Talk:Message Passing Interface

{{Talk header}}

{{WikiProject banner shell|class=C|

{{WikiProject Computing |importance=Low |software=yes |software-importance=Low |free-software=yes |free-software-importance=Low}}

}}

Implementations

I started the "Implementations" section. My goal for it was for programmers looking to get involved in nuts and bolts cluster computing to look up the languages (they already knew or that maybe they should think of learning), determine which langauges have one or more good implementations, and choose their language and MPI implementation based on the NPOV information presented. That's the "necessary and sufficient information" I'd see this section containing in "finished form". Not that it's binding or anything :-P I just thought I'd throw out a "reasonable goal" to maybe inspire some cool edits. - JenniferForUnity 02:15, 2 November 2006 (UTC)

:There, I have been working heavily on the Java version because of my current involvement with it. I also added a "Funcionality overview" section before the implementation one, since the article doesn't explain well what MPI is supposed to do. And no, "handling communication between nodes of computers" is not enough, so I tried hinting at a few examples like reduction/gathering and barriers. EpiVictor 17:34, 5 November 2006 (UTC)

Bad article

Needs pedagogical cleanup, preferably from Quinn's book. This will take time. Help plz. Khazadum 05:05, 12 February 2007 (UTC)

:Trying to help, but what's a Quinn's book? --Earin (talk) 13:17, 25 June 2008 (UTC)

Michael J. Quinn,“Parallel Programming in C with MPI and OpenMP”, Copyright 2004, McGraw-Hill. That's the book I believe they are talking about. 147.133.34.167 (talk) 15:52, 18 October 2010 (UTC)

Has this issue been resolved ?

"Concepts"

These seem a little biased towards the task-farmy end of MPI, and a little too jargonised for an introduction. Perhaps some simplification is in order? --Earin (talk) 13:17, 25 June 2008 (UTC)

Conflicting information

Both Parallel Virtual Machine and Message Passing Interface have conflicting information, and blatant POV.

From Parallel Virtual Machine as of [http://en.wikipedia.org/w/index.php?title=Parallel_Virtual_Machine&oldid=106094982 February 6, 2007]:

: PVM enables users to exploit their existing computer hardware to solve much larger problems at minimal additional cost. Hundreds of sites around the world are using PVM to solve important scientific, industrial, and medical problems in addition to PVM's use as an educational tool to teach parallel programming. With tens of thousands of users, PVM has become the de facto standard for distributed computing world-wide.

From Message Passing Interface as of [http://en.wikipedia.org/w/index.php?title=Message_Passing_Interface&oldid=109292397 February 6, 2007]:

: MPI is a de facto standard for communication among the processes modeling a parallel program on a distributed memory system. Often these programs are mapped to clusters, actual distributed memory supercomputers, and to other environments. However, the principal MPI-1 model has no shared memory concept, and MPI-2 has only a limited distributed shared memory concept used in one portion of that set of extensions.

Since both PVM and MPI are used for distributed parallel applications, more or less for the same purpose, but they are incompatible with each other, it is strange that both are de facto standard.

--Juliano (T) 00:47, 21 February 2007 (UTC)

: the fact remains though, that they both have widespread use, both are a de facto standard - one does not preclude the other. --Ear1grey 03:27, 21 March 2007 (UTC)

163.1.125.49 (talk) 11:24, 31 October 2008 (UTC) : PVM is fringe technology. Sorry, but it's true. In 4 years in parallel computing, I've only ever received one request for PVM-based software, out of hundreds of applications across tens of different disciplines. The PVM page is 101% wrong with the completely slanted words "PVM has become the de facto standard for distributed computing world-wide"

The claim that PVM is a de facto standard was removed by an anonymous editor on 29 November 2007. No reason appears to have been given.

JamesBWatson (talk) 20:43, 31 October 2008 (UTC)

CLI Pure MPI.Net

The text describing this implementation looks like it as been copied from the commercial webpage of this product, and thus makes a very positive presentation of it : "[it] is powerful, yet easy to use for parallel programming" and "[allows] configuration for your environment and performance needs".

I propose to rewrite this paragraph in a more neutral way, citing that many implementation of MPI exists, some of them being commercial.

un_brice (talk) 01:06, 1 June 2009 (UTC)

Commented out inline editor's note

"(Obs: TO FIX: a reference must be included, because at sc94 site it is not possible to find references to support this statement: http://sc94.ameslab.gov/, http://sc94.ameslab.gov/AP/contents.html , http://www.pubzone.org/pages/publications/showVenue.do?venueId=13270)"

I have commented this out. LokiClock (talk) 05:01, 11 February 2010 (UTC)

Copy-edits

{{GOCE}}

Feedback appreciated. Lfstevens (talk) 17:09, 10 July 2010 (UTC)

:I am sorry, but why bother copyediting this list of errors, pretending to be an article. I do not have time to work on this at all, but needs another flag that says "do not read this". It needs a rewrite. History2007 (talk) 15:14, 17 August 2011 (UTC)

Is the word "Implementation" correct?

I'm not sure if the title of the section "Implementations" is correct. What is the difference between an "implementation" and "bindings?" For example, mpi4py is listed on the page as an Python implementation, but the developers call it "MPI bindings." Indeed, when you install mpi4py, you have to install in on top of an implementation such as MPICH or openMPI. What, then, makes mpi4py different from something like openMPI or MPICH? What about the others in this section? Blossomonte (talk) 00:31, 15 May 2012 (UTC)

:: In a loose sense, bindings can be thought of as a special kind of implementation that relies on an existing implementation to work. It translates one interface into another, essentially. However, the official MPI standard is defined only in terms of C and Fortran interfaces, so these are technically speaking not implementations of the official MPI standard, but unofficial extensions to the standard. --Fylwind (talk) 20:06, 22 April 2016 (UTC)