openmpi5.0.7 with Intel2021 can not compile simple MPI program, error #6633

26 views
Skip to first unread message

Miroslav Iliaš

unread,
May 28, 2025, 11:22:08 AMMay 28
to Open MPI users
Dear OpenMPI community,

we have new openmpi5.0.7 library compiled with Intel2021 compilers.

Hower, compiling simple program fails with errors  #6633 .

Here it is: 

and the whole directory :

Best,

Miro



Jeff Squyres (jsquyres)

unread,
May 28, 2025, 1:13:00 PMMay 28
to Open MPI users
I am not a Fortran expert, but you did not declare ierr to be a specific type (e.g., integer).

If you declare it properly, does the problem go away?
From: us...@lists.open-mpi.org <us...@lists.open-mpi.org> on behalf of Miroslav Iliaš <miro....@gmail.com>
Sent: Wednesday, May 28, 2025 11:22 AM
To: Open MPI users <us...@lists.open-mpi.org>
Subject: [OMPI users] openmpi5.0.7 with Intel2021 can not compile simple MPI program, error #6633
 
To unsubscribe from this group and stop receiving emails from it, send an email to users+un...@lists.open-mpi.org.

Miroslav Iliaš

unread,
May 28, 2025, 2:39:37 PMMay 28
to Open MPI users, Jeff Squyres (jsquyres)
Hello,

declaration of ierr does not help.  OpenMPI 5.0.7 has some strange deklaration of "dummy argument", giving this error: 

pi_upd.F90(40): error #6633: The type of the actual argument differs from the type of the dummy argument.   [IERR]
pi_upd.F90(44): error #6633: The type of the actual argument differs from the type of the dummy argument.   [MYID]
pi_upd.F90(45): error #6633: The type of the actual argument differs from the type of the dummy argument.   [NUMPROCS]

Miro

Jeff Squyres (jsquyres)

unread,
May 28, 2025, 2:50:20 PMMay 28
to Open MPI users
How did you build Open MPI?  Did you specify "-i8" for the FCFLAGS, perchance?

What is the output of mpifort --showme?

Open MPI definitely defines the mpi module MPI_INIT interface to take an integer as its parameter: https://github.com/open-mpi/ompi/blob/v5.0.x/ompi/mpi/fortran/use-mpi-ignore-tkr/mpi-ignore-tkr-interfaces.h.in#L2284-L2286

(or, if the ifort 2021 compiler falls into the "we don't support ignore TKR" types of Fortran compilers, then https://github.com/open-mpi/ompi/blob/v5.0.x/ompi/mpi/fortran/use-mpi-tkr/mpi-f90-interfaces.h#L1109-L1111)

Either way, somehow your program is not matching what Open MPI was built with.  One possible way for this is if Open MPI was configured and built with something like -i8, which would make all default integers be 8 bytes instead of 4 bytes.  Or perhaps mpifort is automatically adding something like -i8.  IIRC / I'm pretty sure that neither of these things should have happened unless configure was told to build Open MPI this way, though.
Sent: Wednesday, May 28, 2025 2:39 PM

To: Open MPI users <us...@lists.open-mpi.org>
Cc: Jeff Squyres (jsquyres) <jsqu...@cisco.com>
Subject: Re: [OMPI users] openmpi5.0.7 with Intel2021 can not compile simple MPI program, error #6633
 

Miroslav Iliaš

unread,
May 28, 2025, 4:22:08 PMMay 28
to Open MPI users, Jeff Squyres (jsquyres)
Dear Jeff,

indeed, you are right, the module is not properly built.


mpirun --version
mpirun (Open MPI) 5.0.7

module -v display openmpi/v5.0.7_icc2021
-------------------------------------------------------------------
/cvmfs/hybrilit.jinr.ru/sw/slc7_x86-64/modulefiles/openmpi/v5.0.7_icc2021:

module-whatis    Modulefile for openmpi v5.0.7 with icc 2021
module           load BASE/1.0 intel/oneapi
setenv           OMPI_ROOT /cvmfs/hybrilit.jinr.ru/sw/slc7_x86-64/openmpi/v5.0.7_icc2021
prepend-path     PATH /cvmfs/hybrilit.jinr.ru/sw/slc7_x86-64/openmpi/v5.0.7_icc2021/bin
prepend-path     C_INCLUDE_PATH /cvmfs/hybrilit.jinr.ru/sw/slc7_x86-64/openmpi/v5.0.7_icc2021/include
prepend-path     CPLUS_INCLUDE_PATH /cvmfs/hybrilit.jinr.ru/sw/slc7_x86-64/openmpi/v5.0.7_icc2021/include
prepend-path     LIBRARY_PATH /cvmfs/hybrilit.jinr.ru/sw/slc7_x86-64/openmpi/v5.0.7_icc2021/lib
prepend-path     LD_LIBRARY_PATH /cvmfs/hybrilit.jinr.ru/sw/slc7_x86-64/openmpi/v5.0.7_icc2021/lib

Jeff Squyres (jsquyres)

unread,
May 28, 2025, 4:49:50 PMMay 28
to Miroslav Iliaš, Open MPI users
Does that mean you've fixed the environment module, and now everything is working as expected / your program compiles properly with Open MPI v5.0.x?
From: Miroslav Iliaš <miro....@gmail.com>
Sent: Wednesday, May 28, 2025 4:22 PM

Miroslav Iliaš

unread,
May 29, 2025, 3:01:20 AMMay 29
to Open MPI users, Jeff Squyres (jsquyres), Miroslav Iliaš
No, it is not fixed...when I get someting, I will report here.

M.

Reply all
Reply to author
Forward
0 new messages