March 2017 - discuss - mailman3.cels.anl.gov

Re: [mpich-discuss] MCS lock and MPI RMA problem
by Jeff Hammond 06 Mar '17

06 Mar '17

What processor architecture are you testing? Maybe set lmem to volatile or read it with MPI_Fetch_and_op rather than a load. MPI_Win_sync cannot prevent the compiler from caching *lmem in a register. Jeff On Sat, Mar 4, 2017 at 12:30 AM, Ask Jakobsen <afj(a)qeye-labs.com> wrote: > Hi, > > I have downloaded the source code for the MCS lock from the excellent book > "Using Advanced MPI" from http://www.mcs.anl.gov/research/projects/mpi/ > usingmpi/examples-advmpi/rma2/mcs-lock.c > > I have made a very simple piece of test code for testing the MCS lock but > it works at random and often never escapes the busy loops in the acquire > and release functions (see attached source code). The code appears > semantically correct to my eyes. > > #include <stdio.h> > #include <mpi.h> > #include "mcs-lock.h" > > int main(int argc, char *argv[]) > { > MPI_Win win; > MPI_Init( &argc, &argv ); > > MCSLockInit(MPI_COMM_WORLD, &win); > > int rank, size; > MPI_Comm_rank(MPI_COMM_WORLD, &rank); > MPI_Comm_size(MPI_COMM_WORLD, &size); > > printf("rank: %d, size: %d\n", rank, size); > > > MCSLockAcquire(win); > printf("rank %d aquired lock\n", rank); fflush(stdout); > MCSLockRelease(win); > > > MPI_Win_free(&win); > MPI_Finalize(); > return 0; > } > > > I have tested on several hardware platforms and mpich-3.2 and mpich-3.3a2 > but with no luck. > > It appears that the MPI_Win_Sync are not "refreshing" the local data or I > have a bug I can't spot. > > A simple unfair lock like http://www.mcs.anl.gov/research/projects/mpi/ > usingmpi/examples-advmpi/rma2/ga_mutex1.c works perfectly. > > Best regards, Ask Jakobsen > > > _______________________________________________ > 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 http://jeffhammond.github.io/ _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

MPI_Probe exception from Windows cluster
by Feng Yue 06 Mar '17

06 Mar '17

Hi all, We are experiencing frequent exceptions from MPI_Probe recently. Does anybody know what cause might be for this exception and what I can do to track down the cause? MPI_Probe(113).......................: MPI_Probe(src=MPI_ANY_SOURCE, tag=MPI_ANY_TAG, comm=0x84000002, status=000000033FA217C8) failed MPIDI_CH3I_Progress(402).............: MPID_nem_mpich2_blocking_recv(905)...: MPID_nem_newtcp_module_poll(37)......: MPID_nem_newtcp_module_connpoll(2656): gen_read_fail_handler(1145)..........: read from socket failed - The specified network name is no longer available. Our cluster has MPICH2 1.4.1p1 and Windows Server 2008 R2 installed on 22 slave nodes and 1 master node. Thanks, Feng _______________________________________________ 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] MCS lock and MPI RMA problem
by Ask Jakobsen 05 Mar '17

05 Mar '17

Hi Richard, The window is initialised in MCSLockInit (see book author's code http://www.mcs.anl.gov/research/projects/mpi/usingmpi/examples-advmpi/rma2/…) void MCSLockInit(MPI_Comm comm, MPI_Win *win) { int *lmem, rank; MPI_Aint winsize; MPI_Comm_rank(comm,&rank); if (MCS_LOCKRANK == MPI_KEYVAL_INVALID) MPI_Win_create_keyval(MPI_WIN_NULL_COPY_FN, MPI_WIN_NULL_DELETE_FN, &MCS_LOCKRANK, (void*)0); winsize = 2 * sizeof(int); if (rank == 0) winsize += sizeof(int); MPI_Win_allocate(winsize, sizeof(int), MPI_INFO_NULL, comm, &lmem, win); lmem[nextRank] = -1; lmem[blocked] = 0; if (rank == 0) { lmem[lockTail] = -1; } MPI_Win_set_attr(*win, MCS_LOCKRANK, (void*)(MPI_Aint)rank); MPI_Barrier(comm); } Best regards, Ask Date: Sat, 4 Mar 2017 19:03:13 +0000 > From: Richard Warren <Richard.Warren(a)hdfgroup.org> > To: "discuss(a)mpich.org" <discuss(a)mpich.org> > Subject: Re: [mpich-discuss] MCS lock and MPI RMA problem > Message-ID: > <BY2PR17MB0406FB1E6C8D6DC76CF45279EF2A0@BY2PR17MB0406. > namprd17.prod.outlook.com> > > Content-Type: text/plain; charset="us-ascii" > > I don't see that you've actually created an MPI_Win object anywhere. My > guess is that you need to use MPI_Win_create to initialize your window > before you can do any locks or other RMA operations with it. > > ________________________________ > From: Ask Jakobsen <afj(a)qeye-labs.com> > Sent: Saturday, March 4, 2017 3:30:09 AM > To: discuss(a)mpich.org > Subject: [mpich-discuss] MCS lock and MPI RMA problem > > Hi, > > I have downloaded the source code for the MCS lock from the excellent book > "Using Advanced MPI" from http://www.mcs.anl.gov/research/projects/mpi/ > usingmpi/examples-advmpi/rma2/mcs-lock.c > > I have made a very simple piece of test code for testing the MCS lock but > it works at random and often never escapes the busy loops in the acquire > and release functions (see attached source code). The code appears > semantically correct to my eyes. > > #include <stdio.h> > #include <mpi.h> > #include "mcs-lock.h" > > int main(int argc, char *argv[]) > { > MPI_Win win; > MPI_Init( &argc, &argv ); > > MCSLockInit(MPI_COMM_WORLD, &win); > > int rank, size; > MPI_Comm_rank(MPI_COMM_WORLD, &rank); > MPI_Comm_size(MPI_COMM_WORLD, &size); > > printf("rank: %d, size: %d\n", rank, size); > > > MCSLockAcquire(win); > printf("rank %d aquired lock\n", rank); fflush(stdout); > MCSLockRelease(win); > > > MPI_Win_free(&win); > MPI_Finalize(); > return 0; > } > > > I have tested on several hardware platforms and mpich-3.2 and mpich-3.3a2 > but with no luck. > > It appears that the MPI_Win_Sync are not "refreshing" the local data or I > have a bug I can't spot. > > A simple unfair lock like http://www.mcs.anl.gov/research/projects/mpi/ > usingmpi/examples-advmpi/rma2/ga_mutex1.c works perfectly. > > Best regards, Ask Jakobsen > > _______________________________________________ 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] MCS lock and MPI RMA problem
by Richard Warren 04 Mar '17

04 Mar '17

I don't see that you've actually created an MPI_Win object anywhere. My guess is that you need to use MPI_Win_create to initialize your window before you can do any locks or other RMA operations with it. ________________________________ From: Ask Jakobsen <afj(a)qeye-labs.com> Sent: Saturday, March 4, 2017 3:30:09 AM To: discuss(a)mpich.org Subject: [mpich-discuss] MCS lock and MPI RMA problem Hi, I have downloaded the source code for the MCS lock from the excellent book "Using Advanced MPI" from http://www.mcs.anl.gov/research/projects/mpi/usingmpi/examples-advmpi/rma2/… I have made a very simple piece of test code for testing the MCS lock but it works at random and often never escapes the busy loops in the acquire and release functions (see attached source code). The code appears semantically correct to my eyes. #include <stdio.h> #include <mpi.h> #include "mcs-lock.h" int main(int argc, char *argv[]) { MPI_Win win; MPI_Init( &argc, &argv ); MCSLockInit(MPI_COMM_WORLD, &win); int rank, size; MPI_Comm_rank(MPI_COMM_WORLD, &rank); MPI_Comm_size(MPI_COMM_WORLD, &size); printf("rank: %d, size: %d\n", rank, size); MCSLockAcquire(win); printf("rank %d aquired lock\n", rank); fflush(stdout); MCSLockRelease(win); MPI_Win_free(&win); MPI_Finalize(); return 0; } I have tested on several hardware platforms and mpich-3.2 and mpich-3.3a2 but with no luck. It appears that the MPI_Win_Sync are not "refreshing" the local data or I have a bug I can't spot. A simple unfair lock like http://www.mcs.anl.gov/research/projects/mpi/usingmpi/examples-advmpi/rma2/… works perfectly. Best regards, Ask Jakobsen _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

MCS lock and MPI RMA problem
by Ask Jakobsen 04 Mar '17

04 Mar '17

Hi, I have downloaded the source code for the MCS lock from the excellent book "Using Advanced MPI" from http://www.mcs.anl.gov/research/projects/mpi/usingmpi/examples-advmpi/rma2/… I have made a very simple piece of test code for testing the MCS lock but it works at random and often never escapes the busy loops in the acquire and release functions (see attached source code). The code appears semantically correct to my eyes. #include <stdio.h> #include <mpi.h> #include "mcs-lock.h" int main(int argc, char *argv[]) { MPI_Win win; MPI_Init( &argc, &argv ); MCSLockInit(MPI_COMM_WORLD, &win); int rank, size; MPI_Comm_rank(MPI_COMM_WORLD, &rank); MPI_Comm_size(MPI_COMM_WORLD, &size); printf("rank: %d, size: %d\n", rank, size); MCSLockAcquire(win); printf("rank %d aquired lock\n", rank); fflush(stdout); MCSLockRelease(win); MPI_Win_free(&win); MPI_Finalize(); return 0; } I have tested on several hardware platforms and mpich-3.2 and mpich-3.3a2 but with no luck. It appears that the MPI_Win_Sync are not "refreshing" the local data or I have a bug I can't spot. A simple unfair lock like http://www.mcs.anl.gov/research/projects/mpi/usingmpi/examples-advmpi/rma2/… works perfectly. Best regards, Ask Jakobsen _______________________________________________ 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] Problem compiling mpich on Linux
by Dave Goodell (dgoodell) 03 Mar '17

03 Mar '17

> On Mar 3, 2017, at 2:16 PM, Matthew Brown <mbrown(a)oam.co.nz> wrote: > >>> Is this NFS? Is the clock skewed? I thought extracting cvars was not done in released versions. > >>> ==rob > > Adding 'atime-preserve' option to 'tar zxvf' when extracting mpich-3.2.tar.gz followed by configuring and make on the resulting directory without copying 'cp -r' first allows building to proceed without problems. It seems unlikely that the atime has anything to do with this, the mtime should be what matters. "cp -a" is your friend when copying almost any tree. I basically never use "cp -r" and always use "cp -a". -Dave _______________________________________________ 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] Problem compiling mpich on Linux
by Matthew Brown 03 Mar '17

03 Mar '17

>>Is this NFS? Is the clock skewed? I thought extracting cvars was not done in released versions. >>==rob Adding 'atime-preserve' option to 'tar zxvf' when extracting mpich-3.2.tar.gz followed by configuring and make on the resulting directory without copying 'cp -r' first allows building to proceed without problems. Previously, was using 'tar zxvf' followed by 'cp -r' followed by successful configure. Make would then require aclocal 1.15 which wanted to avoid. This is all done on a vm in hyperv in NZST. Note it all works fine on an Azure linux vm in US time zone. Thank for your pointer about clock. _______________________________________________ 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] (no subject)
by Halim Amer 02 Mar '17

02 Mar '17

I can access that page without any problem, and the mpich.org domain is up and running (http://downforeveryoneorjustme.com/www.mpich.org/static/docs/v3.2/www3/MPI_…) I think the issue is on your side or happened temporary and got resolved. Halim www.mcs.anl.gov/~aamer On 3/2/17 4:56 AM, sergio.rivas(a)um.es wrote: > > > Good afternoon, > > > > I’m not sure if this is the correct place to report this error, but the > documentation website is no longer working. For instance, you can see > that you get a “forbidden” message: > > > > https://www.mpich.org/static/docs/v3.2/www3/MPI_Get_accumulate.html > > > > Is this known already? If so, my apologies in advance. > > > > Thank you for the great effort implementing MPICH! > > > > Kind Regards, > > Sergio. > > > > > > _______________________________________________ > 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

(no subject)
by sergio.rivas＠um.es 02 Mar '17

02 Mar '17

Good afternoon, I’m not sure if this is the correct place to report this error, but the documentation website is no longer working. For instance, you can see that you get a “forbidden” message: https://www.mpich.org/static/docs/v3.2/www3/MPI_Get_accumulate.html Is this known already? If so, my apologies in advance. Thank you for the great effort implementing MPICH! Kind Regards, Sergio. _______________________________________________ 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] Problem compiling mpich on Linux
by Latham, Robert J. 01 Mar '17

01 Mar '17

On Wed, 2017-03-01 at 18:03 +0000, Matthew Brown wrote: > > > ……….. > CC src/util/instrm/lib_libmpi_la-instr.lo > CC src/util/mem/lib_libmpi_la-trmem.lo > CC src/util/mem/lib_libmpi_la-handlemem.lo > CC src/util/mem/lib_libmpi_la-safestr.lo > CC src/util/mem/lib_libmpi_la-argstr.lo > CC src/util/mem/lib_libmpi_la-strerror.lo > CC src/util/other/lib_libmpi_la-outliers.lo > CC src/util/other/lib_libmpi_la-assert.lo > ( cd /home/matthew/mpich-3.2 && /home/matthew/mpich- > 3.2/maint/extractcvars --dirs="`cat /home/matthew/mpich- > 3.2/maint/cvardirs`") > Error: open directory /tmp/wB2927n3sW/mpich-3.2/maint/../src/mpi -- > No such file or directory > make[2]: *** [/home/matthew/mpich-3.2/src/util/cvar/mpich_cvars.c] > Error 2 > make[2]: Leaving directory `/home/matthew/mpich-build' > make[1]: *** [all-recursive] Error 1 > make[1]: Leaving directory `/home/matthew/mpich-build' > make: *** [all] Error 2 > Is this NFS? Is the clock skewed? I thought extracting cvars was not done in released versions. ==rob _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0