<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META http-equiv=Content-Type content="text/html; charset=iso-8859-1">
<META content="MSHTML 6.00.2723.2500" name=GENERATOR>
<STYLE></STYLE>
</HEAD>
<BODY bgColor=#ffffff>
<DIV><FONT face=Arial>Good morning,<BR>I am using a cluster with the following
charactestics:<BR><BR>BEOWULF <BR>MPI OSCAR 3.0<BR>RED HAT 9<BR>11 NODES: 1 NODE
PRINCIPAL<BR> 10 SUBNODES:
EACH SUBNODE HAS 2 PROCESSOR P4 XEON 2.8
GIGA<BR> NO
IPERTRADING<BR> 2 giga RAM
DUAL CHANNEL<BR> EACH
SUBNODE HAS 200 GIGA: EIDE 7200, 8 MEGA BUFFER<BR><BR>I want to use this cluster
to test the scalability of a numerical code<BR>using 1, 2, 4, 8 processors. For
example I would like to test the code<BR>with 4 processors but I am not able to
force it to use 4 subnodes<BR>(that is one processor for each subnode) instead
of 2 subnodes (2<BR>processor for each subnode) as the cluster does. In this way
the<BR>cluster creates local conflict and memory sharing problems in
each<BR>subnode, decreasing the code performances<BR>This is surprising for me
since I use the following script to run the<BR>simulations in which the line 3
prescribes the use of a processor for<BR>each subnode if it
possible<BR>
<BR>#!/bin/csh<BR>#PBS -m e<BR>#PBS -l nodes=4:ppn=1<BR>#PBS -l
walltime=9999:00:00<BR>#PBS -M user@domain<BR>#PBS -j oe<BR>#PBS -o
rb.out<BR>#PBS -N rb<BR>#PBS<BR>limit coredumpsize 0<BR>set NN = `cat
$PBS_NODEFILE | wc -l`<BR>echo "NN = "$NN<BR>#cd $PBS_O_WORKDIR<BR>cd
/home/antonio/test_paper_numerico/RB1E5<BR>pwd<BR>cat $PBS_NODEFILE >
newlist<BR>date<BR>time mpirun -machinefile newlist -np $NN rb >
nav2.log<BR>date<BR><BR>I hope that someone can helps
me<BR>Ciao<BR>Antonio<BR></FONT></DIV></BODY></HTML>