There is a known issue with the PIMD code where it will hang if you
try to use more processors than beads (most of the PIMD tests use 4
beads I think). This bug is currently being worked on.
If you don't need PIMD functionality and you want to test everything
else with >4 processors you could just comment those tests out of the
test Makefile for now. Everything else should be ok.
-Dan
On Wed, Jul 28, 2010 at 1:08 PM, Marc Moreau <jebnor.gmail.com> wrote:
> Sorry to double post, but also in the timing of the 'make
> test.parallel' I got the following
>
> == BEGIN ==
> # time make test.parallel
> ... CUT ...
> <CTRL-C>
> 222 file comparisons passed
> 0 file comparisons failed
> 0 tests experienced errors
> Test log file saved as logs/test_amber_parallel/2010-07-27_15-50-33.log
> No test diffs to save!
>
>
> real 1145m22.168s
> user 1m32.174s
> sys 0m10.422s
> #
> == END ==
>
> The real time is completely off from the user and system time.
> Perhaps there is an MPI message passing loop in the code, or in my MPI
> implementation.
>
> -- Marc
>
> On Wed, Jul 28, 2010 at 11:03 AM, Marc Moreau <jebnor.gmail.com> wrote:
>> Hello all,
>>
>> I'm running test of my parallel build and I am experiencing an odd
>> even. When the test gets to the
>> " cd PIMD/full_cmd_water/equilib && ./Run.full_cmd " test, it seems to
>> take forever. I was running:
>>
>> # export DO_PARALLEL='mpiexec -n 8'
>> # cd $AMBERHOME/test
>> # make test.parallel
>>
>> and it ran for more than 8 hours without returning. It seemed to have
>> run properly with 2 and 4 cores.
>>
>> All the tests up to that point passed without fault.
>>
>> -- Marc
>>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jul 28 2010 - 10:30:06 PDT
