Regarding review of the coll/ml fix: While the fix Gilles worked out overnight proved sufficient on Solaris/SPARC, Linux/PPC64 and Linux/IA64, I had two concerns:
1) As I already voiced on the list, I am concerned with the portability of _SC_PAGESIZE vs _SC_PAGE_SIZE (vs get_pagesize()). 2) Though I have not tried to trace the code, the fact that fixing the alignment prevents a SEGV strongly suggests that there was a mmap (or something else sensitive to page size) call failing. So, there should probably be a check added for failure of that call to produce a cleaner failure than SEGV. Just my USD 0.02. -Paul On Fri, Aug 1, 2014 at 6:39 AM, Ralph Castain <r...@open-mpi.org> wrote: > Okay, I fixed those two and will release rc4 once the coll/ml fix has been > reviewed. Thanks > > On Aug 1, 2014, at 2:46 AM, Mike Dubman <mi...@dev.mellanox.co.il> wrote: > > Also, latest commit into openib (origin/v1.8 > https://svn.open-mpi.org/trac/ompi/changeset/32391) broke something: > > *11:45:01* + timeout -s SIGSEGV 3m > /scrap/jenkins/workspace/OMPI-vendor/label/hpctest/ompi_install1/bin/mpirun > -np 8 -mca pml ob1 -mca btl self,openib > /scrap/jenkins/workspace/OMPI-vendor/label/hpctest/ompi_install1/examples/hello_usempi*11:45:01* > > --------------------------------------------------------------------------*11:45:01* > WARNING: There are more than one active ports on host 'hpctest', but > the*11:45:01* default subnet GID prefix was detected on more than one of > these*11:45:01* ports. If these ports are connected to different physical > IB*11:45:01* networks, this configuration will fail in Open MPI. This > version of*11:45:01* Open MPI requires that every physically separate IB > subnet that is*11:45:01* used between connected MPI processes must have > different subnet ID*11:45:01* values.*11:45:01* *11:45:01* Please see this > FAQ entry for more details:*11:45:01* *11:45:01* > http://www.open-mpi.org/faq/?category=openfabrics#ofa-default-subnet-gid*11:45:01* > *11:45:01* NOTE: You can turn off this warning by setting the MCA > parameter*11:45:01* btl_openib_warn_default_gid_prefix to 0.*11:45:01* > --------------------------------------------------------------------------*11:45:01* > > --------------------------------------------------------------------------*11:45:01* > WARNING: No queue pairs were defined in the > btl_openib_receive_queues*11:45:01* MCA parameter. At least one queue pair > must be defined. The*11:45:01* OpenFabrics (openib) BTL will therefore be > deactivated for this run.*11:45:01* *11:45:01* Local host: > hpctest*11:45:01* > --------------------------------------------------------------------------*11:45:01* > > --------------------------------------------------------------------------*11:45:01* > At least one pair of MPI processes are unable to reach each other > for*11:45:01* MPI communications. This means that no Open MPI device has > indicated*11:45:01* that it can be used to communicate between these > processes. This is*11:45:01* an error; Open MPI requires that all MPI > processes be able to reach*11:45:01* each other. This error can sometimes be > the result of forgetting to*11:45:01* specify the "self" BTL.*11:45:01* > *11:45:01* Process 1 ([[55281,1],1]) is on host: hpctest*11:45:01* > Process 2 ([[55281,1],0]) is on host: hpctest*11:45:01* BTLs attempted: > self*11:45:01* *11:45:01* Your MPI job is now going to abort; > sorry.*11:45:01* > --------------------------------------------------------------------------*11:45:01* > > --------------------------------------------------------------------------*11:45:01* > MPI_INIT has failed because at least one MPI process is > unreachable*11:45:01* from another. This *usually* means that an underlying > communication*11:45:01* plugin -- such as a BTL or an MTL -- has either not > loaded or not*11:45:01* allowed itself to be used. Your MPI job will now > abort.*11:45:01* *11:45:01* You may wish to try to narrow down the > problem;*11:45:01* *11:45:01* * Check the output of ompi_info to see which > BTL/MTL plugins are*11:45:01* available.*11:45:01* * Run your application > with MPI_THREAD_SINGLE.*11:45:01* * Set the MCA parameter btl_base_verbose > to 100 (or mtl_base_verbose,*11:45:01* if using MTL-based communications) > to see exactly which*11:45:01* communication plugins were considered > and/or discarded.*11:45:01* > --------------------------------------------------------------------------*11:45:01* > *** An error occurred in MPI_Init*11:45:01* *** on a NULL > communicator*11:45:01* *** MPI_ERRORS_ARE_FATAL (processes in this > communicator will now abort,*11:45:01* *** and potentially your MPI > job)*11:45:01* [hpctest:2761] Local abort before MPI_INIT completed > successfully; not able to aggregate error messages, and not able to guarantee > that all other processes were killed!*11:45:01* *** An error occurred in > MPI_Init*11:45:01* *** on a NULL communicator*11:45:01* *** > MPI_ERRORS_ARE_FATAL (processes in this communicator will now > abort,*11:45:01* *** and potentially your MPI job)*11:45:01* > [hpctest:2757] Local abort before MPI_INIT completed successfully; not able > to aggregate error messages, and not able to guarantee that all other > processes were killed!*11:45:01* *** An error occurred in MPI_Init*11:45:01* > *** on a NULL communicator*11:45:01* *** MPI_ERRORS_ARE_FATAL (processes in > this communicator will now abort,*11:45:01* *** and potentially your MPI > job)*11:45:01* [hpctest:2751] Local abort before MPI_INIT completed > successfully; not able to aggregate error messages, and not able to guarantee > that all other processes were killed!*11:45:01* *** An error occurred in > MPI_Init*11:45:01* *** on a NULL communicator*11:45:01* *** > MPI_ERRORS_ARE_FATAL (processes in this communicator will now > abort,*11:45:01* *** and potentially your MPI job)*11:45:01* > [hpctest:2752] Local abort before MPI_INIT completed successfully; not able > to aggregate error messages, and not able to guarantee that all other > processes were killed!*11:45:01* *** An error occurred in MPI_Init*11:45:01* > *** on a NULL communicator*11:45:01* *** MPI_ERRORS_ARE_FATAL (processes in > this communicator will now abort,*11:45:01* *** and potentially your MPI > job)*11:45:01* [hpctest:2753] Local abort before MPI_INIT completed > successfully; not able to aggregate error messages, and not able to guarantee > that all other processes were killed!*11:45:01* *** An error occurred in > MPI_Init*11:45:01* *** on a NULL communicator*11:45:01* *** > MPI_ERRORS_ARE_FATAL (processes in this communicator will now > abort,*11:45:01* *** and potentially your MPI job)*11:45:01* > [hpctest:2755] Local abort before MPI_INIT completed successfully; not able > to aggregate error messages, and not able to guarantee that all other > processes were killed!*11:45:01* *** An error occurred in MPI_Init*11:45:01* > *** on a NULL communicator*11:45:01* *** MPI_ERRORS_ARE_FATAL (processes in > this communicator will now abort,*11:45:01* *** and potentially your MPI > job)*11:45:01* [hpctest:2759] Local abort before MPI_INIT completed > successfully; not able to aggregate error messages, and not able to guarantee > that all other processes were killed!*11:45:01* *** An error occurred in > MPI_Init*11:45:01* *** on a NULL communicator*11:45:01* *** > MPI_ERRORS_ARE_FATAL (processes in this communicator will now > abort,*11:45:01* *** and potentially your MPI job)*11:45:01* > [hpctest:2763] Local abort before MPI_INIT completed successfully; not able > to aggregate error messages, and not able to guarantee that all other > processes were killed! > > > > On Fri, Aug 1, 2014 at 11:00 AM, Paul Hargrove <phhargr...@lbl.gov> wrote: > >> Note that the Solaris unresolved alloca problem George fixed in r32388 is >> still present in 1.8.2rc3. >> I have manually confirmed that the same patch resolves the problem in >> 1.8.2rc3. >> >> -Paul >> >> >> On Thu, Jul 31, 2014 at 9:44 PM, Ralph Castain <r...@open-mpi.org> wrote: >> >>> Usual place - this is a last-chance check, so please hit it. Main change >>> from rc2 is the repairs to the Fortran binding config logic >>> >>> http://www.open-mpi.org/software/ompi/v1.8/ >>> >>> Ralph >>> >>> _______________________________________________ >>> devel mailing list >>> de...@open-mpi.org >>> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel >>> Link to this post: >>> http://www.open-mpi.org/community/lists/devel/2014/08/15433.php >>> >> >> >> >> -- >> Paul H. Hargrove phhargr...@lbl.gov >> Future Technologies Group >> Computer and Data Sciences Department Tel: +1-510-495-2352 >> Lawrence Berkeley National Laboratory Fax: +1-510-486-6900 >> >> _______________________________________________ >> devel mailing list >> de...@open-mpi.org >> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel >> Link to this post: >> http://www.open-mpi.org/community/lists/devel/2014/08/15440.php >> > > _______________________________________________ > devel mailing list > de...@open-mpi.org > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel > Link to this post: > http://www.open-mpi.org/community/lists/devel/2014/08/15444.php > > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel > Link to this post: > http://www.open-mpi.org/community/lists/devel/2014/08/15449.php > -- Paul H. Hargrove phhargr...@lbl.gov Future Technologies Group Computer and Data Sciences Department Tel: +1-510-495-2352 Lawrence Berkeley National Laboratory Fax: +1-510-486-6900
