[email protected] wrote:
I'm just reading through the phantastic CTM book and feel Mozart has
lot's of potential when it comes to manycores. Do you think about
spreading the community, publishing new SW releases (ala Haskell e.g.)...?
Mike
Regarding Mozart releases, I can say the following:
- The 1.4.0 release last year was the first implementation of the new
Klintskog/Collet distribution layer (see Raphael Collet's Ph.D. work:
http://www.info.ucl.ac.be/~pvr/raphthesis.pdf). This layer both
increases the power and simplifies the use of the network-transparent
distribution with respect to the previous distribution layer. But
because of the big changes, some bugs were introduced. We are now
working on a 1.4.1 release that fixes these problems. Yves Jaradin is
the main person responsible. We hope to make this release soon. Yves
can say more.
- Network-transparent distribution is the easiest way to exploit
multi-core processors, since an existing program can be made distributed
with almost no changes to the source code. This is our current approach
for exploiting parallelism. It avoids the enormous problems of having
to parallelize the emulator implementation itself. The price is the
overhead of the distribution layer. This means that we cannot do very
fine-grained parallelism, but it is my belief that most realistic
applications will have at least medium-grained parallelism.
- The Beernet transactional peer-to-peer library
(beernet.info.ucl.ac.be) runs on Mozart and is under intense
development. A prototype is available now. We are building a
distributed application using mobile devices based on Beernet. Boris
Mejias is the main person responsible and Jérémie Melchior is building
the application.
- We are also working on adding graph constraints to the Gecode
constraint solving library (www.gecode.org). Gustavo Gutierrez is the
main person responsible. There is related work on a bridge between
Gecode and Mozart in the Avispa group in Colombia. Gecode is directed
by Christian Schulte, who was the main designer of the computation space
abstraction in Mozart.
Peter
_________________________________________________________________________________
mozart-users mailing list
[email protected]
http://www.mozart-oz.org/mailman/listinfo/mozart-users
