allow oversubscription in OpenMPI sanity check

[ocaisa]

Hyperthreading can sometimes cause an unnecessary failure.

Alternative to #2570, closes easybuilders/easybuild-easyconfigs#13945

[ocaisa]

Test report by @ocaisa

Overview of tested easyconfigs (in order)

• SUCCESS OpenMPI-4.1.1-GCC-11.2.0.eb

Build succeeded for 1 out of 1 (1 easyconfigs in total)
login1 - Linux rocky linux 8.4, x86_64, Intel(R) Xeon(R) CPU E5-2690 v3 @ 2.60GHz, Python 3.6.8
See https://gist.github.com/08193ef862eeeffed805539ec08beb58 for a full test report.

[ocaisa]

You can see the results in the trace output (or the logs):

  >> running command 'OMPI_MCA_rmaps_base_oversubscribe=1 mpirun -n 8 /home/ocaisa/.local/easybuild/build/OpenMPI/4.1.1/GCC-11.2.0/mpi_test_hello_c' ...
  >> result for command 'OMPI_MCA_rmaps_base_oversubscribe=1 mpirun -n 8 /home/ocaisa/.local/easybuild/build/OpenMPI/4.1.1/GCC-11.2.0/mpi_test_hello_c': OK

[smoors]

if hyperthreading is the only issue here, maybe it's enough to set OMPI_MCA_hwloc_base_use_hwthreads_as_cpus=1?
see #2596

[boegel]

if hyperthreading is the only issue here, maybe it's enough to set OMPI_MCA_hwloc_base_use_hwthreads_as_cpus=1? see #2596

It's more than that, we also want the OpenMPI sanity check to pass when only a single core is available, for example...

[boegel]

Test report by @boegel

Overview of tested easyconfigs (in order)

• SUCCESS OpenMPI-3.1.1-GCC-7.3.0-2.30.eb
• SUCCESS OpenMPI-4.0.3-GCC-9.3.0.eb
• SUCCESS OpenMPI-4.1.2-GCC-11.2.0.eb

Build succeeded for 3 out of 3 (3 easyconfigs in total)
node2624.swalot.os - Linux centos linux 7.9.2009, x86_64, Intel(R) Xeon(R) CPU E5-2660 v3 @ 2.60GHz (haswell), Python 3.6.8
See https://gist.github.com/34f767f48d21a86b83a916d01694c40a for a full test report.

[boegel]

lgtm