Re: Handling "parfor" as "for"

[Jordi Gutiérrez Hermoso]

On 28 September 2011 21:49, Mark Everitt <address@hidden>
wrote:
> I guess what would be really nice is a true implementation of
> parfor. I'm new to octave, and it seems rather large and complex.
> Can someone point me to where the keywords are actually defined? I
> get lost in the network of files, and I have a short attention
> span...

Yes, getting into Octave is a mess... It's a large and complex
codebase. We have a HACKING file under etc/ that gives broad
description of what's in the code. Start reading that first. I just
added a few things to it, btw:

    http://hg.savannah.gnu.org/hgweb/octave/rev/00c5d8d3ee00

Have you managed to compile from a Mercurial clone first? I would say
that's the first prerequisite. I can help with that. I still haven't
understood the guts of how the interpreter is made.

Note that there's a rather large problem with parfor in that Octave is
horribly non-thread safe, because it was never written with
parallelisation in mind (lots of global state...). A parfor
implementation would probably have to fork processes instead.

> Maybe I can put some code together for this. No guarantees though;
> my coding ability has never been stretched beyond small simulations
> and I don't have a lot of free time. But, I have been looking for an
> open source project to cut my teeth on, and I can think of no
> worthier project than Octave.

Well, you have the right attitude, so you're already halfway there. If
you want to learn, you'll learn. I am quite happy to see someone as
enthusiastic as you stepping up.

Welcome!
- Jordi G. H.