[Openmp-commits] [openmp] [amdgpu] Implement D2D memcpy as HSA call (PR #69955)

[Jan Patrick Lehr via Openmp-commits]

================
@@ -2250,14 +2250,37 @@ struct AMDGPUDeviceTy : public GenericDeviceTy, AMDGenericDeviceTy {
                                           PinnedMemoryManager);
   }
 
-  /// Exchange data between two devices within the plugin. This function is not
-  /// supported in this plugin.
+  /// Exchange data between two devices within the plugin.
   Error dataExchangeImpl(const void *SrcPtr, GenericDeviceTy &DstGenericDevice,
                          void *DstPtr, int64_t Size,
                          AsyncInfoWrapperTy &AsyncInfoWrapper) override {
-    // This function should never be called because the function
-    // AMDGPUPluginTy::isDataExchangable() returns false.
-    return Plugin::error("dataExchangeImpl not supported");
+    AMDGPUDeviceTy &DstDevice = static_cast<AMDGPUDeviceTy &>(DstGenericDevice);
+
+    hsa_agent_t SrcAgent = getAgent();
+    hsa_agent_t DstAgent = DstDevice.getAgent();
+
+    AMDGPUSignalTy Signal;
+    if (auto Err = Signal.init())
+      return Err;
+
+    // The agents need to have access to the corresponding memory
+    // This is presently only true if the pointers were originally
+    // allocated by this runtime or the caller made the appropriate
+    // access calls.
+    hsa_status_t Status = hsa_amd_memory_async_copy(
+        DstPtr, DstAgent, SrcPtr, SrcAgent, (Size > 0) ? (size_t)Size : 0, 0,
+        nullptr, Signal.get());
+    if (auto Err =
+            Plugin::check(Status, "Error in hsa_amd_memory_async_copy: %s"))
----------------
jplehr wrote:

Do we want the error to mention that this happened in the device to device copy, or is this automagically done given the surrounding method of the error?

https://github.com/llvm/llvm-project/pull/69955