blob: b91450f823020cae304cdd89cf10dd73c69ec75b [file] [log] [blame]
// Copyright 2017 The Dawn Authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "dawn_native/d3d12/ShaderModuleD3D12.h"
#include "common/Assert.h"
#include "common/BitSetIterator.h"
#include "common/Log.h"
#include "dawn_native/SpirvUtils.h"
#include "dawn_native/d3d12/BindGroupLayoutD3D12.h"
#include "dawn_native/d3d12/D3D12Error.h"
#include "dawn_native/d3d12/DeviceD3D12.h"
#include "dawn_native/d3d12/PipelineLayoutD3D12.h"
#include "dawn_native/d3d12/PlatformFunctions.h"
#include "dawn_native/d3d12/UtilsD3D12.h"
#include <d3dcompiler.h>
#include <spirv_hlsl.hpp>
#ifdef DAWN_ENABLE_WGSL
// Tint include must be after spirv_hlsl.hpp, because spirv-cross has its own
// version of spirv_headers. We also need to undef SPV_REVISION because SPIRV-Cross
// is at 3 while spirv-headers is at 4.
# undef SPV_REVISION
# include <tint/tint.h>
#endif // DAWN_ENABLE_WGSL
namespace dawn_native { namespace d3d12 {
namespace {
std::vector<const wchar_t*> GetDXCArguments(uint32_t compileFlags, bool enable16BitTypes) {
std::vector<const wchar_t*> arguments;
if (compileFlags & D3DCOMPILE_ENABLE_BACKWARDS_COMPATIBILITY) {
arguments.push_back(L"/Gec");
}
if (compileFlags & D3DCOMPILE_IEEE_STRICTNESS) {
arguments.push_back(L"/Gis");
}
if (compileFlags & D3DCOMPILE_OPTIMIZATION_LEVEL2) {
switch (compileFlags & D3DCOMPILE_OPTIMIZATION_LEVEL2) {
case D3DCOMPILE_OPTIMIZATION_LEVEL0:
arguments.push_back(L"/O0");
break;
case D3DCOMPILE_OPTIMIZATION_LEVEL2:
arguments.push_back(L"/O2");
break;
case D3DCOMPILE_OPTIMIZATION_LEVEL3:
arguments.push_back(L"/O3");
break;
}
}
if (compileFlags & D3DCOMPILE_DEBUG) {
arguments.push_back(L"/Zi");
}
if (compileFlags & D3DCOMPILE_PACK_MATRIX_ROW_MAJOR) {
arguments.push_back(L"/Zpr");
}
if (compileFlags & D3DCOMPILE_PACK_MATRIX_COLUMN_MAJOR) {
arguments.push_back(L"/Zpc");
}
if (compileFlags & D3DCOMPILE_AVOID_FLOW_CONTROL) {
arguments.push_back(L"/Gfa");
}
if (compileFlags & D3DCOMPILE_PREFER_FLOW_CONTROL) {
arguments.push_back(L"/Gfp");
}
if (compileFlags & D3DCOMPILE_RESOURCES_MAY_ALIAS) {
arguments.push_back(L"/res_may_alias");
}
if (enable16BitTypes) {
// enable-16bit-types are only allowed in -HV 2018 (default)
arguments.push_back(L"/enable-16bit-types");
} else {
// Enable FXC backward compatibility by setting the language version to 2016
arguments.push_back(L"-HV");
arguments.push_back(L"2016");
}
return arguments;
}
} // anonymous namespace
ResultOrError<ComPtr<IDxcBlob>> CompileShaderDXC(Device* device,
SingleShaderStage stage,
const std::string& hlslSource,
const char* entryPoint,
uint32_t compileFlags) {
IDxcLibrary* dxcLibrary;
DAWN_TRY_ASSIGN(dxcLibrary, device->GetOrCreateDxcLibrary());
ComPtr<IDxcBlobEncoding> sourceBlob;
DAWN_TRY(CheckHRESULT(dxcLibrary->CreateBlobWithEncodingOnHeapCopy(
hlslSource.c_str(), hlslSource.length(), CP_UTF8, &sourceBlob),
"DXC create blob"));
IDxcCompiler* dxcCompiler;
DAWN_TRY_ASSIGN(dxcCompiler, device->GetOrCreateDxcCompiler());
std::wstring entryPointW;
DAWN_TRY_ASSIGN(entryPointW, ConvertStringToWstring(entryPoint));
std::vector<const wchar_t*> arguments =
GetDXCArguments(compileFlags, device->IsExtensionEnabled(Extension::ShaderFloat16));
ComPtr<IDxcOperationResult> result;
DAWN_TRY(CheckHRESULT(
dxcCompiler->Compile(sourceBlob.Get(), nullptr, entryPointW.c_str(),
device->GetDeviceInfo().shaderProfiles[stage].c_str(),
arguments.data(), arguments.size(), nullptr, 0, nullptr, &result),
"DXC compile"));
HRESULT hr;
DAWN_TRY(CheckHRESULT(result->GetStatus(&hr), "DXC get status"));
if (FAILED(hr)) {
ComPtr<IDxcBlobEncoding> errors;
DAWN_TRY(CheckHRESULT(result->GetErrorBuffer(&errors), "DXC get error buffer"));
std::string message = std::string("DXC compile failed with ") +
static_cast<char*>(errors->GetBufferPointer());
return DAWN_INTERNAL_ERROR(message);
}
ComPtr<IDxcBlob> compiledShader;
DAWN_TRY(CheckHRESULT(result->GetResult(&compiledShader), "DXC get result"));
return std::move(compiledShader);
}
ResultOrError<ComPtr<ID3DBlob>> CompileShaderFXC(Device* device,
SingleShaderStage stage,
const std::string& hlslSource,
const char* entryPoint,
uint32_t compileFlags) {
const char* targetProfile = nullptr;
switch (stage) {
case SingleShaderStage::Vertex:
targetProfile = "vs_5_1";
break;
case SingleShaderStage::Fragment:
targetProfile = "ps_5_1";
break;
case SingleShaderStage::Compute:
targetProfile = "cs_5_1";
break;
}
ComPtr<ID3DBlob> compiledShader;
ComPtr<ID3DBlob> errors;
const PlatformFunctions* functions = device->GetFunctions();
if (FAILED(functions->d3dCompile(hlslSource.c_str(), hlslSource.length(), nullptr, nullptr,
nullptr, entryPoint, targetProfile, compileFlags, 0,
&compiledShader, &errors))) {
std::string message = std::string("D3D compile failed with ") +
static_cast<char*>(errors->GetBufferPointer());
return DAWN_INTERNAL_ERROR(message);
}
return std::move(compiledShader);
}
// static
ResultOrError<ShaderModule*> ShaderModule::Create(Device* device,
const ShaderModuleDescriptor* descriptor,
ShaderModuleParseResult* parseResult) {
Ref<ShaderModule> module = AcquireRef(new ShaderModule(device, descriptor));
DAWN_TRY(module->Initialize(parseResult));
return module.Detach();
}
ShaderModule::ShaderModule(Device* device, const ShaderModuleDescriptor* descriptor)
: ShaderModuleBase(device, descriptor) {
}
MaybeError ShaderModule::Initialize(ShaderModuleParseResult* parseResult) {
DAWN_TRY(InitializeBase(parseResult));
#ifdef DAWN_ENABLE_WGSL
mTintModule = std::move(parseResult->tintModule);
#endif
return {};
}
ResultOrError<std::string> ShaderModule::TranslateToHLSLWithTint(
const char* entryPointName,
SingleShaderStage stage,
PipelineLayout* layout,
std::string* remappedEntryPointName,
FirstOffsetInfo* firstOffsetInfo) const {
ASSERT(!IsError());
#ifdef DAWN_ENABLE_WGSL
std::ostringstream errorStream;
errorStream << "Tint HLSL failure:" << std::endl;
tint::transform::Manager transformManager;
transformManager.append(std::make_unique<tint::transform::BoundArrayAccessors>());
tint::transform::FirstIndexOffset* firstOffsetTransform = nullptr;
if (stage == SingleShaderStage::Vertex) {
auto transformer = std::make_unique<tint::transform::FirstIndexOffset>(
layout->GetFirstIndexOffsetShaderRegister(),
layout->GetFirstIndexOffsetRegisterSpace());
firstOffsetTransform = transformer.get();
transformManager.append(std::move(transformer));
}
tint::ast::Module module;
DAWN_TRY_ASSIGN(module, RunTransforms(&transformManager, mTintModule.get()));
if (firstOffsetTransform != nullptr) {
// Functions are only available after transform has been performed
firstOffsetInfo->usesVertexIndex = firstOffsetTransform->HasVertexIndex();
if (firstOffsetInfo->usesVertexIndex) {
firstOffsetInfo->vertexIndexOffset = firstOffsetTransform->GetFirstVertexOffset();
}
firstOffsetInfo->usesInstanceIndex = firstOffsetTransform->HasInstanceIndex();
if (firstOffsetInfo->usesInstanceIndex) {
firstOffsetInfo->instanceIndexOffset =
firstOffsetTransform->GetFirstInstanceOffset();
}
}
ASSERT(remappedEntryPointName != nullptr);
tint::inspector::Inspector inspector(module);
*remappedEntryPointName = inspector.GetRemappedNameForEntryPoint(entryPointName);
tint::writer::hlsl::Generator generator(std::move(module));
// TODO: Switch to GenerateEntryPoint once HLSL writer supports it.
if (!generator.Generate()) {
errorStream << "Generator: " << generator.error() << std::endl;
return DAWN_VALIDATION_ERROR(errorStream.str().c_str());
}
return generator.result();
#else
return DAWN_VALIDATION_ERROR("Using Tint to generate HLSL is not supported.");
#endif // DAWN_ENABLE_WGSL
}
ResultOrError<std::string> ShaderModule::TranslateToHLSLWithSPIRVCross(
const char* entryPointName,
SingleShaderStage stage,
PipelineLayout* layout) const {
ASSERT(!IsError());
// If these options are changed, the values in DawnSPIRVCrossHLSLFastFuzzer.cpp need to
// be updated.
spirv_cross::CompilerGLSL::Options options_glsl;
// Force all uninitialized variables to be 0, otherwise they will fail to compile
// by FXC.
options_glsl.force_zero_initialized_variables = true;
spirv_cross::CompilerHLSL::Options options_hlsl;
if (GetDevice()->IsExtensionEnabled(Extension::ShaderFloat16)) {
options_hlsl.shader_model = ToBackend(GetDevice())->GetDeviceInfo().shaderModel;
options_hlsl.enable_16bit_types = true;
} else {
options_hlsl.shader_model = 51;
}
// PointCoord and PointSize are not supported in HLSL
// TODO (hao.x.li@intel.com): The point_coord_compat and point_size_compat are
// required temporarily for https://bugs.chromium.org/p/dawn/issues/detail?id=146,
// but should be removed once WebGPU requires there is no gl_PointSize builtin.
// See https://github.com/gpuweb/gpuweb/issues/332
options_hlsl.point_coord_compat = true;
options_hlsl.point_size_compat = true;
options_hlsl.nonwritable_uav_texture_as_srv = true;
spirv_cross::CompilerHLSL compiler(GetSpirv());
compiler.set_common_options(options_glsl);
compiler.set_hlsl_options(options_hlsl);
compiler.set_entry_point(entryPointName, ShaderStageToExecutionModel(stage));
const EntryPointMetadata::BindingInfo& moduleBindingInfo =
GetEntryPoint(entryPointName).bindings;
for (BindGroupIndex group : IterateBitSet(layout->GetBindGroupLayoutsMask())) {
const BindGroupLayout* bgl = ToBackend(layout->GetBindGroupLayout(group));
const auto& bindingOffsets = bgl->GetBindingOffsets();
const auto& groupBindingInfo = moduleBindingInfo[group];
for (const auto& it : groupBindingInfo) {
const EntryPointMetadata::ShaderBindingInfo& bindingInfo = it.second;
BindingNumber bindingNumber = it.first;
BindingIndex bindingIndex = bgl->GetBindingIndex(bindingNumber);
// Declaring a read-only storage buffer in HLSL but specifying a storage buffer in
// the BGL produces the wrong output. Force read-only storage buffer bindings to
// be treated as UAV instead of SRV.
const bool forceStorageBufferAsUAV =
(bindingInfo.buffer.type == wgpu::BufferBindingType::ReadOnlyStorage &&
bgl->GetBindingInfo(bindingIndex).buffer.type ==
wgpu::BufferBindingType::Storage);
uint32_t bindingOffset = bindingOffsets[bindingIndex];
compiler.set_decoration(bindingInfo.id, spv::DecorationBinding, bindingOffset);
if (forceStorageBufferAsUAV) {
compiler.set_hlsl_force_storage_buffer_as_uav(
static_cast<uint32_t>(group), static_cast<uint32_t>(bindingNumber));
}
}
}
return compiler.compile();
}
ResultOrError<CompiledShader> ShaderModule::Compile(const char* entryPointName,
SingleShaderStage stage,
PipelineLayout* layout,
uint32_t compileFlags) {
Device* device = ToBackend(GetDevice());
// Compile the source shader to HLSL.
std::string hlslSource;
std::string remappedEntryPoint;
CompiledShader compiledShader = {};
if (device->IsToggleEnabled(Toggle::UseTintGenerator)) {
DAWN_TRY_ASSIGN(hlslSource, TranslateToHLSLWithTint(entryPointName, stage, layout,
&remappedEntryPoint,
&compiledShader.firstOffsetInfo));
entryPointName = remappedEntryPoint.c_str();
} else {
DAWN_TRY_ASSIGN(hlslSource,
TranslateToHLSLWithSPIRVCross(entryPointName, stage, layout));
// Note that the HLSL will always use entryPoint "main" under
// SPIRV-cross.
entryPointName = "main";
}
// Use HLSL source as the input for the key since it does need to know about the pipeline
// layout. The pipeline layout is only required if we key from WGSL: two different pipeline
// layouts could be used to produce different shader blobs and the wrong shader blob could
// be loaded since the pipeline layout was missing from the key.
// The compiler flags or version used could also produce different HLSL source. HLSL key
// needs both to ensure the shader cache key is unique to the HLSL source.
// TODO(dawn:549): Consider keying from WGSL and serialize the pipeline layout it used.
PersistentCacheKey shaderCacheKey;
DAWN_TRY_ASSIGN(shaderCacheKey,
CreateHLSLKey(entryPointName, stage, hlslSource, compileFlags));
DAWN_TRY_ASSIGN(compiledShader.cachedShader,
device->GetPersistentCache()->GetOrCreate(
shaderCacheKey, [&](auto doCache) -> MaybeError {
if (device->IsToggleEnabled(Toggle::UseDXC)) {
DAWN_TRY_ASSIGN(compiledShader.compiledDXCShader,
CompileShaderDXC(device, stage, hlslSource,
entryPointName, compileFlags));
} else {
DAWN_TRY_ASSIGN(compiledShader.compiledFXCShader,
CompileShaderFXC(device, stage, hlslSource,
entryPointName, compileFlags));
}
const D3D12_SHADER_BYTECODE shader =
compiledShader.GetD3D12ShaderBytecode();
doCache(shader.pShaderBytecode, shader.BytecodeLength);
return {};
}));
return std::move(compiledShader);
}
D3D12_SHADER_BYTECODE CompiledShader::GetD3D12ShaderBytecode() const {
if (cachedShader.buffer != nullptr) {
return {cachedShader.buffer.get(), cachedShader.bufferSize};
} else if (compiledFXCShader != nullptr) {
return {compiledFXCShader->GetBufferPointer(), compiledFXCShader->GetBufferSize()};
} else if (compiledDXCShader != nullptr) {
return {compiledDXCShader->GetBufferPointer(), compiledDXCShader->GetBufferSize()};
}
UNREACHABLE();
return {};
}
ResultOrError<PersistentCacheKey> ShaderModule::CreateHLSLKey(const char* entryPointName,
SingleShaderStage stage,
const std::string& hlslSource,
uint32_t compileFlags) const {
std::stringstream stream;
// Prefix the key with the type to avoid collisions from another type that could have the
// same key.
stream << static_cast<uint32_t>(PersistentKeyType::Shader);
// Provide "guard" strings that the user cannot provide to help ensure the generated HLSL
// used to create this key is not being manufactured by the user to load the wrong shader
// blob.
// These strings can be HLSL comments because Tint does not emit HLSL comments.
// TODO(dawn:549): Replace guards strings with something more secure.
constexpr char kStartGuard[] = "// Start shader autogenerated by Dawn.";
constexpr char kEndGuard[] = "// End shader autogenerated by Dawn.";
ASSERT(hlslSource.find(kStartGuard) == std::string::npos);
ASSERT(hlslSource.find(kEndGuard) == std::string::npos);
stream << kStartGuard << "\n";
stream << hlslSource;
stream << "\n" << kEndGuard;
stream << compileFlags;
// Add the HLSL compiler version for good measure.
// Prepend the compiler name to ensure the version is always unique.
if (GetDevice()->IsToggleEnabled(Toggle::UseDXC)) {
uint64_t dxCompilerVersion;
DAWN_TRY_ASSIGN(dxCompilerVersion, GetDXCompilerVersion());
stream << "DXC" << dxCompilerVersion;
} else {
stream << "FXC" << GetD3DCompilerVersion();
}
// If the source contains multiple entry points, ensure they are cached seperately
// per stage since DX shader code can only be compiled per stage using the same
// entry point.
stream << static_cast<uint32_t>(stage);
stream << entryPointName;
return PersistentCacheKey(std::istreambuf_iterator<char>{stream},
std::istreambuf_iterator<char>{});
}
ResultOrError<uint64_t> ShaderModule::GetDXCompilerVersion() const {
ComPtr<IDxcValidator> dxcValidator;
DAWN_TRY_ASSIGN(dxcValidator, ToBackend(GetDevice())->GetOrCreateDxcValidator());
ComPtr<IDxcVersionInfo> versionInfo;
DAWN_TRY(CheckHRESULT(dxcValidator.As(&versionInfo),
"D3D12 QueryInterface IDxcValidator to IDxcVersionInfo"));
uint32_t compilerMajor, compilerMinor;
DAWN_TRY(CheckHRESULT(versionInfo->GetVersion(&compilerMajor, &compilerMinor),
"IDxcVersionInfo::GetVersion"));
// Pack both into a single version number.
return (uint64_t(compilerMajor) << uint64_t(32)) + compilerMinor;
}
uint64_t ShaderModule::GetD3DCompilerVersion() const {
return D3D_COMPILER_VERSION;
}
}} // namespace dawn_native::d3d12