February 2021 - mpiwg-tools - mailman3.cels.anl.gov

Full Forum Meeting this week -- No Tools WG call on Thu Feb 25, 2021
by Marc-André Hermanns 22 Feb '21

22 Feb '21

Dear all, due to the full forum meeting this week Mon--Thu, we will skip the Tools WG call this week. Cheers, Marc-Andre -- Dr. rer. nat. Marc-André Hermanns IT Center Group: High Performance Computing Division: Computational Science and Engineering RWTH Aachen University Seffenter Weg 23 52074 Aachen Phone: +49 241 80-24381 hermanns(a)itc.rwth-aachen.de www.itc.rwth-aachen.de Social Media Kanäle des IT Centers: https://blog.rwth-aachen.de/itc/ https://www.facebook.com/itcenterrwth https://www.linkedin.com/company/itcenterrwth https://twitter.com/ITCenterRWTH https://www.youtube.com/channel/UCKKDJJukeRwO0LP-ac8x8rQ

1 0

Re: [mpiwg-tools] should there be an MPI_T_source_get_index?
by Marc-André Hermanns 19 Feb '21

19 Feb '21

Hi Ken, sorry for not getting back to you earlier. > (no bug this time __) Oh, thank goodness. (But it's a good thing that you found some of the still existing caveats!) > I am wondering if there should be a function to lookup a source > index by name. There is for categories, cvars, events, and pvars. I > stumbled upon this when modifying a test program > (https://github.com/pmodels/mpich/blob/main/test/mpi/mpi_t/getindex.c) > that gets the names of all available objects (cvars, pvars, > categories), then looks up their index to verify it. I want to do > the same for events and sources, but only events have that > functionality in the text today. The get_index_by_name function are usually convenience functions for the user, as they like to specify the name of a variable (e.g., in an environment variable) to have to tool pick that up (and only that one). Using this, a tool would not have to iterate through all existing variables/events. Now, for sources, we did not anticipate yet that users would like to address sources in the same manner. You'd (the tool) probably just want to iterate over all sources and then grab & save the info to be saved with the performance data. Cheers, Marc-Andre -- Dr. rer. nat. Marc-André Hermanns IT Center Group: High Performance Computing Division: Computational Science and Engineering RWTH Aachen University Seffenter Weg 23 52074 Aachen Phone: +49 241 80-24381 hermanns(a)itc.rwth-aachen.de www.itc.rwth-aachen.de Social Media Kanäle des IT Centers: https://blog.rwth-aachen.de/itc/ https://www.facebook.com/itcenterrwth https://www.linkedin.com/company/itcenterrwth https://twitter.com/ITCenterRWTH https://www.youtube.com/channel/UCKKDJJukeRwO0LP-ac8x8rQ

1 0

No call Feb 18, 2021
by Hermanns, Marc-André 17 Feb '21

17 Feb '21

Dear all, I have a conflicting meeting tomorrow, so I would cancel our call. Talk to you next week. Cheers Marc-André Sent from phone. -- Dr. rer. nat. Marc-André Hermanns IT Center Group: High Performance Computing Division: Computational Science and Engineering RWTH Aachen University Seffenter Weg 23 52074 Aachen Phone: +49 241 80-24381 hermanns(a)itc.rwth-aachen.de www.itc.rwth-aachen.de

1 0

should there be an MPI_T_source_get_index?
by Raffenetti, Kenneth J. 17 Feb '21

17 Feb '21

Hi all, (no bug this time __) I am wondering if there should be a function to lookup a source index by name. There is for categories, cvars, events, and pvars. I stumbled upon this when modifying a test program (https://github.com/pmodels/mpich/blob/main/test/mpi/mpi_t/getindex.c) that gets the names of all available objects (cvars, pvars, categories), then looks up their index to verify it. I want to do the same for events and sources, but only events have that functionality in the text today. Ken

1 0

Re: [mpiwg-tools] question about MPI_T_EVENT_GET_INFO
by Marc-André Hermanns 15 Feb '21

15 Feb '21

Dear Ken, yes, I think you are right in your assessment. I created a ticket https://github.com/mpi-forum/mpi-issues/issues/438 and an accompanying PR https://github.com/mpi-forum/mpi-standard/pull/518 and we should get this out as quickly as possible as an Erratum to 4.0 Cheers, Marc-Andre On 10.02.21 22:48, Raffenetti, Kenneth J. via mpiwg-tools wrote: > Hi all, > > In the latest mpi-4.0 RC pg 758, ln 16 describing MPI_T_EVENT_GET_INFO: > > If the argument to enumtype provided by the user is the MPI_T_ENUM_NULL pointer, no enumeration type is returned. > > This differs from previous get info functions. E.g. MPI_T_PVAR_GET_INFO: > > ... or the argument enumtype is the null pointer, no enumeration type is returned. > > IMO, the user providing MPI_T_ENUM_NULL to enumtype doesn't make sense. The user would pass a null pointer, as it is an output argument. > > Ken > > _______________________________________________ > mpiwg-tools mailing list > mpiwg-tools(a)lists.mpi-forum.org > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-tools > -- Dr. rer. nat. Marc-André Hermanns IT Center Group: High Performance Computing Division: Computational Science and Engineering RWTH Aachen University Seffenter Weg 23 52074 Aachen Phone: +49 241 80-24381 hermanns(a)itc.rwth-aachen.de www.itc.rwth-aachen.de Social Media Kanäle des IT Centers: https://blog.rwth-aachen.de/itc/ https://www.facebook.com/itcenterrwth https://www.linkedin.com/company/itcenterrwth https://twitter.com/ITCenterRWTH https://www.youtube.com/channel/UCKKDJJukeRwO0LP-ac8x8rQ

1 0

Meeting reminder: Feb 11, 2020 (today)
by Marc-André Hermanns 11 Feb '21

11 Feb '21

Hi all, This is a reminder of our regularly scheduled call tomorrow: Feb 11 at 8 am Pacific / 11 am Eastern / 5pm CEST. Meeting Details: https://github.com/mpiwg-tools/mpi-standard/wiki/Meeting-Information *NOTE*: I have updated the meeting details and switched to *ZOOM* for the meeting platform. Let me know if something does not work out for you. Agenda: ------- - Issue #433 [1] and preparation of PRs - Issue on NULL vs. MPI_T_ENUM_NULL (brought up by Ken) I created a PR [2] for the clarification of current behavior that we should discuss and wordsmith so we can move forward quickly. [1] https://github.com/mpi-forum/mpi-issues/issues/433 [2] https://github.com/mpi-forum/mpi-standard/pull/515 Cheers, Marc-Andre -- Dr. rer. nat. Marc-André Hermanns IT Center Group: High Performance Computing Division: Computational Science and Engineering RWTH Aachen University Seffenter Weg 23 52074 Aachen Phone: +49 241 80-24381 hermanns(a)itc.rwth-aachen.de www.itc.rwth-aachen.de

1 0

Re: [mpiwg-tools] question about MPI_T_EVENT_GET_INFO
by Martin Schulz 10 Feb '21

10 Feb '21

Hi Ken, Thanks for pointing this out - I think you are correct. At least on the PVAR side, that's how it was intended. I would assume, the same should be true for the events as well - Marc-Andre, do you agree? Thanks, Martin -- Prof. Dr. Martin Schulz, Chair of Computer Architecture and Parallel Systems Department of Informatics, TU-Munich, Boltzmannstraße 3, D-85748 Garching Member of the Board of Directors at the Leibniz Supercomputing Centre (LRZ) Email: schulzm(a)in.tum.de On 10.02.21, 22:48, "mpiwg-tools on behalf of Raffenetti, Kenneth J. via mpiwg-tools" <mpiwg-tools-bounces(a)lists.mpi-forum.org on behalf of mpiwg-tools(a)lists.mpi-forum.org> wrote: Hi all, In the latest mpi-4.0 RC pg 758, ln 16 describing MPI_T_EVENT_GET_INFO: If the argument to enumtype provided by the user is the MPI_T_ENUM_NULL pointer, no enumeration type is returned. This differs from previous get info functions. E.g. MPI_T_PVAR_GET_INFO: ... or the argument enumtype is the null pointer, no enumeration type is returned. IMO, the user providing MPI_T_ENUM_NULL to enumtype doesn't make sense. The user would pass a null pointer, as it is an output argument. Ken _______________________________________________ mpiwg-tools mailing list mpiwg-tools(a)lists.mpi-forum.org https://lists.mpi-forum.org/mailman/listinfo/mpiwg-tools

1 0

question about MPI_T_EVENT_GET_INFO
by Raffenetti, Kenneth J. 10 Feb '21

10 Feb '21

Hi all, In the latest mpi-4.0 RC pg 758, ln 16 describing MPI_T_EVENT_GET_INFO: If the argument to enumtype provided by the user is the MPI_T_ENUM_NULL pointer, no enumeration type is returned. This differs from previous get info functions. E.g. MPI_T_PVAR_GET_INFO: ... or the argument enumtype is the null pointer, no enumeration type is returned. IMO, the user providing MPI_T_ENUM_NULL to enumtype doesn't make sense. The user would pass a null pointer, as it is an output argument. Ken

1 0

Meeting reminder: Feb 04, 2020 (today)
by Marc-André Hermanns 04 Feb '21

04 Feb '21

Hi all, This is a reminder of our regularly scheduled call tomorrow: Feb 04 at 8 am Pacific / 11 am Eastern / 5pm CEST. Meeting Details: https://github.com/mpiwg-tools/mpi-standard/wiki/Meeting-Information *NOTE*: I have updated the meeting details and switched to *ZOOM* for the meeting platform. Let me know if something does not work out for you. Agenda: ------- - Issue #433 [1] and preparation of PRs In yesterdays Virtual Meeting, we discussed #433 and consensus was to get one (or more) specific PRs on the way as soon as possible. Cheers, Marc-Andre -- Dr. rer. nat. Marc-André Hermanns IT Center Group: High Performance Computing Division: Computational Science and Engineering RWTH Aachen University Seffenter Weg 23 52074 Aachen Phone: +49 241 80-24381 hermanns(a)itc.rwth-aachen.de www.itc.rwth-aachen.de

1 0