saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-24 19:43:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Check status of element-wise op prepare functions. (#4830)

Browse source 
* Check status of BinaryElementwise::Prepare().

* Add additional status checks for BinaryElementwise::Prepare() and UnaryElementwise::Prepare().

* Add status checks for BinaryElementwisePreparation::BinaryElementwiseBroadcastPrepareHelper().
This commit is contained in:
edgchen1 2020-08-17 16:11:30 -07:00 committed by GitHub
parent ef20efe015
commit 32a5f3d5b6
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
7 changed files with 9 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
onnxruntime/contrib_ops/cuda/activation/activations.cc
View file

@ -26,7 +26,7 @@ namespace cuda {
template <> \
Status x<T>::ComputeInternal(OpKernelContext* context) const { \
UnaryElementwisePreparation p; \
UnaryElementwise::Prepare(context, &p); \
ORT_RETURN_IF_ERROR(UnaryElementwise::Prepare(context, &p)); \
Ctx##x func_ctx = MakeFuncCtx(); \
Impl_##x<typename ToCudaType<T>::MappedType>( \
reinterpret_cast<const typename ToCudaType<T>::MappedType*>(p.input_tensor->template Data<T>()), \

2
onnxruntime/contrib_ops/cuda/math/complex_mul.cc
View file

@ -40,7 +40,7 @@ Status ComplexMul<T, is_conj>::ComputeInternal(OpKernelContext* context) const {
}
BinaryElementwisePreparation prepare;
Prepare(context, &prepare);
ORT_RETURN_IF_ERROR(Prepare(context, &prepare));
ComplexMul_Impl<typename ToCudaType<T>::MappedType>(
prepare.output_rank_or_simple_broadcast,
&prepare.lhs_padded_strides,

2
onnxruntime/core/providers/cuda/activation/activations.cc
View file

@ -22,7 +22,7 @@ namespace cuda {
template <> \
Status x<T>::ComputeInternal(OpKernelContext* context) const { \
UnaryElementwisePreparation p; \
UnaryElementwise::Prepare(context, &p); \
ORT_RETURN_IF_ERROR(UnaryElementwise::Prepare(context, &p)); \
Ctx##x func_ctx = MakeFuncCtx(); \
Impl_##x<typename ToCudaType<T>::MappedType>( \
reinterpret_cast<const typename ToCudaType<T>::MappedType*>(p.input_tensor->template Data<T>()), \

4
onnxruntime/core/providers/cuda/math/binary_elementwise_ops.cc
View file

@ -143,7 +143,7 @@ Status BinaryElementwise<ShouldBroadcast>::Prepare(OpKernelContext* context, Bin
template <> \
Status x<T>::ComputeInternal(OpKernelContext* context) const { \
BinaryElementwisePreparation prepare; \
Prepare(context, &prepare); \
ORT_RETURN_IF_ERROR(Prepare(context, &prepare)); \
Impl_##x<typename ToCudaType<T>::MappedType>( \
prepare.output_rank_or_simple_broadcast, \
&prepare.lhs_padded_strides, \
@ -337,7 +337,7 @@ Status DispatchOnFirstArg(const BinaryElementwisePreparation& prepare) {
Status Pow::ComputeInternal(OpKernelContext* context) const {
BinaryElementwisePreparation prepare;
Prepare(context, &prepare);
ORT_RETURN_IF_ERROR(Prepare(context, &prepare));
namespace on = ONNX_NAMESPACE;
using namespace pow12_internal;

2
onnxruntime/core/providers/cuda/math/unary_elementwise_ops.cc
View file

@ -37,7 +37,7 @@ Status UnaryElementwise::Prepare(OpKernelContext* context, UnaryElementwisePrepa
template <> \
Status x<T>::ComputeInternal(OpKernelContext* context) const { \
UnaryElementwisePreparation p; \
UnaryElementwise::Prepare(context, &p); \
ORT_RETURN_IF_ERROR(UnaryElementwise::Prepare(context, &p)); \
Impl_##x( \
reinterpret_cast<const typename ToCudaType<T>::MappedType*>(p.input_tensor->template Data<T>()), \
reinterpret_cast<typename ToCudaType<T>::MappedType*>(p.output_tensor->template MutableData<T>()), \

4
onnxruntime/core/providers/cuda/reduction/reduction_ops.cc
View file

@ -157,7 +157,7 @@ Status ReduceKernel<allow_multi_axes>::ReduceKernelShared(
auto exp_result = GetScratchBuffer<T>(input_count).get();
auto log_sum_result = GetScratchBuffer<T>(output_count).get();
BinaryElementwisePreparation prepare;
prepare.BinaryElementwiseBroadcastPrepareHelper(input_shape, rhs_shape, input_shape);
ORT_RETURN_IF_ERROR(prepare.BinaryElementwiseBroadcastPrepareHelper(input_shape, rhs_shape, input_shape));
Impl_Sub<CudaT>(prepare.output_rank_or_simple_broadcast,
&prepare.lhs_padded_strides,
reinterpret_cast<const CudaT*>(X),
@ -481,7 +481,7 @@ Status ReduceComputeCore(CUDAExecutionProvider& cuda_ep, const Tensor& input, Pr
auto log_sum_result_buffer = cuda_ep.GetScratchBuffer<T>(output_count);
auto log_sum_result = log_sum_result_buffer.get();
BinaryElementwisePreparation prepare;
prepare.BinaryElementwiseBroadcastPrepareHelper(input_shape, output_shape, input_shape);
ORT_RETURN_IF_ERROR(prepare.BinaryElementwiseBroadcastPrepareHelper(input_shape, output_shape, input_shape));
Impl_Sub<CudaT>(prepare.output_rank_or_simple_broadcast,
&prepare.lhs_padded_strides,
reinterpret_cast<const CudaT*>(input.template Data<T>()),

2
orttraining/orttraining/training_ops/cuda/activation/activations_grad.cc
View file

@ -23,7 +23,7 @@ namespace cuda {
template <> \
Status x<T>::ComputeInternal(OpKernelContext* context) const { \
BinaryElementwisePreparation prepare; \
Prepare(context, &prepare); \
ORT_RETURN_IF_ERROR(Prepare(context, &prepare)); \
CudaAsyncBuffer<Ctx##x> func_ctx(this, MakeFuncCtx(), 1); \
if (!std::is_same<CtxNull, Ctx##x>::value) ORT_RETURN_IF_ERROR(func_ctx.CopyToGpu()); \
Impl_##x<typename ToCudaType<T>::MappedType>( \