[Beowulf] MPI & ScaLAPACK: error in MPI_Comm_size: Invalid communicator
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Gabb, Henry henry.gabb at intel.comTue Oct 19 14:21:31 PDT 2004
- Previous message: [Beowulf] Bonding results in 1 HBA for Tx and 1 HBA for Rx.
- Next message: [Beowulf] dual Opteron recommendations
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Hi Clint, The "Null Comm pointer" error that you're seeing is almost always due to a header mismatch. I don't think Intel Cluster MKL 7.0 supports MPICH2 yet. According to the Intel Cluster MKL system requirements (http://www.intel.com/software/products/clustermkl/sysreq.htm), MPICH-1.2.5 is supported. MPICH-1.2.6 should work too. You might post a question to the Intel MKL forum (http://softwareforums.intel.com/ids) about MPICH2 support. I did a quick test of your program on one of my clusters and it ran fine (after fixing the typo in the nocol=3 statement): [henry at castor1 cmkl-test]$ /opt/mpich-1.2.6-gcc/bin/mpif77 -o example example.f \ -L/opt/intel/mkl70cluster/lib/64 -lmkl_scalapack \ -lmkl_blacsF77init_gnu -lmkl_blacs -lmkl_blacsF77init_gnu \ -lmkl_lapack -lmkl -lguide -lpthread [henry at castor1 cmkl-test]$ /opt/mpich-1.2.6-gcc/bin/mpirun -n 6 ./example [henry at castor1 cmkl-test]$ I used a GNU-built MPICH-1.2.6 because I didn't have an Intel-ready MPICH installation handy. Best regards, Henry Gabb Intel Parallel and Distributed Solutions Division
- Previous message: [Beowulf] Bonding results in 1 HBA for Tx and 1 HBA for Rx.
- Next message: [Beowulf] dual Opteron recommendations
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
