Newsgroups: comp.parallel.mpi From: "Umesh Kumar V. Rajasekaran" Subject: Re: MPICH on Red Hat Linux, peculiar behaviour Organization: ECECS Department, University of Cincinnaiti Date: Tue, 09 Jun 1998 03:45:25 +0000 Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <357CAFD5.EA9B93FC@ececs.uc.edu> anand@my-dejanews.com wrote: > In article <357851D9.646866A2@ececs.uc.edu>, > "Umesh Kumar V. Rajasekaran" wrote: > > > Also another behaviour was that, for message sizes 128, 256, 512, the time of > > MPI_Send call is increasing, but after that, for 1024, it suddenly dips around > 1/4th > > of the other time. and then starts increasing for sizes 2048, 4096. > > The standard send mode can use any protocol of communication, may be > depending on the message size, network, etc. The implementor decides how it > should be done based on these factors that could affect performance. MPICH > uses "short", "eager", "rendevous" and may be "get" protocols depending on > the length of the message. See the following web pages for a detailed > description: > > http://www.mcs.anl.gov/mpi/mpich/workingnote/adi2impl/node2.html#Node2 > http://www.mcs.anl.gov/mpi/mpich/workingnote/adi2impl/node15.html#Node15 > > This could very well be what is showing up in your timing results. But this problem does not occur on the Sun Sparcs. thanks, Umesh p.s. dei lenin, chumma lifele MPI vechundu pill udathe - R and L. -- Umesh Kumar V. Rajasekaran Department of Electrical and Computer Science and Computer Engineering University of Cincinnati email: urajasek@ececs.uc.edu URL: http://www.ececs.uc.edu/~urajasek/