Reducing EDR file output

User discussions
,
1

GROMACS version: 2020.6
GROMACS modification: No (But I think a modification is required to do what I want)

I’m running a few very long simulations (1μs) and I need the pressure tensor sampled every 2fs. I only need the pressure tensor. The output .edr file would be about 500GB, and I really don’t need most of it. Is there a way to only save certain parameters to the .edr file?

If there isn’t an in-built way to do this (and I suspect there isn’t), has anyone already made a similar modification that I can use? Or, any tips for what exactly needs to be changed in the gromacs code in order to achieve this?

Thanks so much.

2

Hi,
you can have a look at gmx eneconv
https://manual.gromacs.org/2021/onlinehelp/gmx-eneconv.html
it will allow you to reduce edr dimension after you extract the desired values.
Best regards
Alessandra

3

Hello,

you might have to modify the code to explicitly write the pressure tensor to disk whenever you need it. This would require:

  1. To ensure that it is calculated
  2. To write it

The main MD loop is in src/gromacs/mdrun/md.cpp, you could perhaps have a look at that and where + how the function compute_globals is called? The pressure tensor seems to be stored in the pres variable, but it’s not fully clear to me from a glance exactly when you can be sure that it is updated. Perhaps if you set nstcalcenergy = 1 in your .mdp file it will be correct at the end of the loop, at every step? This might be slow…

Regards,
Petter

4

Hm. I was thinking about going the other way, and just limiting the parameters that actually get saved to the edr file… compute_globals is probably the place for me to start either way, thanks!

5

Well with 5 replicas I don’t have the storage space for the 2.5 TB of data, so I need to do something earlier in the pipeline.

Shea