[CP2K-user] [CP2K:20333] Re: ADMM Method to Speed Up MD Simulation with B3LYP functional

[Frederick Stein]

Dear Jia,
If I compare both outputs, I see that ADMM indeed reduces the computational 
costs of the HFX kernel. What you still see are FFT (fft_wrap_pw1pw2). You 
have two ways to reduce its costs: reduce the cutoff and/or reduce the box 
size. If that reduces the accuracy to much, you may check the keywords 
ALLTOALL_SGL and FFTW_PLAN_TYPE in the GLOBAL section to improve the 
performance of FFTW and the communication algorithms.
HTH,
Frederick

Lim Jia Yang schrieb am Samstag, 15. Juni 2024 um 14:53:09 UTC+2:

> Dear CP2K experts
>
> I attempt to simulate hydrated electron in the gas phase using B3LYP 
> functional. The usage of ADMM does not speed up my simulation. It is slower 
> (~25s per step) than simulation without using ADMM (~21s per step). I have 
> tried to lower EPS_SCHWARZ to 1.0E-6 but it is still slower at ~25s per 
> step.
>
> I would be very grateful for any possible advice with my input, and 
> insight into the usage of ADMM for performance improvement. I am using CP2K 
> v2024.1 compiled with libint-lmax=7 and MAX_CONTR=4. I am new to CP2K.
>
> I have attached the input/output files for the 2 simulations.
>
> Thank you!
> Jia Yang
>

-- 
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/ea2feaa4-6f72-406c-9cf4-dd5cbe53a452n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20240615/72a2437e/attachment.htm>