June 2013 - discuss - mailman3.cels.anl.gov

Re: [mpich-discuss] Problems running MPICH jobs under SLURM
by Pavan Balaji 07 Jun '13

07 Jun '13

FYI, I believe this is now fixed. Please try out the latest nightly snapshot and let us know if you are still running into this issue: http://www.mpich.org/static/tarballs/nightly/master/hydra/ http://www.mpich.org/static/tarballs/nightly/master/mpich/ -- Pavan On 06/06/2013 01:39 AM, Biddiscombe, John A. wrote: > Just FYI. I am also getting the double free error when I run under slurm (mpich 3.0.4). Please don't take correspondence off list as I'm following the thread. > > I can't add anything more useful than Markus has already provided with his stack trace and logs. > > [I did find that if I configure --with-slurm and use srun instead of mpiexec , then all works, as expected, but I need mpiexec to pass env vars to processes using mpmd syntax] > > JB > > -----Original Message----- > From: discuss-bounces(a)mpich.org [mailto:[email protected]] On Behalf Of Markus Geimer > Sent: 03 June 2013 16:36 > To: Pavan Balaji > Cc: discuss(a)mpich.org > Subject: Re: [mpich-discuss] Problems running MPICH jobs under SLURM > > Pavan, > >> 1. Can you run your application processes using "ddd" or some other >> debugger to see where the double free is coming from? You might have >> to build mpich with --enable-g=dbg to get the debug symbols in. > > Here is the full stack backtrace: > > ----- 8< ----- 8< ----- 8< ----- 8< ----- 8< ----- 8< ----- > > #0 0x00007ffff6deb475 in *__GI_raise (sig=<optimized out>) > at ../nptl/sysdeps/unix/sysv/linux/raise.c:64 > #1 0x00007ffff6dee6f0 in *__GI_abort () at abort.c:92 > #2 0x00007ffff6e2652b in __libc_message (do_abort=<optimized out>, > fmt=<optimized out>) at ../sysdeps/unix/sysv/linux/libc_fatal.c:189 > #3 0x00007ffff6e2fd76 in malloc_printerr (action=3, > str=0x7ffff6f081e0 "double free or corruption (fasttop)", > ptr=<optimized out>) at malloc.c:6283 > #4 0x00007ffff6e34aac in *__GI___libc_free (mem=<optimized out>) > at malloc.c:3738 > #5 0x00007ffff7a1d5d9 in populate_ids_from_mapping ( > did_map=<synthetic pointer>, num_nodes=<synthetic pointer>, > mapping=<optimized out>, pg=<optimized out>) > at src/mpid/ch3/src/mpid_vc.c:1063 > #6 MPIDI_Populate_vc_node_ids (pg=pg@entry=0x604910, > our_pg_rank=our_pg_rank@entry=0) at src/mpid/ch3/src/mpid_vc.c:1193 > #7 0x00007ffff7a17dd6 in MPID_Init (argc=argc@entry=0x7fffffffd97c, > argv=argv@entry=0x7fffffffd970, requested=requested@entry=0, > provided=provided@entry=0x7fffffffd8e8, > has_args=has_args@entry=0x7fffffffd8e0, > has_env=has_env@entry=0x7fffffffd8e4) at > src/mpid/ch3/src/mpid_init.c:156 > #8 0x00007ffff7acdf7f in MPIR_Init_thread (argc=argc@entry=0x7fffffffd97c, > argv=argv@entry=0x7fffffffd970, required=required@entry=0, > provided=provided@entry=0x7fffffffd944) at src/mpi/init/initthread.c:431 > #9 0x00007ffff7acd90e in PMPI_Init (argc=0x7fffffffd97c, > argv=0x7fffffffd970) > at src/mpi/init/init.c:136 > #10 0x000000000040086d in main () > > ----- >8 ----- >8 ----- >8 ----- >8 ----- >8 ----- >8 ----- > >> 2. Can you send me the output with the ssh launcher as well? > > See mail sent off-list. > > Thanks, > Markus > > -- > Dr. Markus Geimer > Juelich Supercomputing Centre > Institute for Advanced Simulation > Forschungszentrum Juelich GmbH > 52425 Juelich, Germany > > Phone: +49-2461-61-1773 > Fax: +49-2461-61-6656 > E-mail: m.geimer(a)fz-juelich.de > WWW: http://www.fz-juelich.de/jsc/ > > > > ------------------------------------------------------------------------------------------------ > ------------------------------------------------------------------------------------------------ > Forschungszentrum Juelich GmbH > 52425 Juelich > Sitz der Gesellschaft: Juelich > Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher > Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender), Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt, Prof. Dr. Sebastian M. Schmidt > ------------------------------------------------------------------------------------------------ > ------------------------------------------------------------------------------------------------ > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [mpich-discuss] MPICH problem
by Wesley Bland 06 Jun '13

06 Jun '13

(Re-adding mpich-discuss) Sorry I can't help you more. I'm afraid I'll have to defer to Jayesh on this one he's the only one who still knows anything about Windows support. On Jun 6, 2013, at 11:17 AM, "Miller, Harry CSNR:EX" <Harry.Miller(a)gov.bc.ca> wrote: > I am at a loss as when I even ‘google’ the problem everyone writing in seems to have a similar problem but no final (fix) resolution is ever offered? > > http://trac.mpich.org/projects/mpich/ticket/1151 > http://trac.mpich.org/projects/mpich/ticket/1577 > http://trac.mpich.org/projects/mpich/ticket/1692 > http://lists.mcs.anl.gov/pipermail/mpich-discuss/2011-March/009303.html > http://lists.mcs.anl.gov/pipermail/mpich-discuss/2010-May/007293.html > http://lists.mcs.anl.gov/pipermail/mpich-discuss/2009-July/005324.html > http://lists.mcs.anl.gov/pipermail/mpich-discuss/2008-November/000032.html > http://lists.mcs.anl.gov/pipermail/mpich-discuss/2005-September/000768.html > https://groups.google.com/forum/?fromgroups#!topic/fds-smv/3nSxMLMWRvY > http://auriza.site40.net/notes/mpi/mpich2-on-windows-xp/ > > I can go on ... > > mpiexec -register ONLY works with the admin user credentials (mine). I installed MPICH as ‘admin’ as it indicated one should but it now it doesn’t work if register other regular Citrix user credentials using either wmpiexec or mpiexec. Is wpmiregister.exe the same as mpiexec.exe? mpiexec -register.? Maybe I have to remove the credentials registered with the wmpiexec.exe app. and make sure to register with mpiexec.exe using the mpiexec –register command? > > Confusing. > > Harry > > > H.J. Miller > Spatial Technology Analyst | Infrastructure Services Section > Information Management Branch (IMB) > Phone: 250-356-5217 | FAX: 250-953-3493 | E-Mail: Harry.Miller(a)gov.bc.ca > Corporate Services For The Natural Resources Sector (CSNR) > > > > > > From: Wesley Bland [mailto:[email protected]] > Sent: Wednesday, June 5, 2013 9:05 PM > To: Miller, Harry CSNR:EX > Cc: discuss(a)mpich.org > Subject: Re: [mpich-discuss] MPICH problem > > It sounds like your problem is outside of MPICH's control but just to be sure, you can run a simple MPI application and get expected output as long as you're using a regular Windows user account correct? (You'll have to forgive me if I don't know all the Windowsisms. There aren't many Windows people left on this project as we stopped officially supporting Windows on version 1.4.1p) > > Wesley > > On Jun 5, 2013, at 2:55 PM, "Miller, Harry CSNR:EX" <Harry.Miller(a)gov.bc.ca> wrote: > > Hello, > > We are having problems after having installed the MPICH2 (32-bit) along withTauDEM 5.1 32-bit Install Package on one of our Windows Server 2008 64-bit Citrix 6.5 test severs. When registering users using the wpmiregister.exe it appears that only the admin. credentials will work and NOT Citrix regular user accounts. Is there some way to allow Citrix regular users to be able to run commands that call the TauDEM applications from the server? > > For example, opening up a command prompt on the test server and typing in ... > > mpiexec -n 8 pitremove -z "W:\FOR\RSI\RSI\Projects\wburt\projects\2013\006_HSCM_Boundary\data\Raster\dem25.tif" -fel "T:\test.tif" > > requires users credentials to launch the processes (see screenshot below). Regular Citrix credentials don’t seem to be accepted whilst my admin. ones are? Also, do we add the credentials logged in to the server as ‘admin’ or do individual regular users register their credentials from the Citrix server session? Our users won’t have admin. accounts but they will all log into the same test server with the MPICH2 and TuaDEM 5.1 install ... and yet the processes ask for credentials even though the users are already logged in with their regular IDIR Citrix login creds.??? > > I registered my regular Citrix user creds. and also my admin. ones but ONLy the admin one is accepted. When I installed MPICH2 I also checked off that it be available to everyone. But we still get the Aborting: Unable to connect message??? > > <image001.jpg> > > Thank you in advance for a reply. > > > H.J. Miller > Spatial Technology Analyst | Infrastructure Services Section > Information Management Branch (IMB) > Phone: 250-356-5217 | FAX: 250-953-3493 | E-Mail: Harry.Miller(a)gov.bc.ca > Corporate Services For The Natural Resources Sector (CSNR) > > > > > From: David Tarboton [mailto:[email protected]] > Sent: Tuesday, June 4, 2013 7:23 PM > To: Miller, Harry CSNR:EX > Subject: Re: FW: TauDEM problem > > Harry, > > I do not understand MPICH2 credentials very well so I do not know what to say. I have successfully run on Windows 2008 as a Windows user but not with Citrix. You might want to make sure MPICH2 was installed for all users. Also see http://nick-goodman.blogspot.com/2012/02/using-mpich-from-ms-visual-studio-… some suggestions that may help. Also you might try mpiexec -register with the admin user. If it is running as admin maybe this will work for other users without having to give them admin credentials. Also is wpmiregister.exe the same as mpiexec -register. Sometimes there are different versions of MPI on the same machine from different software and if the wrong one runs it does not work. It is frustrating to me to have to have TauDEM depend on this software that is so hard to install and get configured right. > > Good luck. > > Dave > > On 6/4/2013 5:06 PM, Miller, Harry CSNR:EX wrote: > Hello, > > I have installed the MPICH2 (32-bit) and TauDEM 5.1 32-bit Install Package on one of our Windows Server 2008 64-bit Citrix 6.5 test severs. However, when registering users using the wpmiregister.exe it appears that only the admin. credentials will work and NOT Citrix regular user accounts. Is there some way to allow Citrix regular users to be able to run commands that call the TauDEM applications from the server? > > For example, opening up a command prompt on the test server and typing in ... > > mpiexec -n 8 pitremove -z "W:\FOR\RSI\RSI\Projects\wburt\projects\2013\006_HSCM_Boundary\data\Raster\dem25.tif" -fel "T:\test.tif" > > requires users credentials to launch the processes (see screenshot below). Regular Citrix credentials don’t seem to be accepted whilst my admin. ones are? Our users won’t have admin. accounts but they will all log into the same test server with the MPICH2 and TuaDEM 5.1 install ... and yet the processes ask for credentials even though the users are already logged in with their regular IDIR Citrix login creds.??? > > I registered my regular Citrix user creds. and also my admin. ones. > > Is something still missing here?? > > Thanks for a reply. > > Harry > > > H.J. Miller > Spatial Technology Analyst | Infrastructure Services Section > Information Management Branch (IMB) > Phone: 250-356-5217 | FAX: 250-953-3493 | E-Mail: Harry.Miller(a)gov.bc.ca > Corporate Services For The Natural Resources Sector (CSNR) > Government of British Columbia > > From: Miller, Harry CSNR:EX > Sent: Tuesday, June 4, 2013 3:20 PM > To: Burt, William FLNR:EX > Subject: RE: TauDEM fix > > Will, > > Looks like the creds have to be admin ones. Do you have an admin. account? My admin. creds work but not my regular IDIR creds. > > <ATT00001.png> > > Harry > > > H.J. Miller > Spatial Technology Analyst | Infrastructure Services Section > Information Management Branch (IMB) > Phone: 250-356-5217 | FAX: 250-953-3493 | E-Mail: Harry.Miller(a)gov.bc.ca > Corporate Services For The Natural Resources Sector (CSNR) > > > > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] Linking to personal libraries
by Alejandro Allievi 06 Jun '13

06 Jun '13

Thanks!! On Thu, Jun 6, 2013 at 3:44 PM, Wesley Bland <wbland(a)mcs.anl.gov> wrote: > No, this is the correct way to do it. > > On Jun 6, 2013, at 11:12 AM, Alejandro Allievi < > alejandro.allievi(a)gmail.com> wrote: > > Is there a better way?? > > > On Thu, Jun 6, 2013 at 3:40 PM, Wesley Bland <wbland(a)mcs.anl.gov> wrote: > >> That's true and for most smallish clusters, that's how people use it. >> >> On Jun 6, 2013, at 11:09 AM, Alejandro Allievi < >> alejandro.allievi(a)gmail.com> wrote: >> >> Hi Wesley, >> >> Just to clarify myself: provided our NFS mirrors my home directory and >> that LD_LIBRARY_PATH is set correctly, the dynamic compile/link step for >> the user (me) is the same as for static compile/link?? >> >> Thanks again Wesley!! >> >> Alejandro >> >> >> On Thu, Jun 6, 2013 at 11:15 AM, Wesley Bland <wbland(a)mcs.anl.gov> wrote: >> >>> It depends on whether you are linking your library statically or >>> dynamically. If you are linking statically, the linker puts everything in >>> one executable that is sent to all of the nodes. If you do it dynamically, >>> you need to make sure your libraries are available on all of the nodes you >>> will be using (usually by using something like NFS to mirror your home >>> directory across the cluster. You will also need to make sure your >>> environment is set up correctly to allow the libraries to be found on the >>> remote process, usually via an environment variable such as >>> LD_LIBRARY_PATH. >>> >>> Wesley >>> >>> >>> On Jun 6, 2013, at 6:18 AM, Alejandro Allievi < >>> alejandro.allievi(a)gmail.com> wrote: >>> >>> H >>> i >>> >>> When linking to personal libraries using MPICH2, how does each process >>> access them in a distributed environment?? Does the linker actually >>> combines >>> everything >>> into a single >>> executable >>> program >>> and "sends a copy" of entire executable to each process even though not >>> all processes may use the libraries?? Can somebody shed some light on the >>> entire linking process?? >>> >>> Thanks for any help!! >>> >>> Alejandro >>> >>> _______________________________________________ >>> discuss mailing list discuss(a)mpich.org >>> To manage subscription options or unsubscribe: >>> https://lists.mpich.org/mailman/listinfo/discuss >>> >>> >>> _______________________________________________ >>> discuss mailing list discuss(a)mpich.org >>> To manage subscription options or unsubscribe: >>> https://lists.mpich.org/mailman/listinfo/discuss >>> >> >> >> >> -- >> Alejandro Allievi >> http://www.ace-net.ca/wiki/Alejandro_Allievi >> >> "Tenez, mon ami, si vous y pensez bien, vous trouverez qu'en tout, notre >> véritable sentiment n'est pas celui dans lequel nous n'avons jamais >> vacillé; mais celui auquel nous sommes le plus habituellement revenus". >> Denis Diderot. >> >> Este e-mail esta consignado sólo para el destinatario designado en el >> mismo y puede contener información confidencial y privilegiada. Su >> distribución o copiado está prohibido. Si Usted no fuera el destinatario >> designado, por favor notifíquenos de inmediato y destruya este e-mail en >> forma permanente así como todas las copias del mismo. *Ce courriel peut >> renfermer des renseignements confidentiels et privilégiés et s'adresse au >> destinataire désigné seulement. La distribution ou la copie de ce courriel >> est interdit. Si vous n'êtes pas le destinataire désigné, veuillez nous en >> aviser immédiatement et détruire de façon permanente ce courriel ainsi que >> toute copie de celui-ci*. This e-mail may contain confidential >> information, and is intended only for the named recipient and may be >> privileged. Distribution or copying of this email is prohibited. If you are >> not the named recipient, please notify us immediately and permanently >> destroy this email and all copies of it. >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss >> >> >> >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss >> > > > > -- > Alejandro Allievi > http://www.ace-net.ca/wiki/Alejandro_Allievi > > "Tenez, mon ami, si vous y pensez bien, vous trouverez qu'en tout, notre > véritable sentiment n'est pas celui dans lequel nous n'avons jamais > vacillé; mais celui auquel nous sommes le plus habituellement revenus". > Denis Diderot. > > Este e-mail esta consignado sólo para el destinatario designado en el > mismo y puede contener información confidencial y privilegiada. Su > distribución o copiado está prohibido. Si Usted no fuera el destinatario > designado, por favor notifíquenos de inmediato y destruya este e-mail en > forma permanente así como todas las copias del mismo. *Ce courriel peut > renfermer des renseignements confidentiels et privilégiés et s'adresse au > destinataire désigné seulement. La distribution ou la copie de ce courriel > est interdit. Si vous n'êtes pas le destinataire désigné, veuillez nous en > aviser immédiatement et détruire de façon permanente ce courriel ainsi que > toute copie de celui-ci*. This e-mail may contain confidential > information, and is intended only for the named recipient and may be > privileged. Distribution or copying of this email is prohibited. If you are > not the named recipient, please notify us immediately and permanently > destroy this email and all copies of it. > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > > > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > -- Alejandro Allievi http://www.ace-net.ca/wiki/Alejandro_Allievi "Tenez, mon ami, si vous y pensez bien, vous trouverez qu'en tout, notre véritable sentiment n'est pas celui dans lequel nous n'avons jamais vacillé; mais celui auquel nous sommes le plus habituellement revenus". Denis Diderot. Este e-mail esta consignado sólo para el destinatario designado en el mismo y puede contener información confidencial y privilegiada. Su distribución o copiado está prohibido. Si Usted no fuera el destinatario designado, por favor notifíquenos de inmediato y destruya este e-mail en forma permanente así como todas las copias del mismo. *Ce courriel peut renfermer des renseignements confidentiels et privilégiés et s'adresse au destinataire désigné seulement. La distribution ou la copie de ce courriel est interdit. Si vous n'êtes pas le destinataire désigné, veuillez nous en aviser immédiatement et détruire de façon permanente ce courriel ainsi que toute copie de celui-ci*. This e-mail may contain confidential information, and is intended only for the named recipient and may be privileged. Distribution or copying of this email is prohibited. If you are not the named recipient, please notify us immediately and permanently destroy this email and all copies of it.

1 0

Re: [mpich-discuss] Linking to personal libraries
by Wesley Bland 06 Jun '13

06 Jun '13

No, this is the correct way to do it. On Jun 6, 2013, at 11:12 AM, Alejandro Allievi <alejandro.allievi(a)gmail.com> wrote: > Is there a better way?? > > > On Thu, Jun 6, 2013 at 3:40 PM, Wesley Bland <wbland(a)mcs.anl.gov> wrote: > That's true and for most smallish clusters, that's how people use it. > > On Jun 6, 2013, at 11:09 AM, Alejandro Allievi <alejandro.allievi(a)gmail.com> wrote: > >> Hi Wesley, >> >> Just to clarify myself: provided our NFS mirrors my home directory and that LD_LIBRARY_PATH is set correctly, the dynamic compile/link step for the user (me) is the same as for static compile/link?? >> >> Thanks again Wesley!! >> >> Alejandro >> >> >> On Thu, Jun 6, 2013 at 11:15 AM, Wesley Bland <wbland(a)mcs.anl.gov> wrote: >> It depends on whether you are linking your library statically or dynamically. If you are linking statically, the linker puts everything in one executable that is sent to all of the nodes. If you do it dynamically, you need to make sure your libraries are available on all of the nodes you will be using (usually by using something like NFS to mirror your home directory across the cluster. You will also need to make sure your environment is set up correctly to allow the libraries to be found on the remote process, usually via an environment variable such as LD_LIBRARY_PATH. >> >> Wesley >> >> >> On Jun 6, 2013, at 6:18 AM, Alejandro Allievi <alejandro.allievi(a)gmail.com> wrote: >> >>> Hi >>> >>> When linking to personal libraries using MPICH2, how does each process access them in a distributed environment?? Does the linker actually combines everything into a single executable program and "sends a copy" of entire executable to each process even though not all processes may use the libraries?? Can somebody shed some light on the entire linking process?? >>> >>> Thanks for any help!! >>> >>> Alejandro >>> _______________________________________________ >>> discuss mailing list discuss(a)mpich.org >>> To manage subscription options or unsubscribe: >>> https://lists.mpich.org/mailman/listinfo/discuss >> >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss >> >> >> >> -- >> Alejandro Allievi >> http://www.ace-net.ca/wiki/Alejandro_Allievi >> >> "Tenez, mon ami, si vous y pensez bien, vous trouverez qu'en tout, notre véritable sentiment n'est pas celui dans lequel nous n'avons jamais vacillé; mais celui auquel nous sommes le plus habituellement revenus". Denis Diderot. >> >> Este e-mail esta consignado sólo para el destinatario designado en el mismo y puede contener información confidencial y privilegiada. Su distribución o copiado está prohibido. Si Usted no fuera el destinatario designado, por favor notifíquenos de inmediato y destruya este e-mail en forma permanente así como todas las copias del mismo. Ce courriel peut renfermer des renseignements confidentiels et privilégiés et s'adresse au destinataire désigné seulement. La distribution ou la copie de ce courriel est interdit. Si vous n'êtes pas le destinataire désigné, veuillez nous en aviser immédiatement et détruire de façon permanente ce courriel ainsi que toute copie de celui-ci. This e-mail may contain confidential information, and is intended only for the named recipient and may be privileged. Distribution or copying of this email is prohibited. If you are not the named recipient, please notify us immediately and permanently destroy this email and all copies of it. >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss > > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > > > > -- > Alejandro Allievi > http://www.ace-net.ca/wiki/Alejandro_Allievi > > "Tenez, mon ami, si vous y pensez bien, vous trouverez qu'en tout, notre véritable sentiment n'est pas celui dans lequel nous n'avons jamais vacillé; mais celui auquel nous sommes le plus habituellement revenus". Denis Diderot. > > Este e-mail esta consignado sólo para el destinatario designado en el mismo y puede contener información confidencial y privilegiada. Su distribución o copiado está prohibido. Si Usted no fuera el destinatario designado, por favor notifíquenos de inmediato y destruya este e-mail en forma permanente así como todas las copias del mismo. Ce courriel peut renfermer des renseignements confidentiels et privilégiés et s'adresse au destinataire désigné seulement. La distribution ou la copie de ce courriel est interdit. Si vous n'êtes pas le destinataire désigné, veuillez nous en aviser immédiatement et détruire de façon permanente ce courriel ainsi que toute copie de celui-ci. This e-mail may contain confidential information, and is intended only for the named recipient and may be privileged. Distribution or copying of this email is prohibited. If you are not the named recipient, please notify us immediately and permanently destroy this email and all copies of it. > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] Linking to personal libraries
by Alejandro Allievi 06 Jun '13

06 Jun '13

Is there a better way?? On Thu, Jun 6, 2013 at 3:40 PM, Wesley Bland <wbland(a)mcs.anl.gov> wrote: > That's true and for most smallish clusters, that's how people use it. > > On Jun 6, 2013, at 11:09 AM, Alejandro Allievi < > alejandro.allievi(a)gmail.com> wrote: > > Hi Wesley, > > Just to clarify myself: provided our NFS mirrors my home directory and > that LD_LIBRARY_PATH is set correctly, the dynamic compile/link step for > the user (me) is the same as for static compile/link?? > > Thanks again Wesley!! > > Alejandro > > > On Thu, Jun 6, 2013 at 11:15 AM, Wesley Bland <wbland(a)mcs.anl.gov> wrote: > >> It depends on whether you are linking your library statically or >> dynamically. If you are linking statically, the linker puts everything in >> one executable that is sent to all of the nodes. If you do it dynamically, >> you need to make sure your libraries are available on all of the nodes you >> will be using (usually by using something like NFS to mirror your home >> directory across the cluster. You will also need to make sure your >> environment is set up correctly to allow the libraries to be found on the >> remote process, usually via an environment variable such as >> LD_LIBRARY_PATH. >> >> Wesley >> >> >> On Jun 6, 2013, at 6:18 AM, Alejandro Allievi < >> alejandro.allievi(a)gmail.com> wrote: >> >> H >> i >> >> When linking to personal libraries using MPICH2, how does each process >> access them in a distributed environment?? Does the linker actually >> combines >> everything >> into a single >> executable >> program >> and "sends a copy" of entire executable to each process even though not >> all processes may use the libraries?? Can somebody shed some light on the >> entire linking process?? >> >> Thanks for any help!! >> >> Alejandro >> >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss >> >> >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss >> > > > > -- > Alejandro Allievi > http://www.ace-net.ca/wiki/Alejandro_Allievi > > "Tenez, mon ami, si vous y pensez bien, vous trouverez qu'en tout, notre > véritable sentiment n'est pas celui dans lequel nous n'avons jamais > vacillé; mais celui auquel nous sommes le plus habituellement revenus". > Denis Diderot. > > Este e-mail esta consignado sólo para el destinatario designado en el > mismo y puede contener información confidencial y privilegiada. Su > distribución o copiado está prohibido. Si Usted no fuera el destinatario > designado, por favor notifíquenos de inmediato y destruya este e-mail en > forma permanente así como todas las copias del mismo. *Ce courriel peut > renfermer des renseignements confidentiels et privilégiés et s'adresse au > destinataire désigné seulement. La distribution ou la copie de ce courriel > est interdit. Si vous n'êtes pas le destinataire désigné, veuillez nous en > aviser immédiatement et détruire de façon permanente ce courriel ainsi que > toute copie de celui-ci*. This e-mail may contain confidential > information, and is intended only for the named recipient and may be > privileged. Distribution or copying of this email is prohibited. If you are > not the named recipient, please notify us immediately and permanently > destroy this email and all copies of it. > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > > > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > -- Alejandro Allievi http://www.ace-net.ca/wiki/Alejandro_Allievi "Tenez, mon ami, si vous y pensez bien, vous trouverez qu'en tout, notre véritable sentiment n'est pas celui dans lequel nous n'avons jamais vacillé; mais celui auquel nous sommes le plus habituellement revenus". Denis Diderot. Este e-mail esta consignado sólo para el destinatario designado en el mismo y puede contener información confidencial y privilegiada. Su distribución o copiado está prohibido. Si Usted no fuera el destinatario designado, por favor notifíquenos de inmediato y destruya este e-mail en forma permanente así como todas las copias del mismo. *Ce courriel peut renfermer des renseignements confidentiels et privilégiés et s'adresse au destinataire désigné seulement. La distribution ou la copie de ce courriel est interdit. Si vous n'êtes pas le destinataire désigné, veuillez nous en aviser immédiatement et détruire de façon permanente ce courriel ainsi que toute copie de celui-ci*. This e-mail may contain confidential information, and is intended only for the named recipient and may be privileged. Distribution or copying of this email is prohibited. If you are not the named recipient, please notify us immediately and permanently destroy this email and all copies of it.

1 0

Re: [mpich-discuss] Linking to personal libraries
by Wesley Bland 06 Jun '13

06 Jun '13

That's true and for most smallish clusters, that's how people use it. On Jun 6, 2013, at 11:09 AM, Alejandro Allievi <alejandro.allievi(a)gmail.com> wrote: > Hi Wesley, > > Just to clarify myself: provided our NFS mirrors my home directory and that LD_LIBRARY_PATH is set correctly, the dynamic compile/link step for the user (me) is the same as for static compile/link?? > > Thanks again Wesley!! > > Alejandro > > > On Thu, Jun 6, 2013 at 11:15 AM, Wesley Bland <wbland(a)mcs.anl.gov> wrote: > It depends on whether you are linking your library statically or dynamically. If you are linking statically, the linker puts everything in one executable that is sent to all of the nodes. If you do it dynamically, you need to make sure your libraries are available on all of the nodes you will be using (usually by using something like NFS to mirror your home directory across the cluster. You will also need to make sure your environment is set up correctly to allow the libraries to be found on the remote process, usually via an environment variable such as LD_LIBRARY_PATH. > > Wesley > > > On Jun 6, 2013, at 6:18 AM, Alejandro Allievi <alejandro.allievi(a)gmail.com> wrote: > >> Hi >> >> When linking to personal libraries using MPICH2, how does each process access them in a distributed environment?? Does the linker actually combines everything into a single executable program and "sends a copy" of entire executable to each process even though not all processes may use the libraries?? Can somebody shed some light on the entire linking process?? >> >> Thanks for any help!! >> >> Alejandro >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > > > > -- > Alejandro Allievi > http://www.ace-net.ca/wiki/Alejandro_Allievi > > "Tenez, mon ami, si vous y pensez bien, vous trouverez qu'en tout, notre véritable sentiment n'est pas celui dans lequel nous n'avons jamais vacillé; mais celui auquel nous sommes le plus habituellement revenus". Denis Diderot. > > Este e-mail esta consignado sólo para el destinatario designado en el mismo y puede contener información confidencial y privilegiada. Su distribución o copiado está prohibido. Si Usted no fuera el destinatario designado, por favor notifíquenos de inmediato y destruya este e-mail en forma permanente así como todas las copias del mismo. Ce courriel peut renfermer des renseignements confidentiels et privilégiés et s'adresse au destinataire désigné seulement. La distribution ou la copie de ce courriel est interdit. Si vous n'êtes pas le destinataire désigné, veuillez nous en aviser immédiatement et détruire de façon permanente ce courriel ainsi que toute copie de celui-ci. This e-mail may contain confidential information, and is intended only for the named recipient and may be privileged. Distribution or copying of this email is prohibited. If you are not the named recipient, please notify us immediately and permanently destroy this email and all copies of it. > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] Linking to personal libraries
by Alejandro Allievi 06 Jun '13

06 Jun '13

Hi Wesley, Just to clarify myself: provided our NFS mirrors my home directory and that LD_LIBRARY_PATH is set correctly, the dynamic compile/link step for the user (me) is the same as for static compile/link?? Thanks again Wesley!! Alejandro On Thu, Jun 6, 2013 at 11:15 AM, Wesley Bland <wbland(a)mcs.anl.gov> wrote: > It depends on whether you are linking your library statically or > dynamically. If you are linking statically, the linker puts everything in > one executable that is sent to all of the nodes. If you do it dynamically, > you need to make sure your libraries are available on all of the nodes you > will be using (usually by using something like NFS to mirror your home > directory across the cluster. You will also need to make sure your > environment is set up correctly to allow the libraries to be found on the > remote process, usually via an environment variable such as > LD_LIBRARY_PATH. > > Wesley > > > On Jun 6, 2013, at 6:18 AM, Alejandro Allievi <alejandro.allievi(a)gmail.com> > wrote: > > H > i > > When linking to personal libraries using MPICH2, how does each process > access them in a distributed environment?? Does the linker actually > combines > everything > into a single > executable > program > and "sends a copy" of entire executable to each process even though not > all processes may use the libraries?? Can somebody shed some light on the > entire linking process?? > > Thanks for any help!! > > Alejandro > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > -- Alejandro Allievi http://www.ace-net.ca/wiki/Alejandro_Allievi "Tenez, mon ami, si vous y pensez bien, vous trouverez qu'en tout, notre véritable sentiment n'est pas celui dans lequel nous n'avons jamais vacillé; mais celui auquel nous sommes le plus habituellement revenus". Denis Diderot. Este e-mail esta consignado sólo para el destinatario designado en el mismo y puede contener información confidencial y privilegiada. Su distribución o copiado está prohibido. Si Usted no fuera el destinatario designado, por favor notifíquenos de inmediato y destruya este e-mail en forma permanente así como todas las copias del mismo. *Ce courriel peut renfermer des renseignements confidentiels et privilégiés et s'adresse au destinataire désigné seulement. La distribution ou la copie de ce courriel est interdit. Si vous n'êtes pas le destinataire désigné, veuillez nous en aviser immédiatement et détruire de façon permanente ce courriel ainsi que toute copie de celui-ci*. This e-mail may contain confidential information, and is intended only for the named recipient and may be privileged. Distribution or copying of this email is prohibited. If you are not the named recipient, please notify us immediately and permanently destroy this email and all copies of it.

1 0

Re: [mpich-discuss] Linking to personal libraries
by Wesley Bland 06 Jun '13

06 Jun '13

It depends on whether you are linking your library statically or dynamically. If you are linking statically, the linker puts everything in one executable that is sent to all of the nodes. If you do it dynamically, you need to make sure your libraries are available on all of the nodes you will be using (usually by using something like NFS to mirror your home directory across the cluster. You will also need to make sure your environment is set up correctly to allow the libraries to be found on the remote process, usually via an environment variable such as LD_LIBRARY_PATH. Wesley On Jun 6, 2013, at 6:18 AM, Alejandro Allievi <alejandro.allievi(a)gmail.com> wrote: > Hi > > When linking to personal libraries using MPICH2, how does each process access them in a distributed environment?? Does the linker actually combines everything into a single executable program and "sends a copy" of entire executable to each process even though not all processes may use the libraries?? Can somebody shed some light on the entire linking process?? > > Thanks for any help!! > > Alejandro > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0

Linking to personal libraries
by Alejandro Allievi 06 Jun '13

06 Jun '13

H i When linking to personal libraries using MPICH2, how does each process access them in a distributed environment?? Does the linker actually combines everything into a single executable program and "sends a copy" of entire executable to each process even though not all processes may use the libraries?? Can somebody shed some light on the entire linking process?? Thanks for any help!! Alejandro

1 0

Re: [mpich-discuss] Problems running MPICH jobs under SLURM
by Antonio J. Peña 06 Jun '13

06 Jun '13

Thanks for your inputs JB. Everything related to this issue will be discussed through this thread. You can also check the corresponding ticket: http://trac.mpich.org/projects/mpich/ticket/1871 Antonio On Thursday, June 06, 2013 06:39:20 AM Biddiscombe, John A. wrote: > Just FYI. I am also getting the double free error when I run under slurm > (mpich 3.0.4). Please don't take correspondence off list as I'm following > the thread. > > I can't add anything more useful than Markus has already provided with his > stack trace and logs. > > [I did find that if I configure --with-slurm and use srun instead of mpiexec > , then all works, as expected, but I need mpiexec to pass env vars to > processes using mpmd syntax] > > JB > > -----Original Message----- > From: discuss-bounces(a)mpich.org [mailto:[email protected]] On Behalf > Of Markus Geimer Sent: 03 June 2013 16:36 > To: Pavan Balaji > Cc: discuss(a)mpich.org > Subject: Re: [mpich-discuss] Problems running MPICH jobs under SLURM > > Pavan, > > > 1. Can you run your application processes using "ddd" or some other > > debugger to see where the double free is coming from? You might have > > to build mpich with --enable-g=dbg to get the debug symbols in. > > Here is the full stack backtrace: > > ----- 8< ----- 8< ----- 8< ----- 8< ----- 8< ----- 8< ----- > > #0 0x00007ffff6deb475 in *__GI_raise (sig=<optimized out>) > at ../nptl/sysdeps/unix/sysv/linux/raise.c:64 > #1 0x00007ffff6dee6f0 in *__GI_abort () at abort.c:92 > #2 0x00007ffff6e2652b in __libc_message (do_abort=<optimized out>, > fmt=<optimized out>) at ../sysdeps/unix/sysv/linux/libc_fatal.c:189 > #3 0x00007ffff6e2fd76 in malloc_printerr (action=3, > str=0x7ffff6f081e0 "double free or corruption (fasttop)", > ptr=<optimized out>) at malloc.c:6283 > #4 0x00007ffff6e34aac in *__GI___libc_free (mem=<optimized out>) > at malloc.c:3738 > #5 0x00007ffff7a1d5d9 in populate_ids_from_mapping ( > did_map=<synthetic pointer>, num_nodes=<synthetic pointer>, > mapping=<optimized out>, pg=<optimized out>) > at src/mpid/ch3/src/mpid_vc.c:1063 > #6 MPIDI_Populate_vc_node_ids (pg=pg@entry=0x604910, > our_pg_rank=our_pg_rank@entry=0) at src/mpid/ch3/src/mpid_vc.c:1193 > #7 0x00007ffff7a17dd6 in MPID_Init (argc=argc@entry=0x7fffffffd97c, > argv=argv@entry=0x7fffffffd970, requested=requested@entry=0, > provided=provided@entry=0x7fffffffd8e8, > has_args=has_args@entry=0x7fffffffd8e0, > has_env=has_env@entry=0x7fffffffd8e4) at > src/mpid/ch3/src/mpid_init.c:156 > #8 0x00007ffff7acdf7f in MPIR_Init_thread (argc=argc@entry=0x7fffffffd97c, > argv=argv@entry=0x7fffffffd970, required=required@entry=0, > provided=provided@entry=0x7fffffffd944) at src/mpi/init/initthread.c:431 > #9 0x00007ffff7acd90e in PMPI_Init (argc=0x7fffffffd97c, > argv=0x7fffffffd970) > at src/mpi/init/init.c:136 > #10 0x000000000040086d in main () > > ----- >8 ----- >8 ----- >8 ----- >8 ----- >8 ----- >8 ----- > > > 2. Can you send me the output with the ssh launcher as well? > > See mail sent off-list. > > Thanks, > Markus > > -- > Dr. Markus Geimer > Juelich Supercomputing Centre > Institute for Advanced Simulation > Forschungszentrum Juelich GmbH > 52425 Juelich, Germany > > Phone: +49-2461-61-1773 > Fax: +49-2461-61-6656 > E-mail: m.geimer(a)fz-juelich.de > WWW: http://www.fz-juelich.de/jsc/ > > > > ---------------------------------------------------------------------------- > -------------------- > --------------------------------------------------------------------------- > --------------------- Forschungszentrum Juelich GmbH > 52425 Juelich > Sitz der Gesellschaft: Juelich > Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 > Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher > Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender), Karsten Beneke > (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt, Prof. Dr. Sebastian M. > Schmidt > --------------------------------------------------------------------------- > --------------------- > --------------------------------------------------------------------------- > --------------------- _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0