kohr-h commented on issue #13143: [MXNET-1206] Support NDArray indexing with 
None and Ellipsis
URL: https://github.com/apache/incubator-mxnet/pull/13143#issuecomment-520402215
 
 
   Great to see this PR merged, thanks for the "last mile" effort @reminisce 
and @zoeygxy.
   
   Regarding the performance differences, the trend you observed is somewhat to 
be expected. The checks performed in basic indexing are more comprehensive than 
before, and in more cases than before, views are being returned. Thus, the 
cases where views have always been returned are a bit slower because of extra 
overhead, but the cases that turned from copy to view should amortize easily.
   
   The advanced indexing performance could have gone either way, so I'm happy 
about the improvement there.
   
   To see the effect more clearly, I think it would make sense to test very 
small arrays (pure overhead) and very large arrays (purely copy vs. view).
   
   How much are you worried about the performance hit for the view cases? If 
you really want to be on par with the old speed, you could re-introduce the 
same special-casing as in the old code, but IMO that would make the code more 
complex. My implementation strives to be as generic as possible and takes the 
small performance hit of manipulating tuples and slice objects a bit too much 
for simple cases.

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