[GitHub] [incubator-mxnet] marcoabreu commented on a change in pull request #18921: Einsum cutensor GPU

[GitBox <gi...@apache.org>]

marcoabreu commented on a change in pull request #18921:
URL: https://github.com/apache/incubator-mxnet/pull/18921#discussion_r471601166



##########
File path: 3rdparty/mshadow/CMakeLists.txt
##########
@@ -59,6 +59,9 @@ endif()
 if(USE_CUDNN)
   target_compile_definitions(mshadow INTERFACE MSHADOW_USE_CUDNN)
 endif()
+if(USE_CUTENSOR)
+  target_compile_definitions(mshadow INTERFACE MSHADOW_USE_CUTENSOR)
+endif()

Review comment:
       I'm a bit concerned that we add too many things which depend on optional dependencies that are determined at compile time. This make releasing mxnet more and more difficult. Can we fine a neater way? Like either saying that cutensor is now a required dependency if cudnn and/or cuda are enabled or put them into the external operator space that is currently being developed? We are creating more and more branching within our build system that is impossible to test properly. 
   
   Another topic is of course the whole licensing topic. 




----------------------------------------------------------------
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