- mpiwg-coll - mailman3.cels.anl.gov

Re: [mpiwg-coll] Review implementation of #419+420 in chapter 7 - take 2
by Balaji, Pavan 22 Feb '15

22 Feb '15

The changes look good to me. -- Pavan > On Feb 21, 2015, at 2:56 PM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: > > Dear Chapter Committee (and anybody interested), > > Thanks to all for your review! Puri found some formatting inconsistencies (only missing spaces and line breaks) that took a while to resolve (thanks to Bill and sorry for the delay). > > While resolving them, Bill found some simple ticket-0 changes that I applied based on his comments: > > - removed the 20-year old outdated reference to EXPRESS > - slight wording change in advice to users on MPI_WEIGHTS_EMPTY > - slight wording change in example of graph_neighbors > - slight wording change in first sentence of 7.5.6 > > We can easily back them out (r2007) if we decide that we don't want them. They are rather uncontroversial though. The separate diff of these is at [1]. > > The complete diff (including all changes) is at [2] and the latest version of the chapter is at [3]. Please review this version again, hopefully a last time. > > Thanks & All the Best, > Torsten > > [1]: http://htor.inf.ethz.ch/sec/chap-topol-diff-r2003.diff > [2]: http://htor.inf.ethz.ch/sec/chap-topol-diff-r1806.diff > [3]: http://htor.inf.ethz.ch/sec/chap-topol-mpi-3_1.pdf > > -- > ### qreharg rug ebs fv crryF ------------- http://www.unixer.de/ ----- > Torsten Hoefler | Assistant Professor > Dept. of Computer Science | ETH Zürich > Universitätsstrasse 6 | Zurich-8092, Switzerland > CAB E 64.1 | Phone: +41 76 309 79 29 -- Pavan Balaji ✉️ http://www.mcs.anl.gov/~balaji

1 0

Review implementation of #419+420 in chapter 7 - take 2
by Torsten Hoefler 21 Feb '15

21 Feb '15

Dear Chapter Committee (and anybody interested), Thanks to all for your review! Puri found some formatting inconsistencies (only missing spaces and line breaks) that took a while to resolve (thanks to Bill and sorry for the delay). While resolving them, Bill found some simple ticket-0 changes that I applied based on his comments: - removed the 20-year old outdated reference to EXPRESS - slight wording change in advice to users on MPI_WEIGHTS_EMPTY - slight wording change in example of graph_neighbors - slight wording change in first sentence of 7.5.6 We can easily back them out (r2007) if we decide that we don't want them. They are rather uncontroversial though. The separate diff of these is at [1]. The complete diff (including all changes) is at [2] and the latest version of the chapter is at [3]. Please review this version again, hopefully a last time. Thanks & All the Best, Torsten [1]: http://htor.inf.ethz.ch/sec/chap-topol-diff-r2003.diff [2]: http://htor.inf.ethz.ch/sec/chap-topol-diff-r1806.diff [3]: http://htor.inf.ethz.ch/sec/chap-topol-mpi-3_1.pdf -- ### qreharg rug ebs fv crryF ------------- http://www.unixer.de/ ----- Torsten Hoefler | Assistant Professor Dept. of Computer Science | ETH Zürich Universitätsstrasse 6 | Zurich-8092, Switzerland CAB E 64.1 | Phone: +41 76 309 79 29

1 0

Re: [mpiwg-coll] Review implementation of #419+420 in chapter 7
by Daniel Holmes 11 Feb '15

11 Feb '15

Hi Torsten, I have reviewed the changes for #420 - all looks good to me. My only comments are that the diff shows changes to most of the lines of the example, even lines that have not apparently changed, which makes picking out the actual changes a little more difficult than necessary. Also, the alignment of the first set numbers in the "[s/r]displs = ( xxx + ..." lines is off by one character for the single digit numbers. Minor quibbles that do not need to be fixed unless you have lots of free time! Cheers, Dan. On 07/02/2015 21:35, Torsten Hoefler wrote: > Dear Chapter Committee (and anybody interested), > > Could you please review the implementation of tickets 419 and 420 [1] in > the new pdf [2] or the diff [3] (which includes some more ticket-0 > changes) before February 16. > > I don't have Chulho Kim's email address but I hope he's subscribed here. > > Martin, can you please update the table on > http://meetings.mpi-forum.org/MPI_3.1_main_page.php . I rely on this > table that there is nothing else missing in chapters 5 or 7. Please let > me know if there is still something open. > > Thanks & Best, > Torsten > > [1]: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/419 > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/420 > [2]: http://htor.inf.ethz.ch/sec/chap-topol-mpi-3_1.pdf > [3]: http://htor.inf.ethz.ch/sec/chap-topol-diff-r1806.diff > -- Dan Holmes Applications Consultant in HPC Research EPCC, The University of Edinburgh James Clerk Maxwell Building The Kings Buildings Peter Guthrie Tait Road Edinburgh EH9 3FD T: +44(0)131 651 3465 E: dholmes(a)epcc.ed.ac.uk *Please consider the environment before printing this email.* The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

1 0

Re: [mpiwg-coll] Review implementation of #419+420 in chapter 7
by Jeff Hammond 07 Feb '15

07 Feb '15

I BCC'd Chulho's Lenovo address just in case. Jeff Sent from my iPhone > On Feb 7, 2015, at 1:35 PM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: > > Dear Chapter Committee (and anybody interested), > > Could you please review the implementation of tickets 419 and 420 [1] in > the new pdf [2] or the diff [3] (which includes some more ticket-0 > changes) before February 16. > > I don't have Chulho Kim's email address but I hope he's subscribed here. > > Martin, can you please update the table on > http://meetings.mpi-forum.org/MPI_3.1_main_page.php . I rely on this > table that there is nothing else missing in chapters 5 or 7. Please let > me know if there is still something open. > > Thanks & Best, > Torsten > > [1]: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/419 > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/420 > [2]: http://htor.inf.ethz.ch/sec/chap-topol-mpi-3_1.pdf > [3]: http://htor.inf.ethz.ch/sec/chap-topol-diff-r1806.diff > > -- > ### qreharg rug ebs fv crryF ------------- http://www.unixer.de/ ----- > Torsten Hoefler | Assistant Professor > Dept. of Computer Science | ETH Zürich > Universitätsstrasse 6 | Zurich-8092, Switzerland > CAB E 64.1 | Phone: +41 76 309 79 29 > > > _______________________________________________ > mpiwg-coll mailing list > mpiwg-coll(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll

1 0

Review implementation of #419+420 in chapter 7
by Torsten Hoefler 07 Feb '15

07 Feb '15

Dear Chapter Committee (and anybody interested), Could you please review the implementation of tickets 419 and 420 [1] in the new pdf [2] or the diff [3] (which includes some more ticket-0 changes) before February 16. I don't have Chulho Kim's email address but I hope he's subscribed here. Martin, can you please update the table on http://meetings.mpi-forum.org/MPI_3.1_main_page.php . I rely on this table that there is nothing else missing in chapters 5 or 7. Please let me know if there is still something open. Thanks & Best, Torsten [1]: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/419 https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/420 [2]: http://htor.inf.ethz.ch/sec/chap-topol-mpi-3_1.pdf [3]: http://htor.inf.ethz.ch/sec/chap-topol-diff-r1806.diff -- ### qreharg rug ebs fv crryF ------------- http://www.unixer.de/ ----- Torsten Hoefler | Assistant Professor Dept. of Computer Science | ETH Zürich Universitätsstrasse 6 | Zurich-8092, Switzerland CAB E 64.1 | Phone: +41 76 309 79 29

1 0

Re: [mpiwg-coll] deprecating MPI_GRAPH_CREATE?
by Torsten Hoefler 02 Dec '14

02 Dec '14

Is there an alternative? :-) Torsten PS: I will stop spamming the list with this topic now. On 12/02/2014 08:41 PM, Jeff Hammond wrote: > Does the non-scalable nature of v-collectives prevent people from using them? > > :-) > > Jeff > > On Tue, Dec 2, 2014 at 11:38 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: >> I consider this risk as low because it says it in the standard that it's >> non-scalable as well as in any recent documentation. Yes, some people will >> write bad codes. >> >> All the Best, >> Torsten >> >> >> On 12/02/2014 08:28 PM, Jeff Hammond wrote: >>> >>> I created ticket 462 as a placeholder for whenever someone has time to >>> work on this. >>> >>> MPI_GRAPH_CREATE does harm to users because they might use it. >>> Somebody might naively think that distributed graphs are more >>> complicated to use than replicated ones... >>> >>> Jeff >>> >>> On Tue, Dec 2, 2014 at 1:57 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: >>>> >>>> I am unfortunately too busy these days to drive it (a clean removal will >>>> touch several places). It is ugly but it doesn't do any direct harm to >>>> the >>>> standard or its users. >>>> >>>> Best, >>>> Torsten >>>> >>>> >>>> On 01.12.2014 17:02, Jeff Hammond wrote: >>>>> >>>>> >>>>> But you're not going to write that proposal? You seem more qualified >>>>> than >>>>> most... >>>>> >>>>> Jeff >>>>> >>>>> Sent from my iPhone >>>>> >>>>>> On Dec 1, 2014, at 7:51 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: >>>>>> >>>>>> Jeff, >>>>>> >>>>>> There is no such proposal yet (afaik), I'd support a carefully crafted >>>>>> one and I think I wouldn't be alone. >>>>>> >>>>>> Thanks & Best, >>>>>> Torsten >>>>>> >>>>>>> On 01.12.2014 06:41, Jeff Hammond wrote: >>>>>>> Regarding http://lists.mpi-forum.org/mpiwg-coll/2014/03/0289.php, is >>>>>>> there a proposal to deprecate MPI_GRAPH_CREATE? It seems that there >>>>>>> are many issues with it and they are deliberately not getting fixed >>>>>>> because a subset of the issues with it make it too horrible to fix. >>>>>>> >>>>>>> I find this situation distasteful as it leads to a poor user >>>>>>> experience. Let's do the right thing and put MPI_GRAPH_CREATE out of >>>>>>> our misery sooner rather than later. >>>>>>> >>>>>>> Thanks, >>>>>>> >>>>>>> Jeff >>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> ### qreharg rug ebs fv crryF --- http://htor.inf.ethz.ch/ --- >>>>>> Torsten Hoefler | Assistant Professor >>>>>> Dept. of Computer Science | ETH Zürich >>>>>> Universitätsstrasse 6 | Zurich-8092, Switzerland >>>>>> CAB E 64.1 | Phone: +41 76 309 79 29 >>>>>> _______________________________________________ >>>>>> mpiwg-coll mailing list >>>>>> mpiwg-coll(a)lists.mpi-forum.org >>>>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll >>>>> >>>>> >>>>> _______________________________________________ >>>>> mpiwg-coll mailing list >>>>> mpiwg-coll(a)lists.mpi-forum.org >>>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll >>>>> >>>> >>>> -- >>>> ### qreharg rug ebs fv crryF --- http://htor.inf.ethz.ch/ --- >>>> Torsten Hoefler | Assistant Professor >>>> Dept. of Computer Science | ETH Zürich >>>> Universitätsstrasse 6 | Zurich-8092, Switzerland >>>> CAB E 64.1 | Phone: +41 76 309 79 29 >>>> _______________________________________________ >>>> mpiwg-coll mailing list >>>> mpiwg-coll(a)lists.mpi-forum.org >>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll >>> >>> >>> >>> >> >> -- >> ### qreharg rug ebs fv crryF ------------- http://www.unixer.de/ ----- >> >> Torsten Hoefler | Assistant Professor >> Dept. of Computer Science | ETH Zürich >> Universitätsstrasse 6 | Zurich-8092, Switzerland >> CAB E 64.1 | Phone: +41 76 309 79 29 >> _______________________________________________ >> mpiwg-coll mailing list >> mpiwg-coll(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll > > > -- ### qreharg rug ebs fv crryF ------------- http://www.unixer.de/ ----- Torsten Hoefler | Assistant Professor Dept. of Computer Science | ETH Zürich Universitätsstrasse 6 | Zurich-8092, Switzerland CAB E 64.1 | Phone: +41 76 309 79 29

1 0

Re: [mpiwg-coll] deprecating MPI_GRAPH_CREATE?
by Jeff Hammond 02 Dec '14

02 Dec '14

Does the non-scalable nature of v-collectives prevent people from using them? :-) Jeff On Tue, Dec 2, 2014 at 11:38 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: > I consider this risk as low because it says it in the standard that it's > non-scalable as well as in any recent documentation. Yes, some people will > write bad codes. > > All the Best, > Torsten > > > On 12/02/2014 08:28 PM, Jeff Hammond wrote: >> >> I created ticket 462 as a placeholder for whenever someone has time to >> work on this. >> >> MPI_GRAPH_CREATE does harm to users because they might use it. >> Somebody might naively think that distributed graphs are more >> complicated to use than replicated ones... >> >> Jeff >> >> On Tue, Dec 2, 2014 at 1:57 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: >>> >>> I am unfortunately too busy these days to drive it (a clean removal will >>> touch several places). It is ugly but it doesn't do any direct harm to >>> the >>> standard or its users. >>> >>> Best, >>> Torsten >>> >>> >>> On 01.12.2014 17:02, Jeff Hammond wrote: >>>> >>>> >>>> But you're not going to write that proposal? You seem more qualified >>>> than >>>> most... >>>> >>>> Jeff >>>> >>>> Sent from my iPhone >>>> >>>>> On Dec 1, 2014, at 7:51 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: >>>>> >>>>> Jeff, >>>>> >>>>> There is no such proposal yet (afaik), I'd support a carefully crafted >>>>> one and I think I wouldn't be alone. >>>>> >>>>> Thanks & Best, >>>>> Torsten >>>>> >>>>>> On 01.12.2014 06:41, Jeff Hammond wrote: >>>>>> Regarding http://lists.mpi-forum.org/mpiwg-coll/2014/03/0289.php, is >>>>>> there a proposal to deprecate MPI_GRAPH_CREATE? It seems that there >>>>>> are many issues with it and they are deliberately not getting fixed >>>>>> because a subset of the issues with it make it too horrible to fix. >>>>>> >>>>>> I find this situation distasteful as it leads to a poor user >>>>>> experience. Let's do the right thing and put MPI_GRAPH_CREATE out of >>>>>> our misery sooner rather than later. >>>>>> >>>>>> Thanks, >>>>>> >>>>>> Jeff >>>>> >>>>> >>>>> >>>>> -- >>>>> ### qreharg rug ebs fv crryF --- http://htor.inf.ethz.ch/ --- >>>>> Torsten Hoefler | Assistant Professor >>>>> Dept. of Computer Science | ETH Zürich >>>>> Universitätsstrasse 6 | Zurich-8092, Switzerland >>>>> CAB E 64.1 | Phone: +41 76 309 79 29 >>>>> _______________________________________________ >>>>> mpiwg-coll mailing list >>>>> mpiwg-coll(a)lists.mpi-forum.org >>>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll >>>> >>>> >>>> _______________________________________________ >>>> mpiwg-coll mailing list >>>> mpiwg-coll(a)lists.mpi-forum.org >>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll >>>> >>> >>> -- >>> ### qreharg rug ebs fv crryF --- http://htor.inf.ethz.ch/ --- >>> Torsten Hoefler | Assistant Professor >>> Dept. of Computer Science | ETH Zürich >>> Universitätsstrasse 6 | Zurich-8092, Switzerland >>> CAB E 64.1 | Phone: +41 76 309 79 29 >>> _______________________________________________ >>> mpiwg-coll mailing list >>> mpiwg-coll(a)lists.mpi-forum.org >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll >> >> >> >> > > -- > ### qreharg rug ebs fv crryF ------------- http://www.unixer.de/ ----- > > Torsten Hoefler | Assistant Professor > Dept. of Computer Science | ETH Zürich > Universitätsstrasse 6 | Zurich-8092, Switzerland > CAB E 64.1 | Phone: +41 76 309 79 29 > _______________________________________________ > mpiwg-coll mailing list > mpiwg-coll(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [mpiwg-coll] deprecating MPI_GRAPH_CREATE?
by Torsten Hoefler 02 Dec '14

02 Dec '14

I consider this risk as low because it says it in the standard that it's non-scalable as well as in any recent documentation. Yes, some people will write bad codes. All the Best, Torsten On 12/02/2014 08:28 PM, Jeff Hammond wrote: > I created ticket 462 as a placeholder for whenever someone has time to > work on this. > > MPI_GRAPH_CREATE does harm to users because they might use it. > Somebody might naively think that distributed graphs are more > complicated to use than replicated ones... > > Jeff > > On Tue, Dec 2, 2014 at 1:57 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: >> I am unfortunately too busy these days to drive it (a clean removal will >> touch several places). It is ugly but it doesn't do any direct harm to the >> standard or its users. >> >> Best, >> Torsten >> >> >> On 01.12.2014 17:02, Jeff Hammond wrote: >>> >>> But you're not going to write that proposal? You seem more qualified than >>> most... >>> >>> Jeff >>> >>> Sent from my iPhone >>> >>>> On Dec 1, 2014, at 7:51 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: >>>> >>>> Jeff, >>>> >>>> There is no such proposal yet (afaik), I'd support a carefully crafted >>>> one and I think I wouldn't be alone. >>>> >>>> Thanks & Best, >>>> Torsten >>>> >>>>> On 01.12.2014 06:41, Jeff Hammond wrote: >>>>> Regarding http://lists.mpi-forum.org/mpiwg-coll/2014/03/0289.php, is >>>>> there a proposal to deprecate MPI_GRAPH_CREATE? It seems that there >>>>> are many issues with it and they are deliberately not getting fixed >>>>> because a subset of the issues with it make it too horrible to fix. >>>>> >>>>> I find this situation distasteful as it leads to a poor user >>>>> experience. Let's do the right thing and put MPI_GRAPH_CREATE out of >>>>> our misery sooner rather than later. >>>>> >>>>> Thanks, >>>>> >>>>> Jeff >>>> >>>> >>>> -- >>>> ### qreharg rug ebs fv crryF --- http://htor.inf.ethz.ch/ --- >>>> Torsten Hoefler | Assistant Professor >>>> Dept. of Computer Science | ETH Zürich >>>> Universitätsstrasse 6 | Zurich-8092, Switzerland >>>> CAB E 64.1 | Phone: +41 76 309 79 29 >>>> _______________________________________________ >>>> mpiwg-coll mailing list >>>> mpiwg-coll(a)lists.mpi-forum.org >>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll >>> >>> _______________________________________________ >>> mpiwg-coll mailing list >>> mpiwg-coll(a)lists.mpi-forum.org >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll >>> >> >> -- >> ### qreharg rug ebs fv crryF --- http://htor.inf.ethz.ch/ --- >> Torsten Hoefler | Assistant Professor >> Dept. of Computer Science | ETH Zürich >> Universitätsstrasse 6 | Zurich-8092, Switzerland >> CAB E 64.1 | Phone: +41 76 309 79 29 >> _______________________________________________ >> mpiwg-coll mailing list >> mpiwg-coll(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll > > > -- ### qreharg rug ebs fv crryF ------------- http://www.unixer.de/ ----- Torsten Hoefler | Assistant Professor Dept. of Computer Science | ETH Zürich Universitätsstrasse 6 | Zurich-8092, Switzerland CAB E 64.1 | Phone: +41 76 309 79 29

1 0

Re: [mpiwg-coll] deprecating MPI_GRAPH_CREATE?
by Jeff Hammond 02 Dec '14

02 Dec '14

I created ticket 462 as a placeholder for whenever someone has time to work on this. MPI_GRAPH_CREATE does harm to users because they might use it. Somebody might naively think that distributed graphs are more complicated to use than replicated ones... Jeff On Tue, Dec 2, 2014 at 1:57 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: > I am unfortunately too busy these days to drive it (a clean removal will > touch several places). It is ugly but it doesn't do any direct harm to the > standard or its users. > > Best, > Torsten > > > On 01.12.2014 17:02, Jeff Hammond wrote: >> >> But you're not going to write that proposal? You seem more qualified than >> most... >> >> Jeff >> >> Sent from my iPhone >> >>> On Dec 1, 2014, at 7:51 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: >>> >>> Jeff, >>> >>> There is no such proposal yet (afaik), I'd support a carefully crafted >>> one and I think I wouldn't be alone. >>> >>> Thanks & Best, >>> Torsten >>> >>>> On 01.12.2014 06:41, Jeff Hammond wrote: >>>> Regarding http://lists.mpi-forum.org/mpiwg-coll/2014/03/0289.php, is >>>> there a proposal to deprecate MPI_GRAPH_CREATE? It seems that there >>>> are many issues with it and they are deliberately not getting fixed >>>> because a subset of the issues with it make it too horrible to fix. >>>> >>>> I find this situation distasteful as it leads to a poor user >>>> experience. Let's do the right thing and put MPI_GRAPH_CREATE out of >>>> our misery sooner rather than later. >>>> >>>> Thanks, >>>> >>>> Jeff >>> >>> >>> -- >>> ### qreharg rug ebs fv crryF --- http://htor.inf.ethz.ch/ --- >>> Torsten Hoefler | Assistant Professor >>> Dept. of Computer Science | ETH Zürich >>> Universitätsstrasse 6 | Zurich-8092, Switzerland >>> CAB E 64.1 | Phone: +41 76 309 79 29 >>> _______________________________________________ >>> mpiwg-coll mailing list >>> mpiwg-coll(a)lists.mpi-forum.org >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll >> >> _______________________________________________ >> mpiwg-coll mailing list >> mpiwg-coll(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll >> > > -- > ### qreharg rug ebs fv crryF --- http://htor.inf.ethz.ch/ --- > Torsten Hoefler | Assistant Professor > Dept. of Computer Science | ETH Zürich > Universitätsstrasse 6 | Zurich-8092, Switzerland > CAB E 64.1 | Phone: +41 76 309 79 29 > _______________________________________________ > mpiwg-coll mailing list > mpiwg-coll(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [mpiwg-coll] deprecating MPI_GRAPH_CREATE?
by Torsten Hoefler 02 Dec '14

02 Dec '14

I am unfortunately too busy these days to drive it (a clean removal will touch several places). It is ugly but it doesn't do any direct harm to the standard or its users. Best, Torsten On 01.12.2014 17:02, Jeff Hammond wrote: > But you're not going to write that proposal? You seem more qualified than most... > > Jeff > > Sent from my iPhone > >> On Dec 1, 2014, at 7:51 AM, Torsten Hoefler <htor(a)inf.ethz.ch> wrote: >> >> Jeff, >> >> There is no such proposal yet (afaik), I'd support a carefully crafted one and I think I wouldn't be alone. >> >> Thanks & Best, >> Torsten >> >>> On 01.12.2014 06:41, Jeff Hammond wrote: >>> Regarding http://lists.mpi-forum.org/mpiwg-coll/2014/03/0289.php, is >>> there a proposal to deprecate MPI_GRAPH_CREATE? It seems that there >>> are many issues with it and they are deliberately not getting fixed >>> because a subset of the issues with it make it too horrible to fix. >>> >>> I find this situation distasteful as it leads to a poor user >>> experience. Let's do the right thing and put MPI_GRAPH_CREATE out of >>> our misery sooner rather than later. >>> >>> Thanks, >>> >>> Jeff >> >> -- >> ### qreharg rug ebs fv crryF --- http://htor.inf.ethz.ch/ --- >> Torsten Hoefler | Assistant Professor >> Dept. of Computer Science | ETH Zürich >> Universitätsstrasse 6 | Zurich-8092, Switzerland >> CAB E 64.1 | Phone: +41 76 309 79 29 >> _______________________________________________ >> mpiwg-coll mailing list >> mpiwg-coll(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll > _______________________________________________ > mpiwg-coll mailing list > mpiwg-coll(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-coll > -- ### qreharg rug ebs fv crryF --- http://htor.inf.ethz.ch/ --- Torsten Hoefler | Assistant Professor Dept. of Computer Science | ETH Zürich Universitätsstrasse 6 | Zurich-8092, Switzerland CAB E 64.1 | Phone: +41 76 309 79 29

1 0