August 2004 - PPL-devel - BUGSENG Mailman

ppl/src AskTell.defs.hh
by Roberto Bagnara 19 Aug '04

19 Aug '04

CVSROOT: /cvs/ppl Module name: ppl Branch: ppl-0_6-branch Changes by: roberto(a)cs.unipr.it 2004-08-19 13:43:29 Modified files: src : AskTell.defs.hh Log message: Previous change reverted. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/AskTell.defs.hh.diff?cvsr…

1 0

ppl/src AskTell.defs.hh
by Roberto Bagnara 19 Aug '04

19 Aug '04

CVSROOT: /cvs/ppl Module name: ppl Changes by: roberto(a)cs.unipr.it 2004-08-19 13:43:22 Modified files: src : AskTell.defs.hh Log message: Previous change reverted. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/AskTell.defs.hh.diff?cvsr…

1 0

ppl/doc devref-browse.doxyconf-latex.in devref ...
by Enea Zaffanella 19 Aug '04

19 Aug '04

CVSROOT: /cvs/ppl Module name: ppl Changes by: zaffanella(a)cs.unipr.it 2004-08-19 12:42:30 Modified files: doc : devref-browse.doxyconf-latex.in devref-print.doxyconf-latex.in devref.doxyconf-html.in Log message: Widening_Function and Polyhedra_PowerSet should appear in all the devref manuals. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/doc/devref-browse.doxyconf-la… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/doc/devref-print.doxyconf-lat… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/doc/devref.doxyconf-html.in.d…

1 0

ppl/src AskTell.defs.hh
by Enea Zaffanella 19 Aug '04

19 Aug '04

CVSROOT: /cvs/ppl Module name: ppl Changes by: zaffanella(a)cs.unipr.it 2004-08-19 12:40:52 Modified files: src : AskTell.defs.hh Log message: AskTell should not appear in the user documentation until working. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/AskTell.defs.hh.diff?cvsr…

1 0

ppl/doc devref-browse.doxyconf-latex.in devref ...
by Enea Zaffanella 19 Aug '04

19 Aug '04

CVSROOT: /cvs/ppl Module name: ppl Branch: ppl-0_6-branch Changes by: zaffanella(a)cs.unipr.it 2004-08-19 12:34:23 Modified files: doc : devref-browse.doxyconf-latex.in devref-print.doxyconf-latex.in devref.doxyconf-html.in Log message: Widening_Function should appear in the devref manuals. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/doc/devref-browse.doxyconf-la… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/doc/devref-print.doxyconf-lat… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/doc/devref.doxyconf-html.in.d…

1 0

ppl/doc devref.doxyconf-html.in
by Enea Zaffanella 19 Aug '04

19 Aug '04

CVSROOT: /cvs/ppl Module name: ppl Branch: ppl-0_6-branch Changes by: zaffanella(a)cs.unipr.it 2004-08-19 12:28:26 Modified files: doc : devref.doxyconf-html.in Log message: Polyhedra_PowerSet has to also appear in the developers manual. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/doc/devref.doxyconf-html.in.d…

1 0

ppl/src AskTell.defs.hh
by Enea Zaffanella 19 Aug '04

19 Aug '04

CVSROOT: /cvs/ppl Module name: ppl Branch: ppl-0_6-branch Changes by: zaffanella(a)cs.unipr.it 2004-08-19 12:27:34 Modified files: src : AskTell.defs.hh Log message: AskTell construction should not appear in the user manuals. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/AskTell.defs.hh.diff?cvsr…

1 0

Last problem (hope!)
by Enrico Oliosi 19 Aug '04

19 Aug '04

Hi ppl-gr, I used the function set_output_function() how you suggest me. When I use it in a single program like in your tests example it works correctly, but when I use it in a class I must declare it static; I used it in a my_print function in order to print the ph for each comand. I think that this happens for this reason: since the namespace Variable have static member when I used the functions of this namespace in a class (an object which can be initialized more times) I must declare them static. The problem is in this class: class Cmd{ public: //ph which I have in each statement NNC_Polyhedron ph_in; //these are Hashmap the first element is the dimension //and the second the variable to print static map<dimension_type, string> var_to_print; static map<dimension_type, string>::iterator iter; public: static void Cmd::new_output_function(ostream& s, const Variable v) { iter = var_to_print.find(v.id()); s << iter->second; } void Cmd::print_constraints(const ConSys& cs, const string& intro, ostream& s){ using namespace IO_Operators; // Save the default output function. Variable::Output_Function_Type* p_default_output_function = Variable::get_output_function(); // Install an alternate output function. Variable::set_output_function(new_output_function); /* Your code of print_constraints() */ /* Code to restore the default output function. */ [...] }; when I compile it the g++ gives me these errors: /tmp/ccOzbfS7.o(.gnu.linkonce.t._ZN3Cmd19new_output_functionERSoN23Parma_Polyhedra_Library8VariableE+0x24): In function `Cmd::new_output_function(std::basic_ostream<char, std::char_traits<char> >&, Parma_Polyhedra_Library::Variable)': : undefined reference to `Cmd::var_to_print' [...] : undefined reference to `Cmd::iter' [...] Sorry but I am a java-programmer and sometimes when I used c++ I make some stupid mistake. All the bests, Enrico.

2 4

Re: Help with alias analysis framework
by Roberto Bagnara 19 Aug '04

19 Aug '04

Ber Past wrote: > Looking at the library documentation, it seems it fits exactly my needs. > I have only one question (probably stupid, but my knowledge of the > library is not yet very good): is there any operator/function that > allows to calculate the complement of a Polyhedron? > > Thanks for your help. Dear Ber, I am not sure I understand what you need, so I will give you two alternatives. Our library (PPL) provides an operation called "poly-difference" that yields the smallest convex polyhedron containing the set-theoretic difference of its two arguments. Computing the poly-difference of the universe polyhedron and a polyhedron ph gives an upper approximation of the complement. This approximation can of course be extremely coarse unless your application generates only a very restricted class of polyhedra. Alternatively, the PPL can represent the genuine complement of a given convex polyhedron by a (finite) set of convex polyhedra. Sets of polyhedra are fully supported by version 0.6 of the library, which has just been released. Here is a small program to give you an idea of what you can do: -------------------------------- %< %< %< -------------------------------- #include <ppl.hh> using namespace std; using namespace Parma_Polyhedra_Library; using namespace Parma_Polyhedra_Library::IO_Operators; template <typename PH> Polyhedra_PowerSet<NNC_Polyhedron> complement(const PH& ph) { pair<PH, Polyhedra_PowerSet<NNC_Polyhedron> > partition = linear_partition(ph, PH(ph.space_dimension(), Polyhedron::UNIVERSE)); return partition.second; } int main() { Variable x(0); Variable y(1); C_Polyhedron p(2, Polyhedron::EMPTY); p.add_generator(point(x)); p.add_generator(point(y)); p.add_generator(point(-x)); p.add_generator(point(-y)); cout << "p = " << p << endl; Polyhedra_PowerSet<NNC_Polyhedron> p_c = complement(p); cout << "complement(p) = " << p_c << endl; C_Polyhedron q(2); q.add_constraint(x >= -1); q.add_constraint(x <= 1); q.add_constraint(y >= 1); q.add_constraint(y <= 3); cout << "q = " << q << endl; Polyhedra_PowerSet<NNC_Polyhedron> q_c = complement(q); cout << "complement(q) = " << q_c << endl; return 0; } -------------------------------- %< %< %< -------------------------------- And here is the output of the program (newlines added for readability): p = -A - B >= -1, -A + B >= -1, A - B >= -1, A + B >= -1 complement(p) = { { A + B > 1 }, { -A - B >= -1, A - B > 1 }, { -A - B >= -1, -A + B > 1 }, { -A - B > 1, -A + B >= -1, A - B >= -1 } } q = -A >= -1, -B >= -3, B >= 1, A >= -1 complement(q) = { { A > 1 }, { -A >= -1, B > 3 }, { -A >= -1, -B > -1 }, { -A > 1, -B >= -3, B >= 1 } } Notice that, in general, there are several possible representations of the complement of a polyhedron as a set of polyhedra. This implementation guarantees that the resulting set is a partition of the complement of the given polyhedron (i.e., that the members of the set are pairwise disjoint). It should also be possible to prove that the partition returned has a reasonable number of elements (such as the number of facets of the argument). I hope this helps. Please do not hesitate to come back to us in case it doesn't. All the best, Roberto >> From: Roberto Bagnara <bagnara(a)cs.unipr.it> >> To: BerPast <berpast(a)hotmail.com> >> Subject: Re: Help with alias analysis framework >> Date: Sat, 14 Aug 2004 09:30:12 +0200 >> >> BerPast wrote: >> >>> In the paper "Interprocedural may-alias analysis for pointers: beyond >>> k-limiting", the author presents a parametric framework for the >>> analysis of pointer aliases. >>> The framework is parametrised by a numeric lattice V#. >>> The lattice has to have some abstract operators; the 4th are 5th are >>> for me the most complex: >>> 4) resolution of a linear system: given a system S, I need to compute >>> a member of the lattice which is an upper approximation of the integer >>> solutions to S; >>> 5) intersection with a linear system: if S is a system of linear >>> equations and K is a member of the lattice, I need to compute an upper >>> approximation of the solutions to S that are also in K (no empty >>> intersection with K I guess). >>> >>> I need to implement the lattice operators, but I have no idea of how >>> to do this. Probably using Fourier-Motzkin projection I can solve the >>> problem for the lattice obtained by the smash product of the interval >>> lattice (any other, more efficient method?), but how can this problem >>> be solved for a generic lattice? >>> Can someone point out to me some documentation and (if available code >>> implemenation) explaining how to solve this problem for a generic >>> lattice? >> >> >> I am not sure I understand what you mean by "[solving] this problem [...] >> for a generic lattice." All the numerical abstractions provided by the >> Parma Polyhedra Library (http://www.cs.unipr.it/ppl/) come with the >> operations that a numeric lattice must possess in Alain Deutsch's >> framework. >> All the best, >> >> Roberto Bagnara >> >> -- >> Prof. Roberto Bagnara >> Computer Science Group >> Department of Mathematics, University of Parma, Italy >> http://www.cs.unipr.it/~bagnara/ >> mailto:bagnara@cs.unipr.it -- Prof. Roberto Bagnara Computer Science Group Department of Mathematics, University of Parma, Italy http://www.cs.unipr.it/~bagnara/ mailto:bagnara@cs.unipr.it

1 0

ppl/src Makefile.am
by Roberto Bagnara 18 Aug '04

18 Aug '04

CVSROOT: /cvs/ppl Module name: ppl Branch: ppl-0_6-branch Changes by: roberto(a)cs.unipr.it 2004-08-18 23:07:05 Modified files: src : Makefile.am Log message: AskTell is unfinished and not working: it should not be distributed. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/Makefile.am.diff?cvsroot=…

1 0