December 2009 - Mpich2-core - mailman3.cels.anl.gov

MPICH2 code
by Elad Yom-Tov 04 Jan '10

04 Jan '10

Hello, I am working at IBM and I would like to use the latest version of MPICH2 (version 1.2.1). In the past we used version 1.0.5 and I noticed there were many new authors in the latest version. Do you have a process to check if the authors have authority to contribute the code they submit? Or do you keep track in any way of the contributors? Additionally, I noticed quite a few new references to FSF and GPL code. Is this new version now GPL licensed? And if not, are these hits only references to GPL/FSF and not actual code licensed under GPL? Many thanks, Elad --------------------------------------------------------------------------- Elad Yom-Tov IBM Research Laboratory in Haifa E-mail: yomtov(a)il.ibm.com Web: http://www.haifa.il.ibm.com/dept/svt/simulation_vsml.html Phone: 972-4-8296486, Fax: 972-4-8296114, Mobile +972-54-6399334 Address: IBM Haifa Research Lab, Haifa 31905, Israel

5 9

Use of multiple libraries
by William Gropp 04 Jan '10

04 Jan '10

Is there a reason that all of the necessary routines aren't added to the libmpich (and/or libpmpich when weak symbol aliases aren't available)? We have users who do not use the compilation scripts and expect to provide a single include path for compiling and library for linking (which admittedly doesn't work if they need additional system libraries, such as -lpthread, though such users are often used to adding system libraries). And I just had a bunch of tests fail because libmpl wasn't found (a different problem, caused by an inconsistent setting of prefix). Even if the libraries are built using the so-called convenience library model (which we'll probably need to adopt to make use of the other autotools), the libraries can be merged at the end of the build. 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

3 3

fyi
by Rajeev Thakur 24 Dec '09

24 Dec '09

FYI, Open MPI doesn't yet support fault tolerance... Date: Wed, 23 Dec 2009 05:55:22 -0700 From: Ralph Castain <rhc(a)open-mpi.org> Subject: Re: [OMPI users] fault tolerance in open mpi To: Open MPI Users <users(a)open-mpi.org> Message-ID: <57BB24AE-4DE9-4212-9A5A-D86639BB2B5C(a)open-mpi.org> Content-Type: text/plain; charset="us-ascii" That's just OMPI's default behavior - as Josh said, we are working towards allowing other behaviors, but for now, this is what we have. On Dec 23, 2009, at 5:40 AM, vipin kumar wrote: > Thank you Ralph, > > I did as you said. Programs are running fine, But still killing one process leads to terminate all processes. Am I missing something? Any thing else to be called with MPI::Comm::Disconnect()? > > > Thanks & Regards, > > On Mon, Dec 21, 2009 at 8:00 PM, Ralph Castain <rhc(a)open-mpi.org> wrote: > Disconnect is a -collective- operation. Both parent and child have to call it. Your child process is "hanging" while it waits for the parent. > > On Dec 21, 2009, at 1:37 AM, vipin kumar wrote: > >> Hello folks, >> >> As I explained my problem earlier, I am looking for Fault Tolerance in MPI Programs. I read in Open MPI 2.1 standard document that two DISCONNECTED processes does not affect each other, i.e. they can die or can be killed without whithout affecting other processes. >> >> So, I was trying this to achieve fault tolerance using MPI::Comm::Disconnect() to disconnect the CHILD process with PARENT process, which was spawned by calling MPI::Comm::spawn(). I am calling MPI::Comm::Disconnect() from CHILD process immediatly after calling MPI::Init(). It seems that CHILD process is not returning from this call. >> >> I tried MPI::Comm::Free() too, but this is also not working. Process is not progressing from this point of call. If I comment these statements, everything works fine. Note that I have tried this in Solaris as well as in Linux (fedora core). >> >> My question is, whether Open-mpi suports to disconnect two processes( like child from parent). And if it is, then how? >> >> >> Thanks & Regards, >> >> On Wed, Sep 23, 2009 at 6:41 PM, Josh Hursey <jjhursey(a)open-mpi.org> wrote: >> Unfortunately I cannot provide a precise time frame for availability at this point, but we are targeting the v1.5 release series. There is a handful of core developers working on this issue at the moment. Pieces of this work have already made it into the Open MPI development trunk. If you want to play around with what is available try turning on the resilient mapper: >> -mca rmaps resilient >> >> We will be sure to email the list once this work becomes more stable and available. >> >> -- Josh >> >> >> On Sep 18, 2009, at 2:56 AM, vipin kumar wrote: >> >> Hi Josh, >> >> It is good to hear from you that work is in progress towards resiliency of Open-MPI. I was and I am waiting for this capability in Open-MPI. I have almost finished my development work and waiting for this to happen so that I can test my programs. It will be good if you can tell how long it will take to make Open-MPI a resilient impementation. Here by resiliency I mean abnormal termination or intentionally killing a process should not cause any(parent or sibling) process to be terminated, given that processes are connected. >> >> thanks. >> >> Regards, >> >> On Mon, Aug 3, 2009 at 8:37 PM, Josh Hursey <jjhursey(a)open-mpi.org> wrote: >> Task-farm or manager/worker recovery models typically depend on intercommunicators (i.e., from MPI_Comm_spawn) and a resilient MPI implementation. William Gropp and Ewing Lusk have a paper entitled "Fault Tolerance in MPI Programs" that outlines how an application might take advantage of these features in order to recover from process failure. >> >> However, these techniques strongly depend upon resilient MPI implementations, and behaviors that, some may argue, are non-standard. Unfortunately there are not many MPI implementations that are sufficiently resilient in the face of process failure to support failure in task-farm scenarios. Though Open MPI supports the current MPI 2.1 standard, it is not as resilient to process failure as it could be. >> >> There are a number of people working on improving the resiliency of Open MPI in the face of network and process failure (including myself). We have started to move some of the resiliency work into the Open MPI trunk. Resiliency in Open MPI has been improving over the past few months, but I would not assess it as ready quite yet. Most of the work has focused on the runtime level (ORTE), and there are still some MPI level (OMPI) issues that need to be worked out. >> >> With all of that being said, I would try some of the techniques presented in the Gropp/Lusk paper in your application. Then test it with Open MPI and let us know how it goes. >> >> Best, >> Josh >> >> >> On Aug 3, 2009, at 10:30 AM, Durga Choudhury wrote: >> >> Is that kind of approach possible within an MPI framework? Perhaps a >> grid approach would be better. More experienced people, speak up, >> please? >> (The reason I say that is that I too am interested in the solution of >> that kind of problem, where an individual blade of a blade server >> fails and correcting for that failure on the fly is better than >> taking checkpoints and restarting the whole process excluding the >> failed blade. >> >> Durga >> >> On Mon, Aug 3, 2009 at 9:21 AM, jody<jody.xha(a)gmail.com> wrote: >> Hi >> >> I guess "task-farming" could give you a certain amount of the kind of >> fault-tolerance you want. >> (i.e. a master process distributes tasks to idle slave processors - >> however, this will only work if the slave processes don't need to >> communicate with each other) >> >> Jody >> >> >> On Mon, Aug 3, 2009 at 1:24 PM, vipin kumar<vipinkumar41(a)gmail.com> wrote: >> Hi all, >> >> Thanks Durga for your reply. >> >> Jeff, once you wrote code for Mandelbrot set to demonstrate fault >> tolerance in LAM-MPI. i. e. killing any slave process doesn't affect >> others. Exact behaviour I am looking for in Open MPI. I attempted, >> but no luck. Can you please tell how to write such programs in Open MPI. >> >> Thanks in advance. >> >> Regards, >> On Thu, Jul 9, 2009 at 8:30 PM, Durga Choudhury <dpchoudh(a)gmail.com> wrote: >> >> Although I have perhaps the least experience on the topic in this >> list, I will take a shot; more experienced people, please correct me: >> >> MPI standards specify communication mechanism, not fault tolerance at >> any level. You may achieve network tolerance at the IP level by >> implementing 'equal cost multipath' routes (which means two equally >> capable NIC cards connecting to the same destination and modifying >> the kernel routing table to use both cards; the kernel will >> dynamically load balance.). At the MAC level, you can achieve the >> same effect by trunking multiple network cards. >> >> You can achieve process level fault tolerance by a checkpointing >> scheme such as BLCR, which has been tested to work with OpenMPI (and >> other processes as well) >> >> Durga >> >> On Thu, Jul 9, 2009 at 4:57 AM, vipin kumar<vipinkumar41(a)gmail.com> wrote: >> >> Hi all, >> >> I want to know whether open mpi supports Network and process fault >> tolerance or not? If there is any example demonstrating these >> features that will be best. >> >> Regards, >> -- >> Vipin K.

1 0

TRAC experimentation
by Pavan Balaji 24 Dec '09

24 Dec '09

All, I'm trying some experimentation with the mpich2 trac to see if we can allow users to just enter their email address on the new tickets page and not worry about logging in with OpenID, etc. I'm not sure if it'll be successful, but we'll see. This might result in some random tickets + possibly some spam getting through, but hopefully only for 1-2 days. -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji

3 8

Re: [mpich2-core] maint/updatefiles fails
by chan＠mcs.anl.gov 17 Dec '09

17 Dec '09

To use module, one needs to initialize the module environment by sourcing one of the module init files in ~chan/cmod/Linux/etc/cmod/ that is appropriate for your shell. e.g. in .cshrc, do: if ( -f /homes/chan/cmod/Linux/etc/cmod/csh.init ) then source /homes/chan/cmod/Linux/etc/cmod/csh.init endif If you use bash, source bash.init instead (or you can source path.sh) The init file will define the command module in your environment, then logout and re-login to the machine. The following 4 module commands are probably all you ever need. To see what available modules, do "module avail" To load the default autoconf module, do "module add autoconf" To load autoconf-2.62, do "module add autoconf/2.62" To find out what modules you've loaded, do "module list". To remove module autoconf/2.62, do "module rm autoconf/2.62" To do mpich2's maint/updatefiles, do "module add autoconf automake libtool" The default of these autotools is to build mpich2. The documentation of the tool is ~chan/cmod/Linux/info/cmod.info. Let me know if these is any problem. A.Chan. PS: This "module", cmod, that I used is written in C, i.e. not the original "module" that is written in Tcl/Tk... AFAIK, cmod is a simplified version of the original "module", so some module command is missing in cmod. ----- "William Gropp" <wgropp(a)illinois.edu> wrote: > Anthony, > > Having that module would be nice. I'll be happy to test it for you > :) > > Bill > > On Dec 14, 2009, at 2:23 PM, Anthony Chan wrote: > > > > > Bill, > > > > I have autoconf, automake and libtool in my path, i.e. > > PATH=/homes/chan/libtool/2.2.6/bin:/homes/chan/automake/1.11/bin:/ > > homes/chan/autoconf/2.63/bin:... > > (I set this using command "module add autoconf automake libtool", so > > > I don't > > need to set any of MPICH2_XXX environment variables. If you are > > interested > > in using module, I will set this up so others can use it) > > Then mpich2's maint/updatefiles is able to finish successfully > > (just tested this on grind.mcs.anl.gov) > > > > A.Chan > > > > ----- "William Gropp" <wgropp(a)illinois.edu> wrote: > > > >> I still can't build MPICH2 at Argonne. > >> > >> If I use the stock environment on login3 or elephant, I get a > failure > >> > >> with autoconf being 2.61. > >> If I use the MPICH2_xxx_DIR environment variables that used to > >> work, I > >> > >> get > >> > >> ... > >> Found ./configure.in; executing make configure target > >> (cd . && /home/chan/autoconf/2.63/bin/autoheader -I ./confdb && \ > >> \ > >> /home/chan/autoconf/2.63/bin/autoconf -I ./confdb ) > >> running third-party initialization > >> Remember to add `AC_PROG_LIBTOOL' to `configure.ac'. > >> You should add the contents of `/usr/share/aclocal/libtool.m4' to > >> `aclocal.m4'. > >> Putting files in AC_CONFIG_AUX_DIR, `config'. > >> libtoolize: `config.guess' exists: use `--force' to overwrite > >> libtoolize: `config.sub' exists: use `--force' to overwrite > >> libtoolize: `ltmain.sh' exists: use `--force' to overwrite > >> autoreconf2.50: Entering directory `.' > >> autoreconf2.50: configure.ac: not using Gettext > >> autoreconf2.50: running: /home/chan/automake/1.11/bin/aclocal > --force > >> - > >> I config -I m4 > >> autoreconf2.50: configure.ac: tracing > >> autoreconf2.50: configure.ac: not using Libtool > >> autoreconf2.50: running: /home/chan/autoconf/2.63/bin/autoconf > >> --force > >> configure.ac:417: error: possibly undefined macro: > >> AC_LIBTOOL_WIN32_DLL > >> If this token and others are legitimate, please use > >> m4_pattern_allow. > >> See the Autoconf documentation. > >> autoreconf2.50: /home/chan/autoconf/2.63/bin/autoconf failed with > > >> exit > >> > >> status: 1 > >> login3:/homes/gropp/projects/software/mpich2(5) > >> > >> If I set MPICH_LIBTOOL_DIR to Anthony's libtool, I get the same. > >> > >> There is no MPICH2_LIBTOOLIZE in maint/updatefiles > >> > >> login3:/homes/gropp/projects/software/mpich2(5) grep LIBTOOLIZE > >> maint/ > >> > >> updatefiles > >> LIBTOOLIZE=$libtoolize > >> export LIBTOOLIZE > >> libtoolize=${LIBTOOLIZE:-libtoolize} > >> login3:/homes/gropp/projects/software/mpich2(6) > >> > >> How do you build MPICH2 at Argonne? (and could you put this info > into > >> > >> the instructions on the wiki) > >> > >> Bill > >> > >> > >> > >> On Dec 9, 2009, at 4:15 PM, Pavan Balaji wrote: > >> > >>> Bill, > >>> > >>> We modified all the scripts to accept MPICH2_LIBTOOLIZE. On my > >>> machine, > >>> for example, I don't have LIBTOOL set, and it works correctly for > >> me. > >>> But then, I have all the autotools installed in the same > location; > >> > >>> that > >>> might be a preferred way for autotools anyway, so setting LIBTOOL > >>> might > >>> not be required. > >>> > >>> -- Pavan > >>> > >>> On 12/09/2009 09:48 AM, William Gropp wrote: > >>>> I discovered that the environment variable LIBTOOL needs to be > >>>> properly > >>>> set - though there's no documentation about this, I believe that > >>>> libtoolize may require this. I've made some updates to > updatefiles > >> > >>>> to > >>>> include this and to check on the version of libtool. I'll check > it > >> > >>>> in > >>>> after a few more tests. > >>>> > >>>> Bill > >>>> > >>>> On Nov 18, 2009, at 4:53 PM, Anthony Chan wrote: > >>>> > >>>>> > >>>>> > >>>>>> Oh, Pavan also had me add "libtoolize" to autogen.sh in hwloc: > >>>>>> > >>>>>> % cat src/pm/hydra/tools/bind/hwloc/hwloc/autogen.sh > >>>>>> : > >>>>>> libtoolize > >>>>>> autoreconf -ivf > >>>>> > >>>>> The libtoolize line is not in svn. After adding libtoolize in > >>>>> hwloc/autogen.sh, > >>>>> maint/updatefiles was able to run to completion.... Also, it > >> does > >>>>> require > >>>>> libtool 2.2.6. The build tool failed with libtool 2.2.4.... > >>>>> > >>>>> A.Chan > >>>>> > >>>> > >>>> 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 > >>>> > >>>> > >>>> > >>>> > >>> > >>> -- > >>> Pavan Balaji > >>> http://www.mcs.anl.gov/~balaji > >> > >> 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 > > 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

4 5

[Fwd: [Mpi-forum] MPI survey]
by Darius Buntinas 15 Dec '09

15 Dec '09

Have we spammed anl users with this? -d -------- Original Message -------- Subject: [Mpi-forum] MPI survey Date: Tue, 15 Dec 2009 09:26:20 -0500 From: Jeff Squyres <jsquyres(a)cisco.com> Reply-To: Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org> To: MPI Forum list <mpi-forum(a)lists.mpi-forum.org> The survey is actually doing quite well -- Rolf has been sending "reminder" notices around this morning asking people to take the survey (I've gotten 20 responses so far this morning from all over Europe). *** To date, we have 258 survey responses *** Keep 'em coming! The survey will close at midnight on Dec. 31, 2009. I haven't done any analysis of the responses yet. -- Jeff Squyres jsquyres(a)cisco.com _______________________________________________ mpi-forum mailing list mpi-forum(a)lists.mpi-forum.org http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum

3 5

Re: [mpich2-core] maint/updatefiles fails
by chan＠mcs.anl.gov 14 Dec '09

14 Dec '09

What version does your libtool claim to be. I had 2.2.4 and 2.2.6a on MCS machine and they works with updatefiles's autoreconf. I will try downgrading libtool to see if that works. A.Chan login3.surveyor:~/mpich_work > libtool --version ltmain.sh (GNU libtool) 2.2.6b login3.surveyor:~/mpich_work > libtoolize --version libtoolize (GNU libtool) 2.2.6b ----- "Darius Buntinas" <buntinas(a)mcs.anl.gov> wrote: > Sorry, I missed that. > > Try libtool --version > > For some reason even though I had 2.2.6b installed, it still claimed > to > be the old one. > > -d > > On 11/18/2009 01:26 PM, Anthony Chan wrote: > > I already have libtool 2.2.6b (see below) > > blueprint is AIX though. Surveyor is PPC linux. > > > > ----- "Darius Buntinas" <buntinas(a)mcs.anl.gov> wrote: > > > >> I got this same problem on blueprint. You'll need to install > libtool > >> 2.2.6b. > >> > >> -d > >> > >> On 11/18/2009 01:11 PM, Anthony Chan wrote: > >>> mpich2's maint/updatefiles failed on surveyor. > >>> Any clue what is missing.. ? > >>> > >>> A.Chan > >>> > >>> Found ./configure.in; executing make configure target > >>> (cd . && autoheader -I ./confdb && \ > >>> autoconf -I ./confdb ) > >>> running third-party initialization > >>> autoreconf: Entering directory `.' > >>> autoreconf: configure.ac: not using Gettext > >>> autoreconf: running: aclocal --force -I config -I m4 > >>> autoreconf: configure.ac: tracing > >>> autoreconf: configure.ac: not using Libtool > >>> autoreconf: running: /home/chan/autoconf/2.63/bin/autoconf > --force > >>> configure.ac:417: error: possibly undefined macro: > >> AC_LIBTOOL_WIN32_DLL > >>> If this token and others are legitimate, please use > >> m4_pattern_allow. > >>> See the Autoconf documentation. > >>> autoreconf: /home/chan/autoconf/2.63/bin/autoconf failed with > exit > >> status: 1 > >>> login3.surveyor:~/mpich_work/mpich2 > which autoconf > >>> /home/chan/autoconf/2.63/bin/autoconf > >>> login3.surveyor:~/mpich_work/mpich2 > which automake > >>> /home/chan/automake/1.11/bin/automake > >>> login3.surveyor:~/mpich_work/mpich2 > which libtool > >>> /home/chan/libtool/2.2.6b/bin/libtool

5 8

FW: Converged API documentation
by Rajeev Thakur 09 Dec '09

09 Dec '09

This is the documentation on XMI. We should give them some feedback on this. The information is confidential. Rajeev -----Original Message----- From: Robert Wisniewski [mailto:[email protected]] Sent: Thursday, November 19, 2009 10:34 AM To: seager(a)llnl.gov; bronis(a)llnl.gov; beckman(a)mcs.anl.gov; smc(a)mcs.anl.gov; rbair(a)anlmail.anl.gov; lusk(a)mcs.anl.gov; thakur(a)mcs.anl.gov; KAYelick(a)lbl.gov; JShalf(a)lbl.gov Cc: Sameer Kumar; Brian Smith; Pat McCarthy Subject: Converged API documentation Please find attached a presentation and documentation on the current state of the converged API. This is IBM confidential and may still be changing. If you have feedback please let us know. Also please distribute as appropriate to others at your sites making sure to let them know it is still confidential. Thank you. -bob Robert Wisniewski Manager Blue Gene Software IBM T. J. Watson Research Center 914-945-3181 http://www.research.ibm.com/people/b/bob/ bobww(a)us.ibm.com

3 2

FW: MPI Forum meetings
by Rajeev Thakur 07 Dec '09

07 Dec '09

Would we be able to do this in our new conference facilities? Would we need to get badges for everyone? Rajeev -----Original Message----- From: Graham, Richard L. [mailto:[email protected]] Sent: Monday, December 07, 2009 8:55 AM To: Rajeev Thakur Subject: MPI Forum meetings Rajeev, We have two forum meetings scheduled in Chicago this year - May 3-5 - Oct 11-13 There has been some discussion of holding these meetings at your new facilities ? Would you be willing to do this ? I am trying to get DOE paperwork in place for all the meetings this year ASAP (the approval process has gotten a bit more complicated), so am trying to nail down locations now. Please let me know one way or another. Thanks, Rich

1 0