<?xml version="1.0" encoding="ISO-8859-1"?>

<!DOCTYPE message PUBLIC
  "-//MLarc//DTD MLarc output files//EN"
  "../../mlarc.dtd"[
  <!ATTLIST message
    listname CDATA #REQUIRED
    title CDATA #REQUIRED
  >
]>

  <?xml-stylesheet href="../../mlarc.xsl" type="text/xsl"?>


<message 
  url="2003/07/27aaf0216cdc64b28462ae2fb8011e62"
  from="Hal Daume III &lt;hdaume@I...&gt;"
  author="Hal Daume III"
  date="2003-07-15T16:02:40"
  subject="Re: [Caml-list] ocaml on supercomputer?"
  prev="2003/07/5fed1ff71a77f1003e3d332ed6614452"
  next="2003/07/9dd0cec623926fb5532d7122508b57a1"
  prev-in-thread="2003/07/5fed1ff71a77f1003e3d332ed6614452"
  next-in-thread="2003/07/155910c4eeb09e684f02ea4ae342873b"
  prev-thread="2003/07/090fba18d39b84b18e76264783dfeb3f"
  next-thread="2003/07/9dd0cec623926fb5532d7122508b57a1"
  root="../../"
  period="month"
  listname="caml-list"
  title="Archives of the Caml mailing list">

<thread subject="[Caml-list] ocaml on supercomputer?">
<msg 
  url="2003/07/94b1d15bcfe80899b726ba1495ec2b0f"
  from="Daniel Andor &lt;da209@c...&gt;"
  author="Daniel Andor"
  date="2003-07-15T14:31:26"
  subject="[Caml-list] ocaml on supercomputer?">
<msg 
  url="2003/07/5fed1ff71a77f1003e3d332ed6614452"
  from="Xavier Leroy &lt;xavier.leroy@i...&gt;"
  author="Xavier Leroy"
  date="2003-07-15T15:24:39"
  subject="Re: [Caml-list] ocaml on supercomputer?">
<msg 
  url="2003/07/27aaf0216cdc64b28462ae2fb8011e62"
  from="Hal Daume III &lt;hdaume@I...&gt;"
  author="Hal Daume III"
  date="2003-07-15T16:02:40"
  subject="Re: [Caml-list] ocaml on supercomputer?">
<msg 
  url="2003/07/155910c4eeb09e684f02ea4ae342873b"
  from="Yaron M. Minsky &lt;yminsky@c...&gt;"
  author="Yaron M. Minsky"
  date="2003-07-16T00:23:54"
  subject="Re: [Caml-list] ocaml on supercomputer?">
</msg>
</msg>
</msg>
</msg>
</thread>

<contents>
&gt; OCaml threads do not exploit shared-memory multiprocessing --
&gt; basically, all threads interleave their execution, hence multiple
&gt; processors are not exploited.  Your best bet is MPI or PVM.  Both have
&gt; bindings for OCaml.

I thought I'd just stick in my own props for the Ocaml MPI bindings.  I
had an ~11k line ocaml program which I ported to MPI.  It only required
500 or so more lines of code to MPI-ify it (with minor restructuring so
these parts could be separated out).

It now runs on our supercomputer (56th faster computer in the world,
according to top500.org) on anywhere from 10 to 170 nodes (IBM
Netfinity Cluster P4 Xeon 2 GHz - Myrinet/ 512), and performs incredibly
well.

I am *very* happing with these bindings and would like to formally thank
Xavier for writing them :).

Best of luck.

 - Hal



-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

</contents>

</message>

