itaps-parallel Notes from today's meeting

tautges at mcs.anl.gov tautges at mcs.anl.gov
Mon Apr 19 09:16:39 CDT 2010


I think we mean one part per rank in a communicator, which is  
inherently bound to an mpi model.

- tim


On Apr 19, 2010, at 8:47 AM, Mark Beall <mbeall at simmetrix.com> wrote:

> After reading the various emails here I have a terminology question:  
> when people are referring to a "process" in the context where there  
> may be multiple threads are they saying "thread"="process" or a  
> single "process" may have multiple threads.
>
> The reason for the question is that we do support multiple parts per  
> process, however we really only use this in a multithreaded context  
> where each thread gets a part, thus this could be the same as saying  
> "one part per process" in the case where "process"="thread".
>
> mark
>


More information about the itaps-parallel mailing list