[ase-users] disabling parallelism in ASE MD

[Bernstein, Noam CIV USN NRL (6393) Washington DC (USA)]

On Apr 18, 2020, at 1:33 PM, Noam Bernstein via ase-users <ase-users at listserv.fysik.dtu.dk<mailto:ase-users at listserv.fysik.dtu.dk>> wrote:

Thanks Ask.  I'll try it.  I agree that it'd be nicer if automated mpi usage was more explicit, ideally in a way that let the calling program specify the communicator.

On Apr 17, 2020, at 5:48 PM, Ask Hjorth Larsen <asklarsen at gmail.com<mailto:asklarsen at gmail.com>> wrote:

I think you can apply this hack:

from ase.parallel import world, DummyMPI
world.comm = DummyMPI()


To resurrect a long-dead thread, I tried your hack, and it works.

On a related note, I'm using ASE's VASP calculator in a weird mode - I have an MPI python program and each MPI task runs a serial VASP calculation.  If I do not do this kind of hack to disable parallel I/O, the "vasprun.xml" reads crash. Should the VASP calculator output-file reading commands default to parallel=False?  Or is there some other use case where parallel I/O would be helpful?

Noam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.fysik.dtu.dk/pipermail/ase-users/attachments/20210304/b3a11e30/attachment-0001.htm>