitCastF16U16(EmitContext&); Id EmitBitCastF32U32(EmitContext& ctx, Id value); void EmitBitCastF64U64(EmitContext& ctx); @@ -540,7 +539,8 @@ Id EmitImageGatherDref(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, const IR::Value& offset, const IR::Value& offset2, Id dref); Id EmitImageFetch(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id offset, Id lod, Id ms); -Id EmitImageQueryDimensions(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id lod); +Id EmitImageQueryDimensions(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id lod, + const IR::Value& skip_mips); Id EmitImageQueryLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords); Id EmitImageGradient(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id derivates, Id offset, Id lod_clamp); diff --git a/src/shader_recompiler/frontend/ir/ir_emitter.cpp b/src/shader_recompiler/frontend/ir/ir_emitter.cpp index eb2e49a68..b7caa4246 100644 --- a/src/shader_recompiler/frontend/ir/ir_emitter.cpp +++ b/src/shader_recompiler/frontend/ir/ir_emitter.cpp @@ -704,11 +704,6 @@ IR::U32 IREmitter::BitCast<IR::U32, IR::F32>(const IR::F32& value) { } template <> -IR::S32 IREmitter::BitCast<IR::S32, IR::F32>(const IR::F32& value) { - return Inst<IR::S32>(Opcode::BitCastS32F32, value); -} - -template <> IR::F32 IREmitter::BitCast<IR::F32, IR::U32>(const IR::U32& value) { return Inst<IR::F32>(Opcode::BitCastF32U32, value); } @@ -1851,15 +1846,16 @@ Value IREmitter::ImageFetch(const Value& handle, const Value& coords, const Valu return Inst(op, Flags{info}, handle, coords, offset, lod, multisampling); } -Value IREmitter::ImageQueryDimension(const Value& handle, const IR::U32& lod) { +Value IREmitter::ImageQueryDimension(const Value& handle, const IR::U32& lod, + const IR::U1& skip_mips) { const Opcode op{handle.IsImmediate() ? Opcode::BoundImageQueryDimensions : Opcode::BindlessImageQueryDimensions}; - return Inst(op, handle, lod); + return Inst(op, handle, lod, skip_mips); } Value IREmitter::ImageQueryDimension(const Value& handle, const IR::U32& lod, - TextureInstInfo info) { - return Inst(Opcode::ImageQueryDimensions, Flags{info}, handle, lod); + const IR::U1& skip_mips, TextureInstInfo info) { + return Inst(Opcode::ImageQueryDimensions, Flags{info}, handle, lod, skip_mips); } Value IREmitter::ImageQueryLod(const Value& handle, const Value& coords, TextureInstInfo info) { diff --git a/src/shader_recompiler/frontend/ir/ir_emitter.h b/src/shader_recompiler/frontend/ir/ir_emitter.h index 7aaaa4ab0..f3c81dbe1 100644 --- a/src/shader_recompiler/frontend/ir/ir_emitter.h +++ b/src/shader_recompiler/frontend/ir/ir_emitter.h @@ -320,9 +320,10 @@ public: [[nodiscard]] F32 ImageSampleDrefExplicitLod(const Value& handle, const Value& coords, const F32& dref, const F32& lod, const Value& offset, TextureInstInfo info); - [[nodiscard]] Value ImageQueryDimension(const Value& handle, const IR::U32& lod); [[nodiscard]] Value ImageQueryDimension(const Value& handle, const IR::U32& lod, - TextureInstInfo info); + const IR::U1& skip_mips); + [[nodiscard]] Value ImageQueryDimension(const Value& handle, const IR::U32& lod, + const IR::U1& skip_mips, TextureInstInfo info); [[nodiscard]] Value ImageQueryLod(const Value& handle, const Value& coords, TextureInstInfo info); @@ -408,7 +409,8 @@ private: } template <typename T> - requires(sizeof(T) <= sizeof(u32) && std::is_trivially_copyable_v<T>) struct Flags { + requires(sizeof(T) <= sizeof(u32) && std::is_trivially_copyable_v<T>) + struct Flags { Flags() = default; Flags(T proxy_) : proxy{proxy_} {} diff --git a/src/shader_recompiler/frontend/ir/opcodes.h b/src/shader_recompiler/frontend/ir/opcodes.h index d155afd0f..e300714f3 100644 --- a/src/shader_recompiler/frontend/ir/opcodes.h +++ b/src/shader_recompiler/frontend/ir/opcodes.h @@ -38,7 +38,6 @@ constexpr Type U8{Type::U8}; constexpr Type U16{Type::U16}; constexpr Type U32{Type::U32}; constexpr Type U64{Type::U64}; -constexpr Type S32{Type::S32}; constexpr Type F16{Type::F16}; constexpr Type F32{Type::F32}; constexpr Type F64{Type::F64}; diff --git a/src/shader_recompiler/frontend/ir/opcodes.inc b/src/shader_recompiler/frontend/ir/opcodes.inc index 1fe3749cc..4447d67b0 100644 --- a/src/shader_recompiler/frontend/ir/opcodes.inc +++ b/src/shader_recompiler/frontend/ir/opcodes.inc @@ -175,7 +175,6 @@ OPCODE(SelectF64, F64, U1, OPCODE(BitCastU16F16, U16, F16, ) OPCODE(BitCastU32F32, U32, F32, ) OPCODE(BitCastU64F64, U64, F64, ) -OPCODE(BitCastS32F32, S32, F32, ) OPCODE(BitCastF16U16, F16, U16, ) OPCODE(BitCastF32U32, F32, U32, ) OPCODE(BitCastF64U64, F64, U64, ) @@ -483,7 +482,7 @@ OPCODE(BindlessImageSampleDrefExplicitLod, F32, U32, OPCODE(BindlessImageGather, F32x4, U32, Opaque, Opaque, Opaque, ) OPCODE(BindlessImageGatherDref, F32x4, U32, Opaque, Opaque, Opaque, F32, ) OPCODE(BindlessImageFetch, F32x4, U32, Opaque, Opaque, U32, Opaque, ) -OPCODE(BindlessImageQueryDimensions, U32x4, U32, U32, ) +OPCODE(BindlessImageQueryDimensions, U32x4, U32, U32, U1, ) OPCODE(BindlessImageQueryLod, F32x4, U32, Opaque, ) OPCODE(BindlessImageGradient, F32x4, U32, Opaque, Opaque, Opaque, Opaque, ) OPCODE(BindlessImageRead, U32x4, U32, Opaque, ) @@ -496,7 +495,7 @@ OPCODE(BoundImageSampleDrefExplicitLod, F32, U32, OPCODE(BoundImageGather, F32x4, U32, Opaque, Opaque, Opaque, ) OPCODE(BoundImageGatherDref, F32x4, U32, Opaque, Opaque, Opaque, F32, ) OPCODE(BoundImageFetch, F32x4, U32, Opaque, Opaque, U32, Opaque, ) -OPCODE(BoundImageQueryDimensions, U32x4, U32, U32, ) +OPCODE(BoundImageQueryDimensions, U32x4, U32, U32, U1, ) OPCODE(BoundImageQueryLod, F32x4, U32, Opaque, ) OPCODE(BoundImageGradient, F32x4, U32, Opaque, Opaque, Opaque, Opaque, ) OPCODE(BoundImageRead, U32x4, U32, Opaque, ) @@ -509,7 +508,7 @@ OPCODE(ImageSampleDrefExplicitLod, F32, Opaq OPCODE(ImageGather, F32x4, Opaque, Opaque, Opaque, Opaque, ) OPCODE(ImageGatherDref, F32x4, Opaque, Opaque, Opaque, Opaque, F32, ) OPCODE(ImageFetch, F32x4, Opaque, Opaque, Opaque, U32, Opaque, ) -OPCODE(ImageQueryDimensions, U32x4, Opaque, U32, ) +OPCODE(ImageQueryDimensions, U32x4, Opaque, U32, U1, ) OPCODE(ImageQueryLod, F32x4, Opaque, Opaque, ) OPCODE(ImageGradient, F32x4, Opaque, Opaque, Opaque, Opaque, Opaque, ) OPCODE(ImageRead, U32x4, Opaque, Opaque, ) diff --git a/src/shader_recompiler/frontend/ir/type.h b/src/shader_recompiler/frontend/ir/type.h index 5a7c706ad..04c8c4ddb 100644 --- a/src/shader_recompiler/frontend/ir/type.h +++ b/src/shader_recompiler/frontend/ir/type.h @@ -24,22 +24,21 @@ enum class Type { U16 = 1 << 7, U32 = 1 << 8, U64 = 1 << 9, - S32 = 1 << 10, - F16 = 1 << 11, - F32 = 1 << 12, - F64 = 1 << 13, - U32x2 = 1 << 14, - U32x3 = 1 << 15, - U32x4 = 1 << 16, - F16x2 = 1 << 17, - F16x3 = 1 << 18, - F16x4 = 1 << 19, - F32x2 = 1 << 20, - F32x3 = 1 << 21, - F32x4 = 1 << 22, - F64x2 = 1 << 23, - F64x3 = 1 << 24, - F64x4 = 1 << 25, + F16 = 1 << 10, + F32 = 1 << 11, + F64 = 1 << 12, + U32x2 = 1 << 13, + U32x3 = 1 << 14, + U32x4 = 1 << 15, + F16x2 = 1 << 16, + F16x3 = 1 << 17, + F16x4 = 1 << 18, + F32x2 = 1 << 19, + F32x3 = 1 << 20, + F32x4 = 1 << 21, + F64x2 = 1 << 22, + F64x3 = 1 << 23, + F64x4 = 1 << 24, }; DECLARE_ENUM_FLAG_OPERATORS(Type) diff --git a/src/shader_recompiler/frontend/ir/value.cpp b/src/shader_recompiler/frontend/ir/value.cpp index 30ba12316..346169328 100644 --- a/src/shader_recompiler/frontend/ir/value.cpp +++ b/src/shader_recompiler/frontend/ir/value.cpp @@ -23,8 +23,6 @@ Value::Value(u16 value) noexcept : type{Type::U16}, imm_u16{value} {} Value::Value(u32 value) noexcept : type{Type::U32}, imm_u32{value} {} -Value::Value(s32 value) noexcept : type{Type::S32}, imm_s32{value} {} - Value::Value(f32 value) noexcept : type{Type::F32}, imm_f32{value} {} Value::Value(u64 value) noexcept : type{Type::U64}, imm_u64{value} {} @@ -71,7 +69,6 @@ bool Value::operator==(const Value& other) const { return imm_u16 == other.imm_u16; case Type::U32: case Type::F32: - case Type::S32: return imm_u32 == other.imm_u32; case Type::U64: case Type::F64: diff --git a/src/shader_recompiler/frontend/ir/value.h b/src/shader_recompiler/frontend/ir/value.h index 8b34356fd..22e89dd1b 100644 --- a/src/shader_recompiler/frontend/ir/value.h +++ b/src/shader_recompiler/frontend/ir/value.h @@ -101,9 +101,8 @@ public: TypedValue() = default; template <IR::Type other_type> - requires((other_type & type_) != IR::Type::Void) explicit(false) - TypedValue(const TypedValue<other_type>& value) - : Value(value) {} + requires((other_type & type_) != IR::Type::Void) + explicit(false) TypedValue(const TypedValue<other_type>& value) : Value(value) {} explicit TypedValue(const Value& value) : Value(value) { if ((value.Type() & type_) == IR::Type::Void) { @@ -194,16 +193,16 @@ public: void ReplaceOpcode(IR::Opcode opcode); template <typename FlagsType> - requires(sizeof(FlagsType) <= sizeof(u32) && std::is_trivially_copyable_v<FlagsType>) - [[nodiscard]] FlagsType Flags() const noexcept { + requires(sizeof(FlagsType) <= sizeof(u32) && std::is_trivially_copyable_v<FlagsType>) + [[nodiscard]] FlagsType Flags() const noexcept { FlagsType ret; std::memcpy(reinterpret_cast<char*>(&ret), &flags, sizeof(ret)); return ret; } template <typename FlagsType> - requires(sizeof(FlagsType) <= sizeof(u32) && - std::is_trivially_copyable_v<FlagsType>) void SetFlags(FlagsType value) noexcept { + requires(sizeof(FlagsType) <= sizeof(u32) && std::is_trivially_copyable_v<FlagsType>) + void SetFlags(FlagsType value) noexcept { std::memcpy(&flags, &value, sizeof(value)); } @@ -268,7 +267,6 @@ using U8 = TypedValue<Type::U8>; using U16 = TypedValue<Type::U16>; using U32 = TypedValue<Type::U32>; using U64 = TypedValue<Type::U64>; -using S32 = TypedValue<Type::S32>; using F16 = TypedValue<Type::F16>; using F32 = TypedValue<Type::F32>; using F64 = TypedValue<Type::F64>; diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/texture_query.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/texture_query.cpp index f8cfd4ab6..39af62559 100644 --- a/src/shader_recompiler/frontend/maxwell/translate/impl/texture_query.cpp +++ b/src/shader_recompiler/frontend/maxwell/translate/impl/texture_query.cpp @@ -15,11 +15,13 @@ enum class Mode : u64 { SamplePos = 5, }; -IR::Value Query(TranslatorVisitor& v, const IR::U32& handle, Mode mode, IR::Reg src_reg) { +IR::Value Query(TranslatorVisitor& v, const IR::U32& handle, Mode mode, IR::Reg src_reg, u64 mask) { switch (mode) { case Mode::Dimension: { + const bool needs_num_mips{((mask >> 3) & 1) != 0}; + const IR::U1 skip_mips{v.ir.Imm1(!needs_num_mips)}; const IR::U32 lod{v.X(src_reg)}; - return v.ir.ImageQueryDimension(handle, lod); + return v.ir.ImageQueryDimension(handle, lod, skip_mips); } case Mode::TextureType: case Mode::SamplePos: @@ -46,7 +48,7 @@ void Impl(TranslatorVisitor& v, u64 insn, std::optional<u32> cbuf_offset) { handle = v.X(src_reg); ++src_reg; } - const IR::Value query{Query(v, handle, txq.mode, src_reg)}; + const IR::Value query{Query(v, handle, txq.mode, src_reg, txq.mask)}; IR::Reg dest_reg{txq.dest_reg}; for (int element = 0; element < 4; ++element) { if (((txq.mask >> element) & 1) == 0) { diff --git a/src/shader_recompiler/ir_opt/texture_pass.cpp b/src/shader_recompiler/ir_opt/texture_pass.cpp index 9718c6921..d374c976a 100644 --- a/src/shader_recompiler/ir_opt/texture_pass.cpp +++ b/src/shader_recompiler/ir_opt/texture_pass.cpp @@ -355,21 +355,21 @@ TextureInst MakeInst(Environment& env, IR::Block* block, IR::Inst& inst) { }; } -TextureType ReadTextureType(Environment& env, const ConstBufferAddr& cbuf) { +u32 GetTextureHandle(Environment& env, const ConstBufferAddr& cbuf) { const u32 secondary_index{cbuf.has_secondary ? cbuf.secondary_index : cbuf.index}; const u32 secondary_offset{cbuf.has_secondary ? cbuf.secondary_offset : cbuf.offset}; const u32 lhs_raw{env.ReadCbufValue(cbuf.index, cbuf.offset) << cbuf.shift_left}; const u32 rhs_raw{env.ReadCbufValue(secondary_index, secondary_offset) << cbuf.secondary_shift_left}; - return env.ReadTextureType(lhs_raw | rhs_raw); + return lhs_raw | rhs_raw; +} + +TextureType ReadTextureType(Environment& env, const ConstBufferAddr& cbuf) { + return env.ReadTextureType(GetTextureHandle(env, cbuf)); } TexturePixelFormat ReadTexturePixelFormat(Environment& env, const ConstBufferAddr& cbuf) { - const u32 secondary_index{cbuf.has_secondary ? cbuf.secondary_index : cbuf.index}; - const u32 secondary_offset{cbuf.has_secondary ? cbuf.secondary_offset : cbuf.offset}; - const u32 lhs_raw{env.ReadCbufValue(cbuf.index, cbuf.offset)}; - const u32 rhs_raw{env.ReadCbufValue(secondary_index, secondary_offset)}; - return env.ReadTexturePixelFormat(lhs_raw | rhs_raw); + return env.ReadTexturePixelFormat(GetTextureHandle(env, cbuf)); } class Descriptors { @@ -386,8 +386,10 @@ public: return Add(texture_buffer_descriptors, desc, [&desc](const auto& existing) { return desc.cbuf_index == existing.cbuf_index && desc.cbuf_offset == existing.cbuf_offset && + desc.shift_left == existing.shift_left && desc.secondary_cbuf_index == existing.secondary_cbuf_index && desc.secondary_cbuf_offset == existing.secondary_cbuf_offset && + desc.secondary_shift_left == existing.secondary_shift_left && desc.count == existing.count && desc.size_shift == existing.size_shift && desc.has_secondary == existing.has_secondary; }); @@ -405,15 +407,20 @@ public: } u32 Add(const TextureDescriptor& desc) { - return Add(texture_descriptors, desc, [&desc](const auto& existing) { + const u32 index{Add(texture_descriptors, desc, [&desc](const auto& existing) { return desc.type == existing.type && desc.is_depth == existing.is_depth && desc.has_secondary == existing.has_secondary && desc.cbuf_index == existing.cbuf_index && desc.cbuf_offset == existing.cbuf_offset && + desc.shift_left == existing.shift_left && desc.secondary_cbuf_index == existing.secondary_cbuf_index && desc.secondary_cbuf_offset == existing.secondary_cbuf_offset && + desc.secondary_shift_left == existing.secondary_shift_left && desc.count == existing.count && desc.size_shift == existing.size_shift; - }); + })}; + // TODO: Read this from TIC + texture_descriptors[index].is_multisample |= desc.is_multisample; + return index; } u32 Add(const ImageDescriptor& desc) { @@ -452,7 +459,8 @@ void PatchImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) { const IR::Value coord(inst.Arg(1)); const IR::Value handle(ir.Imm32(0)); const IR::U32 lod{ir.Imm32(0)}; - const IR::Value texture_size = ir.ImageQueryDimension(handle, lod, info); + const IR::U1 skip_mips{ir.Imm1(true)}; + const IR::Value texture_size = ir.ImageQueryDimension(handle, lod, skip_mips, info); inst.SetArg( 1, ir.CompositeConstruct( ir.FPMul(IR::F32(ir.CompositeExtract(coord, 0)), @@ -486,10 +494,10 @@ void PatchTexelFetch(IR::Block& block, IR::Inst& inst, TexturePixelFormat pixel_ const IR::F32 w(ir.CompositeExtract(new_inst, 3)); const IR::F16F32F64 max_value(ir.Imm32(get_max_value())); const IR::Value converted = - ir.CompositeConstruct(ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::S32>(x)), max_value), - ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::S32>(y)), max_value), - ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::S32>(z)), max_value), - ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::S32>(w)), max_value)); + ir.CompositeConstruct(ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::U32>(x)), max_value), + ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::U32>(y)), max_value), + ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::U32>(z)), max_value), + ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::U32>(w)), max_value)); inst.ReplaceUsesWith(converted); } } // Anonymous namespace diff --git a/src/shader_recompiler/object_pool.h b/src/shader_recompiler/object_pool.h index 2b42c4ba2..5d648b159 100644 --- a/src/shader_recompiler/object_pool.h +++ b/src/shader_recompiler/object_pool.h @@ -10,7 +10,7 @@ namespace Shader { template <typename T> -requires std::is_destructible_v<T> + requires std::is_destructible_v<T> class ObjectPool { public: explicit ObjectPool(size_t chunk_size = 8192) : new_chunk_size{chunk_size} { @@ -18,7 +18,7 @@ public: } template <typename... Args> - requires std::is_constructible_v<T, Args...> + requires std::is_constructible_v<T, Args...> [[nodiscard]] T* Create(Args&&... args) { return std::construct_at(Memory(), std::forward<Args>(args)...); } |