Running China+PPL on our benchmark suite I observed
the following phenomenon: when the constraints of polyhedra
are printed I see, e.g., both
B + C + D + E - F = -2,
and
-B - C - D - E + F = 2.
These constraints are equivalent and are also normalized
with our current notion of normalization.
In fact, gcd(1,-1,-2) = gcd(-1,1,2) = 1.
The question is: are we using too weak a notion of normalization?
What happens if, for equalities, we insist having a non-negative
inhomogeneous term?
Notice that, in some places, we are rather syntactic when dealing
with constraints. For instance, in Matrix::merge_rows_assign()
we regard the above constraints as different and we get them
both in the result.
Any idea?
Ciao,
Roberto
--
Roberto Bagnara
Computer Science Group
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara@cs.unipr.it
CVSROOT: /cvs/ppl
Module name: ppl
Changes by: hill(a)cs.unipr.it 2001-11-09 13:19:12
Modified files:
interfaces/Prolog/SICStus: clpq.pl
Log message:
Now there are no calls to the ppl_convex_hull_assign operation.
This version keeps the list of generators and uses that to recover
from failure paths.
Patches:
http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/interfaces/Prolog/SICStus/clp…
CVSROOT: /cvs/ppl
Module name: ppl
Changes by: hill(a)cs.unipr.it 2001-11-09 08:48:45
Modified files:
interfaces/Prolog/SICStus: clpq.pl
Log message:
Now many of the unwanted temporary copies of polyhedra are deleted earlier.
Patches:
http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/interfaces/Prolog/SICStus/clp…
CVSROOT: /cvs/ppl
Module name: ppl
Changes by: hill(a)cs.unipr.it 2001-11-08 22:23:55
Modified files:
interfaces/Prolog: Prolog_interface.dox
Log message:
More detailed specification of the interface predicates with
many examples added.
Introduction includes grammar rules and list of some important
issues wrt using the PPL in Prolog.
Patches:
http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/interfaces/Prolog/Prolog_inte…
