saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-05-26 22:35:43 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Enable custom heaps based on query

Browse source
This commit is contained in:
Jeff Bloomfield 2023-08-23 11:59:57 -07:00
parent d3345f3680
commit 7520974970
4 changed files with 34 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
onnxruntime/core/providers/dml/DmlExecutionProvider/src/DmlGraphFusionHelper.cpp
View file

@ -217,8 +217,7 @@ namespace DmlGraphFusionHelper
{
ComPtr<ID3D12Resource> initializeInputBuffer;
// D3D_FEATURE_LEVEL_1_0_CORE doesn't support Custom heaps
if (providerImpl->IsMcdmDevice())
if (!providerImpl->CustomHeapsSupported())
{
initializeInputBuffer = CreateResource(providerImpl, tensorPtr, tensorByteSize);
}

30
onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.cpp
View file

@ -181,6 +181,31 @@ namespace Dml
}
m_isMcdmDevice = (featureLevels.MaxSupportedFeatureLevel == D3D_FEATURE_LEVEL_1_0_CORE_PRIVATE);
m_areCustomHeapsSupported = !m_isMcdmDevice;
if (m_isMcdmDevice) {
// TODO: Ingest updated header file
typedef struct D3D12_FEATURE_DATA_D3D12_OPTIONS19
{
BOOL MismatchingOutputDimensionsSupported;
UINT SupportedSampleCountsWithNoOutputs;
BOOL PointSamplingAddressesNeverRoundUp;
BOOL RasterizerDesc2Supported;
BOOL NarrowQuadrilateralLinesSupported;
BOOL AnisoFilterWithPointMipSupported;
UINT MaxSamplerDescriptorHeapSize;
UINT MaxSamplerDescriptorHeapSizeWithStaticSamplers;
UINT MaxViewDescriptorHeapSize;
_Out_ BOOL ComputeOnlyCustomHeapSupported;
} D3D12_FEATURE_DATA_D3D12_OPTIONS19;
D3D12_FEATURE_DATA_D3D12_OPTIONS19 options19 = {};
// The call may fail in which case the default value is false
d3d12Device->CheckFeatureSupport((D3D12_FEATURE) 48 /*D3D12_FEATURE_D3D12_OPTIONS19*/, &options19, sizeof(options19));
m_areCustomHeapsSupported = options19.ComputeOnlyCustomHeapSupported;
}
m_context = std::make_shared<ExecutionContext>(m_d3d12Device.Get(), m_dmlDevice.Get(), queue);
@ -1088,6 +1113,11 @@ namespace Dml
return m_isMcdmDevice;
}
bool __stdcall ExecutionProviderImpl::CustomHeapsSupported() const noexcept
{
return m_areCustomHeapsSupported;
}
bool __stdcall ExecutionProviderImpl::MetacommandsEnabled() const noexcept
{
return m_areMetacommandsEnabled;

2
onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.h
View file

@ -148,6 +148,7 @@ namespace Dml
}
STDMETHOD_(bool, IsMcdmDevice)() const noexcept final;
STDMETHOD_(bool, CustomHeapsSupported)() const noexcept final;
STDMETHOD_(bool, MetacommandsEnabled)() const noexcept final;
std::shared_ptr<onnxruntime::IAllocator> GetGpuAllocator();
@ -183,6 +184,7 @@ namespace Dml
ComPtr<ID3D12Device> m_d3d12Device;
ComPtr<IDMLDevice> m_dmlDevice;
bool m_isMcdmDevice = false;
bool m_areCustomHeapsSupported = false;
bool m_areMetacommandsEnabled = true;
bool m_native16BitShaderOpsSupported = false;
std::shared_ptr<ExecutionContext> m_context;

1
onnxruntime/core/providers/dml/DmlExecutionProvider/src/IExecutionProvider.h
View file

@ -69,6 +69,7 @@ namespace Dml
STDMETHOD(AllocatePooledResource(size_t size, AllocatorRoundingMode roundingMode, ID3D12Resource **d3dResource, IUnknown* *pooledResource)) const noexcept = 0;
STDMETHOD_(bool, IsMcdmDevice)() const noexcept = 0;
STDMETHOD_(bool, CustomHeapsSupported)() const noexcept = 0;
STDMETHOD_(bool, MetacommandsEnabled)() const noexcept = 0;
};
} // namespace Dml