[Draft] IC & FD

From: Warwick Harvey <wh_at_icparc.ic.ac.uk>
Date: Tue 26 Nov 2002 05:52:02 PM GMT
Message-ID: <20021126175202.H25998@tempest.icparc.ic.ac.uk>
Hi all,

For those of you using or thinking of using the IC library and its related
libraries, please be aware that you should not mix the use of IC-based
code and FD-based code; IC and FD are two completely independent libraries,
and mixing the two will not "just work".  (The same goes for libraries based
on the Range library, such as the old RIA library.)

If you want to know the whys and wherefores, it's because they use different
domain representations, and in general each solver only updates and looks at
its own representation; hence a change made by one solver will not be seen
by the other solver (at least, not until the variable becomes ground, at
which point it's a normal ground number understood by all solvers).

It is for this reason that we provide both IC and FD (or Range) versions of
many libraries, so that the same functionality can be used in conjunction
with each.  Here's a summary of the current solver-specific libraries and
their equivalences (though the equivalence is not always exact, and not
every library has an IC equivalent):

	IC version	FD version	Range version
	ic		fd		ria
	ic_global	fd_global
	ic_cumulative	cumulative
	ic_edge_finder	edge_finder
	ic_edge_finder3	edge_finder3
	ic_search	fd_search
	ic_sets		fd_sets
	(not available)	conjunto
	ic_sbds		fd_sbds
	ic_eplex			eplex
	(not needed)	fdplex		fdplex

The intention is that this is only a temporary arrangement; when IC is ready
to supersede FD and Range/RIA fully, it is intended that there will only be
one numeric domain representation, used by all numeric solvers, and the
issue will go away completely.

Cheers,
Warwick
Received on Tue Nov 26 17:54:54 2002

This archive was generated by hypermail 2.1.8 : Wed 16 Nov 2005 06:07:18 PM GMT GMT