[CP2K-user] [CP2K:17978] Re: GPU support and ELPA

Ole Schütt ole.schuett at cp2k.org
Wed Nov 2 15:03:56 UTC 2022

Hi Amit,

unfortunately, we never manged to eliminate those block size constraints 
from ELPA.
Going forward our hope rests on the new DLA-Future library which should 
arrive in CP2K soon.


On Wednesday, November 2, 2022 at 1:22:42 AM UTC+1 Amit Gupta wrote:

> I have finally got the GPU version of CP2K working. However when I run, I 
> get the following warnings:
> ```
>  ELPA: Warning, CUDA kernel only works with block size 2^n (n = 1, 2, ..., 
> 10)!
>  The compute kernel will be executed on CPUs!
>  We recommend changing the block size to 2^n
> ```
> There have been discussions here: https://github.com/cp2k/cp2k/issues/464
> I am not sure what was the outcome, but it does say certain commits for 
> closing the issues.
> 1. How to define the block size?
> 2. What are other "good practices" for extracting the most out of GPU?
> CP2K version: 2022.2
> Configuration:  --with-libxc=install --with-libint=install 
> --with-fftw=install --with-mkl=system --with-libxsmm=install 
>  --with-elpa=install   --with-spla=install   --with-cosma=install 
> Thank you--with-intel=system --enable-cuda --gpu-ver=V100 
>  --with-openmpi=system  

You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/9a765bf0-7f6a-4950-bb12-0aaa4ed128d5n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20221102/e0f27cb3/attachment.htm>

More information about the CP2K-user mailing list