I just downloaded the 15-day evaluation version of the Linux amd64 compiler and I’m trying to get our cosmological code to run with OpenMP (the directives were already present and working fine with the Intel compiler).
The code compiles fine with pgf90 with the options:
-O2 -tp amd64 -mp -Minfo=mp -byteswapio
No warnings, just the parallelization diagnostics.
However when I run the code, I can see that only one processor is being used.
I’ve exported NCPUS and OMP_NUM_THREADS=4
Is explicit (OpenMP) parallelization forbidden in the demo version?