On Thu, Aug 15, 2019, Roy Ashworth wrote:
>Looking through the log files it looks like many of the comparison
>failed/errors may not be a problem but I am not in a position to know.
>After export DO_PARALLEL="mpirun -np 2", and make test, the summary output is
>
>988 file comparisons passed
>47 file comparisons failed (0 ignored)
>80 tests experienced errors
>After export DO_PARALLEL="mpirun -np 8", and make test, the summary output is
>
>948 file comparisons passed
>110 file comparisons failed (0 ignored)
>56 tests experienced errors
These are way too many errors to be acceptable. But, indeed, looking at the
logs (especially for errors, but also for comparison diffs) may yield some
information. But I'm guessing taht with hundreds of errors, you won't find
much there that helps.
The thing most likely to help (in my view) is to use the configure_opemnpi
or configure_mpich scripts in AMBERHOME/AmberTools/src to compile your own MPI
stack, which will be installed inside the amber18 directory tree, and will
be automatically found if you source the "amber.sh" script.
Be sure to use the same compiler to build MPI that you used to build the
serial version of Amber.
....good luck....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Aug 15 2019 - 18:30:04 PDT
