From a3632b00a47a6d655ad71e7709ddeca159443398 Mon Sep 17 00:00:00 2001
From: Xinya Zhang <Xinya.Zhang@amd.com>
Date: Mon, 11 Nov 2024 13:55:35 -0600
Subject: [PATCH] Fixes SWDEV-483388 (#74)

* Fix SWDEV-483388: mistakenly invoke hipMemcpy when ORT_ENABLE_STREAM=True

* Fix another hipMemcpy

* Remove MIGRAPHX_STREAM_SYNC guard which does not exists in current version.
---
 onnxruntime/core/providers/migraphx/gpu_data_transfer.cc | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/onnxruntime/core/providers/migraphx/gpu_data_transfer.cc b/onnxruntime/core/providers/migraphx/gpu_data_transfer.cc
index 77c5e18a5878e..a643f6b208f94 100644
--- a/onnxruntime/core/providers/migraphx/gpu_data_transfer.cc
+++ b/onnxruntime/core/providers/migraphx/gpu_data_transfer.cc
@@ -67,6 +67,9 @@ common::Status GPUDataTransfer::CopyTensorAsync(const Tensor& src, Tensor& dst,
     } else if (src_device.Type() == OrtDevice::GPU) {
       // copying between GPU, this is non-blocking
       HIP_CALL_THROW(hipMemcpyAsync(dst_data, src_data, bytes, hipMemcpyDeviceToDevice, static_cast<hipStream_t>(stream.GetHandle())));
+    } else {
+      // copy from other CPU memory to GPU, this is blocking
+      HIP_CALL_THROW(hipMemcpyWithStream(dst_data, src_data, bytes, hipMemcpyHostToDevice, static_cast<hipStream_t>(stream.GetHandle())));
     }
   } else if (src_device.Type() == OrtDevice::GPU) {
     // If dest are not pinned, the memory copy will be performed synchronously.