AMBER Archive (2009)

Subject: Re: [AMBER] how many processors for parallel test

From: michael bane (michael.bane_at_manchester.ac.uk)
Date: Thu Jan 08 2009 - 07:31:31 CST

On 8 Jan 2009, at 12:08, David A. Case wrote:

> On Thu, Jan 08, 2009, michael wrote:
>
>> Morning folks. I'm new to Amber and have been tasked with installing
>> Amber10 on a local machine. The installation appears to have
>> succeeded
>> and I wished to try the provided tests. However, if I use 4
>> processors
>> as suggested in the documentation then for the
>> cd tgtmd/change_target.ntr && ./Run.tgtmd
>> test I get an error:
>> Must have more residues than processors!
>>
>> If I reduce the number of processors to 2 I then get errors such as
>>
>> cd ti_mass/pent_LES_PIMD && ./Run.pentadiene
>> This test not set up for parallel
>> cannot run in parallel with #residues < #pes
>> make[1]: Leaving directory
>> `/software/horace/applications/amber/amber10/test'
>> cd PIMD/full_cmd_water/equilib && ./Run.full_cmd
>> This test case requires a least 4 mpi threads.
>> The number of mpi threads must also be a multiple of 4 and not more
>> than 32.
>> Not running test, exiting.....
>
> The above are not errors. Some of the test cases have limitations on
> the number of processors that can be used, as the messages
> indicate. So
> the test suite just skips these tests.
>

I agree it's not an error, but I was expecting the test to run all
codes ;)

>>
>> Finally, I also see the error
>> ** Error ** : could not determine NUMPROCS in
>> '/software/horace/applications/amber/amber10/test/ncsu/premd'
>
> We would need to know what your DO_PARALLEL variable is set to: see
> $AMBEHROME/test/ncsu/common.sh. NUMPROCS is extracted from this
> variable by a script that may not cover all possibilities; or your
> environment variable might be wrong...

For our IA64 box to run interactively we use 'prun' so my DO_PARALLEL
was set
  export DO_PARALLEL='prun -B4 -n2 -p login'

Thanks, Michael

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber