Hi Ben,
Can you try putting a copy of numprocs.awk in the $AMBERHOME/test directory? I have seen the failures you have, even with two processes, but everything works fine if numprocs.awk is in the right place (I don't know why though).
Tyler
On 2010-04-20, at 9:03 PM, Ben Roberts wrote:
> Hi,
>
> I've just done a run-through of the 18 April RC, patched with Dan's patch.
>
> I also tried RISM MPI this time.
>
> I noticed a couple of things there. First is that until Amber proper is installed, numprocs.awk is missing, with the result that a number of RISM MPI tests complain with the following message (or similar):
>
> awk: can't open file ../../../test/numprocs.awk
> source line number 1 source file ../../../test/numprocs.awk
> context is
> >>> <<<
> ./Run.rism_min: line 11: [: : integer expression expected
>
> I suppose that means a copy of numprocs.awk just needs to be put in the appropriate place.
>
> Also, with 2 CPUs, things pretty much pass, but with 4 or 8, RISM MPI seems to get much more problematic. Maybe it's just rounding errors, but I'm not sure - the diffs are numerous. I've attached them.
>
> I'm in the process of posting a results summary for Tools and Amber to the wiki.
>
> Cheers,
> Ben
>
>
> <4procs.diff><8procs.diff>_______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Apr 21 2010 - 10:00:03 PDT
