xHip stream fix for non cuda env

pytorch · Sep 12, 2024 · 863f596 · 863f596
1 parent 1f166ba
commit 863f596
Show file tree

Hide file tree

Showing 4 changed files with 12 additions and 3 deletions.
diff --git a/fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_blockwise_gemm.hip b/fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_blockwise_gemm.hip
@@ -12,11 +12,13 @@
 #include <numeric>
 
 #include <ATen/ATen.h>
+#if !defined(USE_ROCM)
 #include <c10/cuda/CUDAStream.h>
+#endif
 #include <torch/torch.h>
 
 #if defined(USE_ROCM)
-
+#include <c10/hip/HIPStream.h>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"

diff --git a/fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_rowwise_gemm.hip b/fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_rowwise_gemm.hip
@@ -15,11 +15,13 @@
 #include <unordered_map>
 
 #include <ATen/ATen.h>
+#if !defined(USE_ROCM)
 #include <c10/cuda/CUDAStream.h>
+#endif
 #include <torch/torch.h>
 
 #if defined(USE_ROCM)
-
+#include <c10/hip/HIPStream.h>
 #include "kernels/fp8_rowwise_kernel_manifest.h"
 
 namespace fbgemm_gpu {

diff --git a/fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_tensorwise_gemm.hip b/fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_tensorwise_gemm.hip
@@ -12,11 +12,13 @@
 #include <numeric>
 
 #include <ATen/ATen.h>
+#if !defined(USE_ROCM)
 #include <c10/cuda/CUDAStream.h>
+#endif
 #include <torch/torch.h>
 
 #if defined(USE_ROCM)
-
+#include <c10/hip/HIPStream.h>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"

diff --git a/fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/kernels/fp8_rowwise_common.h b/fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/kernels/fp8_rowwise_common.h
@@ -12,10 +12,13 @@
 #include <numeric>
 
 #include <ATen/ATen.h>
+#if !defined(USE_ROCM)
 #include <c10/cuda/CUDAStream.h>
+#endif
 #include <torch/torch.h>
 
 #if defined(USE_ROCM)
+#include <c10/hip/HIPStream.h>
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"