mpiexec.hydra versus mpirun in Elmer context

General discussion about Elmer
Post Reply
putanowr
Posts: 7
Joined: 14 Dec 2020, 19:53
Antispam: Yes
Location: Cracow, Poland

mpiexec.hydra versus mpirun in Elmer context

Post by putanowr »

Hi Everybody,

I would like to ask if you have some experience in using mpiexec.hydra versus mpirun
in running Elmer?
Apparently on my Linux box (Debian) using mpiexec.hydra makes ElmerSolver_mpi
fail to produce TEST.PASSED_#np.

The relevant part of the source code is in ElmerSolver.F90 between lines 812-827
however I have not jet decoded the link between it and the use of mpiexec.hydra.

IF( FinalizeOnly ) THEN
IF( ParEnv % MyPe == 0 ) THEN
IF( ParEnv % PEs > 1 ) THEN
! Parallel test, add the number of tasks as a suffix
WRITE(PassedMsg, '("TEST.PASSED_",I0)') ParEnv % PEs
OPEN( 10, FILE = PassedMsg )
ELSE
OPEN( 10, FILE = 'TEST.PASSED' )
END IF

Of course it is easy to reconfigure Elmer and force it to use mpirun (did it and all
parallel tests are OK), so the post is more curiosity driven than necessity.

Best regards,

Roman
Post Reply