Am Montag, den 12.05.2008, 14:22 +0100 schrieb Richard Jones:
> This is just barely faster than Jon's OCaml version using message
> passing (12% faster on my test machine[0]).  Which just seems to show
> that the overhead of message passing _isn't_ the problem here[1].

I've just written my own distributed version. You find my comments and
timings here:

http://blog.camlcity.org/blog/parallelmm.html

The code is here:

https://godirepo.camlcity.org/svn/lib-ocamlnet2/trunk/code/examples/rpc/matrixmult/

In this (very unoptimized) multiplier message passing accounts for ~25%
of the runtime. Even for 2 cores there is already a speedup. 10 cores
(over a network) are about 4 times faster than a single core without
message passing.

Gerd

> Perhaps it's the bounds checking in the assignment back to the matrix?
> 
> Anyhow, in real life situations we'd all be using a super-optimized
> hand-coded-in-assembly matrix multiplication library (LAPACK?), so
> this is all very academic.
> 
> Rich.
> 
> [0] Quad core Intel hardware:
> model name      : Intel(R) Core(TM)2 Quad  CPU   Q9450  @ 2.66GHz
> 
> [1] Creation of the result matrix and copying it to shared memory is
> almost instantaneous in my tests.
> 
-- 
------------------------------------------------------------
Gerd Stolpmann * Viktoriastr. 45 * 64293 Darmstadt * Germany 
gerd@gerd-stolpmann.de          http://www.gerd-stolpmann.de
Phone: +49-6151-153855                  Fax: +49-6151-997714
------------------------------------------------------------