Merge pull request #15 from stoneyang/hl_device_func

fixed build issue of double definition of atomicAdd on modern GPUs
PaddlePaddle · Sep 1, 2016 · f2b84d5 · f2b84d5
2 parents df82689 + f2aa274
commit f2b84d5
Showing 1 changed file with 4 additions and 0 deletions.
diff --git a/paddle/cuda/include/hl_device_functions.cuh b/paddle/cuda/include/hl_device_functions.cuh
@@ -16,6 +16,8 @@ limitations under the License. */
 #ifndef HL_DEVICE_FUNCTIONS_CUH_
 #define HL_DEVICE_FUNCTIONS_CUH_
 
+#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 600
+
 namespace hppl {
 
 static __inline__ __device__ double atomicAdd(double* address, double val) {
@@ -38,4 +40,6 @@ static __inline__ __device__ double atomicAdd(double* address, double val) {
 
 using hppl::atomicAdd;
 
+#endif
+
 #endif /* HL_DEVICE_FUNCTIONS_CUH_ */