November 2009 - Mpich2-core - mailman3.cels.anl.gov

FW: [mpich-ms] MPI_COMBINER_ENUM
by Rajeev Thakur 07 Nov '09

07 Nov '09

Should we just delete MPI_UNDEFINED_RANK? It will change mpi.h and hence ABI, but it wasn't supposed to be there in the first place. Rajeev -----Original Message----- From: mpich-ms-bounces(a)mcs.anl.gov [mailto:[email protected]] On Behalf Of Anthony Chan Sent: Friday, November 06, 2009 11:55 PM To: mpich-ms(a)mcs.anl.gov Subject: Re: [mpich-ms] MPI_COMBINER_ENUM The one in mpe2 must be copied from top-level mpi.h when I created the serial version of mpi stub a while back. I notice the MPI_UNDEFINED_RANK exists in all versions of MPICH2 up to at least 1.0.3 and MPICH 1.2.6, it looks like these constants may have been copied from MPICH-1. A.Chan ----- "Rajeev Thakur" <thakur(a)mcs.anl.gov> wrote: > Man, how did that get in! It exists in a few places: > > binding/f77/mpif.h.in: INTEGER MPI_UNDEFINED, > MPI_UNDEFINED_RANK > binding/f77/mpif.h.in: PARAMETER (MPI_UNDEFINED_RANK=(-32766)) > binding/f77/buildiface: print MPIFFD " INTEGER > MPI_UNDEFINED, > MPI_UNDEFINED_RANK\n"; > binding/f77/buildiface: print MPIFFD " PARAMETER > (MPI_UNDEFINED_RANK=$mpidef{'MPI_UNDEFINED'})\n"; > include/mpi.h.in:#define MPI_UNDEFINED_RANK MPI_UNDEFINED > mpe2/src/logging/include/mpi_null.h.in:#define MPI_UNDEFINED_RANK > MPI_UNDEFINED > mpi/errhan/errutil.c: case MPI_UNDEFINED_RANK: > mpi/errhan/errutil.c: MPIU_Strncpy(str, > "MPI_UNDEFINED_RANK", > maxlen); > > > _____ > > From: mpich-ms-bounces(a)mcs.anl.gov > [mailto:[email protected]] > On Behalf Of Erez Haba > Sent: Friday, November 06, 2009 8:44 PM > To: mpich-ms(a)mcs.anl.gov > Subject: Re: [mpich-ms] MPI_COMBINER_ENUM > > > > Same for MPI_UNDEFINED_RANK (not in the standard) > > > > From: mpich-ms-bounces(a)mcs.anl.gov > [mailto:[email protected]] > On Behalf Of Rajeev Thakur > Sent: Friday, November 06, 2009 6:10 AM > To: mpich-ms(a)mcs.anl.gov > Subject: Re: [mpich-ms] MPI_COMBINER_ENUM > > > > Fixed in r5721. > > > > Rajeev > > > > _____ > > From: mpich-ms-bounces(a)mcs.anl.gov > [mailto:[email protected]] > On Behalf Of William Gropp > Sent: Friday, November 06, 2009 7:32 AM > To: mpich-ms(a)mcs.anl.gov > Subject: Re: [mpich-ms] MPI_COMBINER_ENUM > > You are absolutely correct. This name is invalid and should be > something else, probably MPIR_Combiner_enum to match the style of the > other enum. > > > > Bill > > > > On Nov 5, 2009, at 10:53 PM, Erez Haba wrote: > > > > > > Hi all, > > I see the symbol MPI_COMBINER_ENUM defined in mpi.h, but it does not > seem to be defined in the standard. As far as I recall the standard > prohibits defining symbols with the prefix MPI_ that are not part of > the standard. > > Thoughts? > > .Erez > > > > William Gropp > > Deputy Director for Research > > Institute for Advanced Computing Applications and Technologies > > Paul and Cynthia Saylor Professor of Computer Science > > University of Illinois Urbana-Champaign

2 5

mpich2-1.2.1rc1 tomorrow
by Pavan Balaji 06 Nov '09

06 Nov '09

Is the plan for the 1.2.1rc1 release still on for tomorrow? It doesn't look like it from the ticket list: https://trac.mcs.anl.gov/projects/mpich2/report/11 -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji

4 4

BoF
by Darius Buntinas 05 Nov '09

05 Nov '09

Here are the responses I got for for the BoF: Intel: Bill Magro "has it on his radar." This sounds like a "maybe" at best. IBM: Sameer agreed. Cray: No OSU: DK agreed Microsoft: Erez can't make it and I haven't heard from Fab or Eric. I've pinged them again. -d

3 3

PADB
by Rajeev Thakur 02 Nov '09

02 Nov '09

FYI, Ashley Pittman wants to meet at SC sometime to tell us about his open-source PADB tool, which he says works well with MPICH2 and MPD. http://padb.pittman.org.uk/ Rajeev

1 0