larroy commented on issue #14570: add a compiler flag to use int64 as tensor 
size
URL: https://github.com/apache/incubator-mxnet/pull/14570#issuecomment-484582885
 
 
   To be honest, while I understand the rationale of the solution proposed. I 
think it would be better if we could avoid having yet another build flavour if 
possible.  Could we for example, narrow the computation at runtime in case the 
number of dimensions doesn't overflow? Also see my previous comment above to 
see if we can work with memory access to make it almost as fast with the wider 
type in this particular computation. If I understood correctly the regression 
is only in transposing? the other operators are fine? We could try to find a 
solution only for this operator then instead of adding a new build flavour.

----------------------------------------------------------------
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:
[email protected]


With regards,
Apache Git Services

Reply via email to