Specifying the number of processors to run phenix with
Hi- anyone know how to specify the number of processors to run phenix under in linux? Or how to tell what the default is? I'm running linux (fedora) with 4 processors and can't tell whether phenix is using all of them or not. I do the refinement with .csh files and tried doing "myfile.csh --nproc=4". The file ran but no faster- I can't tell if it is using 4 by default or if that's the incorrect command to specify the use of 4 processors. Thanks for your help! -Sam
Hi Sam, phenix.refine cannot use advantage of multiple CPUs. I think the "--nproc=4" keyword in your example below is a valid keyword for phenix.autobuild, but not for phenix.refine. Pavel. On 8/31/09 1:33 PM, Sam Stampfer wrote:
Hi- anyone know how to specify the number of processors to run phenix under in linux? Or how to tell what the default is? I'm running linux (fedora) with 4 processors and can't tell whether phenix is using all of them or not. I do the refinement with .csh files and tried doing "myfile.csh --nproc=4". The file ran but no faster- I can't tell if it is using 4 by default or if that's the incorrect command to specify the use of 4 processors. Thanks for your help! -Sam ------------------------------------------------------------------------
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
On Aug 31, 2009, at 1:41 PM, Pavel Afonine wrote:
phenix.refine cannot use advantage of multiple CPUs. I think the "-- nproc=4" keyword in your example below is a valid keyword for phenix.autobuild, but not for phenix.refine.
Actually, it's just "nproc=4" without the "--". You can use multiple CPUs for the FFTs in phenix.refine if you compile from source with OpenMP support (add "--openmp" to the arguments when running the install script; this requires GCC 4.2 or better, or Intel's compiler, and perhaps others). I think it will automatically use as many CPUs as possible. It's probably not going to help much, though. ------------------- Nathaniel Echols Lawrence Berkeley Lab 510-486-5136 [email protected]
Thanks for the info. I ran two instances of phenix which both were able to
run at 100%, but ran out of memory after awhile My guess is the memory is a
bigger limit than the processors so I think I'll just stick to running on
the one processor. Thanks for everyone's help!
-Sam
PS: I had trouble attempting an installation from source earlier (and don't
have full permissions on this machine). Since you don't think upping the #
of processors will help much, I'm not going to attempt the source
installation again.
On Mon, Aug 31, 2009 at 5:22 PM, Nathaniel Echols
On Aug 31, 2009, at 1:41 PM, Pavel Afonine wrote:
phenix.refine cannot use advantage of multiple CPUs. I think the "-- nproc=4" keyword in your example below is a valid keyword for phenix.autobuild, but not for phenix.refine.
Actually, it's just "nproc=4" without the "--".
You can use multiple CPUs for the FFTs in phenix.refine if you compile from source with OpenMP support (add "--openmp" to the arguments when running the install script; this requires GCC 4.2 or better, or Intel's compiler, and perhaps others). I think it will automatically use as many CPUs as possible. It's probably not going to help much, though.
------------------- Nathaniel Echols Lawrence Berkeley Lab 510-486-5136 [email protected]
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
participants (3)
-
Nathaniel Echols -
Pavel Afonine -
Sam Stampfer