You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mxnet.apache.org by GitBox <gi...@apache.org> on 2020/11/03 13:47:07 UTC

[GitHub] [incubator-mxnet] reminisce edited a comment on issue #19458: Wrong handling of workspace in multiple numpy operators

reminisce edited a comment on issue #19458:
URL: https://github.com/apache/incubator-mxnet/issues/19458#issuecomment-720797018


   @ptrendx If I understand your argument correctly, it seems the mechanism of returning the workspace from the context and resetting the counter similarly leads to the potential problem of memory fragmentation? Also, resetting the temp space in a callee would require copying the data from the original context to the bigger one and consequently would cause performance degradation. I believe it's a decision about the tradeoff between developer experience and ultimate performance by comparing multi-temp-space solution with one lump-sum temp space.
   
   Anyway, I'm just speaking from my past experience of operator development. It's up to the community to bring up some RFC to tackle this.


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



---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org
For additional commands, e-mail: issues-help@mxnet.apache.org