Job submission error after compiling the MFiX solver using the SMP

MFiX
#1

Hi, everyone,

I am encountering an issue when attempting to submit a job after compiling the MFiX solver using the SMP. The command I used for submission is as follows:

env OMP_NUM_THREADS=10 ./mfixsolver_smp -f mfix.dat

However, upon execution, I receive an error indicating that there is an inconsistency between the number of requested processors and the domain decomposition, which leads to the termination of the program.

The specific error message is as follows:

 Number of SMP threads: 10


Using /home/test/test/test4_t/mfixsolver_smp.so
 MFiX (23.4.1) simulation on host: pvecluster
 Run name: test_TFM                Time: 14:42                    Date:  1- 3-2025
 Project version: 20
________________________________________________________________________


 >>>>>-----------------------------------------------------------------
MESSAGE from open_files.f:373
This is a rectangular geometry simulation.
Mesh generation will be performed.
No mesh file will be written, only the [RUN_NAME]_boundary.vtk file will be written.
 <<<<<-----------------------------------------------------------------


 >>>>>#################################################################
ERROR from check_data/check_dmp_prereqs.f:131
Error 1003: The number of requested processors is inconsistent
with the domain decomposition, (NODESi * NODESj * NODESk).
 These numbers must match.
  Number of requested processes:        1
  Domain decomposition :       10
 <<<<<#################################################################


 **********************************************************************
 Fatal error reported on one or more processes. The .LOG file
 may contain additional information about the failure.
 **********************************************************************


 Program Terminated.


ERROR STOP 1
#2

The domain decomposition specified with NODESi , NODESj , NODESk is needed for DMP runs, not SMP. If you want to use SMP, set NODESi=1 , NODESj=1 , NODESk=1

#3

Hi Dr. Dietiker,

Thank you for your solution. The issue has now been resolved.

By the way, I’ve observed that when using an equal number of cores, the parallel computing speed in my SMP setup is slower compared to a DMP setup.

Do you have any suggestions on how I could boost the parallel processing speed in an SMP setup? Any advice or solutions would be greatly appreciated.

Thanks in advance!

Best regards,

#4

SMP performance is going to be slower than DMP. If you are planning to run a lot of parallel simulations, you should consider getting a linux machine. Your university may have a linux cluster you can use.