Gromacs with fast multipole method? (building issues/stability of port)

User discussions
1

GROMACS version: 2019
GROMACS modification: Yes

I’ve been trying to build and run some benchmarks associated with the gromacs FMM electrostatics code (see: https://journals.sagepub.com/doi/full/10.1177/1094342020964857 and SPPEXA | Max Planck Institute for Biophysical Chemistry) by the Grubmüller group but I keep running into segfaults when I build using my standard options (which is the suggested method by the authors).

I was planning on contacting the authors for help, but before I did so, I was wondering if anyone else has had any issues trying to build this version and if they had any advice.

As an aside, does anyone know if/when FMM will become part of the standard gromacs package?

Thank you,

-Micholas

2

Hi Micholas,

do the salt water and multi-droplet benchmark input files run? Have you set the GMX_USE_GPU_BUFFER_OPS environment variable?

Best, Carsten

3

Hi Carsten,

I have been using the GMX_USE_BUFFER_OPS option and the system starts for a moment and then crashes with segfault when I use a test system that I know works with PME and RF electrostatics (10nm radii vesicle in water). I’ll give the salt-water and droplet benchmarks a go and let you know.

-Micholas

4

It could be the memory allocation in the FMM module, which is not yet sophisticated for sparse systems. But we know how it should be fixed. If you want, you can send us your system so we can check where the problem occurs and then hopefully fix it.