First time here? Checkout the FAQ!
0 votes
by (1.2k points)

I have heard from Axel in a previous post that the --np option is not supported out-of-box with bench. This seems true for bench jobs directly executed from the command line. 

However, when I run a bench job through a carputils script with the --np option, it looks like mpiexec is trying to incorporate the multiple processors, as can be seen from the jumbled order of calculations in the output:



9665.000 -8.57911249e+01 +4.29343760e-04 

  9477.000 -8.56104049e+01 +1.91498933e-03 

  9065.000 +2.58231244e+01 +5.72370272e-02 

  9491.000 -8.56355565e+01 +1.68398813e-03 

  9478.000 -8.56123109e+01 +1.89740193e-03 

  9666.000 -8.57915528e+01 +4.26432711e-04 



My question is, is the --np option effectively forcing multiprocessing for bench?

1 Answer

0 votes
by (7.8k points)
selected by
Best answer

While you can run bench on multiple cores, there is nothing beneficial about it. There is no speedup to be gained.

Additionally, you may run into bugs since bench is not regularly tested with MPI nowadays.

Best, Aurel
Welcome to openCARP Q&A. Ask questions and receive answers from other members of the community. For best support, please use appropriate TAGS!
architecture, carputils, documentation, experiments, installation-containers-packages, limpet, slimfem, website, governance