[gecode-users] Gecode 3.0.0 again

Christian Schulte cschulte at kth.se
Tue Feb 10 16:06:37 CET 2009


Dear all,

what has become obvious is that our initial plans for 3.0 had been
ludicrously optimistic. Now, we see much clearer (two of the three main
Gecode people actually got a degree within the last three months).

We have to settle for much less in 3.0 than has been planned. What will be
in:
 - a 50+ page document on modeling with Gecode
 - a complete rework of branching, including randomization and support for
tie-breaking
 - abstractions for cost-based optimization (just define a cost function)
 - the first stable version of Gist as an interactive search tool
 - lots of cleanups (even major) to make Gecode easier to explain
 - crucial abstractions optimized for 64 bit rather than for 32 bit instead
 - new and more efficient memory management
 - reflection API removed (*)
 - non monotonic propagators supported
 - and fixes as always

What will come later (we plan to release more often again in the future as
releasing has become much easier since Gecode/J is not any longer supported
and we do not have to do Java magic any longer) is:
 - integration of parallel search
 - integration of float vars
 - integration of a MDD-based extensional propagator
 - scheduling (edge-finder for disjunctive problems first)

We hope to get it out until beginning of March and will have it out at
latest March 15th.

Cheers
Christian

(*) We decided to remove reflection support as it did not do what we needed,
was to complicated, and had no real killer application. That's not good
enough for something that comprises 10% of the entire code base. Some parts
that are useful for interfacing, etc will be resurrected in simplified form
later. 

--
Christian Schulte, www.ict.kth.se/~cschulte/







More information about the gecode-users mailing list