Re: Parallel processing

[glen e. p. ropella]

Tony,

I'm cc'ing this to the support list because there're probably
others interested.

> 1) Does the Swarm team have any plans to implement coding that
> explicitly takes advantage of parallel processing capabilities in
> Linux?  What is the timelime for this capability?

Well, since we haven't really broached || Swarm explicitly in our
planning meetings, yet, I can't say precisely what mechanisms we will
use.  However, my bet is on message passing as the paradigm.  Since
99.99% of Swarm is already based on message passing between objects,
it would make sense to use message passing between processors and
machines.  Now, with respect to Linux and any peculiar || mechanisms
available within, I can't say whether we'll use those.  We are really
trying to make Swarm a very portable system.  We definitely want it to
work on as many UNIXen as possible.  And we have hints from some users
that they're working on ports to Windows NT.  (mind you,this is only
rumour) So, explicitly requiring capabilities that are peculiar to
Linux and/or Pentium Pros would not be wise.

UNLESS, we end up optimizing Swarm to use OS peculiar features in any
given situation.  I.e. If we had a kind of Swarm backend, like in gcc,
where one codes in standard Swarm-eze (whatever that may be) but
either the libraries were built with platform-specific configuration
or we have compile-time or run-time hooks to use these features.  But,
this would take a tremendous amount of work and I don't think we're
ready to consider this yet.

> [...] Linux can be used on SMP machines (dual or quad pentiums) and
> am thinking about the implications of building a model in Swarm that
> would take advantage of this capability.

Well [grin], since we're trying to foster a transparent network
between users (vs a centrally controlled hub), "Go for it!"  And
anything you end up doing will definitely influence what Swarm
eventually becomes.  I can't say that whatever you ended up doing
would be taken whole-hog (or even at-all); but, I can say that even if
whatever you end up with isn't general purpose enough to get included
in the distribution, it will stand as the first and foremost data
point for a || Swarm.

As another aside, if you buy one of these boards (with more than one
processor on it), you might think about getting a mixed one with
Pentiums and Alphas on it.  Of course, I'm not saying this in any
professional capacity [grin].  I just have this fantasy of owning a
Pentium/Alpha motherboard and running Linux/DEC Unix/Windows NT and
having them all talk to each other. :-)

> 2) Can the project timeline and major tasks be posted in some format
> (here or on the website)?

We're working right now on "formalizing" our task allocation/
completion process.  And I'm not sure that we'll end up with
"honest-to-goodness" schedules with milestones, etc.  But, we will
have some type of timeline.  And we plan to publish it.  But, you have
to remember that we're only staffed at 1.5 to 3.5 people at present.
So, any time taken to do administrative things takes a huge cut out of
programming time.  So, while I respect your desire for a detailed list
of goals, I have to say that it's not that important right now.  It
will become very important after v1.0 comes out.

glen