March 2017 - mpiwg-p2p - mailman3.cels.anl.gov

Re: [mpiwg-p2p] Meeting Today?
by Dan Holmes 27 Mar '17

27 Mar '17

Hi Jim, et al, I was hoping to move the WG on to talking about Freeing-Send and Allocating-Receive (Fsend & Arecv). I’d like to re-boot and refresh caches on that with a goal of presenting something (probably informally) next face-to-face meeting in June. I’ll start the call and see how many turn up. Cheers, Dan. > On 27 Mar 2017, at 15:28, Jim Dinan <james.dinan(a)gmail.com> wrote: > > Hi All, > > The info query proposal seems to be converging. I don't think there's anything to discuss on this topic this week. > > Current status is that Pavan will check with vendors to make sure they are ok with updating MPICH to report the user's info key value instead of the effective value being used by the implementation. Assuming this is ok, it sounds like the proposal that was read at the last meeting is ready to move forward into voting. > > Are there any other topics for discussion? If not, I think we can cancel today's meeting. > > Cheers, > ~Jim. > _______________________________________________ > mpiwg-p2p mailing list > mpiwg-p2p(a)lists.mpi-forum.org > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-p2p -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

1 0

Meeting Today?
by Jim Dinan 27 Mar '17

27 Mar '17

Hi All, The info query proposal seems to be converging. I don't think there's anything to discuss on this topic this week. Current status is that Pavan will check with vendors to make sure they are ok with updating MPICH to report the user's info key value instead of the effective value being used by the implementation. Assuming this is ok, it sounds like the proposal that was read at the last meeting is ready to move forward into voting. Are there any other topics for discussion? If not, I think we can cancel today's meeting. Cheers, ~Jim.

1 0

Re: [mpiwg-p2p] Result of reading of Info Query #63
by Rolf Rabenseifner 02 Mar '17

02 Mar '17

I did not follow the discussion, but I want to remember (for the case that some are not aware of) that MPI_File_get_info returns not what the user set. It returns all used info, system-specific infos plus those set by the user that are not ignored. This works fine with all the parallel file-systems and their system-specific (i.e., non-user supplied?) values. And of course, the user can learn from that and change those values with MPI_File_set_info. The wording for MPI_Win_get_info and MPI_Comm_get_info is the same. See MPI-3.1, Sect. 6.4.4, page 250, lines 9-13, Advice to users. (Same on page 502, lines 1-5) And other wording in Sect. 11.2.7, page 415-416, lines 47ff, Adv. to impl. Rolf ----- Original Message ----- > From: "Daniel Holmes" <dholmes(a)epcc.ed.ac.uk> > To: "MPI WG Point To Point Communications working group" <mpiwg-p2p(a)lists.mpi-forum.org> > Sent: Wednesday, March 1, 2017 11:58:28 PM > Subject: [mpiwg-p2p] Result of reading of Info Query #63 > Exec summary > > The MPI_XXX_GET_INFO text is acceptable to everyone as is. > > The MPI_XXX_SET_INFO text is contentious for two reasons: > 1) (lack of) ability for MPI to pick non-user supplied values. Pavan brought > this up. Resolution: none. Suggestions: continue arguing against this or add > additional MPI_XXX_SET_INFO_INEXACT (or similar), which allows randomness - > lets user choose. > 2) (lack of) transactional behaviour - all key/value updates should succeed or > all should fail. Martin brought this up. Resolution: this is a new API, we’d > love to see your proposal. > > Cheers, > Dan. > -- > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. > > _______________________________________________ > mpiwg-p2p mailing list > mpiwg-p2p(a)lists.mpi-forum.org > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-p2p -- Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307)

1 0

Result of reading of Info Query #63
by Dan Holmes 01 Mar '17

01 Mar '17

Exec summary The MPI_XXX_GET_INFO text is acceptable to everyone as is. The MPI_XXX_SET_INFO text is contentious for two reasons: 1) (lack of) ability for MPI to pick non-user supplied values. Pavan brought this up. Resolution: none. Suggestions: continue arguing against this or add additional MPI_XXX_SET_INFO_INEXACT (or similar), which allows randomness - lets user choose. 2) (lack of) transactional behaviour - all key/value updates should succeed or all should fail. Martin brought this up. Resolution: this is a new API, we’d love to see your proposal. Cheers, Dan. -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

1 0