Specific MPI used in Catalyst build vs Simulation build

I am curious about the ParaView users experience using catalyst and a variety of compilers, specifically when it comes to different MPIs…

My usual practice for software like Catalyst is to install them with the system’s GNU compiler, this usually allows them to get along with my (F90) code no matter what compiler I use with my code.

However Catalyst’s use of MPI seems to throw a wrench in this plan.

It is standard practice to build Catalyst with the same compiler/MPI combo that one intends to use for their simulation?



I’ve always had the least amount of trouble when NOT mixing and matching compilers and MPI libraries. I don’t know if I’ve ever gotten around mixing MPI libraries with Catalyst and the simulation code. I suppose if the MPI libraries are ABI compatible it may work but I never was knowledgeable enough about that or trusted that to go that route.

For non-matching compilers, I’ve been able to go with Intel compilers for the simulation code and GCC for Catalyst but that’s probably because Intel compilers are already meant to work with the GNU system libraries and header files. In practice though I wouldn’t recommend that either.


1 Like