require 'cunn' takes 40s.

[David Gerónimo]

Hi all,

It takes around 40s to load the libary when calling 'require 'cunn'. I've been searching for possible reasons for this but I did not find any solution. Any idea?

Thanks,
david

[soumith]

I think it is related to CUDA's JIT.
This should only happen for the first time you "require 'cunn'", and subsequent times it should be cached properly. If it is being slow every single time, check that ~/.nv/ComputeCache (on Linux) or $HOME/Library/Application Support/NVIDIA/ComputeCache (OSX) exists and has proper permissions.

more info here: http://devblogs.nvidia.com/parallelforall/cuda-pro-tip-understand-fat-binaries-jit-caching/

Warning 

This e-mail is privileged, confidential and contains private information. Any reading, retention, distribution or copying of this communication by any person other than its intended recipient is prohibited.

--
You received this message because you are subscribed to the Google Groups "torch7" group.
To unsubscribe from this group and stop receiving emails from it, send an email to torch7+un...@googlegroups.com.
To post to this group, send email to tor...@googlegroups.com.
Visit this group at http://groups.google.com/group/torch7.
For more options, visit https://groups.google.com/d/optout.

[David Gerónimo]

Thanks for the link. I changed permissions of ~/.nv/ComputeCache (from root to my user) and now the loading is immediate.

On Friday, August 21, 2015 at 5:02:06 PM UTC+2, smth chntla wrote:

I think it is related to CUDA's JIT.
This should only happen for the first time you "require 'cunn'", and subsequent times it should be cached properly. If it is being slow every single time, check that ~/.nv/ComputeCache (on Linux) or $HOME/Library/Application Support/NVIDIA/ComputeCache (OSX) exists and has proper permissions.

more info here: http://devblogs.nvidia.com/parallelforall/cuda-pro-tip-understand-fat-binaries-jit-caching/