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

[Einstein Toolkit]

#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 hinder):

 HPC machines must support the ability to run different numbers of MPI
 processes due to not all codes supporting OpenMP.  At least up to the
 number of cores, there is always a way to run that many MPI processes.
 For a given machine, we will be able to come up with simfactory command-
 line arguments which give the correct number of MPI processes, and then we
 can choose the number of threads separately from this.  Assuming that
 simfactory requests resources sufficient to run the maximum number of MPI
 processes required by any test, there is no limit imposed by the queuing
 system.

 A separate issue (as I think you are referring to) is that if simfactory
 is configured to run a simulation on N processes, the run script will have
 this hard-coded after expansion by simfactory.  i.e. the run script (which
 is used in the "mpirun" command by the simfactory testsuite mechanism,
 http://git.barrywardell.net/simfactory2.git/blob/HEAD:/bin/RunTestSuite)
 doesn't look at the -np option given by the test system.  We should think
 about how we can modify simfactory to run the correct number of MPI
 processes in this case, unless Erik has done so already.

 Erik: I believe the only potential problem is the one related to
 simfactory, and that is definitely a problem unless it has already been
 fixed.

-- 
Ticket URL: <https://trac.einsteintoolkit.org/ticket/1075#comment:20>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit