March 2014 - discuss - mailman3.cels.anl.gov

Re: [mpich-discuss] Fatal error in PMPI_Barrier: A process has failed, error stack:
by Rajeev Thakur 26 Mar '14

26 Mar '14

Does the cpi example run across two machines? Rajeev On Mar 26, 2014, at 7:13 PM, Tony Ladd <tladd(a)che.ufl.edu> wrote: > Rajeev > > Sorry about that. I was switching back and forth from openmpi to mpich. But it does not make a difference. Here is a clean log from a fresh terminal - no mention of openmpi > > Tony > > PS - its a CentOS 6.5install - should have mentioned it before. > > -- > Tony Ladd > > Chemical Engineering Department > University of Florida > Gainesville, Florida 32611-6005 > USA > > Email: tladd-"(AT)"-che.ufl.edu > Web http://ladd.che.ufl.edu > > Tel: (352)-392-6509 > FAX: (352)-392-9514 > > <mpich.log>_______________________________________________ > 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] Fatal error in PMPI_Barrier: A process has failed, error stack:
by Tony Ladd 26 Mar '14

26 Mar '14

Rajeev Sorry about that. I was switching back and forth from openmpi to mpich. But it does not make a difference. Here is a clean log from a fresh terminal - no mention of openmpi Tony PS - its a CentOS 6.5install - should have mentioned it before. -- Tony Ladd Chemical Engineering Department University of Florida Gainesville, Florida 32611-6005 USA Email: tladd-"(AT)"-che.ufl.edu Web http://ladd.che.ufl.edu Tel: (352)-392-6509 FAX: (352)-392-9514

1 0

Re: [mpich-discuss] Fatal error in PMPI_Barrier: A process has failed, error stack:
by Rajeev Thakur 26 Mar '14

26 Mar '14

There are numerous instances of "openmpi" in directory paths in the log file. Is the code entirely compiled and linked with MPICH? Rajeev On Mar 26, 2014, at 3:53 PM, Tony Ladd <tladd(a)che.ufl.edu> wrote: > I get this error when I try to use mpich across two different nodes. My program works on a single node. I realize this is a common error but I have checked all of the issues mentioned in the FAQ and I did not find any further discussion in the archives. > > I followed the standard installation to /global/usr/bin and /global/usr/lib. The file system /global is automounted on the clients which have passwordless ssh configured. > > Running either my network test code or Intel's IMB code produced a similar error (see attached). Both codes run on a single node under mpich and run across the two nodes using openmpi. Also netpipe (using tcp sockets works fine). > > I tried including the LD_LIBRARY_PAT via the -genv option but that did not help. I must have some configuration issue but I do not see what. I have had previous versions of mpich/mvapich working without any trouble but I am stuck here. I would be grateful for any hints. > > Tony > > -- > Tony Ladd > > Chemical Engineering Department > University of Florida > Gainesville, Florida 32611-6005 > USA > > Email: tladd-"(AT)"-che.ufl.edu > Web http://ladd.che.ufl.edu > > Tel: (352)-392-6509 > FAX: (352)-392-9514 > > <mpich.log>_______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0

Fatal error in PMPI_Barrier: A process has failed, error stack:
by Tony Ladd 26 Mar '14

26 Mar '14

I get this error when I try to use mpich across two different nodes. My program works on a single node. I realize this is a common error but I have checked all of the issues mentioned in the FAQ and I did not find any further discussion in the archives. I followed the standard installation to /global/usr/bin and /global/usr/lib. The file system /global is automounted on the clients which have passwordless ssh configured. Running either my network test code or Intel's IMB code produced a similar error (see attached). Both codes run on a single node under mpich and run across the two nodes using openmpi. Also netpipe (using tcp sockets works fine). I tried including the LD_LIBRARY_PAT via the -genv option but that did not help. I must have some configuration issue but I do not see what. I have had previous versions of mpich/mvapich working without any trouble but I am stuck here. I would be grateful for any hints. Tony -- Tony Ladd Chemical Engineering Department University of Florida Gainesville, Florida 32611-6005 USA Email: tladd-"(AT)"-che.ufl.edu Web http://ladd.che.ufl.edu Tel: (352)-392-6509 FAX: (352)-392-9514

1 0

Re: [mpich-discuss] OSX Mavericks lldb+mpich 3.1 debugging problem
by Martin Schatz 26 Mar '14

26 Mar '14

I think the culprit might be that lldb uses the sh shell by default (so it doesn’t use bash). That would explain the problems. I’ll look into how to change that since that’s an lldb issue. Thanks for the help. Martin On Mar 25, 2014, at 3:56 PM, Lu, Huiwei <huiweilu(a)mcs.anl.gov> wrote: > Are you using tcsh? Does bash works for you? > http://lists.mpich.org/pipermail/discuss/2014-January/002079.html > > — > Huiwei Lu > Postdoc Appointee > Mathematics and Computer Science Division, Argonne National Laboratory > http://www.mcs.anl.gov/~huiweilu/ > > On Mar 25, 2014, at 3:50 PM, Martin Schatz <martin.schatz(a)utexas.edu> wrote: > >> Hi, >> >> I’m having an issue debugging an MPI program with lldb on OSX Mavericks. I get the exact behavior as described in >> >> http://lists.mpich.org/pipermail/discuss/2014-January/002074.html >> >> except using lldb instead of gdb as the debugger. I have checked that I am using the bash terminal and unfortunately am still getting the same error. >> >> Details of my installation are as follows: >> >> I have installed the mpich-clang macport (mpich-clang-fortran is active), but when I run the cpi example with >> >> mpiexec -np 1 xterm -e lldb ./cpi >> >> I get the xterm window to spawn but: >> >> Current executable set to './cpi' (x86_64). >> (lldb) r >> Process 56694 launched: './cpi' (x86_64) >> [cli_0]: write_line error; fd=6 buf=:cmd=init pmi_version=1 pmi_subversion=1 >> : >> system msg for write_line failure : Bad file descriptor >> [cli_0]: Unable to write to PMI_fd >> [cli_0]: write_line error; fd=6 buf=:cmd=get_appnum >> : >> system msg for write_line failure : Bad file descriptor >> Fatal error in MPI_Init: Other MPI error, error stack: >> MPIR_Init_thread(467): >> MPID_Init(140).......: channel initialization failed >> MPID_Init(422).......: PMI_Get_appnum returned -1 >> Process 56694 exited with status = 1 (0x00000001) >> (lldb) >> >> >> The lldb version is: >> lldb --version >> lldb-310.2.36 >> >> Any help is appreciated. >> >> Martin >> >> _______________________________________________ >> 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

MPICH in Debian Jessie
by mark 26 Mar '14

26 Mar '14

Hello. I have to imlpement some clustering algorithms for a bachelor essay and i was wondering if the Debian's MPICH package [1] will be enough for this, or will i need to compile MPICH from the source in mpich.org . Also i am new to MPICH and i would like some starting points and tutorials that will guide me with ease into algorithmic impementations (parallel and linear) and mpi programming and functions. Thank you ! [1] https://packages.debian.org/jessie/mpich .

1 0

Re: [mpich-discuss] OSX Mavericks lldb+mpich 3.1 debugging problem
by Martin Schatz 25 Mar '14

25 Mar '14

Sorry if it was not clear but my shell is bash and is working: wireless:build $echo $SHELL echo $SHELL /bin/bash wireless:build $/bin/bash --version /bin/bash --version GNU bash, version 3.2.51(1)-release (x86_64-apple-darwin13) Copyright (C) 2007 Free Software Foundation, Inc. At least, I believe this means I am running with bash. Thanks for the help. Martin On Mar 25, 2014, at 3:56 PM, Lu, Huiwei <huiweilu(a)mcs.anl.gov> wrote: > Are you using tcsh? Does bash works for you? > http://lists.mpich.org/pipermail/discuss/2014-January/002079.html > > — > Huiwei Lu > Postdoc Appointee > Mathematics and Computer Science Division, Argonne National Laboratory > http://www.mcs.anl.gov/~huiweilu/ > > On Mar 25, 2014, at 3:50 PM, Martin Schatz <martin.schatz(a)utexas.edu> wrote: > >> Hi, >> >> I’m having an issue debugging an MPI program with lldb on OSX Mavericks. I get the exact behavior as described in >> >> http://lists.mpich.org/pipermail/discuss/2014-January/002074.html >> >> except using lldb instead of gdb as the debugger. I have checked that I am using the bash terminal and unfortunately am still getting the same error. >> >> Details of my installation are as follows: >> >> I have installed the mpich-clang macport (mpich-clang-fortran is active), but when I run the cpi example with >> >> mpiexec -np 1 xterm -e lldb ./cpi >> >> I get the xterm window to spawn but: >> >> Current executable set to './cpi' (x86_64). >> (lldb) r >> Process 56694 launched: './cpi' (x86_64) >> [cli_0]: write_line error; fd=6 buf=:cmd=init pmi_version=1 pmi_subversion=1 >> : >> system msg for write_line failure : Bad file descriptor >> [cli_0]: Unable to write to PMI_fd >> [cli_0]: write_line error; fd=6 buf=:cmd=get_appnum >> : >> system msg for write_line failure : Bad file descriptor >> Fatal error in MPI_Init: Other MPI error, error stack: >> MPIR_Init_thread(467): >> MPID_Init(140).......: channel initialization failed >> MPID_Init(422).......: PMI_Get_appnum returned -1 >> Process 56694 exited with status = 1 (0x00000001) >> (lldb) >> >> >> The lldb version is: >> lldb --version >> lldb-310.2.36 >> >> Any help is appreciated. >> >> Martin >> >> _______________________________________________ >> 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

Re: [mpich-discuss] OSX Mavericks lldb+mpich 3.1 debugging problem
by Lu, Huiwei 25 Mar '14

25 Mar '14

Are you using tcsh? Does bash works for you? http://lists.mpich.org/pipermail/discuss/2014-January/002079.html — Huiwei Lu Postdoc Appointee Mathematics and Computer Science Division, Argonne National Laboratory http://www.mcs.anl.gov/~huiweilu/ On Mar 25, 2014, at 3:50 PM, Martin Schatz <martin.schatz(a)utexas.edu> wrote: > Hi, > > I’m having an issue debugging an MPI program with lldb on OSX Mavericks. I get the exact behavior as described in > > http://lists.mpich.org/pipermail/discuss/2014-January/002074.html > > except using lldb instead of gdb as the debugger. I have checked that I am using the bash terminal and unfortunately am still getting the same error. > > Details of my installation are as follows: > > I have installed the mpich-clang macport (mpich-clang-fortran is active), but when I run the cpi example with > > mpiexec -np 1 xterm -e lldb ./cpi > > I get the xterm window to spawn but: > > Current executable set to './cpi' (x86_64). > (lldb) r > Process 56694 launched: './cpi' (x86_64) > [cli_0]: write_line error; fd=6 buf=:cmd=init pmi_version=1 pmi_subversion=1 > : > system msg for write_line failure : Bad file descriptor > [cli_0]: Unable to write to PMI_fd > [cli_0]: write_line error; fd=6 buf=:cmd=get_appnum > : > system msg for write_line failure : Bad file descriptor > Fatal error in MPI_Init: Other MPI error, error stack: > MPIR_Init_thread(467): > MPID_Init(140).......: channel initialization failed > MPID_Init(422).......: PMI_Get_appnum returned -1 > Process 56694 exited with status = 1 (0x00000001) > (lldb) > > > The lldb version is: > lldb --version > lldb-310.2.36 > > Any help is appreciated. > > Martin > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0

OSX Mavericks lldb+mpich 3.1 debugging problem
by Martin Schatz 25 Mar '14

25 Mar '14

Hi, I’m having an issue debugging an MPI program with lldb on OSX Mavericks. I get the exact behavior as described in http://lists.mpich.org/pipermail/discuss/2014-January/002074.html except using lldb instead of gdb as the debugger. I have checked that I am using the bash terminal and unfortunately am still getting the same error. Details of my installation are as follows: I have installed the mpich-clang macport (mpich-clang-fortran is active), but when I run the cpi example with mpiexec -np 1 xterm -e lldb ./cpi I get the xterm window to spawn but: Current executable set to './cpi' (x86_64). (lldb) r Process 56694 launched: './cpi' (x86_64) [cli_0]: write_line error; fd=6 buf=:cmd=init pmi_version=1 pmi_subversion=1 : system msg for write_line failure : Bad file descriptor [cli_0]: Unable to write to PMI_fd [cli_0]: write_line error; fd=6 buf=:cmd=get_appnum : system msg for write_line failure : Bad file descriptor Fatal error in MPI_Init: Other MPI error, error stack: MPIR_Init_thread(467): MPID_Init(140).......: channel initialization failed MPID_Init(422).......: PMI_Get_appnum returned -1 Process 56694 exited with status = 1 (0x00000001) (lldb) The lldb version is: lldb --version lldb-310.2.36 Any help is appreciated. Martin

1 0

Re: [mpich-discuss] Is KNEM used for 1-sided comm?
by Balaji, Pavan 25 Mar '14

25 Mar '14

I don’t think we ever investigated KNEM with one-sided communication. — Pavan On Mar 25, 2014, at 12:27 PM, Velickovic Nikola <nikola.velickovic(a)epfl.ch> wrote: > Hello, > > I am using MPICH 3.0.4 with asynchronous progress enabled along with KNEM module on my system. > I have a DMA unit available and it is properly used by KNEM. > When I try a benchmark using non-blocking comm. there is computation-communication overlap, but when I try the same using 1-sided communication there is no overlap. > It seems like MPICH is not using KNEM for 1-sided comm. > > Do you have any idea why this might be? > > Best, > Nikola > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0