clang format off on some cuda .cc file

paddle/cuda/src/hl_cuda_cublas.cc

@@ -56,9 +56,14 @@ void *cublas_dso_handle = nullptr;
 #define DYNAMIC_LOAD_CUBLAS_V2_WRAP(__name) DYNAMIC_LOAD_CUBLAS_WRAP(__name)
 
 // include all needed cublas functions in HPPL
+// clang-format off
 #define CUBLAS_BLAS_ROUTINE_EACH(__macro) \
-  __macro(cublasSgemv) __macro(cublasDgemv) __macro(cublasSgemm) \
-      __macro(cublasDgemm) __macro(cublasSgeam) __macro(cublasDgeam)
+  __macro(cublasSgemv)                    \
+  __macro(cublasDgemv)                    \
+  __macro(cublasSgemm)                    \
+  __macro(cublasDgemm)                    \
+  __macro(cublasSgeam)                    \
+  __macro(cublasDgeam)                    \
 
 DYNAMIC_LOAD_CUBLAS_V2_WRAP(cublasCreate)
 DYNAMIC_LOAD_CUBLAS_V2_WRAP(cublasDestroy)
@@ -81,6 +86,7 @@ CUBLAS_BLAS_ROUTINE_EACH(DYNAMIC_LOAD_CUBLAS_V2_WRAP)
 
 } /* namespace dynload */
 
+// clang-format on
 #ifndef PADDLE_TYPE_DOUBLE
 #define CUBLAS_GEAM dynload::cublasSgeam
 #define CUBLAS_GEMV dynload::cublasSgemv

paddle/cuda/src/hl_cuda_device.cc

@@ -57,10 +57,14 @@ void *curand_dso_handle = nullptr;
 #endif
 
 /* include all needed curand functions in HPPL */
+// clang-format off
 #define CURAND_RAND_ROUTINE_EACH(__macro)    \
-  __macro(curandCreateGenerator) __macro(curandSetStream) \
-      __macro(curandSetPseudoRandomGeneratorSeed)         \
-          __macro(curandGenerateUniform) __macro(curandGenerateUniformDouble)
+  __macro(curandCreateGenerator)             \
+  __macro(curandSetStream)                   \
+  __macro(curandSetPseudoRandomGeneratorSeed)\
+  __macro(curandGenerateUniform)             \
+  __macro(curandGenerateUniformDouble)
+// clang-format on
 
 CURAND_RAND_ROUTINE_EACH(DYNAMIC_LOAD_CURAND_WRAP)
 
@@ -99,25 +103,38 @@ void *cudart_dso_handle = nullptr;
 #endif
 
 /* include all needed cuda functions in HPPL */
+// clang-format off
 #define CUDA_ROUTINE_EACH(__macro)        \
-  __macro(cudaMalloc) __macro(cudaHostAlloc) __macro(cudaFree)                 \
-      __macro(cudaFreeHost) __macro(cudaMemcpy) __macro(cudaMemset) __macro(   \
-          cudaMemcpyAsync) __macro(cudaSetDevice) __macro(cudaGetDevice)       \
-          __macro(cudaGetDeviceCount) __macro(cudaGetDeviceProperties)         \
-              __macro(cudaDeviceSynchronize) __macro(cudaDeviceCanAccessPeer)  \
+  __macro(cudaMalloc)                     \
+  __macro(cudaHostAlloc)                  \
+  __macro(cudaFree)                       \
+  __macro(cudaFreeHost)                   \
+  __macro(cudaMemcpy)                     \
+  __macro(cudaMemset)                     \
+  __macro(cudaMemcpyAsync)                \
+  __macro(cudaSetDevice)                  \
+  __macro(cudaGetDevice)                  \
+  __macro(cudaGetDeviceCount)             \
+  __macro(cudaGetDeviceProperties)        \
+  __macro(cudaDeviceSynchronize)          \
+  __macro(cudaDeviceCanAccessPeer)        \
   __macro(cudaDeviceEnablePeerAccess)     \
-                      __macro(cudaStreamCreate) __macro(cudaStreamDestroy)     \
-                          __macro(cudaStreamSynchronize) __macro(              \
-                              cudaStreamWaitEvent) __macro(cudaEventCreate)    \
-                              __macro(cudaEventRecord) __macro(cudaEventQuery) \
-                                  __macro(cudaEventDestroy) __macro(           \
-                                      cudaEventSynchronize)                    \
-                                      __macro(cudaEventElapsedTime) __macro(   \
-                                          cudaSetDeviceFlags)                  \
-                                          __macro(cudaGetLastError) __macro(   \
-                                              cudaFuncSetCacheConfig)          \
+  __macro(cudaStreamCreate)               \
+  __macro(cudaStreamDestroy)              \
+  __macro(cudaStreamSynchronize)          \
+  __macro(cudaStreamWaitEvent)            \
+  __macro(cudaEventCreate)                \
+  __macro(cudaEventRecord)                \
+  __macro(cudaEventQuery)                 \
+  __macro(cudaEventDestroy)               \
+  __macro(cudaEventSynchronize)           \
+  __macro(cudaEventElapsedTime)           \
+  __macro(cudaSetDeviceFlags)             \
+  __macro(cudaGetLastError)               \
+  __macro(cudaFuncSetCacheConfig)         \
   __macro(cudaRuntimeGetVersion)          \
   __macro(cudaGetErrorString)
+// clang-format on
 
 CUDA_ROUTINE_EACH(DYNAMIC_LOAD_CUDART_WRAP)
 

paddle/cuda/src/hl_cudart_wrap.cc

@@ -47,10 +47,12 @@ extern void *cudart_dso_handle;
   } __name; /* struct DynLoad__##__name */
 
 /* include all needed cuda functions in HPPL */
+// clang-format off
 #define CUDA_ROUTINE_EACH(__macro)          \
-  __macro(cudaLaunch, cudaError_t) __macro(cudaSetupArgument, cudaError_t) \
+  __macro(cudaLaunch, cudaError_t)          \
+  __macro(cudaSetupArgument, cudaError_t)   \
   __macro(cudaConfigureCall, cudaError_t)   \
-          __macro(__cudaRegisterFatBinary, void **)                        \
+  __macro(__cudaRegisterFatBinary, void**)  \
   __macro(__cudaUnregisterFatBinary, void)  \
   __macro(__cudaRegisterFunction, void)     \
   __macro(__cudaRegisterVar, void)          \
@@ -58,6 +60,7 @@ extern void *cudart_dso_handle;
   __macro(__cudaInitModule, char)           \
   __macro(__cudaRegisterTexture, void)      \
   __macro(__cudaRegisterSurface, void)
+// clang-format on
 
 CUDA_ROUTINE_EACH(DYNAMIC_LOAD_CUDART_WRAP)