I have a piece of software that is compiled with the intel mpiicc and mpiicpc compilers. I am running it on a cluster which has compute nodes with both Intel and AMD processors. It works fine on the intel processors but it does not run at all (hangs indefinitely with no error) on the AMD processors.
Cluster support believes it is due to intel specific flags causing issues, but I am not using very many flags in the compilation and the ones I am I don't think are the problem. (-fp-model, -fPIC, -m64)
So presumably if this is the issue, it is something happening automatically. Are there flags I could use to have it compile very generically with no processor specific optimizations?
I was attempting to use the flags recommended here: https://prace-ri.eu/wp-content/uploads/Best-Practice-Guide_AMD.pdf. The recommended ones "-march=core-avx2 -fma -ftz -fomit-frame-pointer" did not fix the issue, and trying the "-axHost" flag which is described there as making lowest common denominator code gave an error stating it was not supported.