August 2011 - mpiwg-hybridpm - mailman3.cels.anl.gov

Re: [Mpi3-hybridpm] Hybrid telecon: 08/30 @ 12 noon central time
by Pavan Balaji 30 Aug '11

30 Aug '11

All, The meeting minutes are uploaded here: https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/notes-2011-08-… I've also updated the MPI_Comm_split_type proposal based on the comments during the telecon and uploaded a new draft on the ticket: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/287 Thanks, -- Pavan On 08/27/2011 11:23 PM, Pavan Balaji wrote: > All, > > We'll have the next hybrid telecon on 08/30 (Tuesday) @ 12 noon central > time. The telecon information is available on the wiki. > > https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid > > The agenda items for the telecon are: > > 1. Discuss the communication creation routine MPI_Comm_split_type. The > proposal text has been uploaded to the ticket. We will have a formal > reading for this at the next Forum. > > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/287 > > 2. Discuss the endpoint proposal slides that will be presented at the > next forum. This will be a plenary session at the next Forum, but not a > formal reading. Marc will send out the slides to the mailing list this > weekend. > > 3. Discuss the helper team proposal and have a straw vote on whether we > want to take this to the Forum for a formal reading. > > We might not be able to cover all of the above items in this telecon and > might need one more telecon before the Forum meeting. > > Also, we should talk about whether we need the Hybrid working group time > on Sep. 22nd at the Forum. We have the slot, but if we can cover > everything we need to during the telecons before the Forum, we can give > up that time at the Forum. > > Thanks, > > -- Pavan > -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

[Mpi3-hybridpm] Fwd: Re: Telecon next week
by Brightwell, Ronald 30 Aug '11

30 Aug '11

-------- Original Message -------- Subject: Re: Telecon next week Date: Sun, 28 Aug 2011 20:59:57 +0000 From: Snir, Marc <snir(a)illinois.edu> To: Pavan Balaji <balaji(a)mcs.anl.gov>, Ron Brightwell <rbbrigh(a)sandia.gov>, "Hoefler, Torsten" <htor(a)illinois.edu>, James Dinan <dinan(a)mcs.anl.gov>, Douglas Miller <dougmill(a)us.ibm.com>, "Bronis R. de Supinski" <bronis(a)llnl.gov>, Brian W Barrett <bwbarre(a)sandia.gov>, Steven Oyanagi <sko(a)cray.com> The shortened Endpoint proposal is now ticket #288 and the communicator merge proposal is ticket #289. I attach slides presenting these 2 proposals.

1 0

Re: [Mpi3-hybridpm] Hybrid telecon: 08/30 @ 12 noon central time
by Pavan Balaji 29 Aug '11

29 Aug '11

Ah, good catch. I want to discuss what we should do for MPI_UNDEFINED anyway. Once we come to a consensus on that, I'll fix this part too. -- Pavan On 08/29/2011 01:32 PM, Rajeev Thakur wrote: >> 1. Discuss the communication creation routine MPI_Comm_split_type. The proposal text has been uploaded to the ticket. We will have a formal reading for this at the next Forum. >> >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/287 > > The text for MPI_Comm_split_type does not explicitly say that "type" must be the same on all processes. I would put that in the last sentence: > > "This is a collective call; each process must provide the same value for type, but can provide different values for key." > > Rajeev > > On Aug 27, 2011, at 11:23 PM, Pavan Balaji wrote: > >> All, >> >> We'll have the next hybrid telecon on 08/30 (Tuesday) @ 12 noon central time. The telecon information is available on the wiki. >> >> https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid >> >> The agenda items for the telecon are: >> >> 1. Discuss the communication creation routine MPI_Comm_split_type. The proposal text has been uploaded to the ticket. We will have a formal reading for this at the next Forum. >> >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/287 >> >> 2. Discuss the endpoint proposal slides that will be presented at the next forum. This will be a plenary session at the next Forum, but not a formal reading. Marc will send out the slides to the mailing list this weekend. >> >> 3. Discuss the helper team proposal and have a straw vote on whether we want to take this to the Forum for a formal reading. >> >> We might not be able to cover all of the above items in this telecon and might need one more telecon before the Forum meeting. >> >> Also, we should talk about whether we need the Hybrid working group time on Sep. 22nd at the Forum. We have the slot, but if we can cover everything we need to during the telecons before the Forum, we can give up that time at the Forum. >> >> Thanks, >> >> -- Pavan >> >> -- >> Pavan Balaji >> http://www.mcs.anl.gov/~balaji >> _______________________________________________ >> Mpi3-hybridpm mailing list >> Mpi3-hybridpm(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm > > > _______________________________________________ > Mpi3-hybridpm mailing list > Mpi3-hybridpm(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [Mpi3-hybridpm] Hybrid telecon: 08/30 @ 12 noon central time
by Rajeev Thakur 29 Aug '11

29 Aug '11

> 1. Discuss the communication creation routine MPI_Comm_split_type. The proposal text has been uploaded to the ticket. We will have a formal reading for this at the next Forum. > > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/287 The text for MPI_Comm_split_type does not explicitly say that "type" must be the same on all processes. I would put that in the last sentence: "This is a collective call; each process must provide the same value for type, but can provide different values for key." Rajeev On Aug 27, 2011, at 11:23 PM, Pavan Balaji wrote: > All, > > We'll have the next hybrid telecon on 08/30 (Tuesday) @ 12 noon central time. The telecon information is available on the wiki. > > https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid > > The agenda items for the telecon are: > > 1. Discuss the communication creation routine MPI_Comm_split_type. The proposal text has been uploaded to the ticket. We will have a formal reading for this at the next Forum. > > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/287 > > 2. Discuss the endpoint proposal slides that will be presented at the next forum. This will be a plenary session at the next Forum, but not a formal reading. Marc will send out the slides to the mailing list this weekend. > > 3. Discuss the helper team proposal and have a straw vote on whether we want to take this to the Forum for a formal reading. > > We might not be able to cover all of the above items in this telecon and might need one more telecon before the Forum meeting. > > Also, we should talk about whether we need the Hybrid working group time on Sep. 22nd at the Forum. We have the slot, but if we can cover everything we need to during the telecons before the Forum, we can give up that time at the Forum. > > Thanks, > > -- Pavan > > -- > Pavan Balaji > http://www.mcs.anl.gov/~balaji > _______________________________________________ > Mpi3-hybridpm mailing list > Mpi3-hybridpm(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm

1 0

Re: [Mpi3-hybridpm] [Mpi3-rma] MPI_Win_Allocate_shared proposal draft
by Rajeev Thakur 29 Aug '11

29 Aug '11

> 6:39 - I would like the sentence to clearly say that it is the user's responsibility to pass a communicator whose processes can share memory. How about changing it to > > "It is the user's responsibility to ensure that the communicator comm represents a group of processes that can create a shared memory segment that can be accessed by all processes in the group." > > Then we should say what happens if the communicator contains processes that cannot share memory. If the new function MPI_Comm_split_type passes, we need to remember to provide a reference here to it, because it is directly relevant for this use case. Rajeev On Aug 29, 2011, at 1:16 PM, Rajeev Thakur wrote: > Looks good. Some comments: > > 6:14 - baseptrs -> baseptr > > 6:26 - DISP_UNIT -> SIZE > > 6:30 - "among all processes" -> "among all processes in comm" > > 6:31 - "returns" -> "and returns" > > 6:32 - "be target" -> "be the target" > > 6:33 - "accessed" -> "accesses" > > 6:33 - change to "remote processes; the base pointers for other processes can be queried..." > > 6:35 - Much of the sentence "The returned memory..." seems to be a repetition of what has been said before. > > 6:39 - I would like the sentence to clearly say that it is the user's responsibility to pass a communicator whose processes can share memory. How about changing it to > > "It is the user's responsibility to ensure that the communicator comm represents a group of processes that can create a shared memory segment that can be accessed by all processes in the group." > > Then we should say what happens if the communicator contains processes that cannot share memory. > > 7:10 - "as info key" -> "as an info key" > > 7:35 - MPI_WIN_FLAVOR_SHM is not defined in 11.2.6 > > > Since the default allocation is supposed to be contiguous, is the intended implementation method supposed to be something like the following: > * wait for everyone to call the function > * gather everyone's local size; add it up > * allocate single, large shared memory > * return local pointers everywhere > > > On Aug 29, 2011, at 12:12 AM, Torsten Hoefler wrote: > >> Hello all, >> >> We just posted a new update to the proposal (thanks to Bill for the >> comments!) to the ticket at >> https://svn.mpi-forum.org/trac/mpi-forum-web/attachment/ticket/284/ >> >> 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 >> _______________________________________________ >> mpi3-rma mailing list >> mpi3-rma(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-rma > > > _______________________________________________ > 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-hybridpm] [Mpi3-rma] MPI_Win_Allocate_shared proposal draft
by Rajeev Thakur 29 Aug '11

29 Aug '11

Looks good. Some comments: 6:14 - baseptrs -> baseptr 6:26 - DISP_UNIT -> SIZE 6:30 - "among all processes" -> "among all processes in comm" 6:31 - "returns" -> "and returns" 6:32 - "be target" -> "be the target" 6:33 - "accessed" -> "accesses" 6:33 - change to "remote processes; the base pointers for other processes can be queried..." 6:35 - Much of the sentence "The returned memory..." seems to be a repetition of what has been said before. 6:39 - I would like the sentence to clearly say that it is the user's responsibility to pass a communicator whose processes can share memory. How about changing it to "It is the user's responsibility to ensure that the communicator comm represents a group of processes that can create a shared memory segment that can be accessed by all processes in the group." Then we should say what happens if the communicator contains processes that cannot share memory. 7:10 - "as info key" -> "as an info key" 7:35 - MPI_WIN_FLAVOR_SHM is not defined in 11.2.6 Since the default allocation is supposed to be contiguous, is the intended implementation method supposed to be something like the following: * wait for everyone to call the function * gather everyone's local size; add it up * allocate single, large shared memory * return local pointers everywhere On Aug 29, 2011, at 12:12 AM, Torsten Hoefler wrote: > Hello all, > > We just posted a new update to the proposal (thanks to Bill for the > comments!) to the ticket at > https://svn.mpi-forum.org/trac/mpi-forum-web/attachment/ticket/284/ > > 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 > _______________________________________________ > 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-hybridpm] [Mpi3-rma] MPI_Win_Allocate_shared proposal draft
by Torsten Hoefler 29 Aug '11

29 Aug '11

Hello all, We just posted a new update to the proposal (thanks to Bill for the comments!) to the ticket at https://svn.mpi-forum.org/trac/mpi-forum-web/attachment/ticket/284/ 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-hybridpm] Hybrid telecon: 08/30 @ 12 noon central time
by Pavan Balaji 27 Aug '11

27 Aug '11

All, We'll have the next hybrid telecon on 08/30 (Tuesday) @ 12 noon central time. The telecon information is available on the wiki. https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid The agenda items for the telecon are: 1. Discuss the communication creation routine MPI_Comm_split_type. The proposal text has been uploaded to the ticket. We will have a formal reading for this at the next Forum. https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/287 2. Discuss the endpoint proposal slides that will be presented at the next forum. This will be a plenary session at the next Forum, but not a formal reading. Marc will send out the slides to the mailing list this weekend. 3. Discuss the helper team proposal and have a straw vote on whether we want to take this to the Forum for a formal reading. We might not be able to cover all of the above items in this telecon and might need one more telecon before the Forum meeting. Also, we should talk about whether we need the Hybrid working group time on Sep. 22nd at the Forum. We have the slot, but if we can cover everything we need to during the telecons before the Forum, we can give up that time at the Forum. Thanks, -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [Mpi3-hybridpm] Ticket 217 - a.k.a. Helper Threads
by Douglas Miller 26 Aug '11

26 Aug '11

I've also updated the helper threads document in ticket #217, to cleanup wording and add a paragraph to the motivation/introduction. https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/217 _______________________________________________ Douglas Miller BlueGene Messaging Development IBM Corp., Rochester, MN USA Bldg 030-2 A401 dougmill(a)us.ibm.com Douglas Miller/Rochester/IBM Douglas Miller/Rochester/ IBM@IBMUS To Sent by: mpi3-hybridpm(a)lists.mpi-forum.org mpi3-hybridpm-bou cc nces(a)lists.mpi-fo rum.org Subject Re: [Mpi3-hybridpm] Ticket 217 - a.k.a. Helper Threads 08/19/2011 05:19 PM Please respond to mpi3-hybridpm@lis ts.mpi-forum.org I have a modified External Interfaces chapter for Ticket 217 "Helper Threads". Both LaTex and PDF are now attached to ticket 217, for review and discussion at the next meeting. The SVN copy of the chapter HAS NOT been updated. https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/217 thanks, _______________________________________________ Douglas Miller BlueGene Messaging Development IBM Corp., Rochester, MN USA Bldg 030-2 A401 dougmill(a)us.ibm.com Douglas Miller/Rochester/IBM _______________________________________________ Mpi3-hybridpm mailing list Mpi3-hybridpm(a)lists.mpi-forum.org http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm

1 0

Re: [Mpi3-hybridpm] MPI_Comm_split_type
by Pavan Balaji 26 Aug '11

26 Aug '11

I have uploaded a new version of the document on the ticket. It corrects the info argument problem and lists an item for discussion for the next telecon. https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/287 -- Pavan On 08/26/2011 02:18 AM, Pavan Balaji wrote: > > On 08/26/2011 02:12 AM, Supalov, Alexander wrote: >> Thanks. The advice to implementers mentioned an Info argument (see >> 23/6). There appears to be no MPI_Info argument in the syntax (see >> 22/26). Is this a remnant of the ol'times, or a glint of the future? >> >> Actually, having an MPI_Info here would be very tempting, as one >> could, e.g., ask the new communicator ranks to be reordered to honour >> cache hierarchy, or latency, etc. With the time we could then also >> agree on some keys/effects common to all implementations. > > Oops, the info argument was always the plan. I somehow messed up the > prototype. Thanks for catching it. > > We did have a detailed discussion on whether we can predefine some info > arguments, but there was no consensus. So we decided to leave it as > implementation specific. If you have any ideas, please feel free to > propose them. > > -- Pavan > -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0