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

The transformer of memcpy is needed for ROCm EP and MIGraphX EP when fallbacking to CPU happens #10522

Merged
merged 1 commit into from
Feb 11, 2022
Merged

The transformer of memcpy is needed for ROCm EP and MIGraphX EP when fallbacking to CPU happens #10522

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
7 changes: 5 additions & 2 deletions onnxruntime/core/optimizer/transformer_memcpy.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -189,7 +189,8 @@ void TransformerMemcpyImpl::ProcessDefs(onnxruntime::Node& node, const KernelReg
InitializedTensorSet& initializers_consumed) {
auto node_provider_type = node.GetExecutionProviderType();
if ((node_provider_type == provider_) ||
(node_provider_type == kCudaExecutionProvider && kTensorrtExecutionProvider == provider_)) {
(node_provider_type == kCudaExecutionProvider && kTensorrtExecutionProvider == provider_) ||
(node_provider_type == kRocmExecutionProvider && kMIGraphXExecutionProvider == provider_)) {
provider_nodes_.insert(&node);
// note KernelCreateInfo might be nullptr for custom kernel
const KernelCreateInfo* kci = nullptr;
Expand Down Expand Up @@ -281,7 +282,9 @@ void TransformerMemcpyImpl::BuildDefsMapping(const onnxruntime::NodeArg* arg, co
if (arg_input_index == -1 && arg_output_index == -1)
continue;
auto node_provider_type = it.GetExecutionProviderType();
if ((node_provider_type == provider_) || (node_provider_type == kCudaExecutionProvider && kTensorrtExecutionProvider == provider_)) {
if ((node_provider_type == provider_) ||
(node_provider_type == kCudaExecutionProvider && kTensorrtExecutionProvider == provider_) ||
(node_provider_type == kRocmExecutionProvider && kMIGraphXExecutionProvider == provider_)) {
const KernelCreateInfo* kci = nullptr;
ORT_IGNORE_RETURN_VALUE(kernel_registries.SearchKernelRegistry(it, &kci));
if (arg_input_index != -1) {
Expand Down