balaji at mcs.anl.gov
Wed Oct 28 14:47:30 CDT 2009
Moving this email to the the MPI3 Hybrid mailing list, so others can
comment as well.
On 10/28/2009 02:29 PM, Pavan Balaji wrote:
> (cc'ing Rajeev)
> On 10/28/2009 02:16 PM, Snir, Marc wrote:
>> good comments -- sorry I missed the phone call. Any other issues that
>> I need to address in the proposal?
> One more comment that Rajeev had pointed out was with respect to the
> footnote on page 4 with respect to dynamic processes. Each dynamic
> process does its own MPI_Init(_thread) call, so it'll specify its own
> thread level. What interaction is missing?
> Btw, MPI-2.2 pg. 385 lines 8-9 explicitly say that each process can have
> a different thread-level in MPI_COMM_WORLD. Maybe we need to modify this
> to "all processes in the MPI application". Will that be sufficient for
> this proposal?
> -- Pavan
More information about the mpiwg-hybridpm