shader: Implement TXQ and fix FragDepth

This commit is contained in:
ReinUsesLisp 2021-03-26 18:45:38 -03:00 committed by ameerj
parent d9c5bd9509
commit 17063d16a3
15 changed files with 264 additions and 21 deletions

View File

@ -126,6 +126,7 @@ add_library(shader_recompiler STATIC
frontend/maxwell/translate/impl/texture_fetch_swizzled.cpp frontend/maxwell/translate/impl/texture_fetch_swizzled.cpp
frontend/maxwell/translate/impl/texture_gather_swizzled.cpp frontend/maxwell/translate/impl/texture_gather_swizzled.cpp
frontend/maxwell/translate/impl/texture_gather.cpp frontend/maxwell/translate/impl/texture_gather.cpp
frontend/maxwell/translate/impl/texture_query.cpp
frontend/maxwell/translate/impl/vote.cpp frontend/maxwell/translate/impl/vote.cpp
frontend/maxwell/translate/impl/warp_shuffle.cpp frontend/maxwell/translate/impl/warp_shuffle.cpp
frontend/maxwell/translate/translate.cpp frontend/maxwell/translate/translate.cpp

View File

@ -244,8 +244,9 @@ void EmitContext::DefineTextures(const Info& info, u32& binding) {
if (desc.count != 1) { if (desc.count != 1) {
throw NotImplementedException("Array of textures"); throw NotImplementedException("Array of textures");
} }
const Id type{TypeSampledImage(ImageType(*this, desc))}; const Id image_type{ImageType(*this, desc)};
const Id pointer_type{TypePointer(spv::StorageClass::UniformConstant, type)}; const Id sampled_type{TypeSampledImage(image_type)};
const Id pointer_type{TypePointer(spv::StorageClass::UniformConstant, sampled_type)};
const Id id{AddGlobalVariable(pointer_type, spv::StorageClass::UniformConstant)}; const Id id{AddGlobalVariable(pointer_type, spv::StorageClass::UniformConstant)};
Decorate(id, spv::Decoration::Binding, binding); Decorate(id, spv::Decoration::Binding, binding);
Decorate(id, spv::Decoration::DescriptorSet, 0U); Decorate(id, spv::Decoration::DescriptorSet, 0U);
@ -254,7 +255,8 @@ void EmitContext::DefineTextures(const Info& info, u32& binding) {
// TODO: Pass count info // TODO: Pass count info
textures.push_back(TextureDefinition{ textures.push_back(TextureDefinition{
.id{id}, .id{id},
.type{type}, .sampled_type{sampled_type},
.image_type{image_type},
}); });
} }
binding += desc.count; binding += desc.count;

View File

@ -31,7 +31,8 @@ private:
struct TextureDefinition { struct TextureDefinition {
Id id; Id id;
Id type; Id sampled_type;
Id image_type;
}; };
struct UniformDefinitions { struct UniformDefinitions {

View File

@ -126,10 +126,10 @@ Id DefineMain(EmitContext& ctx, IR::Program& program) {
return main; return main;
} }
void DefineEntryPoint(Environment& env, EmitContext& ctx, Id main) { void DefineEntryPoint(Environment& env, const IR::Program& program, EmitContext& ctx, Id main) {
const std::span interfaces(ctx.interfaces.data(), ctx.interfaces.size()); const std::span interfaces(ctx.interfaces.data(), ctx.interfaces.size());
spv::ExecutionModel execution_model{}; spv::ExecutionModel execution_model{};
switch (env.ShaderStage()) { switch (program.stage) {
case Shader::Stage::Compute: { case Shader::Stage::Compute: {
const std::array<u32, 3> workgroup_size{env.WorkgroupSize()}; const std::array<u32, 3> workgroup_size{env.WorkgroupSize()};
execution_model = spv::ExecutionModel::GLCompute; execution_model = spv::ExecutionModel::GLCompute;
@ -143,6 +143,9 @@ void DefineEntryPoint(Environment& env, EmitContext& ctx, Id main) {
case Shader::Stage::Fragment: case Shader::Stage::Fragment:
execution_model = spv::ExecutionModel::Fragment; execution_model = spv::ExecutionModel::Fragment;
ctx.AddExecutionMode(main, spv::ExecutionMode::OriginUpperLeft); ctx.AddExecutionMode(main, spv::ExecutionMode::OriginUpperLeft);
if (program.info.stores_frag_depth) {
ctx.AddExecutionMode(main, spv::ExecutionMode::DepthReplacing);
}
break; break;
default: default:
throw NotImplementedException("Stage {}", env.ShaderStage()); throw NotImplementedException("Stage {}", env.ShaderStage());
@ -235,6 +238,7 @@ void SetupCapabilities(const Profile& profile, const Info& info, EmitContext& ct
} }
// TODO: Track this usage // TODO: Track this usage
ctx.AddCapability(spv::Capability::ImageGatherExtended); ctx.AddCapability(spv::Capability::ImageGatherExtended);
ctx.AddCapability(spv::Capability::ImageQuery);
} }
Id PhiArgDef(EmitContext& ctx, IR::Inst* inst, size_t index) { Id PhiArgDef(EmitContext& ctx, IR::Inst* inst, size_t index) {
@ -267,7 +271,7 @@ std::vector<u32> EmitSPIRV(const Profile& profile, Environment& env, IR::Program
u32& binding) { u32& binding) {
EmitContext ctx{profile, program, binding}; EmitContext ctx{profile, program, binding};
const Id main{DefineMain(ctx, program)}; const Id main{DefineMain(ctx, program)};
DefineEntryPoint(env, ctx, main); DefineEntryPoint(env, program, ctx, main);
if (profile.support_float_controls) { if (profile.support_float_controls) {
ctx.AddExtension("SPV_KHR_float_controls"); ctx.AddExtension("SPV_KHR_float_controls");
SetupDenormControl(profile, program, ctx, main); SetupDenormControl(profile, program, ctx, main);

View File

@ -343,6 +343,7 @@ Id EmitBindlessImageSampleDrefExplicitLod(EmitContext&);
Id EmitBindlessImageGather(EmitContext&); Id EmitBindlessImageGather(EmitContext&);
Id EmitBindlessImageGatherDref(EmitContext&); Id EmitBindlessImageGatherDref(EmitContext&);
Id EmitBindlessImageFetch(EmitContext&); Id EmitBindlessImageFetch(EmitContext&);
Id EmitBindlessImageQueryDimensions(EmitContext&);
Id EmitBoundImageSampleImplicitLod(EmitContext&); Id EmitBoundImageSampleImplicitLod(EmitContext&);
Id EmitBoundImageSampleExplicitLod(EmitContext&); Id EmitBoundImageSampleExplicitLod(EmitContext&);
Id EmitBoundImageSampleDrefImplicitLod(EmitContext&); Id EmitBoundImageSampleDrefImplicitLod(EmitContext&);
@ -350,6 +351,7 @@ Id EmitBoundImageSampleDrefExplicitLod(EmitContext&);
Id EmitBoundImageGather(EmitContext&); Id EmitBoundImageGather(EmitContext&);
Id EmitBoundImageGatherDref(EmitContext&); Id EmitBoundImageGatherDref(EmitContext&);
Id EmitBoundImageFetch(EmitContext&); Id EmitBoundImageFetch(EmitContext&);
Id EmitBoundImageQueryDimensions(EmitContext&);
Id EmitImageSampleImplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id EmitImageSampleImplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
Id bias_lc, Id offset); Id bias_lc, Id offset);
Id EmitImageSampleExplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id EmitImageSampleExplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
@ -364,6 +366,7 @@ Id EmitImageGatherDref(EmitContext& ctx, IR::Inst* inst, const IR::Value& index,
const IR::Value& offset, const IR::Value& offset2, Id dref); 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 EmitImageFetch(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id offset,
Id lod, Id ms); Id lod, Id ms);
Id EmitImageQueryDimensions(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id lod);
Id EmitVoteAll(EmitContext& ctx, Id pred); Id EmitVoteAll(EmitContext& ctx, Id pred);
Id EmitVoteAny(EmitContext& ctx, Id pred); Id EmitVoteAny(EmitContext& ctx, Id pred);
Id EmitVoteEqual(EmitContext& ctx, Id pred); Id EmitVoteEqual(EmitContext& ctx, Id pred);

View File

@ -91,7 +91,15 @@ private:
Id Texture(EmitContext& ctx, const IR::Value& index) { Id Texture(EmitContext& ctx, const IR::Value& index) {
if (index.IsImmediate()) { if (index.IsImmediate()) {
const TextureDefinition def{ctx.textures.at(index.U32())}; const TextureDefinition def{ctx.textures.at(index.U32())};
return ctx.OpLoad(def.type, def.id); return ctx.OpLoad(def.sampled_type, def.id);
}
throw NotImplementedException("Indirect texture sample");
}
Id TextureImage(EmitContext& ctx, const IR::Value& index) {
if (index.IsImmediate()) {
const TextureDefinition def{ctx.textures.at(index.U32())};
return ctx.OpImage(def.image_type, ctx.OpLoad(def.sampled_type, def.id));
} }
throw NotImplementedException("Indirect texture sample"); throw NotImplementedException("Indirect texture sample");
} }
@ -149,6 +157,10 @@ Id EmitBindlessImageFetch(EmitContext&) {
throw LogicError("Unreachable instruction"); throw LogicError("Unreachable instruction");
} }
Id EmitBindlessImageQueryDimensions(EmitContext&) {
throw LogicError("Unreachable instruction");
}
Id EmitBoundImageSampleImplicitLod(EmitContext&) { Id EmitBoundImageSampleImplicitLod(EmitContext&) {
throw LogicError("Unreachable instruction"); throw LogicError("Unreachable instruction");
} }
@ -177,6 +189,10 @@ Id EmitBoundImageFetch(EmitContext&) {
throw LogicError("Unreachable instruction"); throw LogicError("Unreachable instruction");
} }
Id EmitBoundImageQueryDimensions(EmitContext&) {
throw LogicError("Unreachable instruction");
}
Id EmitImageSampleImplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id EmitImageSampleImplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
Id bias_lc, Id offset) { Id bias_lc, Id offset) {
const auto info{inst->Flags<IR::TextureInstInfo>()}; const auto info{inst->Flags<IR::TextureInstInfo>()};
@ -241,4 +257,34 @@ Id EmitImageFetch(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id c
Texture(ctx, index), coords, operands.Mask(), operands.Span()); Texture(ctx, index), coords, operands.Mask(), operands.Span());
} }
Id EmitImageQueryDimensions(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id lod) {
const auto info{inst->Flags<IR::TextureInstInfo>()};
const Id image{TextureImage(ctx, index)};
const Id zero{ctx.u32_zero_value};
const auto mips{[&] { return ctx.OpImageQueryLevels(ctx.U32[1], image); }};
switch (info.type) {
case TextureType::Color1D:
case TextureType::Shadow1D:
return ctx.OpCompositeConstruct(ctx.U32[4], ctx.OpImageQuerySizeLod(ctx.U32[1], image, lod),
zero, zero, mips());
case TextureType::ColorArray1D:
case TextureType::Color2D:
case TextureType::ColorCube:
case TextureType::ShadowArray1D:
case TextureType::Shadow2D:
case TextureType::ShadowCube:
return ctx.OpCompositeConstruct(ctx.U32[4], ctx.OpImageQuerySizeLod(ctx.U32[2], image, lod),
zero, mips());
case TextureType::ColorArray2D:
case TextureType::Color3D:
case TextureType::ColorArrayCube:
case TextureType::ShadowArray2D:
case TextureType::Shadow3D:
case TextureType::ShadowArrayCube:
return ctx.OpCompositeConstruct(ctx.U32[4], ctx.OpImageQuerySizeLod(ctx.U32[3], image, lod),
mips());
}
throw LogicError("Unspecified image type {}", info.type.Value());
}
} // namespace Shader::Backend::SPIRV } // namespace Shader::Backend::SPIRV

View File

@ -4,6 +4,7 @@
#include "common/common_types.h" #include "common/common_types.h"
#include "shader_recompiler/program_header.h" #include "shader_recompiler/program_header.h"
#include "shader_recompiler/shader_info.h"
#include "shader_recompiler/stage.h" #include "shader_recompiler/stage.h"
namespace Shader { namespace Shader {
@ -14,6 +15,8 @@ public:
[[nodiscard]] virtual u64 ReadInstruction(u32 address) = 0; [[nodiscard]] virtual u64 ReadInstruction(u32 address) = 0;
[[nodiscard]] virtual TextureType ReadTextureType(u32 cbuf_index, u32 cbuf_offset) = 0;
[[nodiscard]] virtual u32 TextureBoundBuffer() const = 0; [[nodiscard]] virtual u32 TextureBoundBuffer() const = 0;
[[nodiscard]] virtual std::array<u32, 3> WorkgroupSize() const = 0; [[nodiscard]] virtual std::array<u32, 3> WorkgroupSize() const = 0;

View File

@ -1493,6 +1493,12 @@ Value IREmitter::ImageFetch(const Value& handle, const Value& coords, const Valu
return Inst(op, Flags{info}, handle, coords, offset, lod, multisampling); return Inst(op, Flags{info}, handle, coords, offset, lod, multisampling);
} }
Value IREmitter::ImageQueryDimension(const Value& handle, const IR::U32& lod) {
const Opcode op{handle.IsImmediate() ? Opcode::BoundImageQueryDimensions
: Opcode::BindlessImageQueryDimensions};
return Inst(op, handle, lod);
}
U1 IREmitter::VoteAll(const U1& value) { U1 IREmitter::VoteAll(const U1& value) {
return Inst<U1>(Opcode::VoteAll, value); return Inst<U1>(Opcode::VoteAll, value);
} }

View File

@ -239,6 +239,7 @@ public:
const F32& dref, const F32& lod, const F32& dref, const F32& lod,
const Value& offset, const F32& lod_clamp, const Value& offset, const F32& lod_clamp,
TextureInstInfo info); TextureInstInfo info);
[[nodiscard]] Value ImageQueryDimension(const Value& handle, const IR::U32& lod);
[[nodiscard]] Value ImageGather(const Value& handle, const Value& coords, const Value& offset, [[nodiscard]] Value ImageGather(const Value& handle, const Value& coords, const Value& offset,
const Value& offset2, TextureInstInfo info); const Value& offset2, TextureInstInfo info);

View File

@ -356,6 +356,7 @@ OPCODE(BindlessImageSampleDrefExplicitLod, F32, U32,
OPCODE(BindlessImageGather, F32x4, U32, Opaque, Opaque, Opaque, ) OPCODE(BindlessImageGather, F32x4, U32, Opaque, Opaque, Opaque, )
OPCODE(BindlessImageGatherDref, F32x4, U32, Opaque, Opaque, Opaque, F32, ) OPCODE(BindlessImageGatherDref, F32x4, U32, Opaque, Opaque, Opaque, F32, )
OPCODE(BindlessImageFetch, F32x4, U32, Opaque, U32, U32, ) OPCODE(BindlessImageFetch, F32x4, U32, Opaque, U32, U32, )
OPCODE(BindlessImageQueryDimensions, U32x4, U32, U32, )
OPCODE(BoundImageSampleImplicitLod, F32x4, U32, Opaque, Opaque, Opaque, ) OPCODE(BoundImageSampleImplicitLod, F32x4, U32, Opaque, Opaque, Opaque, )
OPCODE(BoundImageSampleExplicitLod, F32x4, U32, Opaque, Opaque, Opaque, ) OPCODE(BoundImageSampleExplicitLod, F32x4, U32, Opaque, Opaque, Opaque, )
@ -364,6 +365,7 @@ OPCODE(BoundImageSampleDrefExplicitLod, F32, U32,
OPCODE(BoundImageGather, F32x4, U32, Opaque, Opaque, Opaque, ) OPCODE(BoundImageGather, F32x4, U32, Opaque, Opaque, Opaque, )
OPCODE(BoundImageGatherDref, F32x4, U32, Opaque, Opaque, Opaque, F32, ) OPCODE(BoundImageGatherDref, F32x4, U32, Opaque, Opaque, Opaque, F32, )
OPCODE(BoundImageFetch, F32x4, U32, Opaque, U32, U32, ) OPCODE(BoundImageFetch, F32x4, U32, Opaque, U32, U32, )
OPCODE(BoundImageQueryDimensions, U32x4, U32, U32, )
OPCODE(ImageSampleImplicitLod, F32x4, U32, Opaque, Opaque, Opaque, ) OPCODE(ImageSampleImplicitLod, F32x4, U32, Opaque, Opaque, Opaque, )
OPCODE(ImageSampleExplicitLod, F32x4, U32, Opaque, Opaque, Opaque, ) OPCODE(ImageSampleExplicitLod, F32x4, U32, Opaque, Opaque, Opaque, )
@ -372,6 +374,7 @@ OPCODE(ImageSampleDrefExplicitLod, F32, U32,
OPCODE(ImageGather, F32x4, U32, Opaque, Opaque, Opaque, ) OPCODE(ImageGather, F32x4, U32, Opaque, Opaque, Opaque, )
OPCODE(ImageGatherDref, F32x4, U32, Opaque, Opaque, Opaque, F32, ) OPCODE(ImageGatherDref, F32x4, U32, Opaque, Opaque, Opaque, F32, )
OPCODE(ImageFetch, F32x4, U32, Opaque, U32, U32, ) OPCODE(ImageFetch, F32x4, U32, Opaque, U32, U32, )
OPCODE(ImageQueryDimensions, U32x4, U32, U32, )
// Warp operations // Warp operations
OPCODE(VoteAll, U1, U1, ) OPCODE(VoteAll, U1, U1, )

View File

@ -373,14 +373,6 @@ void TranslatorVisitor::TXD_b(u64) {
ThrowNotImplemented(Opcode::TXD_b); ThrowNotImplemented(Opcode::TXD_b);
} }
void TranslatorVisitor::TXQ(u64) {
ThrowNotImplemented(Opcode::TXQ);
}
void TranslatorVisitor::TXQ_b(u64) {
ThrowNotImplemented(Opcode::TXQ_b);
}
void TranslatorVisitor::VABSDIFF(u64) { void TranslatorVisitor::VABSDIFF(u64) {
ThrowNotImplemented(Opcode::VABSDIFF); ThrowNotImplemented(Opcode::VABSDIFF);
} }

View File

@ -0,0 +1,76 @@
// Copyright 2021 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <optional>
#include "common/bit_field.h"
#include "common/common_types.h"
#include "shader_recompiler/frontend/ir/modifiers.h"
#include "shader_recompiler/frontend/maxwell/translate/impl/impl.h"
namespace Shader::Maxwell {
namespace {
enum class Mode : u64 {
Dimension = 1,
TextureType = 2,
SamplePos = 5,
};
IR::Value Query(TranslatorVisitor& v, const IR::U32& handle, Mode mode, IR::Reg src_reg) {
switch (mode) {
case Mode::Dimension: {
const IR::U32 lod{v.X(src_reg)};
return v.ir.ImageQueryDimension(handle, lod);
}
case Mode::TextureType:
case Mode::SamplePos:
default:
throw NotImplementedException("Mode {}", mode);
}
}
void Impl(TranslatorVisitor& v, u64 insn, std::optional<u32> cbuf_offset) {
union {
u64 raw;
BitField<49, 1, u64> nodep;
BitField<0, 8, IR::Reg> dest_reg;
BitField<8, 8, IR::Reg> src_reg;
BitField<22, 3, Mode> mode;
BitField<31, 4, u64> mask;
} const txq{insn};
IR::Reg src_reg{txq.src_reg};
IR::U32 handle;
if (cbuf_offset) {
handle = v.ir.Imm32(*cbuf_offset);
} else {
handle = v.X(src_reg);
++src_reg;
}
const IR::Value query{Query(v, handle, txq.mode, src_reg)};
IR::Reg dest_reg{txq.dest_reg};
for (int element = 0; element < 4; ++element) {
if (((txq.mask >> element) & 1) == 0) {
continue;
}
v.X(dest_reg, IR::U32{v.ir.CompositeExtract(query, element)});
++dest_reg;
}
}
} // Anonymous namespace
void TranslatorVisitor::TXQ(u64 insn) {
union {
u64 raw;
BitField<36, 13, u64> cbuf_offset;
} const txq{insn};
Impl(*this, insn, static_cast<u32>(txq.cbuf_offset));
}
void TranslatorVisitor::TXQ_b(u64 insn) {
Impl(*this, insn, std::nullopt);
}
} // namespace Shader::Maxwell

View File

@ -365,7 +365,8 @@ void VisitUsages(Info& info, IR::Inst& inst) {
case IR::Opcode::ImageSampleDrefImplicitLod: case IR::Opcode::ImageSampleDrefImplicitLod:
case IR::Opcode::ImageSampleDrefExplicitLod: case IR::Opcode::ImageSampleDrefExplicitLod:
case IR::Opcode::ImageGather: case IR::Opcode::ImageGather:
case IR::Opcode::ImageGatherDref: { case IR::Opcode::ImageGatherDref:
case IR::Opcode::ImageQueryDimensions: {
const TextureType type{inst.Flags<IR::TextureInstInfo>().type}; const TextureType type{inst.Flags<IR::TextureInstInfo>().type};
info.uses_sampled_1d |= type == TextureType::Color1D || type == TextureType::ColorArray1D || info.uses_sampled_1d |= type == TextureType::Color1D || type == TextureType::ColorArray1D ||
type == TextureType::Shadow1D || type == TextureType::ShadowArray1D; type == TextureType::Shadow1D || type == TextureType::ShadowArray1D;

View File

@ -54,6 +54,9 @@ IR::Opcode IndexedInstruction(const IR::Inst& inst) {
case IR::Opcode::BindlessImageFetch: case IR::Opcode::BindlessImageFetch:
case IR::Opcode::BoundImageFetch: case IR::Opcode::BoundImageFetch:
return IR::Opcode::ImageFetch; return IR::Opcode::ImageFetch;
case IR::Opcode::BoundImageQueryDimensions:
case IR::Opcode::BindlessImageQueryDimensions:
return IR::Opcode::ImageQueryDimensions;
default: default:
return IR::Opcode::Void; return IR::Opcode::Void;
} }
@ -68,6 +71,7 @@ bool IsBindless(const IR::Inst& inst) {
case IR::Opcode::BindlessImageGather: case IR::Opcode::BindlessImageGather:
case IR::Opcode::BindlessImageGatherDref: case IR::Opcode::BindlessImageGatherDref:
case IR::Opcode::BindlessImageFetch: case IR::Opcode::BindlessImageFetch:
case IR::Opcode::BindlessImageQueryDimensions:
return true; return true;
case IR::Opcode::BoundImageSampleImplicitLod: case IR::Opcode::BoundImageSampleImplicitLod:
case IR::Opcode::BoundImageSampleExplicitLod: case IR::Opcode::BoundImageSampleExplicitLod:
@ -76,6 +80,7 @@ bool IsBindless(const IR::Inst& inst) {
case IR::Opcode::BoundImageGather: case IR::Opcode::BoundImageGather:
case IR::Opcode::BoundImageGatherDref: case IR::Opcode::BoundImageGatherDref:
case IR::Opcode::BoundImageFetch: case IR::Opcode::BoundImageFetch:
case IR::Opcode::BoundImageQueryDimensions:
return false; return false;
default: default:
throw InvalidArgument("Invalid opcode {}", inst.Opcode()); throw InvalidArgument("Invalid opcode {}", inst.Opcode());
@ -198,13 +203,20 @@ void TexturePass(Environment& env, IR::Program& program) {
for (TextureInst& texture_inst : to_replace) { for (TextureInst& texture_inst : to_replace) {
// TODO: Handle arrays // TODO: Handle arrays
IR::Inst* const inst{texture_inst.inst}; IR::Inst* const inst{texture_inst.inst};
inst->ReplaceOpcode(IndexedInstruction(*inst));
const auto& cbuf{texture_inst.cbuf};
auto flags{inst->Flags<IR::TextureInstInfo>()};
if (inst->Opcode() == IR::Opcode::ImageQueryDimensions) {
flags.type.Assign(env.ReadTextureType(cbuf.index, cbuf.offset));
inst->SetFlags(flags);
}
const u32 index{descriptors.Add(TextureDescriptor{ const u32 index{descriptors.Add(TextureDescriptor{
.type{inst->Flags<IR::TextureInstInfo>().type}, .type{flags.type},
.cbuf_index{texture_inst.cbuf.index}, .cbuf_index{cbuf.index},
.cbuf_offset{texture_inst.cbuf.offset}, .cbuf_offset{cbuf.offset},
.count{1}, .count{1},
})}; })};
inst->ReplaceOpcode(IndexedInstruction(*inst));
inst->SetArg(0, IR::Value{index}); inst->SetArg(0, IR::Value{index});
} }
} }

View File

@ -25,6 +25,7 @@
#include "video_core/memory_manager.h" #include "video_core/memory_manager.h"
#include "video_core/renderer_vulkan/fixed_pipeline_state.h" #include "video_core/renderer_vulkan/fixed_pipeline_state.h"
#include "video_core/renderer_vulkan/maxwell_to_vk.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h"
#include "video_core/renderer_vulkan/pipeline_helper.h"
#include "video_core/renderer_vulkan/vk_compute_pipeline.h" #include "video_core/renderer_vulkan/vk_compute_pipeline.h"
#include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
#include "video_core/renderer_vulkan/vk_pipeline_cache.h" #include "video_core/renderer_vulkan/vk_pipeline_cache.h"
@ -45,6 +46,10 @@ auto MakeSpan(Container& container) {
return std::span(container.data(), container.size()); return std::span(container.data(), container.size());
} }
u64 MakeCbufKey(u32 index, u32 offset) {
return (static_cast<u64>(index) << 32) | static_cast<u64>(offset);
}
class GenericEnvironment : public Shader::Environment { class GenericEnvironment : public Shader::Environment {
public: public:
explicit GenericEnvironment() = default; explicit GenericEnvironment() = default;
@ -101,15 +106,21 @@ public:
const auto data{std::make_unique<char[]>(code_size)}; const auto data{std::make_unique<char[]>(code_size)};
gpu_memory->ReadBlock(program_base + read_lowest, data.get(), code_size); gpu_memory->ReadBlock(program_base + read_lowest, data.get(), code_size);
const u64 num_texture_types{static_cast<u64>(texture_types.size())};
const u32 texture_bound{TextureBoundBuffer()}; const u32 texture_bound{TextureBoundBuffer()};
file.write(reinterpret_cast<const char*>(&code_size), sizeof(code_size)) file.write(reinterpret_cast<const char*>(&code_size), sizeof(code_size))
.write(reinterpret_cast<const char*>(&num_texture_types), sizeof(num_texture_types))
.write(reinterpret_cast<const char*>(&texture_bound), sizeof(texture_bound)) .write(reinterpret_cast<const char*>(&texture_bound), sizeof(texture_bound))
.write(reinterpret_cast<const char*>(&start_address), sizeof(start_address)) .write(reinterpret_cast<const char*>(&start_address), sizeof(start_address))
.write(reinterpret_cast<const char*>(&read_lowest), sizeof(read_lowest)) .write(reinterpret_cast<const char*>(&read_lowest), sizeof(read_lowest))
.write(reinterpret_cast<const char*>(&read_highest), sizeof(read_highest)) .write(reinterpret_cast<const char*>(&read_highest), sizeof(read_highest))
.write(reinterpret_cast<const char*>(&stage), sizeof(stage)) .write(reinterpret_cast<const char*>(&stage), sizeof(stage))
.write(data.get(), code_size); .write(data.get(), code_size);
for (const auto [key, type] : texture_types) {
file.write(reinterpret_cast<const char*>(&key), sizeof(key))
.write(reinterpret_cast<const char*>(&type), sizeof(type));
}
if (stage == Shader::Stage::Compute) { if (stage == Shader::Stage::Compute) {
const std::array<u32, 3> workgroup_size{WorkgroupSize()}; const std::array<u32, 3> workgroup_size{WorkgroupSize()};
file.write(reinterpret_cast<const char*>(&workgroup_size), sizeof(workgroup_size)); file.write(reinterpret_cast<const char*>(&workgroup_size), sizeof(workgroup_size));
@ -147,10 +158,47 @@ protected:
return std::nullopt; return std::nullopt;
} }
Shader::TextureType ReadTextureTypeImpl(GPUVAddr tic_addr, u32 tic_limit, bool via_header_index,
GPUVAddr cbuf_addr, u32 cbuf_size, u32 cbuf_index,
u32 cbuf_offset) {
const u32 raw{cbuf_offset < cbuf_size ? gpu_memory->Read<u32>(cbuf_addr + cbuf_offset) : 0};
const TextureHandle handle{raw, via_header_index};
const GPUVAddr descriptor_addr{tic_addr + handle.image * sizeof(Tegra::Texture::TICEntry)};
Tegra::Texture::TICEntry entry;
gpu_memory->ReadBlock(descriptor_addr, &entry, sizeof(entry));
const Shader::TextureType result{[&] {
switch (entry.texture_type) {
case Tegra::Texture::TextureType::Texture1D:
return Shader::TextureType::Color1D;
case Tegra::Texture::TextureType::Texture2D:
case Tegra::Texture::TextureType::Texture2DNoMipmap:
return Shader::TextureType::Color2D;
case Tegra::Texture::TextureType::Texture3D:
return Shader::TextureType::Color3D;
case Tegra::Texture::TextureType::TextureCubemap:
return Shader::TextureType::ColorCube;
case Tegra::Texture::TextureType::Texture1DArray:
return Shader::TextureType::ColorArray1D;
case Tegra::Texture::TextureType::Texture2DArray:
return Shader::TextureType::ColorArray2D;
case Tegra::Texture::TextureType::Texture1DBuffer:
throw Shader::NotImplementedException("Texture buffer");
case Tegra::Texture::TextureType::TextureCubeArray:
return Shader::TextureType::ColorArrayCube;
default:
throw Shader::NotImplementedException("Unknown texture type");
}
}()};
texture_types.emplace(MakeCbufKey(cbuf_index, cbuf_offset), result);
return result;
}
Tegra::MemoryManager* gpu_memory{}; Tegra::MemoryManager* gpu_memory{};
GPUVAddr program_base{}; GPUVAddr program_base{};
std::vector<u64> code; std::vector<u64> code;
std::unordered_map<u64, Shader::TextureType> texture_types;
u32 read_lowest = std::numeric_limits<u32>::max(); u32 read_lowest = std::numeric_limits<u32>::max();
u32 read_highest = 0; u32 read_highest = 0;
@ -176,29 +224,45 @@ public:
switch (program) { switch (program) {
case Maxwell::ShaderProgram::VertexA: case Maxwell::ShaderProgram::VertexA:
stage = Shader::Stage::VertexA; stage = Shader::Stage::VertexA;
stage_index = 0;
break; break;
case Maxwell::ShaderProgram::VertexB: case Maxwell::ShaderProgram::VertexB:
stage = Shader::Stage::VertexB; stage = Shader::Stage::VertexB;
stage_index = 0;
break; break;
case Maxwell::ShaderProgram::TesselationControl: case Maxwell::ShaderProgram::TesselationControl:
stage = Shader::Stage::TessellationControl; stage = Shader::Stage::TessellationControl;
stage_index = 1;
break; break;
case Maxwell::ShaderProgram::TesselationEval: case Maxwell::ShaderProgram::TesselationEval:
stage = Shader::Stage::TessellationEval; stage = Shader::Stage::TessellationEval;
stage_index = 2;
break; break;
case Maxwell::ShaderProgram::Geometry: case Maxwell::ShaderProgram::Geometry:
stage = Shader::Stage::Geometry; stage = Shader::Stage::Geometry;
stage_index = 3;
break; break;
case Maxwell::ShaderProgram::Fragment: case Maxwell::ShaderProgram::Fragment:
stage = Shader::Stage::Fragment; stage = Shader::Stage::Fragment;
stage_index = 4;
break; break;
default: default:
UNREACHABLE_MSG("Invalid program={}", program); UNREACHABLE_MSG("Invalid program={}", program);
break;
} }
} }
~GraphicsEnvironment() override = default; ~GraphicsEnvironment() override = default;
Shader::TextureType ReadTextureType(u32 cbuf_index, u32 cbuf_offset) override {
const auto& regs{maxwell3d->regs};
const auto& cbuf{maxwell3d->state.shader_stages[stage_index].const_buffers[cbuf_index]};
ASSERT(cbuf.enabled);
const bool via_header_index{regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex};
return ReadTextureTypeImpl(regs.tic.Address(), regs.tic.limit, via_header_index,
cbuf.address, cbuf.size, cbuf_index, cbuf_offset);
}
u32 TextureBoundBuffer() const override { u32 TextureBoundBuffer() const override {
return maxwell3d->regs.tex_cb_index; return maxwell3d->regs.tex_cb_index;
} }
@ -209,6 +273,7 @@ public:
private: private:
Tegra::Engines::Maxwell3D* maxwell3d{}; Tegra::Engines::Maxwell3D* maxwell3d{};
size_t stage_index{};
}; };
class ComputeEnvironment final : public GenericEnvironment { class ComputeEnvironment final : public GenericEnvironment {
@ -224,6 +289,15 @@ public:
~ComputeEnvironment() override = default; ~ComputeEnvironment() override = default;
Shader::TextureType ReadTextureType(u32 cbuf_index, u32 cbuf_offset) override {
const auto& regs{kepler_compute->regs};
const auto& qmd{kepler_compute->launch_description};
ASSERT(((qmd.const_buffer_enable_mask.Value() >> cbuf_index) & 1) != 0);
const auto& cbuf{qmd.const_buffer_config[cbuf_index]};
return ReadTextureTypeImpl(regs.tic.Address(), regs.tic.limit, qmd.linked_tsc != 0,
cbuf.Address(), cbuf.size, cbuf_index, cbuf_offset);
}
u32 TextureBoundBuffer() const override { u32 TextureBoundBuffer() const override {
return kepler_compute->regs.tex_cb_index; return kepler_compute->regs.tex_cb_index;
} }
@ -278,7 +352,9 @@ class FileEnvironment final : public Shader::Environment {
public: public:
void Deserialize(std::ifstream& file) { void Deserialize(std::ifstream& file) {
u64 code_size{}; u64 code_size{};
u64 num_texture_types{};
file.read(reinterpret_cast<char*>(&code_size), sizeof(code_size)) file.read(reinterpret_cast<char*>(&code_size), sizeof(code_size))
.read(reinterpret_cast<char*>(&num_texture_types), sizeof(num_texture_types))
.read(reinterpret_cast<char*>(&texture_bound), sizeof(texture_bound)) .read(reinterpret_cast<char*>(&texture_bound), sizeof(texture_bound))
.read(reinterpret_cast<char*>(&start_address), sizeof(start_address)) .read(reinterpret_cast<char*>(&start_address), sizeof(start_address))
.read(reinterpret_cast<char*>(&read_lowest), sizeof(read_lowest)) .read(reinterpret_cast<char*>(&read_lowest), sizeof(read_lowest))
@ -286,6 +362,13 @@ public:
.read(reinterpret_cast<char*>(&stage), sizeof(stage)); .read(reinterpret_cast<char*>(&stage), sizeof(stage));
code = std::make_unique<u64[]>(Common::DivCeil(code_size, sizeof(u64))); code = std::make_unique<u64[]>(Common::DivCeil(code_size, sizeof(u64)));
file.read(reinterpret_cast<char*>(code.get()), code_size); file.read(reinterpret_cast<char*>(code.get()), code_size);
for (size_t i = 0; i < num_texture_types; ++i) {
u64 key;
Shader::TextureType type;
file.read(reinterpret_cast<char*>(&key), sizeof(key))
.read(reinterpret_cast<char*>(&type), sizeof(type));
texture_types.emplace(key, type);
}
if (stage == Shader::Stage::Compute) { if (stage == Shader::Stage::Compute) {
file.read(reinterpret_cast<char*>(&workgroup_size), sizeof(workgroup_size)); file.read(reinterpret_cast<char*>(&workgroup_size), sizeof(workgroup_size));
} else { } else {
@ -300,6 +383,14 @@ public:
return code[(address - read_lowest) / sizeof(u64)]; return code[(address - read_lowest) / sizeof(u64)];
} }
Shader::TextureType ReadTextureType(u32 cbuf_index, u32 cbuf_offset) override {
const auto it{texture_types.find(MakeCbufKey(cbuf_index, cbuf_offset))};
if (it == texture_types.end()) {
throw Shader::LogicError("Uncached read texture type");
}
return it->second;
}
u32 TextureBoundBuffer() const override { u32 TextureBoundBuffer() const override {
return texture_bound; return texture_bound;
} }
@ -310,6 +401,7 @@ public:
private: private:
std::unique_ptr<u64[]> code; std::unique_ptr<u64[]> code;
std::unordered_map<u64, Shader::TextureType> texture_types;
std::array<u32, 3> workgroup_size{}; std::array<u32, 3> workgroup_size{};
u32 texture_bound{}; u32 texture_bound{};
u32 read_lowest{}; u32 read_lowest{};