[mpich-discuss] Unstable programs

[Gib Bogle]

I am getting bizarre behaviour with Intel Fortran and Windows on a quad-core machine.  My program 
fails in a wide variety of ways.  Sometimes it is access violation, sometimes a stack overflow. 
Sometimes it runs successfully but I can see that some variables have been clobbered.  I have all 
compiler runtime checking turned on, and in some situations an out-of-range array index is trapped, 
but the reason for the array error is quite obscure, except that it occurs after an MPI call. 
Attempts to pin down these errors by adding write statements usually have the effect of pushing the 
error somewhere else - Heisenbugs.

There is always the possibility - usually a high probability - of operator error, but I've been 
working on this intensively for several days, and the variety and slipperiness of the errors is 
making me wonder whether the fault is perhaps not with my code, which is not very complex.  I'd like 
to know if others are using MPICH2 with Fortran under Windows, and what their experience has been.

Thanks
Gib