kohr-h commented on issue #13040: Import and GPU init extremely slow on Windows
URL: 
https://github.com/apache/incubator-mxnet/issues/13040#issuecomment-434626218
 
 
   > what your gpu?
   
   Titan V
   
   > the import mxnet slow because the need to load many huge dll and 
decompress nvbin.
   > (cudnn about 300mb,mxnet 240mb,mklml 140mb)
   
   Maybe, but MXNet + MKLDNN takes <0.8 seconds, whereas MXNet + CUDA takes 
10-ish seconds. So size alone doesn't explain it.
   
   > nd.array([1, 2, 3], ctx=gpu()) is very slow look like nvbin not include 
you gpu bin,so cuda driver will compile in time.
   
   Okay, so when would that happen? Is that every time a new GPU context is 
used for the first time?
   Can I avoid this issue by building MXNet from source with specific compiler 
flags?

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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