Am 06.09.2007 um 17:58 schrieb Capehart, William J:
We are looking to upgrade our aged version of PGI CDK (Portland
Group Fortran and C family compilers) on our cluster. However,
since we only compile our models on the master node we would like
to get a single node license for the master node without using the
CDK option and paying the CDK price (you can only enter our cluster
environment through the master node anyway rather than some
clusters where you can join and do your dogwork on any node). We
would then compile MPICH and move onward.
The single node license is fine. With any parallel lib you can run as
many processes as you want in the cluster. The limit is in the number
of used cores inside a machine, if you use threads created by OpenMP.
But they just raised the limit for OpenMP also.
http://www.pgroup.com/pricing/index.htm
-- Reuti
We have heard some chatter that we can do this, and we have also
heard chatter that PGI single node compilers (even when run though
MPICH) are hobbled so that they can only be applied to 4 processors
(accounting for dual core systems).
Obviously we’d rather not shell out big bucks for a PGI CDK kit
compiler that will only be compiling on one node.
Does anyone have any experience or insight on this problem.
Cheers and Thanks much!
================================================================
Bill Capehart <[EMAIL PROTECTED]> Associate Professor
Institute of Atmospheric Sciences Land Surface Processes
213 Mineral Industries Building Hydrometeorology
South Dakota School of Mines and Technology Ph: +1-605-394-1994
501 East Saint Joseph Street Fax: +1-605-394-6061
Rapid City, SD 57701-3995 Mobile: +1-605-484-5692
=================== http://capehart.sdsmt.edu ==================
_______________________________________________
Beowulf mailing list, Beowulf@beowulf.org
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf
_______________________________________________
Beowulf mailing list, Beowulf@beowulf.org
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf
