ADMM Method to Speed Up MD Simulation with B3LYP functional
69 views
Skip to first unread message
Lim Jia Yang
unread,
Jun 15, 2024, 8:53:09 AMJun 15
Reply to author
Sign in to reply to author
Forward
Sign in to forward
Delete
You do not have permission to delete messages in this group
Copy link
Report message
Show original message
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to cp2k
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.
You do not have permission to delete messages in this group
Copy link
Report message
Show original message
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to cp2k
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
Jürg Hutter
unread,
Jun 17, 2024, 3:50:21 AMJun 17
Reply to author
Sign in to reply to author
Forward
Sign in to forward
Delete
You do not have permission to delete messages in this group
Copy link
Report message
Show original message
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to cp...@googlegroups.com
Hi
the use of aug- ADMM Basis sets is not recommended (because it
is against the basic idea for performance in ADMM). Only use those
basis sets if tests show that they are absolutely necessary to
reach the required accuracy.
regards
JH
________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Lim Jia Yang <limjiay...@gmail.com>
Sent: Saturday, June 15, 2024 1:12 PM
To: cp2k
Subject: [CP2K:20331] ADMM Method to Speed Up MD Simulation with B3LYP functional