godotengine/godot-angle-static
1
0
Fork 0
mirror of https://github.com/godotengine/godot-angle-static.git synced 2026-01-06 02:09:55 +03:00
Code Issues Packages Projects Releases Wiki Activity

Metal: Increase the size of the default uniform block

Browse Source 
Use a buffer pool per shader type and size it to be a nice multiple of
the block that will be allocated for each uniform update.

Set the buffer pool size to unbounded. When the pool runs out of
buffers it currently does a full CPU/GPU sync which is catastrophic
for performance when uniforms are updated every draw call.

Bug: angleproject:8301
Change-Id: I7f826feb4e766fd1cf1b2db01954af7089e3fe90
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4850950
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>
This commit is contained in:
Geoff Lang
2023-09-07 16:48:22 -04:00
committed by Angle LUCI CQ
parent a412b149d5
commit 99d39241ad
2 changed files with 32 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/libANGLE/renderer/metal/ProgramExecutableMtl.h
View File

@@ -166,7 +166,7 @@ class ProgramExecutableMtl : public ProgramExecutableImpl
const gl::ShaderMap<size_t> &requiredBufferSize);
void initUniformBlocksRemapper(const gl::SharedCompiledShaderState &shader);
mtl::BufferPool *getBufferPool(ContextMtl *context);
mtl::BufferPool *getBufferPool(ContextMtl *context, gl::ShaderType shaderType);
angle::Result getSpecializedShader(ContextMtl *context,
gl::ShaderType shaderType,
@@ -234,7 +234,7 @@ class ProgramExecutableMtl : public ProgramExecutableImpl
uint32_t mShadowCompareModes[mtl::kMaxShaderSamplers];
mtl::BufferPool *mAuxBufferPool;
gl::ShaderMap<std::unique_ptr<mtl::BufferPool>> mDefaultUniformBufferPools;
};
angle::Result CreateMslShaderLib(ContextMtl *context,

59
src/libANGLE/renderer/metal/ProgramExecutableMtl.mm
View File

@@ -317,10 +317,7 @@ DefaultUniformBlockMtl::DefaultUniformBlockMtl() {}
DefaultUniformBlockMtl::~DefaultUniformBlockMtl() = default;
ProgramExecutableMtl::ProgramExecutableMtl(const gl::ProgramExecutable *executable)
: ProgramExecutableImpl(executable),
mProgramHasFlatAttributes(false),
mShadowCompareModes{},
mAuxBufferPool(nullptr)
: ProgramExecutableImpl(executable), mProgramHasFlatAttributes(false), mShadowCompareModes{}
{}
ProgramExecutableMtl::~ProgramExecutableMtl() {}
@@ -328,12 +325,6 @@ ProgramExecutableMtl::~ProgramExecutableMtl() {}
void ProgramExecutableMtl::destroy(const gl::Context *context)
{
auto contextMtl = mtl::GetImpl(context);
if (mAuxBufferPool)
{
mAuxBufferPool->destroy(contextMtl);
delete mAuxBufferPool;
mAuxBufferPool = nullptr;
}
reset(contextMtl);
}
@@ -362,14 +353,12 @@ void ProgramExecutableMtl::reset(ContextMtl *context)
var.reset(context);
}
if (mAuxBufferPool)
for (gl::ShaderType shaderType : gl::kAllGLES2ShaderTypes)
{
if (mAuxBufferPool->reset(context, mtl::kDefaultUniformsMaxSize * 2,
mtl::kUniformBufferSettingOffsetMinAlignment,
3) != angle::Result::Continue)
if (mDefaultUniformBufferPools[shaderType])
{
mAuxBufferPool->destroy(context);
SafeDelete(mAuxBufferPool);
mDefaultUniformBufferPools[shaderType]->destroy(context);
mDefaultUniformBufferPools[shaderType].reset();
}
}
}
@@ -827,15 +816,27 @@ void ProgramExecutableMtl::initUniformBlocksRemapper(const gl::SharedCompiledSha
mUniformBlockConversions.insert(conversionMap.begin(), conversionMap.end());
}
mtl::BufferPool *ProgramExecutableMtl::getBufferPool(ContextMtl *context)
mtl::BufferPool *ProgramExecutableMtl::getBufferPool(ContextMtl *context, gl::ShaderType shaderType)
{
if (mAuxBufferPool == nullptr)
auto &pool = mDefaultUniformBufferPools[shaderType];
if (pool == nullptr)
{
mAuxBufferPool = new mtl::BufferPool(true);
mAuxBufferPool->initialize(context, mtl::kDefaultUniformsMaxSize * 2,
mtl::kUniformBufferSettingOffsetMinAlignment, 3);
DefaultUniformBlockMtl &uniformBlock = mDefaultUniformBlocks[shaderType];
// Size each buffer to hold 10 draw calls worth of uniform updates before creating extra
// buffers. This number was chosen loosely to balance the size of buffers versus the total
// number allocated. Without any sub-allocation, the total buffer count can reach the
// thousands when many draw calls are issued with the same program.
size_t bufferSize =
std::max(uniformBlock.uniformData.size() * 10, mtl::kDefaultUniformsMaxSize * 2);
pool.reset(new mtl::BufferPool(false));
// Allow unbounded growth of the buffer count. Doing a full CPU/GPU sync waiting for new
// uniform uploads has catastrophic performance cost.
pool->initialize(context, bufferSize, mtl::kUniformBufferSettingOffsetMinAlignment, 0);
}
return mAuxBufferPool;
return pool.get();
}
angle::Result ProgramExecutableMtl::setupDraw(const gl::Context *glContext,
@@ -1039,10 +1040,7 @@ angle::Result ProgramExecutableMtl::commitUniforms(ContextMtl *context,
{
continue;
}
if (mAuxBufferPool)
{
mAuxBufferPool->releaseInFlightBuffers(context);
}
// If we exceed the default inline max size, try to allocate a buffer
bool needsCommitUniform = true;
if (needsCommitUniform && uniformBlock.uniformData.size() <= mtl::kInlineConstDataMaxSize)
@@ -1054,17 +1052,20 @@ angle::Result ProgramExecutableMtl::commitUniforms(ContextMtl *context,
}
else if (needsCommitUniform)
{
mtl::BufferPool *bufferPool = getBufferPool(context, shaderType);
bufferPool->releaseInFlightBuffers(context);
ASSERT(uniformBlock.uniformData.size() <= mtl::kDefaultUniformsMaxSize);
mtl::BufferRef mtlBufferOut;
size_t offsetOut;
uint8_t *ptrOut;
// Allocate a new Uniform buffer
ANGLE_TRY(getBufferPool(context)->allocate(context, uniformBlock.uniformData.size(),
&ptrOut, &mtlBufferOut, &offsetOut));
ANGLE_TRY(bufferPool->allocate(context, uniformBlock.uniformData.size(), &ptrOut,
&mtlBufferOut, &offsetOut));
// Copy the uniform result
memcpy(ptrOut, uniformBlock.uniformData.data(), uniformBlock.uniformData.size());
// Commit
ANGLE_TRY(getBufferPool(context)->commit(context));
ANGLE_TRY(bufferPool->commit(context));
// Set buffer
cmdEncoder->setBuffer(shaderType, mtlBufferOut, (uint32_t)offsetOut,
mtl::kDefaultUniformsBindingIndex);