February 2011 - mpiwg-rma - mailman3.cels.anl.gov

Re: [Mpi3-rma] current version of proposal
by Torsten Hoefler 28 Feb '11

28 Feb '11

Hello all, I updated the version again (hopefully for a last time) with Brian's cleanup of the examples (Brian, I edited the style a bit). See https://svn.mpi-forum.org/trac/mpi-forum-web/attachment/wiki/mpi3-rma-propo… All the Best, Torsten -- bash$ :(){ :|:&};: --------------------- http://www.unixer.de/ ----- Torsten Hoefler | Performance Modeling and Simulation Lead Blue Waters Directorate | University of Illinois (UIUC) 1205 W Clark Street | Urbana, IL, 61801 NCSA Building | +01 (217) 244-7736

1 0

Re: [Mpi3-rma] RMA Model Figure
by Torsten Hoefler 15 Feb '11

15 Feb '11

Hi Jim, > I've redrawn Figure 11.5 in Xfig. I tried to match the original figure > as much as possible. Here are the things I (intentionally) changed: > > * Changed PUT/GET/LOAD/STORE to Helvetica to match the font used for > these things in the rest of the document. > > * Changed the labels "public window copy" to "Window 0/1 public copy" > > I don't think I have SVN access. Could someone set me up or check this > in for me (please also check in the fig file so we don't lose it again) > and let me know if there are any problems (dimensions, etc) integrating > this with the document? > > Please send any feedback on the changes I made or other changes we might > want to make to this figure. I think we also wanted to update the > caption to indicate that this figure illustrates the separate memory > model. Thank you! It looks good to me for now. I added the figure to svn and it will replace the old one in the next build and upload. All the Best, Torsten -- bash$ :(){ :|:&};: --------------------- http://www.unixer.de/ ----- Torsten Hoefler | Performance Modeling and Simulation Lead Blue Waters Directorate | University of Illinois (UIUC) 1205 W Clark Street | Urbana, IL, 61801 NCSA Building | +01 (217) 244-7736

1 0

Re: [Mpi3-rma] RMA Model Figure
by William Gropp 15 Feb '11

15 Feb '11

Thanks, James. Bill On Feb 15, 2011, at 11:27 AM, James Dinan wrote: > Hi All, > > I've redrawn Figure 11.5 in Xfig. I tried to match the original > figure > as much as possible. Here are the things I (intentionally) changed: > > * Changed PUT/GET/LOAD/STORE to Helvetica to match the font used for > these things in the rest of the document. > > * Changed the labels "public window copy" to "Window 0/1 public copy" > > I don't think I have SVN access. Could someone set me up or check > this > in for me (please also check in the fig file so we don't lose it > again) > and let me know if there are any problems (dimensions, etc) > integrating > this with the document? > > Please send any feedback on the changes I made or other changes we > might > want to make to this figure. I think we also wanted to update the > caption to indicate that this figure illustrates the separate memory > model. > > Cheers, > ~Jim. > <rma_model.fig><rma_model.pdf><ATT00001.txt> 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

1 0

Re: [Mpi3-rma] current version of proposal
by William Gropp 15 Feb '11

15 Feb '11

I've updated these with the most recent version checked into the svn repository. There are still more updates to do, but it is slightly more up to date. Bill On Feb 15, 2011, at 11:28 AM, Howard Pritchard wrote: > Hi Folks (esp. Pavan), > > Does the 'current version' pdf at > > https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/mpi3-rma-proposal1 > > include the various corrections/changes discussed last week > at the forum? The date says Feb. 7 so I wasn't sure. > > Thanks, > > Howard > > -- > Howard Pritchard > Software Engineering > Cray, Inc. > _______________________________________________ > mpi3-rma mailing list > mpi3-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-rma 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

1 0

[Mpi3-rma] current version of proposal
by Howard Pritchard 15 Feb '11

15 Feb '11

Hi Folks (esp. Pavan), Does the 'current version' pdf at https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/mpi3-rma-proposal1 include the various corrections/changes discussed last week at the forum? The date says Feb. 7 so I wasn't sure. Thanks, Howard -- Howard Pritchard Software Engineering Cray, Inc.

1 0

[Mpi3-rma] RMA Model Figure
by James Dinan 15 Feb '11

15 Feb '11

Hi All, I've redrawn Figure 11.5 in Xfig. I tried to match the original figure as much as possible. Here are the things I (intentionally) changed: * Changed PUT/GET/LOAD/STORE to Helvetica to match the font used for these things in the rest of the document. * Changed the labels "public window copy" to "Window 0/1 public copy" I don't think I have SVN access. Could someone set me up or check this in for me (please also check in the fig file so we don't lose it again) and let me know if there are any problems (dimensions, etc) integrating this with the document? Please send any feedback on the changes I made or other changes we might want to make to this figure. I think we also wanted to update the caption to indicate that this figure illustrates the separate memory model. Cheers, ~Jim.

1 0

Re: [Mpi3-rma] Updated proposal 1
by Torsten Hoefler 07 Feb '11

07 Feb '11

On Sun, Feb 06, 2011 at 06:00:01PM -0600, Rajeev Thakur wrote: > In Example 11.12, don't you want the unlock to be after the read value on process A? > > In Example 11.15, why is a Win_sync needed after the lock_all? > > In Example 11.16, why is a flush needed immediately before the unlock_all? > > In Example 11.17, why is the sync needed after the lock_all, and why is the flush needed before the unlock_all? Correct, I simply added the lock_all, unlock_all because it was defined to be inside such an epoch. This is is correct but might be simplified as Rajeev pointed out :-). All instances are fixed now (see new attached documents). I also fixed a bug in the MPI_Get_accumulate() C interface which Adam pointed out to me a second ago (we need more reviews). Thanks & All the Best, Torsten -- bash$ :(){ :|:&};: --------------------- http://www.unixer.de/ ----- Torsten Hoefler | Performance Modeling and Simulation Lead Blue Waters Directorate | University of Illinois (UIUC) 1205 W Clark Street | Urbana, IL, 61801 NCSA Building | +01 (217) 244-7736

1 0

[Mpi3-rma] Final plenary RMA presentation uploaded
by William Gropp 07 Feb '11

07 Feb '11

I made a few more changes to the presentation, mostly to improve readability, and uploaded it to the wiki. Bill 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

1 0

Re: [Mpi3-rma] Updated proposal 1
by Rajeev Thakur 06 Feb '11

06 Feb '11

In Example 11.12, don't you want the unlock to be after the read value on process A? In Example 11.15, why is a Win_sync needed after the lock_all? In Example 11.16, why is a flush needed immediately before the unlock_all? In Example 11.17, why is the sync needed after the lock_all, and why is the flush needed before the unlock_all? Rajeev On Feb 1, 2011, at 10:04 PM, Torsten Hoefler wrote: > Hello All, > > I updated the proposal as discussed in our last telecon (including > Rajeev's comments -- thanks!). Significant changes: > > - moved all info keys to win_create > - added info key about operation > - changed ordering info key to be more flexible > - fixed examples > - added comment for Pavan to fix a comment > - changed the color of the proposal 2 merges (as they are still being > discussed and need a better letter than "n"). > > The complete diff is attached to this email and the documents on the > wiki are updated. Please review! > > https://svn.mpi-forum.org/trac/mpi-forum-web/attachment/wiki/mpi3-rma-propo… > > Thanks & All the Best, > Torsten > > -- > bash$ :(){ :|:&};: --------------------- http://www.unixer.de/ ----- > Torsten Hoefler | Performance Modeling and Simulation Lead > Blue Waters Directorate | University of Illinois (UIUC) > 1205 W Clark Street | Urbana, IL, 61801 > NCSA Building | +01 (217) 244-7736 > <diff-01022011.txt>_______________________________________________ > mpi3-rma mailing list > mpi3-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-rma

1 0

Re: [Mpi3-rma] Draft Plenary Presentation for MPI RMA
by Torsten Hoefler 04 Feb '11

04 Feb '11

Hi all, I just uploaded an updated version to the wiki that includes all fixes discussed here. >>> On 02/04/2011 01:14 PM, Torsten Hoefler wrote: >>>>> Slide 27: ARMCI has per-operation non-blocking calls. Blocking calls >>>>> and waits give you local completion and only fence gives you remote >>>>> completion. >>>> Yes, ARMCI has per operation remote completion which we don't support at >>>> this point. Proposal 2 had this a while ago but we didn't find a strong >>>> use-case (as opposed to the per-operation local completion). Added a >>>> comment. >>> >>> No, ARMCI doesn't have per operation remote completion; only local >>> completion. Attached the slides I had distributed earlier about this for >>> your reference. >> Ok, I got this impression from our previous discussion. So Jim, what did >> your comment then mean? > > > I meant the same thing as what Pavan wrote, but had a language fail: > ARMCI gives per-operation local completion for non-blocking operations. > Blocking operations are locally complete when they return. Bulk remote > completion is accomplished through fence. Ok, this should all be in there. Best, Torsten -- bash$ :(){ :|:&};: --------------------- http://www.unixer.de/ ----- A man can't be too careful in the choice of his enemies. --Oscar Wilde

1 0