MPI Rattle Segfaults

[johnemmanuelbond]

Hi hoomd devs,

Since version 3.9 I have been struggling getting domain decomposition to work with molecular dynamics sims of particles confined to a surface using rattle. Until recently I was just ineffeciently running sims on one core, but with recent changes to my university's supercomputing cluster I need to figure this out before I waste more resources than just my own time.

I've attached below a folder with a simple script that recreates the segfault errors in hoomd version 4.0.1. It simply runs md.methods.rattle.Brownian on a spherical surface with some circular DLVO particles. When I run it on one core it works, when I try to run it on 8 it always segfaults. There's more info in the readme, but any help resolving this issue would be greatly appreciated.

Thanks!
- Jack Bond

hoomd_rattle_mpi_segfault.zip

[joaander]

Thanks for the complete test case. I was able to reproduce the error and fix it: #1748.

[johnemmanuelbond]

Excellent, Thanks! Is this in a release version I can clone and build?

[joaander]

Not at this time. It will be included in the 4.7.0 release.