Use mpiexec to run unit tests

[quantumsteve]

No description provided.

[quantumsteve]

test this please

[quantumsteve]

Test this please

[quantumsteve]

Test this please

[idigs]

Test this please

[idigs]

Test this please

[idigs]

Test this please

[quantumsteve]

Test this please

[quantumsteve]

@cianciosa mars_test passes on one process but is failing failing on two or more processes. Are there random number initialization differrences?

[mbeidler3]

@quantumsteve I've been trying to create gold data for multiple processor runs. However, when running on my machine, each processor has rank=0 and nmpi=1. Also, it's even more confusing because only one processor prints the output.korc file even though that task is done by rank=0. I compiled the recent version on Perlmutter using their module system (cray-mpich for MPI, rather than openmpi on my system) and it initialized the MPI processes as expected.

[mbeidler3]

@quantumsteve, I'm getting the same result for a build with mpich. On my machine, I see it compiling with gfortran and including the MPI library directoy, while on Perlmutter I see it compiling with the wrapper compiler ftn. Could the issue be that my system is not compiling with the correct MPI-wrapped compilers?

[quantumsteve]

@quantumsteve, I'm getting the same result for a build with mpich. On my machine, I see it compiling with gfortran and including the MPI library directoy, while on Perlmutter I see it compiling with the wrapper compiler ftn. Could the issue be that my system is not compiling with the correct MPI-wrapped compilers?

Something is going on with the wrapper. If I add export FC=mpifort to build.sh, I get one output file per process. Not setting FC and I only get a single file for ranks 1, 2, and 4. With rank 8, I get 8 files in both cases 😕

[quantumsteve]

Test this please

[idigs]

I approve these changes