I am using Intel MPI 3.1 build 26 with OFED 1.3 beta2 on RHEL4 x86_64. Intel MPI works as before with dapl1, but I see very slow performance with dapl2. Are there extra command-line params I need to use dapl2? For example: $ /data/software/qa/MPI/intel_mpi/intelmpi-3.1-`uname -m`/bin/mpiexec -genv I_MPI_DEBUG 3 -n 2 osu_latency.x [1] MPI startup(): DAPL provider <NULL on rank 0:svbu-qa1850-1 differs from <NUL L string> on rank 1:svbu-qa1850-2 [0] MPI startup(): socket data transfer mode [0] MPI Startup(): process is pinned to CPU00 on node svbu-qa1850-1 [1] MPI startup(): socket data transfer mode [1] MPI Startup(): process is pinned to CPU00 on node svbu-qa1850-2 [0] Rank Pid Node name [0] 0 9605 svbu-qa1850-1 [0] 1 8547 svbu-qa1850-2 # OSU MPI Latency Test (Version 2.1) # Size Latency (us) 0 124.96 1 124.98 2 99.62 4 63.04 8 62.98
Scott Weitzenkamp SQA and Release Manager Server Virtualization Business Unit Cisco Systems
_______________________________________________ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg