October 2020 - mpi-forum - mailman3.cels.anl.gov

Re: [Mpi-forum] Progress Question
by Thakur, Rajeev 11 Oct '20

11 Oct '20

Does it mean that in the following program, although all processes have called barrier, some process may not exit the barrier for 100 days? MPI_Init MPI_Barrier sleep(100 days) MPI_Finalize Rajeev From: Jim Dinan <james.dinan(a)gmail.com> Date: Sunday, October 11, 2020 at 10:31 AM To: "Thakur, Rajeev" <thakur(a)anl.gov> Cc: Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org> Subject: Re: [Mpi-forum] Progress Question Hi Rajeev, Yes, that's the question and my initial answer was the same as yours. However, we then started talking about the implementation of the barrier, which led to the second example. For example, consider a situation where there is an error in transmission and the implementation needs to enter the progress engine to retry a send operation in software. ~Jim. On Sat, Oct 10, 2020 at 5:10 PM Thakur, Rajeev <thakur(a)anl.gov<mailto:[email protected]>> wrote: Jim, I don’t fully understand your question. Is it “If all processes reach MPI_Barrier, are they guaranteed to exit the barrier without the need for any other MPI function to be called on any process?” I would say yes. Rajeev From: mpi-forum <mpi-forum-bounces(a)lists.mpi-forum.org<mailto:[email protected]>> on behalf of Jim Dinan via mpi-forum <mpi-forum(a)lists.mpi-forum.org<mailto:[email protected]>> Reply-To: Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org<mailto:[email protected]>> Date: Saturday, October 10, 2020 at 12:31 PM To: Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org<mailto:[email protected]>> Cc: Jim Dinan <james.dinan(a)gmail.com<mailto:[email protected]>> Subject: [Mpi-forum] Progress Question Hi All, A colleague recently asked a question that I wasn't able to answer definitively. Is the following code guaranteed to make progress? MPI_Barrier(); if rank == 1 create_file("test") if rank == 0 while not_exists("test") sleep(1); That is, can rank 1 require rank 0 to make MPI calls after its return from the barrier, in order for rank 1 to complete the barrier? If the code were written as follows: isend(..., other_rank, &req[0]) irecv(..., other_rank, &req[1]) waitall(2, req) if rank == 1 create_file("test") if rank == 0 while not_exists("test") sleep(1); I think it would clearly not guarantee progress since the send data can be buffered. Is the same true for barrier? Cheers, ~Jim.

1 0

Re: [Mpi-forum] Progress Question
by Jim Dinan 11 Oct '20

11 Oct '20

Hi Rajeev, Yes, that's the question and my initial answer was the same as yours. However, we then started talking about the implementation of the barrier, which led to the second example. For example, consider a situation where there is an error in transmission and the implementation needs to enter the progress engine to retry a send operation in software. ~Jim. On Sat, Oct 10, 2020 at 5:10 PM Thakur, Rajeev <thakur(a)anl.gov> wrote: > Jim, > > I don’t fully understand your question. Is it “If all processes > reach MPI_Barrier, are they guaranteed to exit the barrier without the need > for any other MPI function to be called on any process?” I would say yes. > > > > Rajeev > > > > > > *From: *mpi-forum <mpi-forum-bounces(a)lists.mpi-forum.org> on behalf of > Jim Dinan via mpi-forum <mpi-forum(a)lists.mpi-forum.org> > *Reply-To: *Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org> > *Date: *Saturday, October 10, 2020 at 12:31 PM > *To: *Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org> > *Cc: *Jim Dinan <james.dinan(a)gmail.com> > *Subject: *[Mpi-forum] Progress Question > > > > Hi All, > > > > A colleague recently asked a question that I wasn't able to answer > definitively. Is the following code guaranteed to make progress? > > > > MPI_Barrier(); > > if rank == 1 > > create_file("test") > > if rank == 0 > > while not_exists("test") > > sleep(1); > > > > That is, can rank 1 require rank 0 to make MPI calls after its return from > the barrier, in order for rank 1 to complete the barrier? If the code were > written as follows: > > > > isend(..., other_rank, &req[0]) > > irecv(..., other_rank, &req[1]) > > waitall(2, req) > > if rank == 1 > > create_file("test") > > if rank == 0 > > while not_exists("test") > > sleep(1); > > > > I think it would clearly not guarantee progress since the send data can be > buffered. Is the same true for barrier? > > > > Cheers, > > ~Jim. >

1 0

Re: [Mpi-forum] Progress Question
by Thakur, Rajeev 10 Oct '20

10 Oct '20

Jim, I don’t fully understand your question. Is it “If all processes reach MPI_Barrier, are they guaranteed to exit the barrier without the need for any other MPI function to be called on any process?” I would say yes. Rajeev From: mpi-forum <mpi-forum-bounces(a)lists.mpi-forum.org> on behalf of Jim Dinan via mpi-forum <mpi-forum(a)lists.mpi-forum.org> Reply-To: Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org> Date: Saturday, October 10, 2020 at 12:31 PM To: Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org> Cc: Jim Dinan <james.dinan(a)gmail.com> Subject: [Mpi-forum] Progress Question Hi All, A colleague recently asked a question that I wasn't able to answer definitively. Is the following code guaranteed to make progress? MPI_Barrier(); if rank == 1 create_file("test") if rank == 0 while not_exists("test") sleep(1); That is, can rank 1 require rank 0 to make MPI calls after its return from the barrier, in order for rank 1 to complete the barrier? If the code were written as follows: isend(..., other_rank, &req[0]) irecv(..., other_rank, &req[1]) waitall(2, req) if rank == 1 create_file("test") if rank == 0 while not_exists("test") sleep(1); I think it would clearly not guarantee progress since the send data can be buffered. Is the same true for barrier? Cheers, ~Jim.

1 0

Re: [Mpi-forum] [EXT]: Progress Question
by Skjellum, Anthony 10 Oct '20

10 Oct '20

Jim, OK, my attempt at answering below. See if you agree with my annotations. -Tony Anthony Skjellum, PhD Professor of Computer Science and Chair of Excellence Director, SimCenter University of Tennessee at Chattanooga (UTC) tony-skjellum(a)utc.edu [or skjellum(a)gmail.com] cell: 205-807-4968 ________________________________ From: mpi-forum <mpi-forum-bounces(a)lists.mpi-forum.org> on behalf of Jim Dinan via mpi-forum <mpi-forum(a)lists.mpi-forum.org> Sent: Saturday, October 10, 2020 1:31 PM To: Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org> Cc: Jim Dinan <james.dinan(a)gmail.com> Subject: [EXT]: [Mpi-forum] Progress Question External Email Hi All, A colleague recently asked a question that I wasn't able to answer definitively. Is the following code guaranteed to make progress? MPI_Barrier(); -- everything is uncertain to within one message, if layered on pt2pt; --- let's assume a power of 2, and recursive doubling (RD). --- At each stage, it posts an irecv and isend to its corresponding element in RD --- All stages must complete to get to the last stage. --- At the last stage, it appears like your example below for N/2 independent process pairs, which appears always to complete. Oif rank == 1 create_file("test") if rank == 0 while not_exists("test") sleep(1); That is, can rank 1 require rank 0 to make MPI calls after its return from the barrier, in order for rank 1 to complete the barrier? If the code were written as follows: isend(..., other_rank, &req[0]) irecv(..., other_rank, &req[1]) waitall(2, req) --- Assume both isends buffer on the send-side and return immediately--valid. --- Both irecvs are posted, but unmatched as yet. Nothing has transferred on network. --- Waitall would mark the isends done at once, and work to complete the irecvs; in that process, each would have to progress the isends across the network. On this comm and all comms, incidentally. --- When waitall returns, the data has transferred to the receiver, otherwise the irecvs aren't done. if rank == 1 create_file("test") if rank == 0 while not_exists("test") sleep(1); I think it would clearly not guarantee progress since the send data can be buffered. Is the same true for barrier? Cheers, ~Jim. This message is not from a UTC.EDU address. Caution should be used in clicking links and downloading attachments from unknown senders or unexpected email.

1 0

Progress Question
by Jim Dinan 10 Oct '20

10 Oct '20

Hi All, A colleague recently asked a question that I wasn't able to answer definitively. Is the following code guaranteed to make progress? MPI_Barrier(); if rank == 1 create_file("test") if rank == 0 while not_exists("test") sleep(1); That is, can rank 1 require rank 0 to make MPI calls after its return from the barrier, in order for rank 1 to complete the barrier? If the code were written as follows: isend(..., other_rank, &req[0]) irecv(..., other_rank, &req[1]) waitall(2, req) if rank == 1 create_file("test") if rank == 0 while not_exists("test") sleep(1); I think it would clearly not guarantee progress since the send data can be buffered. Is the same true for barrier? Cheers, ~Jim.

1 0

Re: [Mpi-forum] [EXTERNAL] Last Call for MPI 4.0 Issues
by Wesley Bland 07 Oct '20

07 Oct '20

Yes. Sorry I was unclear on the title there. That’s correct. > On Oct 7, 2020, at 12:34 PM, Pritchard Jr., Howard <howardp(a)lanl.gov> wrote: > > HI Wesley, > > By project do you mean the MPI 4.0 ratification? > > Howard > > From: mpi-forum <mpi-forum-bounces(a)lists.mpi-forum.org <mailto:[email protected]>> on behalf of mpi-forum <mpi-forum(a)lists.mpi-forum.org <mailto:[email protected]>> > Reply-To: mpi-forum <mpi-forum(a)lists.mpi-forum.org <mailto:[email protected]>> > Date: Wednesday, October 7, 2020 at 11:11 AM > To: mpi-forum <mpi-forum(a)lists.mpi-forum.org <mailto:[email protected]>> > Cc: Wesley Bland <work(a)wesbland.com <mailto:[email protected]>> > Subject: [EXTERNAL] [Mpi-forum] Last Call for MPI 4.0 Issues > > Hi all, > > (Action Item for everyone in bold below) > > For those of you who couldn’t attend the last day of last week’s MPI Forum meeting, I’d highly recommend that you want the presentation I gave about the process we’ll be going through over the next few months to ratify MPI 4.0. The video is available here: > > https://youtu.be/zX1XneiVTUg?t=2249 <https://youtu.be/zX1XneiVTUg?t=2249> (starts at 37:40) > > You can also find the slides here: > > https://github.com/mpi-forum/mpi-forum.github.io/blob/master/slides/2020/09… <https://github.com/mpi-forum/mpi-forum.github.io/blob/master/slides/2020/09…> > > As part of the virtual meeting next week, I plan to hold “office hours” to answer any questions people may have about that. I’ll also be going over the details of exactly how some of these pieces will work where I left some gaps in the original presentation (e.g., What will the Git branches look like? How do we make PRs? Where do the issues go?). Everyone who will be helping review the Standard before December should plan to attend (that’s everyone who’s attended 3 meetings this year), chapter committees in particular. > > In anticipation of that, each person should go through any issues or pull requests that they’ve opened on GitHub and see if there are any last minute items for MPI 4.0. I made a pass through every PR today and cleaned up quite a few, but there may still be some that need to be added to the agenda. To do this, go find all of the issues where you are either an author or an assignee (you can find how to do that in the attached image) and decide whether it should be for MPI 4.0, MPI 4.1, MPI 5.0, or closed. There are labels for the first three and you can just close it for the last one. As usual, if you have questions, just ask me. If you want it to be in MPI 4.0, please also add it to the MPI 4.0 “project” so we’ll be able to track it. > > <image001.png> > > Please note that anything that is going into MPI 4.0 really needs to be called out in the next few weeks. Bill will be generating the release candidate at the beginning of November and everything needs to be merged before that happens. If something gets missed this month, it will probably have to wait for MPI 4.1. > > Thanks, > Wes

1 0

Re: [Mpi-forum] [EXTERNAL] Last Call for MPI 4.0 Issues
by Pritchard Jr., Howard 07 Oct '20

07 Oct '20

HI Wesley, By project do you mean the MPI 4.0 ratification? Howard From: mpi-forum <mpi-forum-bounces(a)lists.mpi-forum.org> on behalf of mpi-forum <mpi-forum(a)lists.mpi-forum.org> Reply-To: mpi-forum <mpi-forum(a)lists.mpi-forum.org> Date: Wednesday, October 7, 2020 at 11:11 AM To: mpi-forum <mpi-forum(a)lists.mpi-forum.org> Cc: Wesley Bland <work(a)wesbland.com> Subject: [EXTERNAL] [Mpi-forum] Last Call for MPI 4.0 Issues Hi all, (Action Item for everyone in bold below) For those of you who couldn’t attend the last day of last week’s MPI Forum meeting, I’d highly recommend that you want the presentation I gave about the process we’ll be going through over the next few months to ratify MPI 4.0. The video is available here: https://youtu.be/zX1XneiVTUg?t=2249 (starts at 37:40) You can also find the slides here: https://github.com/mpi-forum/mpi-forum.github.io/blob/master/slides/2020/09… As part of the virtual meeting next week, I plan to hold “office hours” to answer any questions people may have about that. I’ll also be going over the details of exactly how some of these pieces will work where I left some gaps in the original presentation (e.g., What will the Git branches look like? How do we make PRs? Where do the issues go?). Everyone who will be helping review the Standard before December should plan to attend (that’s everyone who’s attended 3 meetings this year), chapter committees in particular. In anticipation of that, each person should go through any issues or pull requests that they’ve opened on GitHub and see if there are any last minute items for MPI 4.0. I made a pass through every PR today and cleaned up quite a few, but there may still be some that need to be added to the agenda. To do this, go find all of the issues where you are either an author or an assignee (you can find how to do that in the attached image) and decide whether it should be for MPI 4.0, MPI 4.1, MPI 5.0, or closed. There are labels for the first three and you can just close it for the last one. As usual, if you have questions, just ask me. If you want it to be in MPI 4.0, please also add it to the MPI 4.0 “project” so we’ll be able to track it. [cid:[email protected]] Please note that anything that is going into MPI 4.0 really needs to be called out in the next few weeks. Bill will be generating the release candidate at the beginning of November and everything needs to be merged before that happens. If something gets missed this month, it will probably have to wait for MPI 4.1. Thanks, Wes

1 0

Last Call for MPI 4.0 Issues
by Wesley Bland 07 Oct '20

07 Oct '20

Hi all, (Action Item for everyone in bold below) For those of you who couldn’t attend the last day of last week’s MPI Forum meeting, I’d highly recommend that you want the presentation I gave about the process we’ll be going through over the next few months to ratify MPI 4.0. The video is available here: https://youtu.be/zX1XneiVTUg?t=2249 <https://youtu.be/zX1XneiVTUg?t=2249> (starts at 37:40) You can also find the slides here: https://github.com/mpi-forum/mpi-forum.github.io/blob/master/slides/2020/09… <https://github.com/mpi-forum/mpi-forum.github.io/blob/master/slides/2020/09…> As part of the virtual meeting next week, I plan to hold “office hours” to answer any questions people may have about that. I’ll also be going over the details of exactly how some of these pieces will work where I left some gaps in the original presentation (e.g., What will the Git branches look like? How do we make PRs? Where do the issues go?). Everyone who will be helping review the Standard before December should plan to attend (that’s everyone who’s attended 3 meetings this year), chapter committees in particular. In anticipation of that, each person should go through any issues or pull requests that they’ve opened on GitHub and see if there are any last minute items for MPI 4.0. I made a pass through every PR today and cleaned up quite a few, but there may still be some that need to be added to the agenda. To do this, go find all of the issues where you are either an author or an assignee (you can find how to do that in the attached image) and decide whether it should be for MPI 4.0, MPI 4.1, MPI 5.0, or closed. There are labels for the first three and you can just close it for the last one. As usual, if you have questions, just ask me. If you want it to be in MPI 4.0, please also add it to the MPI 4.0 “project” so we’ll be able to track it. Please note that anything that is going into MPI 4.0 really needs to be called out in the next few weeks. Bill will be generating the release candidate at the beginning of November and everything needs to be merged before that happens. If something gets missed this month, it will probably have to wait for MPI 4.1. Thanks, Wes

1 0

Re: [Mpi-forum] Virtual Meeting Tomorrow
by Wesley Bland 07 Oct '20

07 Oct '20

I didn’t hear any requests so today’s virtual meeting is cancelled. Thanks, Wes > On Oct 6, 2020, at 6:43 PM, Wesley Bland <work(a)wesbland.com> wrote: > > Hi all, > > We have a virtual meeting scheduled for tomorrow at the usual time 10am US Central. The agenda is anything from the various chapter committees to start getting ready for MPI 4.0. Are any committees ready to present anything? > > If I don’t hear anything tonight, I’ll assume that we should start up next week. I ought to have some more detailed guidance about how to do the git reviews then. > > Thanks, > Wes

1 0

Virtual Meeting Tomorrow
by Wesley Bland 06 Oct '20

06 Oct '20

Hi all, We have a virtual meeting scheduled for tomorrow at the usual time 10am US Central. The agenda is anything from the various chapter committees to start getting ready for MPI 4.0. Are any committees ready to present anything? If I don’t hear anything tonight, I’ll assume that we should start up next week. I ought to have some more detailed guidance about how to do the git reviews then. Thanks, Wes

1 0