wrong GM board number

[Garrick Staples]

Typical line of output from mpiexec -v:

read_gm_startup_ports: id 198 port 5 board 96 gm_node_id 0xdd483118
  numanode 0 pid 29494 remote_port 34172

board number 96?


The mpirun perl script seems to disagree slightly with the boot
parameters from the nodes:

      if ($incoming_data !~ /^<<<(\d+):(\d+):(\d+):(\d+):(\d+):(\d+):(\d+)::(\d+)>>>$/) {
        warn ("Received invalid data format !\n");
      ...
      $port_board_ids[$2] = $3;
      $unique_high_ids[$2] = $4;
      $unique_low_ids[$2] = $5;
      $numanodes[$2] = $6;
      $remote_pids[$2] = $7;
      $remote_ports[$2] = $8;
      ...
      printf "MPI Id %d is using GM port %d, board %d (MAC %04x%06x).\n",
          $2, ($3 & 0xFFFF), ($3 >> 16), $4, $5;


Compared to the equivalent mpiexec code:
        if (sscanf(s, "<<<%d:%d:%d:%d:%u:%d:%d::%d>>>", &magic, &id,
          &port, &board, &node, &numanode, &pid, &remote_port) == 8) {
        ...
        printf("%s: id %d port %d board %d gm_node_id 0x%08x\n",
          __func__, id, port, board, node);
        printf("  numanode %d pid %5d remote_port %5d\n",
          numanode, pid, remote_port);


Note the perl script combines the port and board info in the 3rd field.
I haven't a clue what "high_ids" and "low_ids" mean.


-- 
Garrick Staples, Linux/HPCC Administrator
University of Southern California
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://email.osc.edu/pipermail/mpiexec/attachments/20051110/78b5ceb8/attachment.bin