running roms on a grid with e-w periodic boundary conditions and compilation done with mpich1 works fine (32-bit).
when I compile it using mpich2 (and 64 bit, on another cluster), it generates negative salinities in two spots. I ran it on yet another cluster (64 bit) without problem.
Anyone had this problem with mpich2 + periodic conditions? suggestion, comments welcome. thank you.
Guillermo.