Files
UnrealEngineUWP/Engine/Source/Runtime/RenderCore/Private/ShaderResource.cpp
kenzo terelst 5cffab66a5 PSO Precache Base Changes (disabled by default)
Add PSOPrecache file which wraps engine level PSO structs and functions
- FPSOPrecacheParams used to wrap certain parameters which drive the PSO collection (mostly set from components)
- IPSOCollector interface used as engine entry to the mesh pass processors in renderer project
- FPSOCollectorCreateManager to wrap all statically registered IPSOCollector CreateFuncions
- PSOCollectorStats structs and helper functions

Add PSOPrecache support to PipelineStateCache:
- functions to precache compute and graphics PSOs
- functions to check current precache state
- FPrecacheGraphicsPipelineCache which wraps all currently precached PSOs

IMaterial entry function to PrecachePSOs with given vertex factories and params (implemented by UMaterial & UMaterialInstance) which forwards call to FMaterial and finally the FMaterialShaderMap
- FMaterialShaderMap will iterate all registered IPSOCollector to collect all possible PSOs used and forward them to the PipelineStateCache for actual (async) precaching

RHI Changes:
- Add RHI specific functions to retrieve & compare the FRHIVertexDeclaration and FGraphicsPipelineStateInitializer hash for each RHI
- Add helper functions to match RHI*State data by retrieving initializer

Minimal vertex factory changes for PSO precaching as prep - actual changes in upcoming CLs

MeshPassProcessor base functions to collect PSOs which can be used by all mesh pass processor implemention (actual changes in upcoming CLs)
- Cache complete precache data hash of stored minimal PSO so it can be used for fast actual PSO initializer hash computation (full PSO initializer is build at runtime)
- Cache if minimal PSO was successfully precache and option to skip draw when PSO is still precaching (disabled by default and doesn't work yet via GPU scene based rendering)
- Add helper functions to setup the per mesh pass render target data used during PSO collection

Refactor SceneTextureConfig so it can be initialized from an init struct and move all function to setup the internal members from SceneTexture to SceneTextureConfig using the data from the init struct (so it's available in Engine project)

Fix GBufferInfo format for velocity target (was always floating point while it should be unorm when not android

Make certain helper functions definitions to h file so they can be used everywhere (SupportsNaniteRendering, FTextureRenderTarget2DResource::GetFormat & CreateFlags)

#preflight 63189a5a967ffc68fb9044a5
#jira UE-139584
#rb Mihnea.Balta

[CL 21907508 by kenzo terelst in ue5-main branch]
2022-09-08 19:34:08 -04:00

703 lines
23 KiB
C++

// Copyright Epic Games, Inc. All Rights Reserved.
/*=============================================================================
ShaderResource.cpp: ShaderResource implementation.
=============================================================================*/
#include "Shader.h"
#include "Misc/CoreMisc.h"
#include "Misc/StringBuilder.h"
#include "Stats/StatsMisc.h"
#include "Serialization/MemoryWriter.h"
#include "VertexFactory.h"
#include "ProfilingDebugging/DiagnosticTable.h"
#include "Interfaces/ITargetPlatform.h"
#include "Interfaces/ITargetPlatformManagerModule.h"
#include "Interfaces/IShaderFormat.h"
#include "ShaderCodeLibrary.h"
#include "ShaderCore.h"
#include "RenderUtils.h"
#include "Misc/ConfigCacheIni.h"
#include "Misc/ScopeLock.h"
#include "UObject/RenderingObjectVersion.h"
#include "UObject/FortniteMainBranchObjectVersion.h"
#include "ProfilingDebugging/LoadTimeTracker.h"
#include "Misc/MemStack.h"
#include "ShaderCompilerCore.h"
#include "Compression/OodleDataCompression.h"
#if WITH_EDITORONLY_DATA
#include "Interfaces/IShaderFormat.h"
#endif
DECLARE_LOG_CATEGORY_CLASS(LogShaderWarnings, Log, Log);
int32 GShaderCompressionFormatChoice = 2;
static FAutoConsoleVariableRef CVarShaderCompressionFormatChoice(
TEXT("r.Shaders.CompressionFormat"),
GShaderCompressionFormatChoice,
TEXT("Select the compression methods for the shader code.\n")
TEXT(" 0: None (uncompressed)\n")
TEXT(" 1: LZ4\n")
TEXT(" 2: Oodle (default)\n")
TEXT(" 3: ZLib\n"),
ECVF_ReadOnly);
int32 GShaderCompressionOodleAlgo = 2;
static FAutoConsoleVariableRef CVarShaderCompressionOodleAlgo(
TEXT("r.Shaders.CompressionFormat.Oodle.Algo"),
GShaderCompressionOodleAlgo,
TEXT("Oodle compression method for the shader code, from fastest to slowest to decode.\n")
TEXT(" 0: None (invalid setting)\n")
TEXT(" 1: Selkie (fastest to decode)\n")
TEXT(" 2: Mermaid\n")
TEXT(" 3: Kraken\n")
TEXT(" 4: Leviathan (slowest to decode)\n"),
ECVF_ReadOnly);
int32 GShaderCompressionOodleLevel = 6;
static FAutoConsoleVariableRef CVarShaderCompressionOodleAlgoChoice(
TEXT("r.Shaders.CompressionFormat.Oodle.Level"),
GShaderCompressionOodleLevel,
TEXT("Oodle compression level. This mostly trades encode speed vs compression ratio, decode speed is determined by r.Shaders.CompressionFormat.Oodle.Algo\n")
TEXT(" -4 : HyperFast4\n")
TEXT(" -3 : HyperFast3\n")
TEXT(" -2 : HyperFast2\n")
TEXT(" -1 : HyperFast1\n")
TEXT(" 0 : None\n")
TEXT(" 1 : SuperFast\n")
TEXT(" 2 : VeryFast\n")
TEXT(" 3 : Fast\n")
TEXT(" 4 : Normal\n")
TEXT(" 5 : Optimal1\n")
TEXT(" 6 : Optimal2\n")
TEXT(" 7 : Optimal3\n")
TEXT(" 8 : Optimal4\n"),
ECVF_ReadOnly);
static int32 GShaderCompilerEmitWarningsOnLoad = 0;
static FAutoConsoleVariableRef CVarShaderCompilerEmitWarningsOnLoad(
TEXT("r.ShaderCompiler.EmitWarningsOnLoad"),
GShaderCompilerEmitWarningsOnLoad,
TEXT("When 1, shader compiler warnings are emitted to the log for all shaders as they are loaded."),
ECVF_Default
);
FName GetShaderCompressionFormat(const FName& ShaderFormat)
{
// support an older developer-only CVar for compatibility and make it preempt
#if !(UE_BUILD_SHIPPING || UE_BUILD_TEST)
static const IConsoleVariable* CVarSkipCompression = IConsoleManager::Get().FindConsoleVariable(TEXT("r.Shaders.SkipCompression"));
static bool bSkipCompression = (CVarSkipCompression && CVarSkipCompression->GetInt() != 0);
if (UNLIKELY(bSkipCompression))
{
return NAME_None;
}
#endif
static FName Formats[]
{
NAME_None,
NAME_LZ4,
NAME_Oodle,
NAME_Zlib
};
//GShaderCompressionFormatChoice = (GShaderCompressionFormatChoice < 0) ? 0 : GShaderCompressionFormatChoice;
GShaderCompressionFormatChoice = FMath::Clamp<int32>(GShaderCompressionFormatChoice, 0, UE_ARRAY_COUNT(Formats) - 1);
return Formats[GShaderCompressionFormatChoice];
}
void GetShaderCompressionOodleSettings(FOodleDataCompression::ECompressor& OutCompressor, FOodleDataCompression::ECompressionLevel& OutLevel, const FName& ShaderFormat)
{
// support an older developer-only CVar for compatibility and make it preempt
#if !(UE_BUILD_SHIPPING || UE_BUILD_TEST)
static const IConsoleVariable* CVarSkipCompression = IConsoleManager::Get().FindConsoleVariable(TEXT("r.Shaders.SkipCompression"));
static bool bSkipCompression = (CVarSkipCompression && CVarSkipCompression->GetInt() != 0);
if (UNLIKELY(bSkipCompression))
{
OutCompressor = FOodleDataCompression::ECompressor::Selkie;
OutLevel = FOodleDataCompression::ECompressionLevel::None;
return;
}
#endif
GShaderCompressionOodleAlgo = FMath::Clamp(GShaderCompressionOodleAlgo, static_cast<int32>(FOodleDataCompression::ECompressor::NotSet), static_cast<int32>(FOodleDataCompression::ECompressor::Leviathan));
OutCompressor = static_cast<FOodleDataCompression::ECompressor>(GShaderCompressionOodleAlgo);
GShaderCompressionOodleLevel = FMath::Clamp(GShaderCompressionOodleLevel, static_cast<int32>(FOodleDataCompression::ECompressionLevel::HyperFast4), static_cast<int32>(FOodleDataCompression::ECompressionLevel::Optimal4));
OutLevel = static_cast<FOodleDataCompression::ECompressionLevel>(GShaderCompressionOodleLevel);
}
bool FShaderMapResource::ArePlatformsCompatible(EShaderPlatform CurrentPlatform, EShaderPlatform TargetPlatform)
{
bool bFeatureLevelCompatible = CurrentPlatform == TargetPlatform;
if (!bFeatureLevelCompatible && IsPCPlatform(CurrentPlatform) && IsPCPlatform(TargetPlatform))
{
bFeatureLevelCompatible = GetMaxSupportedFeatureLevel(CurrentPlatform) >= GetMaxSupportedFeatureLevel(TargetPlatform);
bool const bIsTargetD3D = IsD3DPlatform(TargetPlatform);
bool const bIsCurrentPlatformD3D = IsD3DPlatform(CurrentPlatform);
// For Metal in Editor we can switch feature-levels, but not in cooked projects when using Metal shader librariss.
bool const bIsCurrentMetal = IsMetalPlatform(CurrentPlatform);
bool const bIsTargetMetal = IsMetalPlatform(TargetPlatform);
bool const bIsMetalCompatible = (bIsCurrentMetal == bIsTargetMetal)
#if !WITH_EDITOR // Static analysis doesn't like (|| WITH_EDITOR)
&& (!IsMetalPlatform(CurrentPlatform) || (CurrentPlatform == TargetPlatform))
#endif
;
bool const bIsCurrentOpenGL = IsOpenGLPlatform(CurrentPlatform);
bool const bIsTargetOpenGL = IsOpenGLPlatform(TargetPlatform);
bFeatureLevelCompatible = bFeatureLevelCompatible && (bIsCurrentPlatformD3D == bIsTargetD3D && bIsMetalCompatible && bIsCurrentOpenGL == bIsTargetOpenGL);
}
return bFeatureLevelCompatible;
}
#if RHI_RAYTRACING
class FRayTracingShaderLibrary
{
public:
uint32 AddShader(FRHIRayTracingShader* Shader)
{
FScopeLock Lock(&CS);
if (UnusedIndicies.Num() != 0)
{
uint32 Index = UnusedIndicies.Pop(false);
checkSlow(Shaders[Index] == nullptr);
Shaders[Index] = Shader;
return Index;
}
else
{
Shaders.Add(Shader);
return Shaders.Num() - 1;
}
}
void RemoveShader(uint32 Index)
{
if (Index != ~0u)
{
FScopeLock Lock(&CS);
UnusedIndicies.Push(Index);
Shaders[Index] = nullptr;
}
}
void GetShaders(TArray<FRHIRayTracingShader*>& OutShaders, FRHIRayTracingShader* DefaultShader)
{
FScopeLock Lock(&CS);
OutShaders = Shaders;
for (uint32 Index : UnusedIndicies)
{
OutShaders[Index] = DefaultShader;
}
}
private:
TArray<uint32> UnusedIndicies;
TArray<FRHIRayTracingShader*> Shaders;
FCriticalSection CS;
};
static FRayTracingShaderLibrary GlobalRayTracingHitGroupLibrary;
static FRayTracingShaderLibrary GlobalRayTracingCallableShaderLibrary;
static FRayTracingShaderLibrary GlobalRayTracingMissShaderLibrary;
void FShaderMapResource::GetRayTracingHitGroupLibrary(TArray<FRHIRayTracingShader*>& RayTracingShaders, FRHIRayTracingShader* DefaultShader)
{
GlobalRayTracingHitGroupLibrary.GetShaders(RayTracingShaders, DefaultShader);
}
void FShaderMapResource::GetRayTracingCallableShaderLibrary(TArray<FRHIRayTracingShader*>& RayTracingCallableShaders, FRHIRayTracingShader* DefaultShader)
{
GlobalRayTracingCallableShaderLibrary.GetShaders(RayTracingCallableShaders, DefaultShader);
}
void FShaderMapResource::GetRayTracingMissShaderLibrary(TArray<FRHIRayTracingShader*>& RayTracingMissShaders, FRHIRayTracingShader* DefaultShader)
{
GlobalRayTracingMissShaderLibrary.GetShaders(RayTracingMissShaders, DefaultShader);
}
#endif // RHI_RAYTRACING
static void ApplyResourceStats(FShaderMapResourceCode& Resource)
{
#if STATS
INC_DWORD_STAT_BY(STAT_Shaders_ShaderResourceMemory, Resource.GetSizeBytes());
for (const FShaderMapResourceCode::FShaderEntry& Shader : Resource.ShaderEntries)
{
INC_DWORD_STAT_BY_FName(GetMemoryStatType(Shader.Frequency).GetName(), Shader.Code.Num());
}
#endif // STATS
}
static void RemoveResourceStats(FShaderMapResourceCode& Resource)
{
#if STATS
DEC_DWORD_STAT_BY(STAT_Shaders_ShaderResourceMemory, Resource.GetSizeBytes());
for (const FShaderMapResourceCode::FShaderEntry& Shader : Resource.ShaderEntries)
{
DEC_DWORD_STAT_BY_FName(GetMemoryStatType(Shader.Frequency).GetName(), Shader.Code.Num());
}
#endif // STATS
}
FShaderMapResourceCode::FShaderMapResourceCode(const FShaderMapResourceCode& Other)
{
ResourceHash = Other.ResourceHash;
ShaderHashes = Other.ShaderHashes;
ShaderEntries = Other.ShaderEntries;
#if WITH_EDITORONLY_DATA
PlatformDebugData = Other.PlatformDebugData;
PlatformDebugDataHashes = Other.PlatformDebugDataHashes;
#endif // WITH_EDITORONLY_DATA
}
FShaderMapResourceCode::~FShaderMapResourceCode()
{
RemoveResourceStats(*this);
}
void FShaderMapResourceCode::Finalize()
{
FSHA1 Hasher;
Hasher.Update((uint8*)ShaderHashes.GetData(), ShaderHashes.Num() * sizeof(FSHAHash));
Hasher.Final();
Hasher.GetHash(ResourceHash.Hash);
ApplyResourceStats(*this);
#if WITH_EDITORONLY_DATA
LogShaderCompilerWarnings();
#endif
}
uint32 FShaderMapResourceCode::GetSizeBytes() const
{
uint64 Size = sizeof(*this) + ShaderHashes.GetAllocatedSize() + ShaderEntries.GetAllocatedSize();
for (const FShaderEntry& Entry : ShaderEntries)
{
Size += Entry.Code.GetAllocatedSize();
}
check(Size <= TNumericLimits<uint32>::Max());
return static_cast<uint32>(Size);
}
int32 FShaderMapResourceCode::FindShaderIndex(const FSHAHash& InHash) const
{
return Algo::BinarySearch(ShaderHashes, InHash);
}
void FShaderMapResourceCode::AddShaderCompilerOutput(const FShaderCompilerOutput& Output)
{
#if WITH_EDITORONLY_DATA
AddPlatformDebugData(Output.PlatformDebugData);
for (const FShaderCompilerError& Error : Output.Errors)
{
CompilerWarnings.Add(Error.GetErrorString());
}
#endif
AddShaderCode(Output.Target.GetFrequency(), Output.OutputHash, Output.ShaderCode);
}
void FShaderMapResourceCode::AddShaderCode(EShaderFrequency InFrequency, const FSHAHash& InHash, const FShaderCode& InCode)
{
TRACE_CPUPROFILER_EVENT_SCOPE(FShaderMapResourceCode::AddShaderCode);
const int32 Index = Algo::LowerBound(ShaderHashes, InHash);
if (Index >= ShaderHashes.Num() || ShaderHashes[Index] != InHash)
{
ShaderHashes.Insert(InHash, Index);
FShaderEntry& Entry = ShaderEntries.InsertDefaulted_GetRef(Index);
Entry.Frequency = InFrequency;
const TArray<uint8>& ShaderCode = InCode.GetReadAccess();
FName ShaderCompressionFormat = GetShaderCompressionFormat();
if (ShaderCompressionFormat != NAME_None)
{
Entry.UncompressedSize = InCode.GetUncompressedSize();
// we trust that SCWs also obeyed by the same CVar, so we expect a compressed shader code at this point
// However, if we see an uncompressed shader, it perhaps means that SCW tried to compress it, but the result was worse than uncompressed.
// Because of that we special-case NAME_None here
if (ShaderCompressionFormat != InCode.GetCompressionFormat())
{
if (InCode.GetCompressionFormat() != NAME_None)
{
UE_LOG(LogShaders, Fatal, TEXT("Shader %s is expected to be compressed with %s, but it is compressed with %s instead."),
*InHash.ToString(),
*ShaderCompressionFormat.ToString(),
*InCode.GetCompressionFormat().ToString()
);
// unreachable
return;
}
// assume uncompressed due to worse ratio than the compression
Entry.UncompressedSize = ShaderCode.Num();
UE_LOG(LogShaders, Verbose, TEXT("Shader %s is expected to be compressed with %s, but it arrived uncompressed (size=%d). Assuming compressing made it longer and storing uncompressed."),
*InHash.ToString(),
*ShaderCompressionFormat.ToString(),
ShaderCode.Num()
);
}
else if (ShaderCompressionFormat == NAME_Oodle)
{
// check if Oodle-specific settings match
FOodleDataCompression::ECompressor OodleCompressor;
FOodleDataCompression::ECompressionLevel OodleLevel;
GetShaderCompressionOodleSettings(OodleCompressor, OodleLevel);
if (InCode.GetOodleCompressor() != OodleCompressor || InCode.GetOodleLevel() != OodleLevel)
{
UE_LOG(LogShaders, Fatal, TEXT("Shader %s is expected to be compressed with Oodle compressor %d level %d, but it is compressed with compressor %d level %d instead."),
*InHash.ToString(),
static_cast<int32>(OodleCompressor),
static_cast<int32>(OodleLevel),
static_cast<int32>(InCode.GetOodleCompressor()),
static_cast<int32>(InCode.GetOodleLevel())
);
// unreachable
return;
}
}
}
else
{
Entry.UncompressedSize = ShaderCode.Num();
}
Entry.Code = ShaderCode;
}
}
#if WITH_EDITORONLY_DATA
void FShaderMapResourceCode::AddPlatformDebugData(TConstArrayView<uint8> InPlatformDebugData)
{
if (InPlatformDebugData.Num() == 0)
{
return;
}
FSHAHash Hash;
{
FSHA1 Hasher;
Hasher.Update(InPlatformDebugData.GetData(), InPlatformDebugData.Num());
Hasher.Final();
Hasher.GetHash(Hash.Hash);
}
const int32 Index = Algo::LowerBound(PlatformDebugDataHashes, Hash);
if (Index >= PlatformDebugDataHashes.Num() || PlatformDebugDataHashes[Index] != Hash)
{
PlatformDebugDataHashes.Insert(Hash, Index);
PlatformDebugData.EmplaceAt(Index, InPlatformDebugData.GetData(), InPlatformDebugData.Num());
}
}
void FShaderMapResourceCode::LogShaderCompilerWarnings()
{
if (CompilerWarnings.Num() > 0 && GShaderCompilerEmitWarningsOnLoad != 0)
{
// Emit all the compiler warnings seen whilst serializing/loading this shader to the log.
// Since successfully compiled shaders are stored in the DDC, we'll get the compiler warnings
// even if we didn't compile the shader this run.
for (const FString& CompilerWarning : CompilerWarnings)
{
UE_LOG(LogShaderWarnings, Warning, TEXT("%s"), *CompilerWarning);
}
}
}
#endif // WITH_EDITORONLY_DATA
void FShaderMapResourceCode::ToString(FStringBuilderBase& OutString) const
{
OutString.Appendf(TEXT("Shaders: Num=%d\n"), ShaderHashes.Num());
for (int32 i = 0; i < ShaderHashes.Num(); ++i)
{
const FShaderEntry& Entry = ShaderEntries[i];
OutString.Appendf(TEXT(" [%d]: { Hash: %s, Freq: %s, Size: %d, UncompressedSize: %d }\n"),
i, *ShaderHashes[i].ToString(), GetShaderFrequencyString(Entry.Frequency), Entry.Code.Num(), Entry.UncompressedSize);
}
}
void FShaderMapResourceCode::Serialize(FArchive& Ar, bool bLoadedByCookedMaterial)
{
Ar << ResourceHash;
Ar << ShaderHashes;
Ar << ShaderEntries;
check(ShaderEntries.Num() == ShaderHashes.Num());
#if WITH_EDITORONLY_DATA
const bool bSerializeEditorOnlyData = !bLoadedByCookedMaterial && (!Ar.IsCooking() || Ar.CookingTarget()->HasEditorOnlyData());
if (bSerializeEditorOnlyData)
{
Ar << PlatformDebugDataHashes;
Ar << PlatformDebugData;
Ar << CompilerWarnings;
}
#endif // WITH_EDITORONLY_DATA
ApplyResourceStats(*this);
#if WITH_EDITORONLY_DATA
if (Ar.IsLoading())
{
LogShaderCompilerWarnings();
}
#endif
}
#if WITH_EDITORONLY_DATA
void FShaderMapResourceCode::NotifyShadersCompiled(FName FormatName)
{
#if WITH_ENGINE
// Notify the platform shader format that this particular shader is being used in the cook.
// We discard this data in cooked builds unless Ar.CookingTarget()->HasEditorOnlyData() is true.
if (PlatformDebugData.Num())
{
if (const IShaderFormat* ShaderFormat = GetTargetPlatformManagerRef().FindShaderFormat(FormatName))
{
for (const TArray<uint8>& Entry : PlatformDebugData)
{
ShaderFormat->NotifyShaderCompiled(Entry, FormatName);
}
}
}
#endif // WITH_ENGINE
}
void FShaderMapResourceCode::NotifyShadersCooked(const ITargetPlatform* TargetPlatform)
{
#if WITH_ENGINE
TArray<FName> ShaderFormatNames;
TargetPlatform->GetAllTargetedShaderFormats(ShaderFormatNames);
for (FName FormatName : ShaderFormatNames)
{
NotifyShadersCompiled(FormatName);
}
#endif
}
#endif // WITH_EDITORONLY_DATA
FShaderMapResource::FShaderMapResource(EShaderPlatform InPlatform, int32 NumShaders)
: NumRHIShaders(NumShaders)
, Platform(InPlatform)
, NumRefs(0)
{
RHIShaders = MakeUnique<std::atomic<FRHIShader*>[]>(NumRHIShaders); // this MakeUnique() zero-initializes the array
#if RHI_RAYTRACING
if (GRHISupportsRayTracing && GRHISupportsRayTracingShaders)
{
RayTracingLibraryIndices.AddUninitialized(NumShaders);
FMemory::Memset(RayTracingLibraryIndices.GetData(), 0xff, NumShaders * RayTracingLibraryIndices.GetTypeSize());
}
#endif // RHI_RAYTRACING
}
FShaderMapResource::~FShaderMapResource()
{
ReleaseShaders();
check(NumRefs.load(std::memory_order_relaxed) == 0);
}
void FShaderMapResource::AddRef()
{
NumRefs.fetch_add(1, std::memory_order_relaxed);
}
void FShaderMapResource::Release()
{
check(NumRefs.load(std::memory_order_relaxed) > 0);
if (NumRefs.fetch_sub(1, std::memory_order_release) - 1 == 0 && TryRelease())
{
//check https://www.boost.org/doc/libs/1_55_0/doc/html/atomic/usage_examples.html for explanation
std::atomic_thread_fence(std::memory_order_acquire);
// Send a release message to the rendering thread when the shader loses its last reference.
BeginReleaseResource(this);
BeginCleanup(this);
DEC_DWORD_STAT_BY(STAT_Shaders_ShaderResourceMemory, GetSizeBytes());
}
}
void FShaderMapResource::ReleaseShaders()
{
if (RHIShaders)
{
for (int32 Idx = 0; Idx < NumRHIShaders; ++Idx)
{
if (FRHIShader* Shader = RHIShaders[Idx].load(std::memory_order_acquire))
{
Shader->Release();
DEC_DWORD_STAT(STAT_Shaders_NumShadersUsedForRendering);
}
}
RHIShaders = nullptr;
NumRHIShaders = 0;
}
}
void FShaderMapResource::ReleaseRHI()
{
#if RHI_RAYTRACING
if (GRHISupportsRayTracing && GRHISupportsRayTracingShaders)
{
check(NumRHIShaders == RayTracingLibraryIndices.Num());
for (int32 Idx = 0; Idx < NumRHIShaders; ++Idx)
{
if (FRHIShader* Shader = RHIShaders[Idx].load(std::memory_order_acquire))
{
int32 IndexInLibrary = RayTracingLibraryIndices[Idx];
switch (Shader->GetFrequency())
{
case SF_RayHitGroup:
GlobalRayTracingHitGroupLibrary.RemoveShader(IndexInLibrary);
break;
case SF_RayCallable:
GlobalRayTracingCallableShaderLibrary.RemoveShader(IndexInLibrary);
break;
case SF_RayMiss:
GlobalRayTracingMissShaderLibrary.RemoveShader(IndexInLibrary);
break;
default:
break;
}
}
}
}
RayTracingLibraryIndices.Empty();
#endif // RHI_RAYTRACING
ReleaseShaders();
}
void FShaderMapResource::BeginCreateAllShaders()
{
FShaderMapResource* Resource = this;
ENQUEUE_RENDER_COMMAND(InitCommand)(
[Resource](FRHICommandListImmediate& RHICmdList)
{
for (int32 ShaderIndex = 0; ShaderIndex < Resource->GetNumShaders(); ++ShaderIndex)
{
Resource->GetShader(ShaderIndex);
}
});
}
FRHIShader* FShaderMapResource::CreateShader(int32 ShaderIndex)
{
check(!RHIShaders[ShaderIndex].load(std::memory_order_acquire));
TRefCountPtr<FRHIShader> RHIShader = CreateRHIShader(ShaderIndex);
#if RHI_RAYTRACING
if (GRHISupportsRayTracing && GRHISupportsRayTracingShaders && RHIShader.IsValid())
{
switch (RHIShader->GetFrequency())
{
case SF_RayHitGroup:
RayTracingLibraryIndices[ShaderIndex] = GlobalRayTracingHitGroupLibrary.AddShader(static_cast<FRHIRayTracingShader*>(RHIShader.GetReference()));
break;
case SF_RayCallable:
RayTracingLibraryIndices[ShaderIndex] = GlobalRayTracingCallableShaderLibrary.AddShader(static_cast<FRHIRayTracingShader*>(RHIShader.GetReference()));
break;
case SF_RayMiss:
RayTracingLibraryIndices[ShaderIndex] = GlobalRayTracingMissShaderLibrary.AddShader(static_cast<FRHIRayTracingShader*>(RHIShader.GetReference()));
break;
default:
break;
}
}
#endif // RHI_RAYTRACING
// keep the reference alive (the caller will release)
if (RHIShader.IsValid())
{
RHIShader->AddRef();
}
return RHIShader.GetReference();
}
TRefCountPtr<FRHIShader> FShaderMapResource_InlineCode::CreateRHIShader(int32 ShaderIndex)
{
TRACE_CPUPROFILER_EVENT_SCOPE(FShaderMapResource_InlineCode::CreateRHIShader);
#if STATS
double TimeFunctionEntered = FPlatformTime::Seconds();
ON_SCOPE_EXIT
{
if (IsInRenderingThread())
{
double ShaderCreationTime = FPlatformTime::Seconds() - TimeFunctionEntered;
INC_FLOAT_STAT_BY(STAT_Shaders_TotalRTShaderInitForRenderingTime, ShaderCreationTime);
}
};
#endif
// we can't have this called on the wrong platform's shaders
if (!ArePlatformsCompatible(GMaxRHIShaderPlatform, GetPlatform()))
{
if (FPlatformProperties::RequiresCookedData())
{
UE_LOG(LogShaders, Fatal, TEXT("FShaderMapResource_InlineCode::InitRHI got platform %s but it is not compatible with %s"),
*LegacyShaderPlatformToShaderFormat(GetPlatform()).ToString(), *LegacyShaderPlatformToShaderFormat(GMaxRHIShaderPlatform).ToString());
}
return TRefCountPtr<FRHIShader>();
}
FMemStackBase& MemStack = FMemStack::Get();
const FShaderMapResourceCode::FShaderEntry& ShaderEntry = Code->ShaderEntries[ShaderIndex];
const uint8* ShaderCode = ShaderEntry.Code.GetData();
FMemMark Mark(MemStack);
if (ShaderEntry.Code.Num() != ShaderEntry.UncompressedSize)
{
void* UncompressedCode = MemStack.Alloc(ShaderEntry.UncompressedSize, 16);
bool bSucceed = FCompression::UncompressMemory(GetShaderCompressionFormat(), UncompressedCode, ShaderEntry.UncompressedSize, ShaderCode, ShaderEntry.Code.Num());
check(bSucceed);
ShaderCode = (uint8*)UncompressedCode;
}
const auto ShaderCodeView = MakeArrayView(ShaderCode, ShaderEntry.UncompressedSize);
const FSHAHash& ShaderHash = Code->ShaderHashes[ShaderIndex];
const EShaderFrequency Frequency = ShaderEntry.Frequency;
TRefCountPtr<FRHIShader> RHIShader;
switch (Frequency)
{
case SF_Vertex: RHIShader = RHICreateVertexShader(ShaderCodeView, ShaderHash); break;
case SF_Mesh: RHIShader = RHICreateMeshShader(ShaderCodeView, ShaderHash); break;
case SF_Amplification: RHIShader = RHICreateAmplificationShader(ShaderCodeView, ShaderHash); break;
case SF_Pixel: RHIShader = RHICreatePixelShader(ShaderCodeView, ShaderHash); break;
case SF_Geometry: RHIShader = RHICreateGeometryShader(ShaderCodeView, ShaderHash); break;
case SF_Compute: RHIShader = RHICreateComputeShader(ShaderCodeView, ShaderHash); break;
case SF_RayGen: case SF_RayMiss: case SF_RayHitGroup: case SF_RayCallable:
#if RHI_RAYTRACING
if (GRHISupportsRayTracing && GRHISupportsRayTracingShaders)
{
RHIShader = RHICreateRayTracingShader(ShaderCodeView, ShaderHash, Frequency);
}
#endif // RHI_RAYTRACING
break;
default:
checkNoEntry();
break;
}
if (RHIShader)
{
INC_DWORD_STAT(STAT_Shaders_NumShadersUsedForRendering);
RHIShader->SetHash(ShaderHash);
}
return RHIShader;
}