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

Re: [mpich-discuss] mpiexec crash
by Jain, Rohit 28 Oct '13

28 Oct '13

Pavan, We retried the runs again. There is no ENOENT error now. But, MPI is still failing consistently with same error: > [proxy:0:0@gretel] HYD_pmcd_pmip_control_cmd_cb (</PATH/TO>/src/pm/hydra/pm/pmiserv/pmip_cb.c:934): assert (!closed) failed > [proxy:0:0@gretel] HYDT_dmxu_poll_wait_for_event (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/tools/demux/demux_poll.c:77): callback returned error status > [proxy:0:0@gretel] main (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/pm/pmiserv/pmip.c:210): demux engine error waiting for event > [mpiexec@gretel] control_cb (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/pm/pmiserv/pmiserv_cb.c:201): assert (!closed) failed > [mpiexec@gretel] HYDT_dmxu_poll_wait_for_event (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/tools/demux/demux_poll.c:77): callback returned error status > [mpiexec@gretel] HYD_pmci_wait_for_completion (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/pm/pmiserv/pmiserv_pmci.c:196): error waiting for event > [mpiexec@gretel] main (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/ui/mpich/mpiexec.c:325): process manager error waiting for completion We are running it on same machine as: mpiexec -n 1 <exec> : -n 1 <exec> :..... What would cause such error to appear? How do we debug such issues? Regards, Rohit -----Original Message----- From: discuss-bounces(a)mpich.org [mailto:[email protected]] On Behalf Of Pavan Balaji Sent: Wednesday, October 23, 2013 7:22 PM To: discuss(a)mpich.org Subject: Re: [mpich-discuss] mpiexec crash On Oct 23, 2013, at 5:29 PM, Cherukumilli, Vasu <Vasu_Cherukumilli(a)mentor.com> wrote: > Crash that we are seeing: > > [proxy:0:0@gretel] HYD_pmcd_pmip_control_cmd_cb (</PATH/TO>/src/pm/hydra/pm/pmiserv/pmip_cb.c:934): assert (!closed) failed > [proxy:0:0@gretel] HYDT_dmxu_poll_wait_for_event (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/tools/demux/demux_poll.c:77): callback returned error status > [proxy:0:0@gretel] main (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/pm/pmiserv/pmip.c:210): demux engine error waiting for event > [mpiexec@gretel] control_cb (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/pm/pmiserv/pmiserv_cb.c:201): assert (!closed) failed > [mpiexec@gretel] HYDT_dmxu_poll_wait_for_event (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/tools/demux/demux_poll.c:77): callback returned error status > [mpiexec@gretel] HYD_pmci_wait_for_completion (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/pm/pmiserv/pmiserv_pmci.c:196): error waiting for event > [mpiexec@gretel] main (</PATH/TO>/src/mpich2-1.5/src/pm/hydra/ui/mpich/mpiexec.c:325): process manager error waiting for completion These are cleanup messages. You should have gotten an output which says so. > No such file or directory. (errno = ENOENT) This is the real error message. Did you make sure your executables are located on all the nodes in the same location? -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji _______________________________________________ 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 Hello world Program failes for large no. of process
by Jeff Hammond 28 Oct '13

28 Oct '13

Running anything that isn't system administration as root is a really bad idea. It's up there with smoking a cigarette while handling an open container of gasoline. You might not always get burned, but... Jeff On Mon, Oct 28, 2013 at 9:16 AM, heshsham basit <hf.basit1(a)gmail.com> wrote: > Hi, > > Whenever I launch more than 400 processes I get the following error. But > for less than 400 process every thing goes perfectly fine. > > I am running my code on two machines: hesh and Ubuntu (from which I am > launching the jobs) > > root@ubuntu:/home# mpiexec -f hosts.cfg -n 500 ./hello > > ===================================================================================== > = BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES > > = EXIT CODE: 139 > = CLEANING UP REMAINING PROCESSES > = YOU CAN IGNORE THE BELOW CLEANUP MESSAGES > ===================================================================================== > [proxy:0:0@hesh] HYD_pmcd_pmip_control_cmd_cb (./pm/pmiserv/pmip_cb.c:928): > assert (!closed) failed > > [proxy:0:0@hesh] HYDT_dmxu_poll_wait_for_event > (./tools/demux/demux_poll.c:77): callback returned error status > [proxy:0:0@hesh] main (./pm/pmiserv/pmip.c:226): demux engine error waiting > for event > [mpiexec@ubuntu] HYDT_bscu_wait_for_completion > (./tools/bootstrap/utils/bscu_wait.c:70): one of the processes terminated > badly; aborting > > [mpiexec@ubuntu] HYDT_bsci_wait_for_completion > (./tools/bootstrap/src/bsci_wait.c:23): launcher returned error waiting for > completion > [mpiexec@ubuntu] HYD_pmci_wait_for_completion > (./pm/pmiserv/pmiserv_pmci.c:191): launcher returned error waiting for > completion > > [mpiexec@ubuntu] main (./ui/mpich/mpiexec.c:405): process manager error > waiting for completion > > > Also some I time I got the following error: > > root@ubuntu:/home# mpiexec -f hosts.cfg -n 400 ./hello > > [proxy:0:1@ubuntu] send_cmd_downstream (./pm/pmiserv/pmip_pmi_v1.c:80): > assert (!closed) failed > [proxy:0:1@ubuntu] fn_get (./pm/pmiserv/pmip_pmi_v1.c:349): error sending > PMI response > [proxy:0:1@ubuntu] pmi_cb (./pm/pmiserv/pmip_cb.c:327): PMI handler returned > error > > [proxy:0:1@ubuntu] HYDT_dmxu_poll_wait_for_event > (./tools/demux/demux_poll.c:77): callback returned error status > [proxy:0:1@ubuntu] main (./pm/pmiserv/pmip.c:226): demux engine error > waiting for event > [mpiexec@ubuntu] control_cb (./pm/pmiserv/pmiserv_cb.c:215): assert > (!closed) failed > > [mpiexec@ubuntu] HYDT_dmxu_poll_wait_for_event > (./tools/demux/demux_poll.c:77): callback returned error status > [mpiexec@ubuntu] HYD_pmci_wait_for_completion > (./pm/pmiserv/pmiserv_pmci.c:181): error waiting for event > > [mpiexec@ubuntu] main (./ui/mpich/mpiexec.c:405): process manager error > waiting for completion > > > > Also I raised the open file limit on both the PCs by using ulimit -n 2048. > But still the same result. > > The code is the simple one: > > /* C Example */ > #include <mpi.h> > #include <stdio.h> > #include <stddef.h> > #include <stdlib.h> > > int main (int argc, char* argv[]) > { > int rank, size; > int buffer_length = MPI_MAX_PROCESSOR_NAME; > char hostname[buffer_length]; > > MPI_Init (&argc, &argv); /* starts MPI */ > MPI_Comm_rank (MPI_COMM_WORLD, &rank); /* get current process id */ > MPI_Comm_size (MPI_COMM_WORLD, &size); /* get number of processes > */ > > MPI_Get_processor_name(hostname, &buffer_length); /* get hostname */ > > printf( "Hello world from process %d running on %s of %d\n", rank, > hostname, size ); > MPI_Finalize(); > return 0; > } > > For 10000 process the program is unresponsive and I have to close the > terminal to exit.t > > How do i solve this issue? > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss -- Jeff Hammond jeff.science(a)gmail.com

1 0

Re: [mpich-discuss] coupling
by Gus Correa 28 Oct '13

28 Oct '13

Hi Fereshteh 1) I would look for log or error messages in the model output *before* the final one ("MPI_Abort ..."). Although it aborted with 2 processors, that was a graceful termination by the program. So what is before the last error message may shed some light on why the model aborted, and what it requires to run correctly. ** 2) MCT is also used for coupling climate models that we use here. Some of these models it run in SPMD mode (i.e. a single executable is launched by mpiexec). Others run in MPMD mode (i.e. several executables are launched by mpiexec, which requires a mpiexec command line with more parameters). Your mpiexec suggests that the model relies on a single executable (oceanG). However, this may/may not be the case, as you mentioned three separate items also: ROMS, SWAN, and the coupler. Is it a single executable or more than one that your model uses? [If you describe how you compile the model it may help.] ** 3) Are you trying to run it in hybrid mode, i.e. using both MPI and OpenMP? I am not familiar to this model, so I am just guessing. The conventional wisdom is to use parameters names like "Nthreads" for OpenMP threads, although parameter names are elusive. OpenMP would add another twist to your processor configuration, as you would need to provision additional processors for OpenMP, besides those for MPI. ** 4) Does the domain decomposition (Ntilei and Ntilej) have to be the same for both models? Have you tried something like: Ntilei=2 Ntilej=2 i.e. 2*2=4 processors for the ocean plus one processor for SWANN, a total of 5 processors, then: mpiexec -np 5 ... In some models we run here (e.g. MITgcm) you need to match precisely the number of processors to the tiles used to decompose your domain (and possibly also the OpenMP threads). When you tried mpiexec -np 2, did you choose Ntilei=Ntilej=Nthreads=1 for both models? Do you have perhaps to add one processor (or more) specifically for the coupler? ** 5) Do your input files (describing the ocean initial state, and perhaps the wave state), have to be organized based on the "tile" configuration, or can they span the whole domain? This could be another source of error. 6) Does ROMS and SWANN have mailing lists that can perhaps help you more than the generic MPICH list? I hope this helps, Gus Correa On 10/27/2013 06:33 AM, fereshteh komijani wrote: > Sure. > > ROMS and SWAN models are models for ocean and wave models, > respectively. Coupling of them require 3 input files: coupling_test.in > <http://coupling_test.in>, swan_test.in <http://swan_test.in> and > roms_test.in <http://roms_test.in> which first one is coupled input file > and also one build.bash file which in it user called requested > libraries, cpp optins, header file, compilers (for me gcc and gfortran). > For example in build.bash file I set > > USE_MPI=on = on > USE_MPIF90 =on > > WHICH_MPI=mpich2 > > FORT = gfortran > > also with regard that for coupling model, MCT (model coupling toolkit) > is necessary I have installed it and set its include and lib > directories in build.bash file. > > In coupling_test.in <http://coupling_test.in> user call swan_test.in > <http://swan_test.in>and roms_test.in <http://roms_test.in> as wave and > ocean model's inputs. > > In roms_test.in <http://roms_test.in> some coefficients, solving > technique and some input forcing file determined. > > for choosing number of nodes for each model there are NtileI and Ntilej > in roms_test.in <http://roms_test.in> and Nthreads (ocean) and Nthreads > (wave) in coupling_test.in <http://coupling_test.in> file ( Nthreads > (ocean)=NtileI *Ntilej ) and total nodes are equal to Nthreads > (ocean)+Nthreads (wave). > > whenever i set one node for wave model (SWAN) (Nthreads (wave)=1) after > running by > > mpirun -np 2 ./oceanG coupling_inlet-test.in > <http://coupling_inlet-test.in/>>mpi.log > > it replies: > > application called MPI_Abort(comm=0x84000002, 4) - process 0 > > But when setNthreads (wave)>1 (and everything for Nthreads (ocean)) for > example > > mpirun -np 8 ./oceanG coupling_inlet-test.in > <http://coupling_inlet-test.in/>>mpi.log > > mpi.log file (attachï¿½ file) shows that roms model does not have problems > and its nodes are active but nothing happen for swan model after 2 weeks > than running. > > I hope that this information be sufficient. > > All the best > > fereshte > > > > _______________________________________________ > 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 Hello world Program failes for large no. of process
by Wesley Bland 28 Oct '13

28 Oct '13

If you’re only using two machines, you might be running out of resources when you try to use that many processes. If you want to run hundred of processes, you need hundreds of cores. On Oct 28, 2013, at 9:16 AM, heshsham basit <hf.basit1(a)gmail.com> wrote: > Hi, > > Whenever I launch more than 400 processes I get the following error. But for less than 400 process every thing goes perfectly fine. > > I am running my code on two machines: hesh and Ubuntu (from which I am launching the jobs) > > root@ubuntu:/home# mpiexec -f hosts.cfg -n 500 ./hello > > ===================================================================================== > = BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES > > = EXIT CODE: 139 > = CLEANING UP REMAINING PROCESSES > = YOU CAN IGNORE THE BELOW CLEANUP MESSAGES > ===================================================================================== > [proxy:0:0@hesh] HYD_pmcd_pmip_control_cmd_cb (./pm/pmiserv/pmip_cb.c:928): assert (!closed) failed > > [proxy:0:0@hesh] HYDT_dmxu_poll_wait_for_event (./tools/demux/demux_poll.c:77): callback returned error status > [proxy:0:0@hesh] main (./pm/pmiserv/pmip.c:226): demux engine error waiting for event > [mpiexec@ubuntu] HYDT_bscu_wait_for_completion (./tools/bootstrap/utils/bscu_wait.c:70): one of the processes terminated badly; aborting > > [mpiexec@ubuntu] HYDT_bsci_wait_for_completion (./tools/bootstrap/src/bsci_wait.c:23): launcher returned error waiting for completion > [mpiexec@ubuntu] HYD_pmci_wait_for_completion (./pm/pmiserv/pmiserv_pmci.c:191): launcher returned error waiting for completion > > [mpiexec@ubuntu] main (./ui/mpich/mpiexec.c:405): process manager error waiting for completion > > > Also some I time I got the following error: > > root@ubuntu:/home# mpiexec -f hosts.cfg -n 400 ./hello > > [proxy:0:1@ubuntu] send_cmd_downstream (./pm/pmiserv/pmip_pmi_v1.c:80): assert (!closed) failed > [proxy:0:1@ubuntu] fn_get (./pm/pmiserv/pmip_pmi_v1.c:349): error sending PMI response > [proxy:0:1@ubuntu] pmi_cb (./pm/pmiserv/pmip_cb.c:327): PMI handler returned error > > [proxy:0:1@ubuntu] HYDT_dmxu_poll_wait_for_event (./tools/demux/demux_poll.c:77): callback returned error status > [proxy:0:1@ubuntu] main (./pm/pmiserv/pmip.c:226): demux engine error waiting for event > [mpiexec@ubuntu] control_cb (./pm/pmiserv/pmiserv_cb.c:215): assert (!closed) failed > > [mpiexec@ubuntu] HYDT_dmxu_poll_wait_for_event (./tools/demux/demux_poll.c:77): callback returned error status > [mpiexec@ubuntu] HYD_pmci_wait_for_completion (./pm/pmiserv/pmiserv_pmci.c:181): error waiting for event > > [mpiexec@ubuntu] main (./ui/mpich/mpiexec.c:405): process manager error waiting for completion > > > Also I raised the open file limit on both the PCs by using ulimit -n 2048. But still the same result. > The code is the simple one: > > /* C Example */ > #include <mpi.h> > #include <stdio.h> > #include <stddef.h> > #include <stdlib.h> > > int main (int argc, char* argv[]) > { > int rank, size; > int buffer_length = MPI_MAX_PROCESSOR_NAME; > char hostname[buffer_length]; > > MPI_Init (&argc, &argv); /* starts MPI */ > MPI_Comm_rank (MPI_COMM_WORLD, &rank); /* get current process id */ > MPI_Comm_size (MPI_COMM_WORLD, &size); /* get number of processes */ > > MPI_Get_processor_name(hostname, &buffer_length); /* get hostname */ > > printf( "Hello world from process %d running on %s of %d\n", rank, hostname, size ); > MPI_Finalize(); > return 0; > } > > For 10000 process the program is unresponsive and I have to close the terminal to exit.t > How do i solve this issue? > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0

MPI Hello world Program failes for large no. of process
by heshsham basit 28 Oct '13

28 Oct '13

Hi, Whenever I launch more than 400 processes I get the following error. But for less than 400 process every thing goes perfectly fine. I am running my code on two machines: hesh and Ubuntu (from which I am launching the jobs) root@ubuntu:/home# mpiexec -f hosts.cfg -n 500 ./hello ===================================================================================== = BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES = EXIT CODE: 139 = CLEANING UP REMAINING PROCESSES = YOU CAN IGNORE THE BELOW CLEANUP MESSAGES ===================================================================================== [proxy:0:0@hesh] HYD_pmcd_pmip_control_cmd_cb (./pm/pmiserv/pmip_cb.c:928): assert (!closed) failed [proxy:0:0@hesh] HYDT_dmxu_poll_wait_for_event (./tools/demux/demux_poll.c:77): callback returned error status [proxy:0:0@hesh] main (./pm/pmiserv/pmip.c:226): demux engine error waiting for event [mpiexec@ubuntu] HYDT_bscu_wait_for_completion (./tools/bootstrap/utils/bscu_wait.c:70): one of the processes terminated badly; aborting [mpiexec@ubuntu] HYDT_bsci_wait_for_completion (./tools/bootstrap/src/bsci_wait.c:23): launcher returned error waiting for completion [mpiexec@ubuntu] HYD_pmci_wait_for_completion (./pm/pmiserv/pmiserv_pmci.c:191): launcher returned error waiting for completion [mpiexec@ubuntu] main (./ui/mpich/mpiexec.c:405): process manager error waiting for completion Also some I time I got the following error: root@ubuntu:/home# mpiexec -f hosts.cfg -n 400 ./hello [proxy:0:1@ubuntu] send_cmd_downstream (./pm/pmiserv/pmip_pmi_v1.c:80): assert (!closed) failed [proxy:0:1@ubuntu] fn_get (./pm/pmiserv/pmip_pmi_v1.c:349): error sending PMI response [proxy:0:1@ubuntu] pmi_cb (./pm/pmiserv/pmip_cb.c:327): PMI handler returned error [proxy:0:1@ubuntu] HYDT_dmxu_poll_wait_for_event (./tools/demux/demux_poll.c:77): callback returned error status [proxy:0:1@ubuntu] main (./pm/pmiserv/pmip.c:226): demux engine error waiting for event [mpiexec@ubuntu] control_cb (./pm/pmiserv/pmiserv_cb.c:215): assert (!closed) failed [mpiexec@ubuntu] HYDT_dmxu_poll_wait_for_event (./tools/demux/demux_poll.c:77): callback returned error status [mpiexec@ubuntu] HYD_pmci_wait_for_completion (./pm/pmiserv/pmiserv_pmci.c:181): error waiting for event [mpiexec@ubuntu] main (./ui/mpich/mpiexec.c:405): process manager error waiting for completion Also I raised the open file limit on both the PCs by using ulimit -n 2048. But still the same result. The code is the simple one: /* C Example */ #include <mpi.h> #include <stdio.h> #include <stddef.h> #include <stdlib.h> int main (int argc, char* argv[]) { int rank, size; int buffer_length = MPI_MAX_PROCESSOR_NAME; char hostname[buffer_length]; MPI_Init (&argc, &argv); /* starts MPI */ MPI_Comm_rank (MPI_COMM_WORLD, &rank); /* get current process id */ MPI_Comm_size (MPI_COMM_WORLD, &size); /* get number of processes */ MPI_Get_processor_name(hostname, &buffer_length); /* get hostname */ printf( "Hello world from process %d running on %s of %d\n", rank, hostname, size ); MPI_Finalize(); return 0; } For 10000 process the program is unresponsive and I have to close the terminal to exit.t How do i solve this issue?

1 0

Re: [mpich-discuss] MPICH2 on heterogeneous platforms
by heshsham basit 28 Oct '13

28 Oct '13

Ok Pavan. Got it ;) Rgrds, On 27 October 2013 19:45, Pavan Balaji <balaji(a)mcs.anl.gov> wrote: > > On Oct 27, 2013, at 12:54 AM, heshsham basit <hf.basit1(a)gmail.com> wrote: > > Wondering if it Is possible to use MPICH2 on heterogeneous platforms > such as one where OS and Processor make are different on different machines? > > It’ll work correctly as long as the datatype representations are > consistent. For example, if both are little endian machines with the same > size for integers, doubles, etc., it’ll be fine. For example, if you have > x86_64 processors on both machines, but are using two different variants of > 64-bit linux (and the same compiler on both machines), it should be OK. > > Nevertheless, we don’t really support such configurations. So if things > break, you are on your own. :-) > > —- Pavan > > -- > Pavan Balaji > http://www.mcs.anl.gov/~balaji > > _______________________________________________ > 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] coupling
by Huiwei Lu 27 Oct '13

27 Oct '13

Hi, Fereshte, I am not familiar with oceanG application. So the following suggestions are based on my guesses or experiences: 1. As Toni and I have suggested, try to use the most recent version of MPICH in http://www.mpich.org/, to see if the bug still occurs. 2. Have you tried to run it under gdb or ddd to catch the bug? You can follow the steps in http://wiki.mpich.org/mpich/index.php/Building_MPICH_For_Development#Use_gdb. Try ‘bt’ in gdb to show where things goes wrong. mpiexec -n [number of processes] xterm -e gdb ./your_app 3. What’s the difference between 'mpirun -np 2 ./oceanG coupling_inlet-test.in>mpi.log’ and 'mpirun -np 8 ./oceanG coupling_inlet-test.in>mpi.log’? Only the number of processes? Maybe the difference of these two runs is the place you can start to find the bug. 4. As you said, the swan model may go wrong. Have you tried to run swan model only to see if it is broken? -- Huiwei Lu http://www.mcs.anl.gov/~huiweilu/ On Oct 27, 2013, at 5:33 AM, fereshteh komijani <fereshtehkomijani(a)gmail.com> wrote: > Sure. > ROMS and SWAN models are models for ocean and wave models, respectively. Coupling of them require 3 input files: coupling_test.in, swan_test.in and roms_test.in which first one is coupled input file and also one build.bash file which in it user called requested libraries, cpp optins, header file, compilers (for me gcc and gfortran). For example in build.bash file I set > USE_MPI=on = on > USE_MPIF90 =on > WHICH_MPI=mpich2 > FORT = gfortran > also with regard that for coupling model, MCT (model coupling toolkit) is necessary I have installed it and set its include and lib directories in build.bash file. > > In coupling_test.in user call swan_test.in and roms_test.in as wave and ocean model's inputs. > In roms_test.in some coefficients, solving technique and some input forcing file determined. > > for choosing number of nodes for each model there are NtileI and Ntilej in roms_test.in and Nthreads (ocean) and Nthreads (wave) in coupling_test.in file ( Nthreads (ocean)=NtileI *Ntilej ) and total nodes are equal to Nthreads (ocean)+Nthreads (wave). > whenever i set one node for wave model (SWAN) (Nthreads (wave)=1) after running by > mpirun -np 2 ./oceanG coupling_inlet-test.in>mpi.log > it replies: > application called MPI_Abort(comm=0x84000002, 4) - process 0 > But when set Nthreads (wave)>1 (and everything for Nthreads (ocean)) for example > mpirun -np 8 ./oceanG coupling_inlet-test.in>mpi.log > mpi.log file (attaché file) shows that roms model does not have problems and its nodes are active but nothing happen for swan model after 2 weeks than running. > I hope that this information be sufficient. > All the best > fereshte > > <mpi.txt>_______________________________________________ > 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] coupling
by Antonio J. Peña 27 Oct '13

27 Oct '13

Hi Fereshte, There is the possibility that your application is failing due to a bug in MPICH which could have been already solved. I do a +1 in Huiwei's suggestion to try with the current stable release. Antonio On Sunday, October 27, 2013 02:03:54 PM fereshteh komijani wrote: Sure. ROMS and SWAN models are models for ocean and wave models, respectively. Coupling of them require 3 input files: coupling_test.in[1], swan_test.in[2] and roms_test.in[3] which first one is coupled input file and also one build.bash file which in it user called requested libraries, cpp optins, header file, compilers (for me gcc and gfortran). For example in build.bash file I set USE_MPI=on = on USE_MPIF90 =on WHICH_MPI=mpich2 FORT = gfortran also with regard that for coupling model, MCT (model coupling toolkit) is necessary I have installed it and set its include and lib directories in build.bash file. In coupling_test.in[1] user call swan_test.in[2] and roms_test.in[3] as wave and ocean model's inputs. In _roms_test.in_ some coefficients, solving technique and some input forcing file determined. for choosing number of nodes for each model there are NtileI and Ntilej in _roms_test.in_ and Nthreads (ocean) and Nthreads (wave) in coupling_test.in[1] file ( Nthreads (ocean)=NtileI *Ntilej ) and total nodes are equal to Nthreads (ocean)+Nthreads (wave). whenever i set one node for wave model (SWAN) (Nthreads (wave)=1) after running by mpirun -np 2 ./oceanG coupling_inlet-test.in[4]>mpi.log it replies: application called MPI_Abort(comm=0x84000002, 4) - process 0 But when set Nthreads (wave)>1 (and everything for Nthreads (ocean)) for example mpirun -np 8 ./oceanG coupling_inlet-test.in>mpi.log mpi.log file (attaché file) shows that roms model does not have problems and its nodes are active but nothing happen for swan model after 2 weeks than running. I hope that this information be sufficient. All the best fereshte -------- [1] http://coupling_test.in [2] http://swan_test.in [3] http://roms_test.in [4] http://coupling_inlet-test.in/

1 0

Re: [mpich-discuss] MPICH2 on heterogeneous platforms
by Pavan Balaji 27 Oct '13

27 Oct '13

On Oct 27, 2013, at 12:54 AM, heshsham basit <hf.basit1(a)gmail.com> wrote: > Wondering if it Is possible to use MPICH2 on heterogeneous platforms such as one where OS and Processor make are different on different machines? It’ll work correctly as long as the datatype representations are consistent. For example, if both are little endian machines with the same size for integers, doubles, etc., it’ll be fine. For example, if you have x86_64 processors on both machines, but are using two different variants of 64-bit linux (and the same compiler on both machines), it should be OK. Nevertheless, we don’t really support such configurations. So if things break, you are on your own. :-) —- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [mpich-discuss] coupling
by fereshteh komijani 27 Oct '13

27 Oct '13

Sure. ROMS and SWAN models are models for ocean and wave models, respectively. Coupling of them require 3 input files: coupling_test.in, swan_test.in and roms_test.in which first one is coupled input file and also one build.bash file which in it user called requested libraries, cpp optins, header file, compilers (for me gcc and gfortran). For example in build.bash file I set USE_MPI=on = on USE_MPIF90 =on WHICH_MPI=mpich2 FORT = gfortran also with regard that for coupling model, MCT (model coupling toolkit) is necessary I have installed it and set its include and lib directories in build.bash file. In coupling_test.in user call swan_test.in and roms_test.in as wave and ocean model's inputs. In roms_test.in some coefficients, solving technique and some input forcing file determined. for choosing number of nodes for each model there are NtileI and Ntilej in roms_test.in and Nthreads (ocean) and Nthreads (wave) in coupling_test.infile ( Nthreads (ocean)=NtileI *Ntilej ) and total nodes are equal to Nthreads (ocean)+Nthreads (wave). whenever i set one node for wave model (SWAN) (Nthreads (wave)=1) after running by mpirun -np 2 ./oceanG coupling_inlet-test.in>mpi.log it replies: application called MPI_Abort(comm=0x84000002, 4) - process 0 But when set Nthreads (wave)>1 (and everything for Nthreads (ocean)) for example mpirun -np 8 ./oceanG coupling_inlet-test.in>mpi.log mpi.log file (attaché file) shows that roms model does not have problems and its nodes are active but nothing happen for swan model after 2 weeks than running. I hope that this information be sufficient. All the best fereshte

1 0