faster simulation by adding more cores?

Salome-Meca platform Salome-Meca installation
1

Hi everyone,

I managed to run a simulation with one contact and around 250,000 elements. but it took more than 28 hours to complete. Is there any way I can run the simulation faster in salome-meca by adding more core (like cpus and gpus); similar like Abaqus?

Thanks

Saleh

2

[upl-image-preview uuid=b07ed43a-41cd-4397-83bb-ecd7a758ca50 url=

image.png
image.png418×441 15.5 KB
alt={TEXT?}]

I am asking how to edit the number of MPI CPU here

3

Hello,

You need to compile a parallel version of code_aster inside singularity container. See for example GitHub - bursica/Code_Aster-MPI-in-Singularity-of-SM2022 · GitHub

Or use docker container.

4

Hi,

As Vishwas mentioned, one option for obtaining an MPI-compiled version is to use a Docker container, such as those available on Simvia’s Docker Hub: https://hub.docker.com/r/simvia/code_aster. We have set up a utility to make switching between versions even easier: cave.

Once you have access to a version that includes MPI, you can modify the .export file, as demonstrated in the case of the perf018a benchmark https://gitlab.com/codeaster/perf/-/blob/main/astest/perf018a.export

5

Hello,

If you just want to exploit more than one core of your single CPU you can use OpenMP by simply specifying the number of threads here:

[upl-image-preview uuid=8540517d-c71b-4ba3-8a03-dd5677f42b48 url=

multi-thread.png
multi-thread.png526×567 60.5 KB
alt={TEXT?}]

Regards,

Pier