Dear all,
Does anyone know how to run UMAT written for ABAQUS in parallel using MPI? I know that ABAQUS takes care of parallel decomposition for FE model using PMPI, IMPI or CPI. But has anyone been able to actually bring the parallel execution to the UMAT itself? I have a SA code which runs parallel using MPI itself and was wondering if its UMAT version can be run in parallel using ABAQUS MPI. I appreciate any inputs. Thanks.
Best
Adnan
why not using UMAT to call your MPI program? like 'call system(executable program)'
Dear @Qingge,
Thanks for your input. I appreciate it.
I have never tried that, but not sure if it works unless you have done it yourself. Have you? If yes, could you please be more specific?
I'm afraid, since the code runs with PGI compiler and OpenMPI, the MPI libraries btw the UMAT and ABAQUS may not match? I realized ABAQUS is using C version of mpich and it's making MPI linking issues btw different types of MPI. Also, calling to ABAQUS executable should accompany the definition of input and UMAT itself. Is that going to work? To call the mother solver from the child?
Thanks
Adnan
Previously, I did like the following. For each integration point(IP), during running the UMAT, it first generates input files in its own folder (each IP has one particular file folder). then UMAT calls an executable program which was implemented by MPI (it also runs in the folder corresponding to the right IP.). After the executable program finished, UMAT reads the output from the executable program.
ABAQUS itself runs parallelly. For each IP, the MPI program runs also parallel and it seems not conflicting with ABAQUS.
Previously I just used this method to calculate the texture and the yield function for each IP. In my model, each IP was represented by several thousands grain orientations.
Hi Qingge,
Great. Thank you very much for your kindly details. That sounds great. Knowing you have managed to run UMAT utilizing the way you suggested is promising. Thanks again!
Best
Adnan
I am giving ABAQUS Subroutine/UMAT Lectures here:
https://youtube.com/playlist?list=PL4qG7zaS9ome3teA3h3dMMEwEToKloDlv
https://www.researchgate.net/project/Creating-SUBROUTINES-In-Abaqus-Lectures-by-Dr-Ronald-Wagner
future lectures will come later this year
Dear Ronald,
Many thanks for letting us know. Sounds good.
Best,
Adnan