singa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From GitBox <...@apache.org>
Subject [GitHub] [singa] chrishkchris commented on a change in pull request #716: SINGA-510 Distributed Training Time Profiling
Date Sat, 06 Jun 2020 15:31:25 GMT

chrishkchris commented on a change in pull request #716:
URL: https://github.com/apache/singa/pull/716#discussion_r436277284



##########
File path: include/singa/core/common.h
##########
@@ -100,13 +100,21 @@ typedef struct _Context {
   std::mt19937 random_generator;
 #ifdef USE_CUDA
   cublasHandle_t cublas_handle;
-  cudaStream_t stream;
-  curandGenerator_t curand_generator;
+  cudaStream_t stream; 
+  curandGenerator_t curand_generator; 
+
 #ifdef USE_CUDNN
   cudnnHandle_t cudnn_handle;
 #endif
 #endif  // USE_CUDA
 
+#ifdef USE_DIST

Review comment:
       In the CPU only version, there is no cuda library. It will not have those cuda related
headers.
   In the building process, the default USE_DIST is 0. Note that the USE_DIST and USE_CUDA
is determined at build, not at runtime.
   So the Travis CI build of CPU version didn't include this part, you may check the CI CPU
version build result
   My concern not to use USE_CUDA instead is that it could add three more useless cuda stream
into the single GPU version.




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



Mime
View raw message