[ROCm] define C10_WARP_SIZE to warpSize HIP constant (#64302)

Summary:
warpSize is defined as a constexpr in HIP headers.  It is incorrect to assume warpSize 64.  This change fixes the C10_WARP_SIZE definition in torch sources similar to [how it was done in caffe2](https://github.com/pytorch/pytorch/blob/master/caffe2/utils/GpuDefs.cuh#L10-L14).

cc jeffdaily sunway513 jithunnair-amd ROCmSupport

Pull Request resolved: https://github.com/pytorch/pytorch/pull/64302

Reviewed By: mrshenli

Differential Revision: D30785975

Pulled By: malfet

fbshipit-source-id: 68f8333182ad4d02bd0c8d02f1751a50bc5bafa7

diff --git a/c10/macros/Macros.h b/c10/macros/Macros.h
index 4df7dfc..6bb3b76 100644 (file)
--- a/c10/macros/Macros.h
+++ b/c10/macros/Macros.h
@@ -302,7 +302,7 @@ constexpr uint32_t CUDA_THREADS_PER_BLOCK_FALLBACK = 256;
 #endif
 
 #ifdef __HIP_PLATFORM_HCC__
-#define C10_WARP_SIZE 64
+#define C10_WARP_SIZE warpSize // = 64 or 32 (Defined in hip_runtime.h)
 #else
 #define C10_WARP_SIZE 32
 #endif