[CP2K-user] [CP2K:21040] Re: Resolving MPI limitations with the wavelet Poisson solver for a large system

[Frederick Stein]

Dear Xing,
Sorry for the late replay. I think the error message is a bit misleading. I 
checked the code and my impression is that this error apparently occurred 
under the prescribed circumstances but nobody got along to actually 
investigate/fix the related issue. I am also not sure what actually 
triggers this error. Can you try a different poisson solver than the 
Wavelet solver?
Best,
Frederick

Xing Wang schrieb am Dienstag, 17. Dezember 2024 um 02:37:57 UTC+1:

> Hi community,
>
> I am running a calculation on a system with 480 atoms in a unit cell of 
> size approximately 37×37×37 Å. I used 128 CPUs with the wavelet Poisson 
> solver, but encountered the following error:
>
>
> *******************************************************************************
> *   ___                                                                   
>     *
> *  /   \                                                                   
>    *
> * [ABORT]  Index out of range in wavelet solver. Try decreasing the number 
> of *
> *  \___/      MPI processors, or adjust the PW_CUTOFF or cell size so that 
>    *
> *    |       2*(number of RS grid points) matches the allowed FFT 
> dimensions  *
> *  O/|                      (see ps_wavelet_fft3d.F) exactly.             
>     *
> * /| |                                                                     
>    *
> * / \                                                 
> ps_wavelet_kernel.F:996 *
>
> *******************************************************************************
>
> I tried reducing the number of CPUs to 48, and the calculation ran 
> successfully. However, given the large system size, the calculation is very 
> slow with fewer CPUs.
>
> My Questions:
> - Is there a way to use more CPUs with the Wavelet solver without 
> encountering the FFT grid error?
> - Are there specific strategies (e.g., adjusting PW_CUTOFF, hybrid 
> MPI+OpenMP) to ensure FFT grid compatibility while improving 
> parallelization efficiency?
>
> Any suggestions would be greatly appreciated!
>
> Best regards,
> Xing
>

-- 
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 visit https://groups.google.com/d/msgid/cp2k/3c19c672-4ffe-4be3-ac3a-dc5bb2e62b1dn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20250113/ca196170/attachment.htm>