[molpro-user] molpro using more than 100% CPU with em64t
jlane at alkali.otago.ac.nz
Wed Jul 25 22:14:57 BST 2007
I am getting some unusual behaviour when running jobs on our new Xeon
multicore cluster (8 cores per node). When run in parallel (using tcgmsg)
molpro correctly creates 'n' execution threads and 1 shepherd thread.
However, the 'n' execution threads use all available CPU resources. If I
request molpro -n4, each thread uses 200% CPU and so the 'load' on the
system is seen as 8 not 4 as it should be. Is this a consequence of using
tcgmsg and will mpi solve this problem? Or is the problem more to do with
architecture and if so how can I stop molpro jobs from using more than
their allocated CPU.
As always, any help would be most appreciated.
Department of Chemistry
University of Otago
P.O. Box 56
Dunedin, New Zealand
ph +64 3 479 5378
email : jlane at alkali.otago.ac.nz
More information about the Molpro-user