April 2017 - PPL-devel - BUGSENG Mailman

join-like operation
by Piotr Mardziel 26 Apr '17

26 Apr '17

Hello PPL developers; I'm finding myself in need of a particular join-like operation but I'm having trouble finding the necessary operations in PPL to make it work. I figure that the need for this operation would be common especially when implementing inter-procedural program analyses so I'm suspicious that I'm just missing the right terminology for it. Let me explain this operation: The setup has two polyhedra that model two sets of program variables with some variables shared but some variables not shared. For example: polyhedron A has the constraint x ≤ y, w = 1 polyhedron B has the constraint y ≤ z, w = 2 Note that each of those two only has 3 space dimensions but the are not intended to represent the same 3 variables, hence I gave them names to distinguish them. Thus we see that y and w is shared between the two polyhedra, whereas x or z are only represented in A or B respectively. Want I need is to join these two (thus requiring to first expand their dimensions to include all 4 dimensions) such that the result is a polyhedron C that "effectively" joins shared dimensions but preserves constraints over unshared dimensions, in this case I would like outcome C to have constraints x ≤ y ≤ z, 1 ≤ w ≤ 2 . The problem is that to expand each input polyhedron to 4 dimensions, one either has to embed or project and then follow it up with either join or meet. The join doesn't handle the x,y,z portion of this problem correctly whereas meet doesn't handle the w portion correctly. More technical, I need an operation over A, B to return a C with γ(A) ⊆ { π_{x,y,w} (e) | e ∈ γ(C) } γ(B) ⊆ { π_{y,z,w} (e) | e ∈ γ(C) } Where γ is concretization, and π_{...}(e) is a projection of a tuple e to a subset of its dimensions. Top trivially satisfies this requirement but the polyhedron {x ≤ y ≤ z, 1 ≤ w ≤ 2} also satisfies it and does so much more precisely (is minimal in terms of a concretized subset relation). Am I missing something obvious here? Any help would be greatly appreciated. Best, Piotr (Peter) Mardziel

2 1

[GIT] ppl/ppl(master): Work around the fact that __attribute__((weak)) does not work on 64-bit Cygwin.
by Roberto Bagnara 08 Apr '17

08 Apr '17

Module: ppl/ppl Branch: master Commit: 4e3d897f08b63ac7abc55d6a3621f84cdf1a69fe URL: http://www.cs.unipr.it/git/gitweb.cgi?p=ppl/ppl.git;a=commit;h=4e3d897f08b6… Author: Roberto Bagnara <roberto.bagnara(a)bugseng.com> Date: Sat Apr 8 16:38:17 2017 +0200 Work around the fact that __attribute__((weak)) does not work on 64-bit Cygwin. Thanks to Erik M. Bray. --- src/assertions.hh | 5 ++++- 1 files changed, 4 insertions(+), 1 deletions(-) diff --git a/src/assertions.hh b/src/assertions.hh index cceb7c9..63afbc1 100644 --- a/src/assertions.hh +++ b/src/assertions.hh @@ -98,7 +98,10 @@ site: http://bugseng.com/products/ppl/ . */ namespace Parma_Polyhedra_Library { -#if PPL_CXX_SUPPORTS_ATTRIBUTE_WEAK +// Note: __attribute__((weak)) does not work on 64-bit Cygwin. +// See https://trac.sagemath.org/ticket/16152 and links therein. +#if PPL_CXX_SUPPORTS_ATTRIBUTE_WEAK \ + && !(defined(__CYGWIN__) && defined(__x86_64__)) #define PPL_WEAK_NORETURN __attribute__((weak, noreturn)) #else #define PPL_WEAK_NORETURN __attribute__((noreturn))

1 0

Need help
by Amit Gurung 04 Apr '17

04 Apr '17

Dear Sir/Madam I see that there is very little example programs for beginners to start coding using PPL library, It would be of great help if you can kindly help me in this direction. I managed to write a simple code to create a PPL NNC_Polyhedron object named myPoly and assigned linear constraints (values of the form Ax <= b) where A is the coefficient matrix (boost library) and vector b is the bound values of the linear constraints. PPL_Polyhedron(math::matrix<double> A, std::vector<double> b, int sgn){ > int noOfCons = A.size1(); > int dim = A.size2(); > PPL::NNC_Polyhedron myPoly = PPL:: NNC_Polyhedron(dim); > for(int j=0;j<noOfCons;j++) > { > PPL :: Linear_Expression expr; > PPL :: Linear_Expression ex; > for (int i=0;i<dim;i++) > { > PPL :: Variable var(i); > PPL :: Linear_Expression ep=var; > ep.set_coefficient(var,A(j,i)); > if(i==0) > expr=ep; > else { > PPL :: Linear_Expression e(expr+ep); > expr=e; > } > } > if(sgn==1){ //sgn==1 for <= > PPL :: Constraint c= operator<=(expr, b[j]); > myPoly.add_constraint(c); > } else if(sgn==2){ //sgn==2 for >= > PPL :: Constraint c= operator>=(expr, b[j]); > myPoly.add_constraint(c); > } else if(sgn==0){ //sgn==0 for = > PPL :: Constraint c= operator==(expr, b[j]); > myPoly.add_constraint(c); > } > } > } I see that after this I was able to perform some of the operation such as *contains, poly_hull_assign, etc.* However, I observe that this works only for *INTEGER* coefficients values. *But it is incorrect when the supplied A and b values are of type double.* Could you please direct what am I missing? Or am I going on the wrong direction in this creation? I would appreciates for any help in this direction. Thanking you Regards -- Praying for your good health, *Amit Gurung*

1 0