it seems that generation of species (inside a real job, using species keywords) does not work in parallel with MPI.
Probably you won't see the problem on single machine,
but you will see it from time to time with asynchronous nfs.
Should be easy to fix: generate the species on on the master rank.
Thanks for pointing that out. I'll fix that for the next release.
I looks like the problem with running jobs in parallel/mpi with species keyword is still there (elk-2.2.10). In the elk.in:
1 0 1 2
Using MPI, number of processes : 2
Info(genspecies): running Z = 76, (osmium)
Error(readinput): invalid block name : 76
Log in to post a comment.