[PPL-devel] Re: PPL and parameters

[Roberto Bagnara]

Sven Verdoolaege wrote:
 > I'm working on counting the number of points in parametrized polyhedra.
 > (http://www.cs.kuleuven.ac.be/publicaties/rapporten/cw/CW376.abs.html)
 > Currently, I'm using PolyLib because I could reuse a lot from the
 > previous implementation.
 >
 > But PolyLib has some problems.
 > For one, each time a polytope is created, it creates new mpz_ts,
 > which results in a lot of calls to mpz_init and mpz_clear.
 > Does PPL handle this differently ?

Dear Sven,

I am not sure I understand what you mean.  The current version of the
PPL is also based on GMP (even though we have a development branch to
implement support for native integers and other kinds of coefficients, see
http://www.cs.unipr.it/pipermail/ppl-devel/2004-February/004115.html).
So when you create a polyhedron, new GMP objects will be created.  How
many of them, depends on the polyhedron.  What you can count on is
that the PPL uses lazy and incremental computation techniques wherever
possible, so that it will not create unnecessary objects (i.e., if the
polyhedron being created is "simple", few memory allocation will be
required).  I realize I am probably not answering your question: if
you can be more specific we will do our best.

 > A quick glance at the documentation suggests that PPL only
 > supports what Omega would call "set variables".
 > Are there any plans for supporting parameters and/or
 > existential variables ?

What do you mean by "existential variables"?

I think we will support parameterized polyhedra in the future.
But this is something that will not happen tomorrow.  There is
much activity now but the focus is elsewhere at the moment
(specialized classes of polyhedra, powersets, different kinds of
coefficients, new interfaces, incorporation of the simplex method,
even more memory efficiency, improved NNC polyhedra, grids and
Z-polyhedra, ...).
All the best,

     Roberto

-- 
Prof. Roberto Bagnara
Computer Science Group
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara at cs.unipr.it