June 2015 - devel - mailman3.cels.anl.gov

Re: [mpich-devel] about MPI hydra process management
by Balaji, Pavan 30 Jun '15

30 Jun '15

Hydra supports a bunch of daemon processes, including ssh. There is a daemon process included with MPICH called "persist", which you could use for this purpose, but it's not really maintained. You could try it. -- Pavan From: unix haha Reply-To: "devel(a)mpich.org<mailto:[email protected]>" Date: Tuesday, June 30, 2015 at 6:18 PM To: "devel(a)mpich.org<mailto:[email protected]>" Subject: [mpich-devel] about MPI hydra process management Hi, all: I'm working on port the MPICH2 3.1.4 to an embedded environment. I met some problems for hydra process management, Could anyone help me? 1. For the old MPD, before the mpirun, there's a daemon process running on all computer nodes. But I don't see what's hydra's daemon process in the current MPI document. Could you give me some hints? 2. As we run the MPICH2 in a resource constraint environment, the SSH or the other linux-based launcher may not be used. Does it exist any simple launcher running on the computer nodes before the mpiexec? 3. What's my understanding for hydra is: once the Launcher process running on all computer nodes, I can use the mpiexec on the control node to start the parallel jobs? If so, the problem is: Could I start the Launcher process manually if there's no linux-based launcher? Thanks!

1 0

about MPI hydra process management
by unix haha 30 Jun '15

30 Jun '15

Hi, all: I'm working on port the MPICH2 3.1.4 to an embedded environment. I met some problems for hydra process management, Could anyone help me? 1. For the old MPD, before the mpirun, there's a daemon process running on all computer nodes. But I don't see what's hydra's daemon process in the current MPI document. Could you give me some hints? 2. As we run the MPICH2 in a resource constraint environment, the SSH or the other linux-based launcher may not be used. Does it exist any simple launcher running on the computer nodes before the mpiexec? 3. What's my understanding for hydra is: once the Launcher process running on all computer nodes, I can use the mpiexec on the control node to start the parallel jobs? If so, the problem is: Could I start the Launcher process manually if there's no linux-based launcher? Thanks!

1 0

Re: [mpich-devel] MPI_Init triggers Windows Firewall
by Anh Vo 30 Jun '15

30 Jun '15

Hi Colin, If this is MS-MPI you can set the environment variable MSMPI_DISABLE_SOCK to 1 before running the job, that should prevent the Windows Firewall from popping up Anh On Thu, Jun 25, 2015 at 12:41 PM, Antonio J. Peña <apenya(a)mcs.anl.gov> wrote: > > Sorry, MPICH on Windows is unsupported (and so it is the use of MPI > without MPI_Init). > > Please refer to our FAQ: > > https://wiki.mpich.org/mpich/index.php/Frequently_Asked_Questions#Q:_Why_ca… > > > On 06/25/2015 02:26 PM, Colin Potter wrote: > > Is there any way to get the MPI_Init call to not trigger the Windows > Firewall dialog? It asks for TCP and UDP access. This is in an application > being run standalone, without mpiexec. > > If I comment out MPI_Init, it does not trigger the dialog. > > > _______________________________________________ > To manage subscription options or unsubscribe:https://lists.mpich.org/mailman/listinfo/devel > > > > -- > Antonio J. Peña > Postdoctoral Appointee > Mathematics and Computer Science Division > Argonne National Laboratory > 9700 South Cass Avenue, Bldg. 240, Of. 3148 > Argonne, IL 60439-4847apenya(a)mcs.anl.govwww.mcs.anl.gov/~apenya > > > _______________________________________________ > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/devel >

1 0

Re: [mpich-devel] MPI_Init triggers Windows Firewall
by Antonio J. Peña 25 Jun '15

25 Jun '15

Sorry, MPICH on Windows is unsupported (and so it is the use of MPI without MPI_Init). Please refer to our FAQ: https://wiki.mpich.org/mpich/index.php/Frequently_Asked_Questions#Q:_Why_ca… On 06/25/2015 02:26 PM, Colin Potter wrote: > Is there any way to get the MPI_Init call to not trigger the Windows > Firewall dialog? It asks for TCP and UDP access. This is in an > application being run standalone, without mpiexec. > > If I comment out MPI_Init, it does not trigger the dialog. > > > _______________________________________________ > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/devel -- Antonio J. Peña Postdoctoral Appointee Mathematics and Computer Science Division Argonne National Laboratory 9700 South Cass Avenue, Bldg. 240, Of. 3148 Argonne, IL 60439-4847 apenya(a)mcs.anl.gov www.mcs.anl.gov/~apenya

1 0

MPI_Init triggers Windows Firewall
by Colin Potter 25 Jun '15

25 Jun '15

Is there any way to get the MPI_Init call to not trigger the Windows Firewall dialog? It asks for TCP and UDP access. This is in an application being run standalone, without mpiexec. If I comment out MPI_Init, it does not trigger the dialog.

1 0

Re: [mpich-devel] MPI_Win leaking memory
by Zhao, Xin 08 Jun '15

08 Jun '15

Hi Lisandro, You can track the progress of this bug here: https://trac.mpich.org/projects/mpich/ticket/2273 Thanks, Xin ________________________________________ From: Zhao, Xin [xinzhao3(a)illinois.edu] Sent: Friday, June 05, 2015 10:14 AM To: devel(a)mpich.org Subject: Re: [mpich-devel] MPI_Win leaking memory Hi Lisandro, Thanks. I will look into this. Xin ________________________________________ From: Lisandro Dalcin [dalcinl(a)gmail.com] Sent: Friday, June 05, 2015 10:02 AM To: mpich-devel Subject: Re: [mpich-devel] MPI_Win leaking memory On 5 June 2015 at 15:01, Balaji, Pavan <balaji(a)anl.gov> wrote: > > Huh. I wonder why our testing is not catching this. Can you send us the exact test you used? > This is the minimal Python code able to reproduce the leak, translation to C is obvious. I would not be surprised if none of your tests have such a code path. from mpi4py import MPI win = MPI.Win.Create(MPI.BOTTOM, 1, MPI.INFO_NULL, MPI.COMM_SELF) info = win.Get_info() win.Set_info(info) info.Free() win.Free() -- Lisandro Dalcin ============ Research Scientist Computer, Electrical and Mathematical Sciences & Engineering (CEMSE) Numerical Porous Media Center (NumPor) King Abdullah University of Science and Technology (KAUST) http://numpor.kaust.edu.sa/ 4700 King Abdullah University of Science and Technology al-Khawarizmi Bldg (Bldg 1), Office # 4332 Thuwal 23955-6900, Kingdom of Saudi Arabia http://www.kaust.edu.sa Office Phone: +966 12 808-0459 _______________________________________________ To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/devel _______________________________________________ To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/devel

1 0

Re: [mpich-devel] MPI_Win leaking memory
by Zhao, Xin 05 Jun '15

05 Jun '15

Hi Lisandro, Thanks. I will look into this. Xin ________________________________________ From: Lisandro Dalcin [dalcinl(a)gmail.com] Sent: Friday, June 05, 2015 10:02 AM To: mpich-devel Subject: Re: [mpich-devel] MPI_Win leaking memory On 5 June 2015 at 15:01, Balaji, Pavan <balaji(a)anl.gov> wrote: > > Huh. I wonder why our testing is not catching this. Can you send us the exact test you used? > This is the minimal Python code able to reproduce the leak, translation to C is obvious. I would not be surprised if none of your tests have such a code path. from mpi4py import MPI win = MPI.Win.Create(MPI.BOTTOM, 1, MPI.INFO_NULL, MPI.COMM_SELF) info = win.Get_info() win.Set_info(info) info.Free() win.Free() -- Lisandro Dalcin ============ Research Scientist Computer, Electrical and Mathematical Sciences & Engineering (CEMSE) Numerical Porous Media Center (NumPor) King Abdullah University of Science and Technology (KAUST) http://numpor.kaust.edu.sa/ 4700 King Abdullah University of Science and Technology al-Khawarizmi Bldg (Bldg 1), Office # 4332 Thuwal 23955-6900, Kingdom of Saudi Arabia http://www.kaust.edu.sa Office Phone: +966 12 808-0459 _______________________________________________ To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/devel

1 0

Re: [mpich-devel] MPI_Win leaking memory
by Lisandro Dalcin 05 Jun '15

05 Jun '15

On 5 June 2015 at 15:01, Balaji, Pavan <balaji(a)anl.gov> wrote: > > Huh. I wonder why our testing is not catching this. Can you send us the exact test you used? > This is the minimal Python code able to reproduce the leak, translation to C is obvious. I would not be surprised if none of your tests have such a code path. from mpi4py import MPI win = MPI.Win.Create(MPI.BOTTOM, 1, MPI.INFO_NULL, MPI.COMM_SELF) info = win.Get_info() win.Set_info(info) info.Free() win.Free() -- Lisandro Dalcin ============ Research Scientist Computer, Electrical and Mathematical Sciences & Engineering (CEMSE) Numerical Porous Media Center (NumPor) King Abdullah University of Science and Technology (KAUST) http://numpor.kaust.edu.sa/ 4700 King Abdullah University of Science and Technology al-Khawarizmi Bldg (Bldg 1), Office # 4332 Thuwal 23955-6900, Kingdom of Saudi Arabia http://www.kaust.edu.sa Office Phone: +966 12 808-0459

1 0

Re: [mpich-devel] MPI_Win leaking memory
by William Gropp 05 Jun '15

05 Jun '15

That’s a good question. If the coverage testing hadn’t been broken, we could check to see if there is RMA code that our tests aren’t executing. Bill On Jun 5, 2015, at 7:01 AM, Balaji, Pavan <balaji(a)anl.gov> wrote: > > Huh. I wonder why our testing is not catching this. Can you send us the exact test you used? > > -- Pavan > > > -- > Pavan Balaji > http://www.mcs.anl.gov/~balaji > > > > > > On 6/5/15, 3:42 AM, "Lisandro Dalcin" <dalcinl(a)gmail.com> wrote: > >> I configured 3.2b3 with --enable-g=all --enable-error-messages=all >> --enable-error-checking=all >> >> Then I get: >> >> $ python test/runtests.py -q >> ---------------------------------------------------------------------- >> Ran 1018 tests in 7.253s >> >> OK >> [0] 20480 at [0x0000000002b3ed58], src/mpid/ch3/src/mpid_rma.c[382] >> [0] 20480 at [0x0000000002b38118], src/mpid/ch3/src/mpid_rma.c[382] >> [0] 20480 at [0x0000000002b70ad8], src/mpid/ch3/src/mpid_rma.c[382] >> [0] 20480 at [0x0000000002b6ba28], src/mpid/ch3/src/mpid_rma.c[382] >> [0] 20480 at [0x0000000002b60978], src/mpid/ch3/src/mpid_rma.c[382] >> [0] 20480 at [0x0000000002b5b8c8], src/mpid/ch3/src/mpid_rma.c[382] >> [0] 20480 at [0x0000000002b56818], src/mpid/ch3/src/mpid_rma.c[382] >> [0] 20480 at [0x0000000002b4c1c8], src/mpid/ch3/src/mpid_rma.c[382] >> >> So it seems MPI_Win's are leaking memory [ (*win_ptr)->lock_entry_pool_start ] >> >> -- >> Lisandro Dalcin >> ============ >> Research Scientist >> Computer, Electrical and Mathematical Sciences & Engineering (CEMSE) >> Numerical Porous Media Center (NumPor) >> King Abdullah University of Science and Technology (KAUST) >> http://numpor.kaust.edu.sa/ >> >> 4700 King Abdullah University of Science and Technology >> al-Khawarizmi Bldg (Bldg 1), Office # 4332 >> Thuwal 23955-6900, Kingdom of Saudi Arabia >> http://www.kaust.edu.sa >> >> Office Phone: +966 12 808-0459 >> _______________________________________________ >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/devel > _______________________________________________ > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/devel

1 0

Re: [mpich-devel] MPI_Win leaking memory
by Balaji, Pavan 05 Jun '15

05 Jun '15

Huh. I wonder why our testing is not catching this. Can you send us the exact test you used? -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji On 6/5/15, 3:42 AM, "Lisandro Dalcin" <dalcinl(a)gmail.com> wrote: >I configured 3.2b3 with --enable-g=all --enable-error-messages=all >--enable-error-checking=all > >Then I get: > >$ python test/runtests.py -q >---------------------------------------------------------------------- >Ran 1018 tests in 7.253s > >OK >[0] 20480 at [0x0000000002b3ed58], src/mpid/ch3/src/mpid_rma.c[382] >[0] 20480 at [0x0000000002b38118], src/mpid/ch3/src/mpid_rma.c[382] >[0] 20480 at [0x0000000002b70ad8], src/mpid/ch3/src/mpid_rma.c[382] >[0] 20480 at [0x0000000002b6ba28], src/mpid/ch3/src/mpid_rma.c[382] >[0] 20480 at [0x0000000002b60978], src/mpid/ch3/src/mpid_rma.c[382] >[0] 20480 at [0x0000000002b5b8c8], src/mpid/ch3/src/mpid_rma.c[382] >[0] 20480 at [0x0000000002b56818], src/mpid/ch3/src/mpid_rma.c[382] >[0] 20480 at [0x0000000002b4c1c8], src/mpid/ch3/src/mpid_rma.c[382] > >So it seems MPI_Win's are leaking memory [ (*win_ptr)->lock_entry_pool_start ] > >-- >Lisandro Dalcin >============ >Research Scientist >Computer, Electrical and Mathematical Sciences & Engineering (CEMSE) >Numerical Porous Media Center (NumPor) >King Abdullah University of Science and Technology (KAUST) >http://numpor.kaust.edu.sa/ > >4700 King Abdullah University of Science and Technology >al-Khawarizmi Bldg (Bldg 1), Office # 4332 >Thuwal 23955-6900, Kingdom of Saudi Arabia >http://www.kaust.edu.sa > >Office Phone: +966 12 808-0459 >_______________________________________________ >To manage subscription options or unsubscribe: >https://lists.mpich.org/mailman/listinfo/devel

1 0