May 2016 - discuss - mailman3.cels.anl.gov

Re: [mpich-discuss] MPICH2 running problem
by Seo, Sangmin 09 May '16

09 May '16

Hello, We no longer support the Windows environment including Cygwin. Please refer to https://wiki.mpich.org/mpich/index.php/Frequently_Asked_Questions#Q:_Why_ca… Best regards, Sangmin On May 4, 2016, at 4:34 PM, Iraj Ghofrani <iraj.ghofrani(a)yahoo.com<mailto:[email protected]>> wrote: Hi I want to run convergecfd software with MPICH2 but i can't do that.... Can you please help me? the programs guide is: -------------------------------------------------------------------------------------------------------------------------------------------- Before you run an MPICH parallel job, an administrator must launch the smpd service by running: smpd.exe Prior to running your first MPICH2 job, register your user credentials with the MPICH2 user management system by running:   mpiexec -register Now you can launch an MPICH2 job by running the following set of commands from a DOS prompt:   chdir <location of CONVERGE input files> mpiexec –n <number of processors> converge-<version>-mpich2_windows-<architecture> --------------------------------------------------------------------------------------------------------------------------------------------------- I attached my path <2016-05-05_1-45-32.png>_______________________________________________ discuss mailing list discuss(a)mpich.org<mailto:[email protected]> 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] Enable Debugging MPICH
by Halim Amer 09 May '16

09 May '16

I don't see why you couldn't step into MPI_Get_processor_name, other than you are linking to the wrong MPI implementation that doesn't have debug symbols. I tested your example with --enable-g=all when building MPICH and it was sufficient to step into MPI_Get_processor_name. --Halim On 5/8/16 6:56 AM, 冰 wrote: > Hi, > I want to use |gdb| to step into the code to see what's going on inside > the code, for example my code as below, I set a break at the line > MPI_Get_processor_name (processor_name, &namelen), I want to step into > MPI_Get_processor_name, I configure with |--enable-g=all and > |--enable-fast=O0, but gdb cann't step into MPI_Get_processor_name, it > steps into fprintf, what should I do? > > #include "mpi.h" > #include <stdio.h> > int main (int argc, char **argv) > { > int myid, numprocs; > int namelen; > char processor_name[MPI_MAX_PROCESSOR_NAME]; > > MPI_Init (&argc, &argv); > MPI_Comm_rank (MPI_COMM_WORLD, &myid); > MPI_Comm_size (MPI_COMM_WORLD, &numprocs); > MPI_Get_processor_name (processor_name, &namelen); > fprintf (stderr, "Hello World! Process %d of %d on %s\n", myid, > numprocs, processor_name); > MPI_Finalize (); > return 0; > } > > > > > > > _______________________________________________ > 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] MPI_Comm_accept() pins cpu core 100%
by Balaji, Pavan 09 May '16

09 May '16

For now, the only workaround for this is to configure with --with-device=ch3:sock -- Pavan > On May 6, 2016, at 5:49 PM, Dmitriy Lyubimov <dlieu.7(a)gmail.com> wrote: > > Hi, > > I have a situation when a process my start up and execute MPI_Comm_accept(), while waiting for others to join, which may take some time. > > I have observed that all processes that wait in MPI_Comm_accept(), pin 1 cpu core exactly 100%. > > Which is unfortunate in my situation since if i launch enough processes, i pin the entire machine node very quickly -- and it does do other things that run into a crawl. > > Is this a known problem? is there a workaround for this? > > I tried on ubuntu 14-04 LTS which installs mpich 3.0.4 by default but i also tested 3.2 Nov/2015 and they both exhibit this very behavior. > > Thank you very much in advance. > -Dmitriy > _______________________________________________ > 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

Enable Debugging MPICH
by 冰 08 May '16

08 May '16

Hi, I want to use gdb to step into the code to see what's going on inside the code, for example my code as below, I set a break at the line MPI_Get_processor_name (processor_name, &namelen), I want to step into MPI_Get_processor_name, I configure with --enable-g=all and --enable-fast=O0, but gdb cann't step into MPI_Get_processor_name, it steps into fprintf, what should I do? #include "mpi.h" #include <stdio.h> int main (int argc, char **argv) { int myid, numprocs; int namelen; char processor_name[MPI_MAX_PROCESSOR_NAME]; MPI_Init (&argc, &argv); MPI_Comm_rank (MPI_COMM_WORLD, &myid); MPI_Comm_size (MPI_COMM_WORLD, &numprocs); MPI_Get_processor_name (processor_name, &namelen); fprintf (stderr, "Hello World! Process %d of %d on %s\n", myid, numprocs, processor_name); MPI_Finalize (); return 0; } _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

MPI_Comm_accept() pins cpu core 100%
by Dmitriy Lyubimov 06 May '16

06 May '16

Hi, I have a situation when a process my start up and execute MPI_Comm_accept(), while waiting for others to join, which may take some time. I have observed that all processes that wait in MPI_Comm_accept(), pin 1 cpu core exactly 100%. Which is unfortunate in my situation since if i launch enough processes, i pin the entire machine node very quickly -- and it does do other things that run into a crawl. Is this a known problem? is there a workaround for this? I tried on ubuntu 14-04 LTS which installs mpich 3.0.4 by default but i also tested 3.2 Nov/2015 and they both exhibit this very behavior. Thank you very much in advance. -Dmitriy _______________________________________________ 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] cb_nodes accurate number and list
by Bob Cernohous 06 May '16

06 May '16

Are you asking about MPICH_MPIIO_HINTS_DISPLAY? That should display the hints at file open. > man intro_mpi MPICH_MPIIO_HINTS_DISPLAY If set, causes rank 0 in the participating communicator to display the names and values of all MPI-IO hints that are set for the file being opened with the MPI_File_open call. Default: not enabled. The MPICH_MPIIO_HINTS is used to set hints. > MPICH_MPIIO_HINTS_DISPLAY=1 MPICH_MPIIO_HINTS=*:romio_no_indep_rw=true ....run a program PE 0: MPIIO hints for testdir/file: ... cb_nodes = 1 romio_no_indep_rw = true cb_config_list = *:* romio_filesystem_type = CRAY ADIO: On Thu, May 5, 2016 at 8:35 PM, Rob Latham <robl(a)mcs.anl.gov> wrote: > > > On 05/05/2016 07:27 PM, Jaln wrote: > >> Hi, >> I'm using the environmental variable MPICH_MPIIO_HINTS, to dump the >> mpiio configuration, >> but apparently, many layers(HDF5, Filesystems, etc) can overwrite the >> hints that passed in by the user. I think MPICH_MPIIO_HINTS can only >> grab the static information, not the runtime info. I'm wondering if >> there is anyway to accurately print out the MPIIO information without >> hacking the source codes? Thanks, >> > > First off, I don't have any idea what Cray is doing with their MPI > implementation. They don't share their code with us, so you'd better > contact them for support. > > > I'm going to guess that MPICH_MPIIO_HINTS is really doing what you want -- > if it's not, that's a bug and you should contact cray about that. > > ==rob > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > -- A conclusion is the place where you got tired of thinking. -- Steven Wright _______________________________________________ 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] cb_nodes accurate number and list
by Rob Latham 05 May '16

05 May '16

On 05/05/2016 07:27 PM, Jaln wrote: > Hi, > I'm using the environmental variable MPICH_MPIIO_HINTS, to dump the > mpiio configuration, > but apparently, many layers(HDF5, Filesystems, etc) can overwrite the > hints that passed in by the user. I think MPICH_MPIIO_HINTS can only > grab the static information, not the runtime info. I'm wondering if > there is anyway to accurately print out the MPIIO information without > hacking the source codes? Thanks, First off, I don't have any idea what Cray is doing with their MPI implementation. They don't share their code with us, so you'd better contact them for support. I'm going to guess that MPICH_MPIIO_HINTS is really doing what you want -- if it's not, that's a bug and you should contact cray about that. ==rob _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

cb_nodes accurate number and list
by Jaln 05 May '16

05 May '16

Hi, I'm using the environmental variable MPICH_MPIIO_HINTS, to dump the mpiio configuration, but apparently, many layers(HDF5, Filesystems, etc) can overwrite the hints that passed in by the user. I think MPICH_MPIIO_HINTS can only grab the static information, not the runtime info. I'm wondering if there is anyway to accurately print out the MPIIO information without hacking the source codes? Thanks, Best, Jialin _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

MPICH2 running problem
by Iraj Ghofrani 04 May '16

04 May '16

HiI want to run convergecfd software with MPICH2 but i can't do that....Can you please help me? the programs guide is:--------------------------------------------------------------------------------------------------------------------------------------------Before you run an MPICH parallel job, an administrator must launch the smpd service by running:smpd.exePrior to running your first MPICH2 job, register your user credentials with the MPICH2 user management system by running:  mpiexec -registerNow you can launch an MPICH2 job by running the following set of commands from a DOS prompt:  chdir <location of CONVERGE input files> mpiexec –n <number of processors> converge-<version>-mpich2_windows-<architecture>---------------------------------------------------------------------------------------------------------------------------------------------------I attached my path _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0