Parallel Languages (was RE: [Algorithms] Multi core/cell processing and game algorithms...)

[George W. <ge...@ap...>]

On Wed, 6 Apr 2005 15:13:00 -0700, "Tony Cox" <to...@mi...> wrote:
> 2) How we got started on this thread initially was the observation that
> the new hardware platforms may introduce some new problems. What does
> *not* seem obvious to me is that any of the fancy new language
> approaches we've talked about actually make things any better on those
> platforms - especially with regards to memory access patterns (in fact,
> I contend that they may be *worse*).

I agree. I was reluctant to mention it because the thread was of interest
and I didn't want to kill it. So what I'm doing here is forking it. ;-)

So far we've discussed imperative vs. functional languages. I'd also suggest
comparing procedural vs. data flow languages. It's been my experience that
data flow languages optimize much better than procedure languages do. Plus
for real parallization (multi-processors vs. just time sliced threads on a
uni-processor) they avoid many of the synchronization pit falls (live & dead
locks, race conditions, etc.) of procedure languages.

Your opinion may vary; Feel free to discuss. ;-)
-- 
Enjoy,
George Warner,
Schizophrenic Optimization Scientist
Apple Developer Technical Support (DTS)