PROPTI is an interface tool that couples simulation models with algorithms to solve the inverse problem of material parameter estimation in a modular way. It is designed with flexibility in mind and can communicate with arbitrary algorithm libraries and simulation models. Furthermore, it provides basic means of pre- and post-processing.
During our BESKID team meeting at the weekend, we realised that propti cannot yet be executed in parallel within the Docker container using mpiexec.
I have solved the problem and would be very grateful if you could merge this pull request so that we can run propti faster at BCL.
It was due to the Python package mpi4py, which is provided by default via PIP for OpenMPI and not for Intel MPI. I have therefore now replaced ‘libmpich-dev’ with ‘libopenmpi-dev’. The FDS simulation itself is still parallelised by Intel MPI.
The propti images will then be rebuilt automatically.
TristanHehnen
commented
4 months ago
Hi @TristanHehnen,
During our BESKID team meeting at the weekend, we realised that propti cannot yet be executed in parallel within the Docker container using mpiexec.
I have solved the problem and would be very grateful if you could merge this pull request so that we can run propti faster at BCL.
It was due to the Python package mpi4py, which is provided by default via PIP for OpenMPI and not for Intel MPI. I have therefore now replaced ‘libmpich-dev’ with ‘libopenmpi-dev’. The FDS simulation itself is still parallelised by Intel MPI.
The propti images will then be rebuilt automatically.
Kind regards Robert