Newsgroups: comp.parallel.mpi
From: Junglas@tu-harburg.d400.de (Peter Junglas)
Subject: Problem with ANL/MSU implementation of MPI on HP-UX
Organization: Technical University Hamburg-Harburg, Germany
Date: 8 Aug 1994 11:44:53 GMT
Message-ID: <3255rl$mva@rztsun.tu-harburg.de>

I tried to install the ANL/MSU implementation of MPI on a cluster of HP 9000/735 machines running HP-UX 9.01. I used the P4 device ("ch_p4) and enabled the shared memory option of P4. It works fine as long as I only use two machines (and a lot of processes on each). But if I try to use three or more machines, MPI_Init hangs somewhere in the P4 initialisation producing the following error messages:

  rm_17595:  p4_error: net_recv read:  EOF on socket: 5
  rm_l_2063809316_17596:  p4_error: interrupt SIGINT: 2
  p0_7996:  p4_error: net_recv read:  EOF on socket: 5
  bm_list_7997:  p4_error: interrupt SIGINT: 2
  rm_l_0_8058:  p4_error: interrupt SIGINT: 2
  p2_8057:  p4_error: interrupt SIGINT: 2

Has anybody seen a similar behaviour or even found a solution?

--
Peter Junglas                           Tel.:  040/7718-3193
TU Hamburg-Harburg,                     Fax.:  040/7718-2803
  Rechenzentrum                         email: Junglas@tu-harburg.d400.de
Denickestr. 17    
21071 Hamburg





