Facing problems to run an openmp fortran program with gfortran 8.2.0 that is shipped with OM LX 4.0 alpha1. This does not happen in OM LX 3.03 with gfortran 7.3 and in RHEL with a compilation of gcc 8.1.
Can’t say this is not caused by a combination of my own mistakes and bug fixes in gcc-gfortran.
Just in case, let me ask if, in OM LX 4.0 alpha1, gcc was build with --disable-libmpx option. This libmpx option is known to cause problems and it is said to be removed from gcc in next releases.
In search for the culprit for my problems with segmentation fault and Nan, inf values when running my code compiled with gfortran 8.2.0, given that no problem seems to happen running at OM LX 3.03 and at RHEL with gcc 8.1.X (compiled by my job’s IT people), I’ve found this,
This switch tells GCC to not build mpx (Memory Protection Extensions) that can cause problems on some processors. It has been removed from the next version of gcc.
I wonder if this is not a kind of GNU’s itself recommendation for building gcc 8.2.0 …
I already had this version. Anyway, I’ve reinstalled it (dnf reinstall gcc …). And no luck. Still don’t know what can be wrong! Could be my own mistake with openmp rules (most likely), could be some gcc bug since different gcc releases yield different outcomes, and, of course, could be a combination of everything.
Beginning to consider a customized installation of gcc …