Solving PDEs in parallel (extra-node, MPI, PETSc) using Docker on HPC?

Question

Are there any experiences about openCARP - Docker Installation on HPC - Systems?

parallele Features

openCARP has been written to run in either of two parallel computation modes, a shared memory model, or a distributed memory model. A large common code base is maintained between the two versions and only low-level wrapper functions implement the specifc memory model. Parallelization is realized in the shared memory model based on OpenMP directives and native numerical libraries. For distributed memory parallelization, extensive use is made of the PETSc parallel library as well as MPI function calls.

iterative solver of elliptic PDE: ellip_options.file; parabolic PDE: parab_options.file

Is there a need for Singularity? https://sylabs.io/guides/3.5/user-guide/introduction.html

further read

https://arxiv.org/pdf/1905.08415.pdf

Answer 1

Hi! Singularity supports more than Docker to run the container on HPC systems, but from the developers' side we have no experience so far. Currently, we provide the docker installation for personal use, and for HPC systems we suggest to compile from source.