@cjolivier01 

> Using the llvm OpenMP gives a significant performance improvement over gcc's 
> libgomp.

You can see at [this 
line](https://github.com/apache/incubator-mxnet/pull/12160/files#diff-af3b638bc2a3e6c650974192a53c7291L384)
 that not GNU OpenMP was replaced, but Intel OpenMP. I have reasons to believe 
that llvm OpenMP is not significantly faster (if at all) than Intel OpenMP.

Can you provide measurements you did on performance? What build configurations 
you were using?

[ Full content available at: 
https://github.com/apache/incubator-mxnet/pull/12160 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to