Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[NVIDIA] Set proper math type for convolution #702

Merged
merged 1 commit into from
Aug 10, 2023
+4 −2
Merged

[NVIDIA] Set proper math type for convolution #702

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions modules/nvidia_plugin/src/ops/convolution_components/convolution_cudnn_components.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,7 +74,7 @@ CUDA::DnnConvolutionDescriptor ConvolutionParamsCuDnn::MakeConvolutionDescriptor

// Enable computations on Tensor Core hardware which requires at least Volta GPU
// (compute capability 7.0).
const cudnnMathType_t math_type = CUDNN_TENSOR_OP_MATH_ALLOW_CONVERSION;
const cudnnMathType_t math_type = CUDNN_TENSOR_OP_MATH;
throwIfError(::cudnnSetConvolutionMathType(conv_desc.get(), math_type));
throwIfError(::cudnnSetConvolutionGroupCount(conv_desc.get(), groups_));

Expand All @@ -96,6 +96,7 @@ ConvolutionDescriptorsCuDnn::ConvolutionDescriptorsCuDnn(const CreationContext&
} else {
GetAlgo(dnnHandle);
}
throwIfError(::cudnnSetConvolutionMathType(conv_.get(), algo_perf_.mathType));
}

void ConvolutionDescriptorsCuDnn::BenchmarkOptimalAlgo(const CUDA::DnnHandle& dnnHandle,
Expand Down Expand Up @@ -305,7 +306,7 @@ CUDA::DnnConvolutionDescriptor ConvolutionBackpropDataParamsCuDnn::MakeConvoluti

// Enable computations on Tensor Core hardware which requires at least Volta GPU
// (compute capability 7.0).
const cudnnMathType_t math_type = CUDNN_TENSOR_OP_MATH_ALLOW_CONVERSION;
const cudnnMathType_t math_type = CUDNN_TENSOR_OP_MATH;
throwIfError(::cudnnSetConvolutionMathType(conv_desc.get(), math_type));
throwIfError(::cudnnSetConvolutionGroupCount(conv_desc.get(), groups_));

Expand All @@ -327,6 +328,7 @@ ConvolutionBackpropDataDescriptorCuDnn::ConvolutionBackpropDataDescriptorCuDnn(
} else {
GetAlgo(dnnHandle);
}
throwIfError(::cudnnSetConvolutionMathType(conv_.get(), algo_perf_.mathType));
}

void ConvolutionBackpropDataDescriptorCuDnn::BenchmarkOptimalAlgo(const CUDA::DnnHandle& dnnHandle) {
Expand Down
Loading