Improving the computational efficiency of votca for relative entropy minimization

[dd ding]

Hello developers，I want to know the reason why the calculation speed of relative entropy minimization in Votca is relatively slow. I use CBSPL potential in the calculation. A total of 344 parameters need to be optimized.

Run a total of 500000 steps, outputting a frame every 1000 steps.

By reviewing the log file, I found that for an iteration step, the total calculation time is 547 seconds

Among them, MD used 158.753 seconds,

Equivalent to REM using 547-158.753=388.247 seconds.

Can it be understood as: MD took 158.753 seconds to calculate 500000 steps, and REM took 388.247 seconds to process 500 frames output from MD, resulting in a difference in computational efficiency of more than 2400 times? I would like to know the reason for this computational efficiency gap, whether it is due to my installation environment and methods, or the VOTCA's own calculation method? Is it possible to improve computational efficiency through certain methods?

looking forward to your reply

best wishes

D

[Christoph Junghans]

On Sun, Jul 7, 2024, 09:44 dd ding <winter...@gmail.com> wrote:

Hello developers，I want to know the reason why the calculation speed of relative entropy minimization in Votca is relatively slow. I use CBSPL potential in the calculation. A total of 344 parameters need to be optimized.

Run a total of 500000 steps, outputting a frame every 1000 steps.

By reviewing the log file, I found that for an iteration step, the total calculation time is 547 seconds

Among them, MD used 158.753 seconds,

Equivalent to REM using 547-158.753=388.247 seconds.

Can it be understood as: MD took 158.753 seconds to calculate 500000 steps, and REM took 388.247 seconds to process 500 frames output from MD, resulting in a difference in computational efficiency of more than 2400 times? I would like to know the reason for this computational efficiency gap, whether it is due to my installation environment and methods, or the VOTCA's own calculation method?

MD and the RE update are two very different things, so it is hard to compare the timings.

Did you run the RE update on multiple threads?

And on how many threads did you run the MD?

Also see https://doi.org/10.1021/acs.jctc.2c00665 for discussion on that.

Is it possible to improve computational efficiency through certain methods?

Yes, our implementation isn't the fastest, but patches are always welcome if you have ideas to make it faster.

Christoph

looking forward to your reply

best wishes

D

--
Join us on Slack: https://join.slack.com/t/votca/signup
---
You received this message because you are subscribed to the Google Groups "votca" group.
To unsubscribe from this group and stop receiving emails from it, send an email to votca+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/votca/16aa638d-2f45-4ee7-aae9-56d11d664c98n%40googlegroups.com.