March 2005 - PPL-devel - BUGSENG Mailman

ppl/tests print.hh
by Matthew Mundell 21 Mar '05

21 Mar '05

CVSROOT: /cvs/ppl Module name: ppl Branch: grids Changes by: mundell(a)cs.unipr.it 2005-03-21 19:24:32 Modified files: tests : print.hh Log message: Set NOISY and VERY_NOISY from the associated directives. Create the nout and vnout streams. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/print.hh.diff?cvsroot=p…

1 0

ppl/tests Makefile.am
by Matthew Mundell 21 Mar '05

21 Mar '05

CVSROOT: /cvs/ppl Module name: ppl Branch: grids Changes by: mundell(a)cs.unipr.it 2005-03-21 19:23:00 Modified files: tests : Makefile.am Log message: Add Grid to SUBDIRS. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/Makefile.am.diff?cvsroo…

1 0

ppl/tests/Grid writecongruencesystem.cc grid1. ...
by Matthew Mundell 21 Mar '05

21 Mar '05

CVSROOT: /cvs/ppl Module name: ppl Branch: grids Changes by: mundell(a)cs.unipr.it 2005-03-21 19:21:18 Added files: tests/Grid : writecongruencesystem.cc grid1.cc congruencesystem2.cc congruencesystem1.cc congruence1.cc Makefile.in Log message: Initial revision. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/Grid/writecongruencesys… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/Grid/grid1.cc?cvsroot=p… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/Grid/congruencesystem2.… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/Grid/congruencesystem1.… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/Grid/congruence1.cc?cvs… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/Grid/Makefile.in?cvsroo…

1 0

ppl/tests/Grid
by Matthew Mundell 21 Mar '05

21 Mar '05

CVSROOT: /cvs/ppl Module name: ppl Changes by: mundell(a)cs.unipr.it 2005-03-21 19:15:50 ppl/tests/Grid Update of /cvs/ppl/ppl/tests/Grid In directory spartacus.cs.unipr.it:/tmp/cvs-serv29411/tests/Grid Log Message: Directory /cvs/ppl/ppl/tests/Grid added to the repository --> Using per-directory sticky tag `grids'

1 0

ppl/src Grid_simplify.cc Grid_public.cc Grid_n ...
by Matthew Mundell 21 Mar '05

21 Mar '05

1 0

ppl configure.ac
by Matthew Mundell 21 Mar '05

21 Mar '05

CVSROOT: /cvs/ppl Module name: ppl Branch: grids Changes by: mundell(a)cs.unipr.it 2005-03-21 19:11:01 Modified files: . : configure.ac Log message: Add tests/Grid/Makefile to AC_CONFIG_FILES. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/configure.ac.diff?cvsroot=ppl…

1 0

ppl STANDARDS
by Matthew Mundell 21 Mar '05

21 Mar '05

CVSROOT: /cvs/ppl Module name: ppl Branch: grids Changes by: mundell(a)cs.unipr.it 2005-03-21 19:09:40 Modified files: . : STANDARDS Log message: Add "Standards for the Test Programs" section. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/STANDARDS.diff?cvsroot=ppl&on…

1 0

Re: Widening operation in SICStus
by Roberto Bagnara 19 Mar '05

19 Mar '05

V Panumong wrote: > I've just tried a simple program on SICStus with ppl, but it failed. > Is it something I've done wrongly? > > The program is: > test(CS, GS):-I='$VAR'(0), J='$VAR'(1), > ppl_new_Polyhedron_from_generators(nnc, [point(2*I+0*J), > point(6*I+0*J), > point(4*I+1*J)], PL1), > ppl_new_Polyhedron_from_generators(nnc, [point(2*I+0*J), > point(10*I+0*J), > point(6*I+0*J)], PL2), > ppl_Polyhedron_BHRZ03_widening_assign(PL1, PL2), > ppl_Polyhedron_get_constraints(PL1, CS), > ppl_Polyhedron_get_generators(PL1, GS). > > The error returned is: > ppl_sicstus: Polyhedron_widenings.cc:667: void > Parma_Polyhedra_Library::Polyhedron::BHRZ03_widening_assign(const > Parma_Polyhedra_Library::Polyhedron&, unsigned int*): Assertion > `x_copy.contains(y_copy)' failed. > ./ps: line 1: 29556 Aborted > > The program also fails if I change the predicate to > ppl_Polyhedron_H79_widening_assign. > > I took this example program from "Automatic Discovery of Linear > Constraints Among Variables of a Program" by Cousot and Halwachs '78 > p.94 Dear Vaji, an important requirement of widening predicates such as ppl_Polyhedron_BHRZ03_widening_assign/2 is that the first polyhedron must contain (or be equal to) the second polyhedron. So what you want to do is probably the following: test(CS, GS):-I='$VAR'(0), J='$VAR'(1), ppl_new_Polyhedron_from_generators(nnc, [point(2*I+0*J), point(6*I+0*J), point(4*I+1*J)], PL1), ppl_new_Polyhedron_from_generators(nnc, [point(2*I+0*J), point(10*I+0*J), point(6*I+0*J)], PL2), ppl_Polyhedron_poly_hull_assign(PL1, PL2), ppl_Polyhedron_BHRZ03_widening_assign(PL1, PL2), ppl_Polyhedron_get_constraints(PL1, CS), ppl_Polyhedron_get_generators(PL1, GS). Notice that this is actually the _only_ interface invariant that is only checked in versions of the library compiled with assertions enabled. Admittedly, this is a limitation and we would love to find a cheap way to always do this test. As things stand now, I think we should mitigate this problem as much as possible by means of documentation. The general library documentation explains the containment prerequisite of extrapolation and widening methods; the documentation of the C++ and of the C interfaces do recall it; the Prolog interface seems to say nothing on the subject. Pat, can you please update the Prolog interface documentation so as to make it less likely that users stumble on this kind of problem? Thank you Vaji: as you can see this kind of user feedback is crucial for improving the library. Cheers, Roberto -- 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

[Fwd: Re: About PPL]
by Roberto Bagnara 18 Mar '05

18 Mar '05

-------- Original Message -------- Subject: Re: About PPL Date: Fri, 18 Mar 2005 06:05:50 -0800 (PST) From: gulsah yilmaz <gulsahyilmaz20(a)yahoo.com> To: Roberto Bagnara <bagnara(a)cs.unipr.it> Mr. Bagnara, First of all, thanks for your attention. Since Polylib uses this representation I think you will be familiar with the representation. I will exlplain what I want to do: I have inequalities like: x+M+N+5 >= 0 y+z+1 >= 0 x+y >= 0 x+z+1 >= 0 z+M+2>=0 and an inequality which include unknown coefficients like: (a-b)x+cz+(d+e)M+N >= 0 and by combining these inequalities I want to know the relation between unknowns like: a+b>1 d-e+c>M .... Thanks Gulsah Yilmaz --- Roberto Bagnara <bagnara(a)cs.unipr.it> wrote: > Dear Gulsah, > > if the file `Makefile' is missing, it is quite > likely > you did not configure your build directory. Please > follow > the instructions in the files `README.configure' and > `INSTALL' (they are part of the library's > documentation). > > > An example to my problem: > > The numbers are faik. > > > > x y z M N Constant > > 1 1 0 0 1 1 5 > > 1 0 1 1 0 0 1 > > 1 1 1 0 0 0 0 > > 1 1 0 1 0 0 1 > > 1 0 0 1 1 0 2 > > > > and > > 1 a-b 0 c d+e 1 0 > > > > I want to know the relation between the unknown > > coefficient a,b,c,d,e. I ask you whether this is > > possible or not by using PPL. > > Sorry, but the table above means nothing to me: > I can imagine the columns named `x', `y' and so > forth contain coefficients of the variables > with the same names; I can imagine the column > named `Constant' gives the inhomogeneous term; > but I cannot make any sense of the first, unnamed > column; then I don't know if that is meant to > represent a set of constraints (equalities? > non-strict inequalities? <=? >=? something else?) > or a set of generators. > > Can you please explain your problem in the language > of linear algebra? Chances are that we will be > able to answer your question then. > Cheers, > __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com -- 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

Re: About PPL
by Roberto Bagnara 18 Mar '05

18 Mar '05

gulsah yilmaz wrote: > Mr. Bagnara > > I read library's documentation, but I could not run > the test programs since compilation failure. I > download the source codes from ftp site and execute > make command. It fails since the MakeFile document is > missing. Dear Gulsah, if the file `Makefile' is missing, it is quite likely you did not configure your build directory. Please follow the instructions in the files `README.configure' and `INSTALL' (they are part of the library's documentation). > An example to my problem: > The numbers are faik. > > x y z M N Constant > 1 1 0 0 1 1 5 > 1 0 1 1 0 0 1 > 1 1 1 0 0 0 0 > 1 1 0 1 0 0 1 > 1 0 0 1 1 0 2 > > and > 1 a-b 0 c d+e 1 0 > > I want to know the relation between the unknown > coefficient a,b,c,d,e. I ask you whether this is > possible or not by using PPL. Sorry, but the table above means nothing to me: I can imagine the columns named `x', `y' and so forth contain coefficients of the variables with the same names; I can imagine the column named `Constant' gives the inhomogeneous term; but I cannot make any sense of the first, unnamed column; then I don't know if that is meant to represent a set of constraints (equalities? non-strict inequalities? <=? >=? something else?) or a set of generators. Can you please explain your problem in the language of linear algebra? Chances are that we will be able to answer your question then. Cheers, Roberto -- 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