Newsgroups: comp.parallel.pvm,comp.parallel.mpi From: gleeson@erc.msstate.edu (Matthew P. Gleeson) Subject: Re: PVM 2 MPI? Organization: Mississippi State University Date: 29 Apr 1998 08:58:10 -0500 Message-ID: <3saf94d91p.fsf@Titan.ERC.MsState.Edu> There was a project here at Mississippi State a few years ago called "Unify" that let you transition your PVM code to MPI. It might be what you're looking for. It's still available on our ftp server at ftp.erc.msstate.edu/unify. I found the original announcement, inserting it below. Newsgroups: comp.parallel,comp.parallel.pvm From: tony@aurora.cs.msstate.edu (Tony Skjellum) Subject: UNIFY System 0.9.1 Available Summary: UNIFY supports MPI calls in the PVM environment Keywords: MPI, PVM, Message Passing, Portability Organization: Mississippi State University Date: Wed, 29 Jun 1994 12:53:06 GMT Message-ID: Availability of UNIFY Version 0.9.1 The UNIFY system, which patches PVM 3.2, 3.2.6, or 3.3, provides a subset of MPI within the PVM environment, without sacrificing the PVM calls already available. This system currently works with the network versions of PVM, and has been tested with Sun and SGI platforms. It is easily installed. The motivations for UNIFY are as follows: * Permit an easy learning curve for MPI to existing PVM users * Provide a dual-API mechanism for mixed PVM and MPI applications * Aid with the migration of codes from PVM syntax to MPI syntax * Provide a testbed for further MPI development UNIFY is based on completely independent MPI code. Version 0.9.1 provides a subset. Subsequent version will provide the entire MPI specification, and make use of the Argonne/Mississippi State implementation. Currently, UNIFY has the following limits: * SPMD model only * Static process creation * C Interface only Who will use UNIFY? * Anyone who wants to learn MPI without reading the entire standards document * Anyone who wants to migrate applications to MPI * Anyone who wants to learn parallel programming, and is interested in comparing strategies for message-passing syntax and semantics We will aggressively update Unify as users send their requests, bug reports, and suggestions. The following mailing addresses may be used: unify@erc.msstate.edu (suggestions, questions) unify-bugs@erc.msstate.edu (problem reports) The software is available free, for all uses, from the following ftp site: site: ftp.erc.msstate.edu directory: unify/ Read the file "README" to get started. UNIFY comes with a tutorial manual that explains the MPI calls that it supports, explains installation, and provides a number of figures to help with intuition about message-passing. Sincerely, Paula Vaughan Anthony Skjellum Donna Reese NSF Engineering Research Center for Computational Field Simulation Mississippi State University bjo@wins.uva.nl (Benno J. Overeinder) writes: > "William T. Rankin" writes: > > >This may be wild goose chase, but I was wondering if anyone has > >implemented a "translation library" that implements simple PVM3 > >calls but uses MPI primitives on the back end? > > >My research applications utilize PVM (been using it since '92) > >but I would like to get a quick/dirty first cut at an MPI port. > > We have played with the idea, but after a second look rejected the idea. > The idea is not bad, but if you want to do it right, it is quite complicated. > In particular the way PVM allows you to play around with buffers, makes > it very difficult to translate "directly" into MPI calls. One has to > "so-to-speak" compile parts of PVM to figure out what is intended to be > packed into a buffer and send to its destination. > > Ciao, > > -- Benno > -- > Benno Overeinder Department of Computer Science > voice: (+31) 20 525 7536 University of Amsterdam > fax: (+31) 20 525 7490 Kruislaan 403, 1098 SJ Amsterdam > e-mail: bjo@wins.uva.nl The Netherlands -- Matt Gleeson http://www.erc.msstate.edu/~gleeson/