Sven Verdoolaege wrote:
On Mon, Aug 18, 2008 at 05:17:01PM +0200, Roberto Bagnara wrote:
Albert Cohen wrote:
The concept is called "gist" in Omega, and simplification under a context in PolyLib. You'd like to run the simplification with the assumption that some constraints are already enforced, and only retain the additional constraints (minimized) to define a given sub-polyhedron.
Hi Albert,
we need to have a more precise specification of the required operation. In particular, for the simplification of { A <= 0 } under the context
I assume you mean { A = 0 } here (as { A <= 0 } is not a subset of {A >= 0}).
{ A >= 0 }, what do you consider simpler?
Possibility 1: { A <= 0 } Possibility 2: { A = 0 }.
I asked the same question on the polylib mailing list many years ago but got no answer.
Dear Sven,
if you run the attached input through PolyLib's polytestgmp, you will obtain:
$ ./polytestgmp </tmp/simplify_bug.in 1 3 1 0 1
This would seem to indicate that the simplification of the domain { {A>=0} } (i.e., the singleton set of polyhedra containing only the polyhedron defined by the single equality A >= 0) in the context { {A<=0} } gives the universe as the result. Can you confirm that this is a bug in PolyLib? Cheers,
Roberto