From: vfl@angewandte-geophysik.ruhr-uni-bochum.de (Tony Stoecker)
Newsgroups: comp.parallel.mpi
Subject: LAM question: what does mpirun -c2c ?
Date: 17 May 1999 12:28:06 GMT
Organization: Ruhr-Universitaet Bochum, Rechenzentrum
Message-Id: <7hp20m$gat$1@sunu789.rz.ruhr-uni-bochum.de>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Xref: ukc comp.parallel.mpi:5092


Hi ,

I just tested my first MPI program on our new 5 Node Dual-PentiumII Cluster
running Linux-2.2.3 with 100Mbit-Ethernet.
The program should benchmark the system. It is a master-slave application -
the slaves wait x seconds and then send back y bytes and the master
measures transfer-time and total-process-time for 10 tasks. This happens
for some (x,y) (x=2^n, n=10,11,...,21) .

I started the program on the server with

>>  mpirun -O appschema 

and the appschema file is

#
master h
slave  n1-5 -c 10 -s h

. The program executed without erros and then i tried

>> mpirun -O -c2c appschema

to reduce the transfer-times. (but I really did not understand
why this option should lead to much improved performance, as
it is stated in the mpirun-manpage.)
The following happened:
The transfer-time are little smaller for x <= 2^15 byte , but 
when each slave send back >= 2^16 byte the transfer-time
increased strongly.  (factor 13 larger than without the -c2c option)

Can anybody please help me what is going on?
Maybe I have to adjust mpi-packet-sizes or something like that?

please email me.
Thanks,
__
--------------------------------------------------------- 
 Dipl.-Geophys. Tony Stoecker 
 Institut fuer Geophysik 
 Ruhr-Universitaet Bochum 
 44780 Bochum  
 Tel.: ++49-234-700-3197 
 email: vfl@angewandte-geophysik.ruhr-uni-bochum.de
---------------------------------------------------------