On Dec 13, 2010, at 6:08 PM, David Fiala wrote: > I noticed that I can get the openib transport to work successfully > under version 1.4.3 when configured with: --with-openib > --enable-openib-ibcm
I'm surprised; the IBCM connection manager support in OMPI is not complete. You probably shouldn't be using it. > When I configure 1.5 or 1.5.1 I used: --with-openib (noting the > absence of the ibmc flag) Good. > However, when I actually try to use openib on a basic MPI program I > get a segfault such as the one copied below. > > Our IB hardware is: > InfiniBand: Mellanox Technologies MT26428 [ConnectX VPI PCIe 2.0 5GT/s > - IB QDR / 10GigE] (rev b0) Ick. Mellanox, can you reply? > dfiala@compute-0-2 ~]$ mpirun -mca btl openib,self ./mpitest/mpitest > [compute-0-2:07582] *** Process received signal *** > [compute-0-2:07582] Signal: Segmentation fault (11) > [compute-0-2:07582] Signal code: Address not mapped (1) > [compute-0-2:07582] Failing at address: 0x2 > [compute-0-2:07582] [ 0] /lib64/libpthread.so.0 [0x3ed2e0eb10] > [compute-0-2:07582] [ 1] /usr/lib64/libmlx4-rdmav2.so [0x2aaaab0de5d1] > [compute-0-2:07582] [ 2] > /home/dfiala/openmpi/install-1.5.1/lib/openmpi/mca_btl_openib.so > [0x2b1637155f15] > [compute-0-2:07582] [ 3] > /home/dfiala/openmpi/install-1.5.1/lib/openmpi/mca_bml_r2.so > [0x2b163691b4b2] > [compute-0-2:07582] [ 4] > /home/dfiala/openmpi/install-1.5.1/lib/openmpi/mca_pml_ob1.so > [0x2b1636d3844f] > [compute-0-2:07582] [ 5] /home/dfiala/openmpi/install/lib/libmpi.so.1 > [0x2b16347afe37] > [compute-0-2:07582] [ 6] > /home/dfiala/openmpi/install/lib/libmpi.so.1(MPI_Init+0xf0) > [0x2b16347c46d0] > [compute-0-2:07582] [ 7] ./mpitest/mpitest(main+0x2b) [0x4008d3] > [compute-0-2:07582] [ 8] /lib64/libc.so.6(__libc_start_main+0xf4) > [0x3ed261d994] > [compute-0-2:07582] [ 9] ./mpitest/mpitest [0x4007f9] > [compute-0-2:07582] *** End of error message *** > -------------------------------------------------------------------------- > mpirun noticed that process rank 0 with PID 7582 on node > compute-0-2.local exited on signal 11 (Segmentation fault). > -------------------------------------------------------------------------- > > Any ideas? > > Thanks for your help, > David Fiala > > North Carolina State University > _______________________________________________ > devel mailing list > de...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/devel -- Jeff Squyres jsquy...@cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
