[ET Trac] [Einstein Toolkit] #1075: Test cases with specific number of processes

Einstein Toolkit trac-noreply at einsteintoolkit.org
Wed Sep 19 13:57:30 CDT 2012

#1075: Test cases with specific number of processes
  Reporter:  eschnett     |       Owner:        
      Type:  enhancement  |      Status:  review
  Priority:  major        |   Milestone:        
 Component:  Cactus       |     Version:        
Resolution:               |    Keywords:        

Comment (by knarf):

 I agree that simfactory usually handles/parses/changes these lists.
 However, simfactory's wouldn't know here that one particular testsuite
 should be run with a different number of processes than specified (even to
 simfactory). The problem isn't really made easier because of simfactory.

 If an mpi implementation wraps around there should be no problem, yes.
 However, some don't because especially when asking for 1 process (and,
 e.g., one openmp thread, you might actually get only part of a node and
 running two processes there would oversubscribe the system and hurt other
 users (granted, using openmp would as well). In any case: what do you
 propose to do on systems that are not wrapping around? Currently the test
 would be marked as failure.

 Could we not avoid this issue by only overriding the users' choice if the
 number of processes specified by the testsuite is smaller than what the
 user specified?

Ticket URL: <https://trac.einsteintoolkit.org/ticket/1075#comment:18>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit

More information about the Trac mailing list