From: Axel vom Endt <Axel.vom.Endt@linmpi.mpg.de>
Newsgroups: comp.parallel.mpi
Subject: Re: Why my MPICH/LAM under linux give low float point performance
Date: Fri, 28 Aug 1998 12:37:51 +0200
Organization: MPI f. Aeronomie
Message-Id: <35E6887F.B986CD83@linmpi.mpg.de>
References: <6s3nt4$8kd$1@nnrp1.dejanews.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit


Hi,

could you please post your plain version of your little example? I tried
your program and an obviuos replacement of MPI_Wtime by gettimeofday.
Compiling with 

gcc -o mpi_speed.mpi.x -O5 -I/usr/local/mpi/include/ mpi_speed.mpi.c
-L/usr/local/mpi/lib/LINUX/ch_p4/ -lmpi

and 

gcc -o mpi_speed.plain.x -O5 mpi_speed.plain.c

gave the following results:

12:28> time
./mpi_speed.plain.x                                              The my
final e is 2.718282
wall clock time = 3.358648
3.110u 0.000s 0:03.37 92.2%     0+0k 0+0io 60pf+0w

12:28> time ./mpi_speed.mpi.x
The my final e is 2.718282
wall clock time = 3.448278
3.090u 0.010s 0:03.48 89.0%     0+0k 0+0io 106pf+0w

12:28> time mpirun -np 1 ./mpi_speed.mpi.x
The my final e is 2.718282
wall clock time = 3.361181
3.600u 0.370s 0:04.67 85.0%     0+0k 0+0io 2213pf+0w

Pretty much the same time.

Axel

-- 
  Axel vom Endt                      Office: A2_95
  Max-Planck-Institut f. Aeronomie   Phone:  +49-(0)5556-979-481
  Postfach 20                        Fax:    +49-(0)5556-979-240
  D-37191 Katlenburg-Lindau          Email:  endt@linmpi.mpg.de