RE: [Repast-developer] Controller hierarchy

[Andy C. <ac...@ll...>]

At 12:52 PM 12/13/2002 -0500, Nick Collier wrote:
>Hi,
>
>
>I agree, that is my head tells me, that #1 would be more useful given
>our current user base. As for which would be "cooler" to work on, I'd
>have to go with #2.

[snipped: #1 discussed "embarrassingly parallel" ABM, while #2 was a 
"massively parallel" ABM]

An FYI: my group at Livermore has a modest amount of funding to develop #2 
(not necessarily for Repast but for "entity based simulation" in general) 
that started in October (which is why I am just beginning to show up on 
this list)... I've had a preliminary discussion with Mike North about this. 
I think we are all in agreement that the uses for #2 are probably much more 
specialized than #1, but to be perfectly frank, I wouldn't have gotten 
funding except for #2 because our lab is generally interested in "macho 
computing" which is #2, and that's where most of our expertise lies.

At this point, we've settled on working with Repast for our initial work, 
though it's still a little unclear exactly what form we will ultimately 
deliver it (there are some legitimate if unfortunate language concerns when 
one talks about the big parallel supercomputers: if Java is a choice at 
all, it may not always be a very maintained choice, there are no parallel 
debuggers, parallel java code generally requires nonstandard layers like 
JavaMPI that may not be all that bulletproof, etc). I'd certainly like to 
hold on to the idea of being able to do it in a form that would be 
relatively easy to plug into Repast as long as I can, though.

Anyways, the point is that we'd be glad to trade notes or ideas on this 
topic with any of you who have spent some time thinking along these lines...

p.s. Is this a good place to solicit potential applicants for positions, 
particularly postdocs, in this area? We don't have money yet but I am 
hopeful, and my first choice would be to bring in someone with direct R&D 
experience in this field... So I'd like to ask those with students in the 
field first...

Best,
Andy
============================================
Andrew J. Cleary, Software Engineer/Computational Scientist
Lawrence Livermore National Labs, 925-424-5890.
============================================