Again: Floating point exception (8) Floating point divide-by-zero (3)

Hi Paul,

Pulling an old thread here. I could run my jobs by fixing a patch for this but am curious, has this been fixed and released in Gromacs 2021 version? I think probably it is planned to be released as 2021.2 end of this month. It will be helpful to know (for common cluster computing installations). Thanks a lot for fixing it.

Thanks.
Dave

Hello Dave,

we are still looking into this. The issue is that the fix we have right now means we no longer follow the reference implementation for the gamma distribution from stdlibc++. One solution that is being explored now is replacing the code altogether with a different reference implementation that doesn’t run into those issues.

Cheers

Paul

Thanks Paul.

Hi everyone

I am getting the same error as discussed in this topic in Gromacs version 2020.5. As far as I understand from here the 2020.5 version is a fixed version. This issue only pertains with runs longer than 500 ns. Moreover even if append and rerun the the simulation using .cpt file, the mdrun runs for approximately 20 mins and stops at the same step as before. I am attaching the screenshot of the error here for reference.

This was fixed in version 2021.4, so 2020.5 is still affected. Unfortunately a release note seems to be missing.

Thanks