From: Lars Rzymianowicz Newsgroups: comp.parallel.pvm,comp.parallel.mpi Subject: Re: MPI vs. PVM: What to choose? Date: Thu, 18 Mar 1999 08:26:16 +0100 Organization: Dept. of Computer Engineering, University of Mannheim, Germany Message-Id: <36F0AA98.F8194DA1@mufasa.informatik.uni-mannheim.de> References: <36EFAD85.EB4ABB4@postech.ac.kr> <36F00274.1D2880C5@scl.ameslab.gov> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Xref: ukc comp.parallel.pvm:8141 comp.parallel.mpi:4774 Vasiliu Bogdan wrote: > I think you should start with PVM because it is simpler (or at least > it was at the time when I learned it) and later move to MPI. Hmm, there is a subset of MPI (6 functions: init, exit, send, recv, bcast, barrier) which is enough for the first programs. A lot of extended stuff is for mathematics (form a grid, shift rows of data etc.) > MPI is intended to be a standard in parallel computing... This is an important point. In fact, the PVM people helped to define the MPI standard. MPI tries to gather the best from all message passing libs. (but this need not to be the best choice...the MPI-2 standard doc is about 360 pages. That might be a reason, why a full implementation of MPI-2 is not available ;-) > I believe people will keep developing the MPI standard, > whereas the PVM project has been stopped. Yip, another good point. Currently, PVM is a stable MP lib, but if you take a look into the future, there is only MPI left... BTW, the LAM-MPI implementation has dynamic process creation, if this is a must for you. Lars -- Homepage: http://mufasa.informatik.uni-mannheim.de/lsra/persons/lars/