Hello !
I am running the example Confined bi-compartment poromechanical problem in fenicsX (code in section 4 of the following github: GitHub - Th0masLavigne/Dolfinx_Porous_Media: Multi-compartment porous modeling using FEniCSx: from the theory to the implementation) The tutorial of the example is available in this paper https://www.sciencedirect.com/science/article/pii/S1751616123002552
I am also running a similar problem with a more complicated geometry, which will require parallelisation. In the HPC, I started running my code on 1 node 1 core (no parallelisation), 1 node 60 cores and 20 nodes 1 core (which I understand is not optimal, but still faster than 1 node 1 core). I notice that my code on 1 node 60 cores takes the same time as 1 node 1 core.
For the sake of example, I run the same tests on the Confined bi-compartment poromechanical problem GitHub example code:
- 1 node 1 core : 33 minutes
- 1 node 60 cores : 33 minutes
- 20 nodes 1 core : 11 minutes
Is there something I am missing to parallelise the code in fenicsX ? I run the code using mpirun.
Thank you very much !