[GitHub] [incubator-mxnet] ymjiang edited a comment on issue #14485: Any suggestion to accelerate parameter update on PS for distributed training?

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

ymjiang edited a comment on issue #14485: Any suggestion to accelerate parameter update on PS for distributed training?
URL: https://github.com/apache/incubator-mxnet/issues/14485#issuecomment-489387686
 
 
   Hi @eric-haibin-lin , thank you for the note of extra pull copy. Actually I solve it by directly casting the pointer to response vals, instead of using the `CopyFrom`:
   
   `response->vals = ps::SArray<char>(static_cast<char*>(stored.data().dptr_), len, false);`
   
   The performance indeed increases noticeably without the memory copy. And with proper `WaitToRead` I believe this will not affect the  correctness (I have confirmed the convergence on Resnet50 with ImageNet dataset).

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


With regards,
Apache Git Services