Thanks, Bob and Ross. This simple configure+compile worked
perfectly. The performance seems pretty good at about 12ns/day/64
cores/ 27000 atoms compared to the same job on another machine and the
Amber 9 JAC benchmark for the XT3. Much appreciated!
Patrick
On Oct 29, 2008, at 10:31 PM, Robert Duke wrote:
> Hello Patrick,
> You probably want to run pmemd for most of your amber simulations on
> an xt3. You could probably also configure and run the rest of amber,
> but I don't have that information readily available, as my account
> on an xt3 is inactive at the moment. However, for pmemd, just do:
>
> cd $AMBERHOME/src/pmemd
> ./configure cray_xt3 pgf90 mpi pubfft
> make install
>
> The last time I checked, it was better to not use the binary
> trajectory file capability (which would have been made available by
> adding "bintraj" at the end of the configure line) on this type of
> machine, as the lustre file system did not really do all that well
> in that mode (and you would have to either find netcdf on the
> machine or get it built as part of amber). I'll have to check into
> this again for you all if file i/o is an issue on sapphire (I don't
> know this particular machine).
>
> Let me know if you have any problems, and give my best to Ken!
>
> Best Regards - Bob Duke
>
> ----- Original Message ----- From: "Patrick McCarren" <mccarren.chem.ucla.edu
> >
> To: <amber.scripps.edu>
> Sent: Thursday, October 30, 2008 12:30 AM
> Subject: AMBER: Optimal compiling of amber on XT3
>
>
>> I'm trying to compile AMBER on a Cray XT3 at the ERDC MSRC
>> (sapphire). What are the optimal settings for libraries, mpi/mpich/
>> mpich2, compiler, etc. for AMBER? Am I right that the test
>> programs in the configure script will not run on the login node?
>> I can switch to pathscale or whichever compiler is best I think.
>> If I could get a configure script or configure diff from someone,
>> I would be really happy. Thanks,
>>
>> Patrick
>>
>>
>> Patrick McCarren
>> Graduate Student Researcher
>> Houk Research Group
>> UCLA Dept of Chem and Biochem
>> 607 Charles E Young Dr E
>> Los Angeles, CA 90095-1569
>>
>> -----------------------------------------------------------------------
>> The AMBER Mail Reflector
>> To post, send mail to amber.scripps.edu
>> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
>> to majordomo.scripps.edu
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
> to majordomo.scripps.edu
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
to majordomo.scripps.edu
Received on Sun Nov 02 2008 - 05:09:57 PST
