Copyright (C) 2001-2010 Roberto Bagnara
Copyright (C) 2010-2012 BUGSENG srl (http://bugseng.com)
Verbatim copying and distribution of this entire article is permitted
in any medium, provided this notice is preserved.
Parma Polyhedra Library NEWS -- history of user-visible changes
===============================================================
--------------------------------------------------------------------------
NEWS for version 1.0 (released on June 28, 2012)
--------------------------------------------------------------------------
New and Changed Features
========================
o Significant improvements have been obtained in both time and space
resource usage by the definition of data structures and algorithms
for the case of "sparse rows", i.e., sequences of coefficients
where most of the values are zero.
o The library fully supports two different representations for rows:
the "dense" representation is an array-like representation tailored
to sequences having most of their coefficients different from zero;
the "sparse" representation saves memory space (as well as CPU
cycles) when most of the coefficients in the sequence are zero.
o A generic interface allows for a seamless interaction between the
dense and the sparse row representation. Most library entities
(linear expressions, constraints, generators, congruences, and
their systems) can be built using either representation, specified
as a constructor's argument.
o As a by-product of this sparse/dense refactoring work, efficiency
improvements have been obtained even for those computations that
are still based on the dense row representation.
o Reasonable default values for the row representation are provided
for each library entity, automatically leading to significant
memory space savings even in old client/library code, e.g., when
dealing with constraint systems describing weakly relational
abstractions such as boxes and octagonal shapes.
o If desired, these default values can be customized to user's needs
by changing just a few lines of library code. For instance, the
constraint systems stored inside C_Polyhedron and NNC_Polyhedron
objects can be made to use the sparse representation by just
changing the following line in Polyhedron.defs.hh:
static const Representation default_con_sys_repr = DENSE;
to become
static const Representation default_con_sys_repr = SPARSE;
Bugfixes
========
o Fixed a bug affecting methods
bool BD_Shape::contains(const BD_Shape& y) const;
bool Octagonal_Shape::contains(const Octagonal_Shape& y) const;
whereby the wrong result was obtained when *this is an empty
weakly-relational shape and y is not empty.
o Fixed a bug affecting the PIP solver whereby a wrong result could have
been obtained if the input constraint system contained multiple linear
equality constraints.
--------------------------------------------------------------------------
NEWS for version 0.12.1 (released on April 16, 2012)
--------------------------------------------------------------------------
New and Changed Features
========================
o In the C, Java, OCaml and Prolog interfaces, modified the signature
of the function/method/predicate for setting the deterministic timeout
threshold. The new interfaces take two input values, named `unscaled'
and `scale', that are used to compute the threshold value as
`unscaled * 2^scale'.
o Added new Box methods
bool has_upper_bound(Variable var,
Coefficient& n, Coefficient& d, bool& closed) const;
bool has_lower_bound(Variable var,
Coefficient& n, Coefficient& d, bool& closed) const;
to query a non-empty box for the existence and value of its upper/lower
bound on variable `var'. The methods have been also added to all the
available language interfaces.
o Two BibTeX databases of papers related to the Parma Polyhedra Library
have been added to the distribution (in the `doc' directory).
Bugfixes
========
o Restored the support for deterministic timeouts in the PIP solver
(it was removed by accident in PPL 0.12).
o Minor documentation fixes.
o Portability improved.
--------------------------------------------------------------------------
NEWS for version 0.12 (released on February 27, 2012)
--------------------------------------------------------------------------
New and Changed Features
========================
o New configure options `--with-gmp=DIR', `--with-gmp-include=DIR' and
`--with-gmp-lib=DIR' supersede the (now removed) option
`--with-gmp-prefix'. (The old option never really worked; hopefully
this is the last change in this area.)
o New configuration option `--disable-documentation'. When specified
no new documentation is built: only the documentation already present
in the source tree is installed upon `make install'.
o The resolution process for PIP_Problem now better exploits the
integrality of parameters to simplify the newly generated tautological
constraints, the splitting constraints of decision nodes, and the
expressions defining artificial parameters.
o The implementations of the MIP and PIP solvers are based on a new
data structure leading to significant space and time savings when
the tableau matrix is sparse; the benchmarks of the ppl_lpsol demo
show an improvement on the average case, that grows when the toughest
tests in the benchmark suite are considered.
o When the `--check' option is used, the input data for demo ppl_lpsol
is perturbed the same way as GLPK does, thereby allowing for a
meaningful comparison of the results obtained.
o The input routine for PPL numeric datatypes has been extended to
accept the ISO9899 (C99) hexadecimal floating constant syntax.
o The Parma Watchdog Library has been merged into the
Parma Polyhedra Library.
Bugfixes
========
o Corrected a precision bug in methods
Box::upper_bound_assign(const Box&)
Box::upper_bound_assign_if_exact(const Box&)
whereby, provided any argument is an empty box and under other rather
specific conditions, the computed result was correct but unnecessarily
imprecise.
o Corrected a bug in method
Grid::relation_with(const Constraint&) const
whereby, under specific conditions, the method was creating invalid
Grid_Generator objects and providing an incorrect result.
--------------------------------------------------------------------------
NEWS for version 0.11.2 (released on February 27, 2011)
--------------------------------------------------------------------------
Bugfixes
========
o Fixed the semantics of the `--disable-fpmath' configure option
(which is equivalent to `--enable-fpmath=no'). It now disables all
floating point computations and, consequently, all numerical
abstractions based on floating point numbers.
o The PPL no longer overwrites the SIGILL signal handler.
o Minor documentation fixes.
o Portability improved.
--------------------------------------------------------------------------
NEWS for version 0.11.1 (released on February 20, 2011)
--------------------------------------------------------------------------
Bugfixes
========
o Corrected a problem in the simplification of PIP_Problem solution trees
whereby, under specific conditions, the node merging process produced
decision nodes that did not satisfy their class invariant.
o Corrected a precision bug in method
Octagonal_Shape::affine_image()
whereby in the case of an invertible affine transformation implementing
a variable sign symmetry (and optional translation), the computed result
was correct but unnecessarily imprecise.
o Corrected a problem in the input method for checked integers whereby,
under specific conditions, the input stream state bits were not updated.
The bug was only affecting builds using checked integer coefficients.
o Corrected a bug in the OCaml interface, which was affecting functions
ppl_Pointset_Powerset__get_disjunct.
o Corrected a couple of resource (re-)allocation problems that, under
specific conditions, could affect the correctness of Grid constructor
Grid::Grid(const Box& box)
and NNC_Polyhedron method
Polyhedron::generalized_affine_image().
o Corrected an efficiency bug in the C language interface function
ppl_Linear_Expression_add_to_coefficient().
o Corrected an efficiency bug in method
MIP_Problem::compute_generator().
o Corrected a bug affecting the input routine of ppl_lpsol, whereby
the inhomogeneous term of the objective function was disregarded.
o Corrected a bug affecting methods
Box::CC76_widening_assign(const T&, Iterator, Iterator)
Interval::CC76_widening_assign(const From&, Iterator, Iterator)
whereby a lower bound would not be computed correctly when the two
iterators specify an empty list of stop points.
o Fixed a bug affecting
Interval::Interval(const char* s)
whereby a wrong interval would be constructed if `s' denotes a number
that can only be represented as an infinity.
o Fixed a bug whereby the argument of all the methods
unconstrain(Variable var)
was not checked correctly for space dimension compatibility.
o Portability improved.
--------------------------------------------------------------------------
NEWS for version 0.11 (released on August 2, 2010)
--------------------------------------------------------------------------
New and Changed Features
========================
o New class PIP_Problem provides a Parametric Integer Programming
(PIP) problem solver (mainly based on P. Feautrier's
specification). The implementation combines a parametric dual
simplex algorithm using exact arithmetic with Gomory's cut
generation.
o New "deterministic" timeout computation facilities: it is now
possible to set computational bounds (on the library calls taking
exponential time) that do not depend on the actual elapsed time and
hence are independent from the actual computation environment (CPU,
operating system, etc.).
o New support for termination analysis via the automatic synthesis of
linear ranking functions. Given a sound approximation of a loop,
the PPL provides methods to decide whether that approximation
admits a linear ranking function (possibly obtaining one as a
witness for termination) and to compute the space of all such
functions. In addition, methods are provided to obtain the space
of all linear quasi-ranking functions, for use in conditional
termination analysis.
o New support for approximating computations involving (bounded)
machine integers. A general wrapping operator is provided that is
parametric with respect to the set of space dimensions (variables)
to be wrapped, the width, representation and overflow behavior of
all these variables. An optional constraint system can, when
given, improve the precision.
o All the PPL semantic objects provide new methods
void drop_some_non_integer_points(Complexity_Class)
void drop_some_non_integer_points(const Variables_Set&,
Complexity_Class)
which possibly tighten the object by dropping some points with
non-integer coordinates (for the space dimensions corresponding to
`vars'), within a certain computational complexity bound.
o New Linear_Expression methods
bool is_zero() const
bool all_homogeneous_terms_are_zero() const
return true if and only if `*this' is 0, and if and only if all the
homogeneous terms of `*this' are 0, respectively.
o New Linear_Expression methods
void add_mul_assign(Coefficient_traits::const_reference c, Variable v)
void sub_mul_assign(Coefficient_traits::const_reference c, Variable v)
assign linear expression *this + c * v (resp., *this - c * v) to *this,
while avoiding the allocation of a temporary Linear_Expression.
o For the PPL semantic objects, other than the Pointset_Powerset and
Partially_Reduced Product, there is a new method:
bool frequency(const Linear_Expression& expr,
Coefficient& freq_n, Coefficient& freq_d,
Coefficient& val_n, Coefficient& val_d)
This operator computes both the "frequency" (f = freq_n/freq_d)
and a value (v = val_n/val_d) closest to zero so that every point
in the object satisfies the congruence (expr %= v) / f.
o New reduction operator "Shape_Preserving_Reduction" has been added
to the Partially_Reduced_Product abstraction. This operator is
aimed at the product of a grid and a shape domain, allowing the
bounds of the shape to shrink to touch the points of the grid,
such that the new bounds are parallel to the old bounds.
o The Java interface has to be explicitly initialized before use by
calling static method Parma_Polyhedra_Library.initialize_library().
Initialization makes more explicit the exact point where PPL
floating point rounding mode is set; it also allows for the caching
of Java classes and field/method IDs, thereby reducing the overhead
of native method callbacks.
o The C and Java interfaces now support timeout computation facilities.
o Implementation of general (C and NNC) polyhedra speeded up.
o Implementation of the MIP solver speeded up.
o When the PPL has been configured with
CPPFLAGS="-DPPL_ARM_CAN_CONTROL_FPU=1", the library initialization
procedure checks that the FPU can indeed be controlled and fails if
that is not the case.
o New configure option `--with-gmp-prefix' supersedes the (now removed)
options `--with-libgmp-prefix' and `--with-libgmpxx-prefix'.
o New configuration option `--with-gmp-build=DIR' allows to use a
non-installed build of GMP in DIR.
Deprecated and Removed Methods
==============================
o All methods having a name ending in `_and_minimize' (e.g.,
add_constraints_and_minimize, poly_hull_assign_and_minimize, ...)
have been removed (they were deprecated in version 0.10).
Bugfixes
========
o Corrected a bug in maximize and mininimize optimization methods of
class template Pointset_Powerset, whereby the Boolean value true
(indicating successful optimization) was returned for empty powersets.
o Corrected a bug in method
bool NNC_Polyhedron::poly_hull_assign_if_exact(const NNC_Polyhedron&);
whereby some inexact NNC hulls were incorrectly flagged as exact.
--------------------------------------------------------------------------
NEWS for version 0.10.2 (released on April 18, 2009)
--------------------------------------------------------------------------
Bugfixes
========
o Correctly detect GMP 4.3.0.
o Fixed the C interface library version information.
o Test program tests/Polyhedron/memory1 disabled on the zSeries s390x
platform.
o Makefiles fixed so as to avoid failure of `make -n check'.
--------------------------------------------------------------------------
NEWS for version 0.10.1 (released on April 14, 2009)
--------------------------------------------------------------------------
New and Changed Features
========================
o Added support for cross compilation.
o The configuration script now explicitly checks that a recent enough
version of GNU M4 is available if at least one non-C++ interface is
enabled (in previous versions this check was not performed and
building the library could fail in a mysterious way).
o Robustness improved.
o Some packaging issues have been fixed.
o New macro PPL_DIRTY_TEMP_COEFFICIENT allows users of the C++
interface to decrease memory allocation overhead and to improve
locality whenever they need a temporary variable of type
`Coefficient'.
o The C++, C, Java and OCaml interfaces now provide utility functions
to format the textual representations of constraints, congruences
and so on. This makes it easy to code debugging prints with line
indentation and wrapping.
o The C interface now provides functions of the form
int ppl_io_asprint_Polyhedron(char** strp, P x)
where `P' is any opaque pointer to a const PPL object. These
functions print `x' to a malloc-allocated string, a pointer to
which is returned via `strp'.
o The OCaml interface can now be compiled to native code using ocamlopt.
o New configuration option `--with-mlgmp=DIR' allows to specify the
installation directory of the ML GMP package.
o The OCaml interface now supports timeout computation facilities
through functions ppl_set_timeout and ppl_reset_timeout. Moreover,
new functions ppl_Coefficient_is_bounded, ppl_Coefficient_min,
ppl_Coefficient_max and ppl_max_space_dimension have been added.
o The Prolog interfaces are no longer enabled by default in the
release tarballs (they are enabled by default in the Git versions).
Bugfixes
========
o Fixed a bug whereby `make check' was failing when the library was
configured with the `--disable-watchdog' option.
o Fixed a bug in method
bool Polyhedron::contains_integer_point() const;
whereby, under very specific conditions, an empty polyhedron is
incorrectly said to contain an integer point.
o Fixed a bug in method
Partially_Reduced_Product::time_elase_assign(y)
whereby, if the product y was not already reduced, the operation could
lose precision.
o Fixed a bug in the OCaml interface, which was affecting functions
ppl_Grid_generalized_affine_image_with_congruence
and
ppl_Grid_generalized_affine_preimage_with_congruence.
o Fixed a bug in the Grid class that affected the methods
Grid::bounds_from_above(), Grid::bounds_from_below(),
Grid::maximize() and Grid::minimize();
causing all of them to wrongly return true in certain cases where
the grid generators were not minimized.
o Fixed a bug whereby big-endian architectures were not properly
recognized by the configuration script.
o Fixed a bug in the Java/OCaml/Prolog interfaces, whereby
the method/function/predicate for dropping a disjunct from a
Pointset_Powerset object were returning an invalid iterator.
o Fixed a bug in method Octagonal_Shape::affine_image(var, expr)
whereby a wrong result was computed under specific conditions.
o Fixed a bug in the OCaml interface, whereby functions of form
ppl_..._widening_assign_with_tokens
and
ppl_..._extrapolation_assign_with_tokens
could return a wrong number of tokens.
o Fixed a bug in the OCaml interface, whereby functions that returned
an OCaml 'unit' type were returning the wrong value.
o Fixed several garbage collection related bugs in the OCaml interface.
--------------------------------------------------------------------------
NEWS for version 0.10 (released on November 4, 2008)
--------------------------------------------------------------------------
New and Changed Features
========================
The license
-----------
o The Parma Polyhedra Library is now released under the terms of the
version 3 (or later) of the GNU General Public License.
New and renamed classes
-----------------------
o The new class Octagonal_Shape provides an implementation of the domain
of octagonal shapes (including optimized algorithms and a provably
correct widening) as proposed by Roberto Bagnara, Patricia Hill,
Elena Mazzi and Enea Zaffanella in their SAS 2005 paper.
o A new abstraction called Box has been added. Geometrically
speaking, a Box represents a not necessarily closed, iso-oriented
hyperrectangle. This can also be seen as the smash product of `n'
not necessarily closed and possibly unbounded intervals, where `n'
is the space dimension of the box. The Box template class is
parametric with respect to a class of intervals.
o A generic implementation of intervals has been added. The template
class Interval is parametric on the type used to represent the
interval boundaries (all native integer and floating point types
are supported as well as unbounded integers and rational numbers
provided by GMP). Another class template type parameter allows for
the control of a number of other features of the class (such as the
ability to represent: open as well as closed boundaries, empty
intervals in addition to nonempty ones, intervals of extended
number families that contain positive and negative infinities,
plain intervals of real numbers and intervals of integer numbers).
The Interval class still needs a lot of work and both its
implementation and its interface are likely to change significantly:
it is released now because it is needed for the Box class and as a
kind of technology preview.
o The class LP_Problem has been renamed MIP_Problem and now supports
the solution of Mixed Integer (Linear) Programming problems.
Support has been added for the incremental solution of MIP
problems: it is now possible to add new space dimensions or new
constraints to the feasible region, as well as change the objective
function and the optimization mode, while still exploiting some of
the computational work done before these changes. Support has also
been added to change control parameters for the pricing method.
This allows a choice between the steepest edge pricing method,
either implemented with floating point numbers (default) or with
integer coefficients, and the textbook pricing method.
o The PPL semantic object Polyhedra_Powerset has been replaced by the
templatic object template Pointset_Powerset that can
take any (simple) PPL semantic object for the domain of its
disjuncts. In addition to the methods common to all the PPL
semantic objects, methods specific to this domain include:
void add_disjunct(const PS&),
void pairwise_reduce(),
void omega_reduce() const,
bool geometrically_covers(const Pointset_Powerset&) const,
bool geometrically_equals(const Pointset_Powerset&) const, and
bool simplify_using_context_assign(const Pointset_Powerset&).
o A new abstraction called Partially_Reduced_Product (PRP) has been
added. A PRP is a pair of two PPL semantic objects that is
parametric on the component domains and on a reduction operator.
The PPL currently provides three reduction operators and hence,
three different kinds of products:
- a Direct_Product where the reduction operator is the identity;
- a Smash_Product where the reduction operator shares emptiness
information between the components; and
- a Constraints_Product where the reduction operator refines each
component with the constraints satisfied by the other component.
The PRP class still needs a lot of work and both its implementation
and its interface are likely to change significantly: it is released
now as a kind of technology preview and any feedback is welcome.
New and renamed methods
-----------------------
o All PPL semantic objects can now be constructed from other simple
PPL semantic objects. All these constructors have an optional complexity
argument with default value allowing algorithms with any complexity to be
used.
o New methods
void restore_pre_PPL_rounding() and
void set_rounding_for_PPL()
allow the FPU rounding mode to be set to what it was before the
initialization of the PPL, and to set it (again) so that the PPL
abstractions based on floating point numbers work correctly, respectively.
o All PPL semantic objects now provide methods
void refine_with_constraint(const Constraint&),
void refine_with_congruence(const Congruence&),
void refine_with_constraints(const Constraint_System&), and
void refine_with_congruences(const Congruence_System&).
These methods are similar to the corresponding `add_' methods.
The difference is in the reaction policy when the argument
constraint/congruence is not optimally supported by the semantic
domain: while the `add_' methods will throw an exception, the
`refine_with_' methods will apply an upward approximation semantics.
o Default widening operators of the form:
void widening_assign(const ABSTRACTION&, unsigned*)
are now provided for all abstractions except for the Pointset_Powerset
abstractions.
o All PPL semantic objects now provide the method
int32_t hash_code() const
returning a 32-bit hash code for *this. If x and y are such that
x == y evaluates to true, so does x.hash_code() == y.hash_code().
o All PPL semantic objects now provide the methods
memory_size_type total_memory_in_bytes() const
memory_size_type external_memory_in_bytes() const
returning, respectively, the total size in bytes of the memory
occupied by the PPL object and the size in bytes of the memory
managed by the PPL object.
o For all the PPL semantic objects there are new methods:
static bool can_recycle_constraint_systems() and
static bool can_recycle_congruence_systems()
that indicate whether or not a PPL semantic object is able to recycle
constraints and congruences, respectively.
o For all PPL semantic objects there is a new method:
bool contains_integer_point() const
which checks if a PPL semantic object contains an integer point;
Note that this is not currently provided for the Partially_Reduced_Product
classes.
o For all PPL semantic objects there is a new method:
bool constrains(Variable) const
which checks if a dimension is constrained by a PPL semantic object;
o For all PPL semantic objects there are new methods:
void unconstrain(Variable)
void unconstrain(const Variables_Set&)
which assign, to a PPL semantic object, the cylindrification
of the object with respect to one (resp., a set) of its dimensions,
as defined by L. Henkin, J. D. Monk, and A. Tarski in Cylindric Algebras:
Part I (published by North-Holland in 1971).
o Several methods
bool is_topologically_closed() const
void topological_closure_assign()
that were provided for just some of the PPL semantic objects are now
uniformly available for all the objects.
o Methods using the Congruence and Congruence_System classes
such as
Congruence_System congruences() const,
Congruence_System minimized_congruences() const,
void add_congruence(const Congruence&),
void add_congruences(const Congruence_System&),
void add_recycled_congruences(const Congruence_System&), and
Poly_Con_Relation relation_with(const Congruence&).
that were just provided for the Grid domain are now provided for
all the PPL semantic objects.
o For the Grid class, as it is not always possible to obtain a
Pointset_Powerset object that is a finite linear partition of
the difference of two grids, we have added the method:
std::pair >
approximate_partition(const Grid&, const Grid&, bool&)
where the third argument is set to false if there is no
finite linear partition.
o In the Congruence class, for consistency with the Constraint class,
the methods is_trivial_true() and is_trivial_false() have been renamed
as is_tautological() and is_inconsistent(), respectively.
o The methods
bool Constraint_System::empty() const,
bool Generator_System::empty() const,
bool Congruence_System::empty() const, and
bool Grid_Generator_System::empty() const
return true if and only if the system in question is empty
(i.e., it has no constraints, generators, congruences or grid-generators,
respectively).
Deprecated and Removed Methods
==============================
o As all PPL semantic objects can now be constructed from boxes,
the constructors
template C_Polyhedron(const Box&, From_Bounding_Box),
template NNC_Polyhedron(const Box&, From_Bounding_Box),
template Grid(const Box&, From_Bounding_Box)
have been removed. Similarly, as boxes can be constructed from other
PPL semantic objects, the method
template
void shrink_bounding_box(Box&, Complexity_Class) const
has been removed from all the classes.
o The use of methods having a name ending in `_and_minimize' (e.g.,
add_constraints_and_minimize, poly_hull_assign_and_minimize, ...)
is now deprecated (see the core library documentation for an
explanation); their complete removal is planned for version 0.11.
o Class BD_Shape and Grid no longer provide methods such as
bds_hull_*, join_*, bds_difference_* and grid_difference_*. The
uniformly named methods upper_bound_* and difference_assign should
be used instead. For (C and NNC) polyhedra, the poly_hull_* and
poly_difference_assign methods have been kept for backward
compatibility (users should anyway prefer adopting the uniformly
named variants).
o For Grids, the PPL no longer supports covering boxes; hence the constructor
template Grid(const Box&, From_Covering_Box)
and also the method
template void get_covering_box(Box&) const
have been removed.
Other changes for the C++ interface
-----------------------------------
o All identifiers containing the strings `less_than_or_equal' or
`greater_than_or_equal', any case, have been renamed so as to contain
`less_or_equal' or `greater_or_equal', respectively.
A similar change also applies to the C interface (see below).
o The `ppl.hh' header file no longer defines macros not prefixed
by "PPL_".
o Users of the C++ interface of the library can now decide to disable
the automatic initialization mechanism of the PPL. To do so, the
preprocessor symbol PPL_NO_AUTOMATIC_INITIALIZATION should be
defined before including the header file. When automatic
initialization is disabled it is imperative to explicitly call the
new function
void Parma_Polyhedra_Library::initialize()
before using the library. The new function
void Parma_Polyhedra_Library::finalize() and
should also be called (to release a small amount of memory) when
done with the library.
Changes to the other language interfaces
----------------------------------------
o Support for language interfaces has been expanded to include
OCaml and Java. Thus the PPL now supports interfaces to
C++, C, Java, OCaml, Ciao Prolog, GNU Prolog, SICStus Prolog,
SWI Prolog, XSB Prolog and YAP Prolog.
o Most of the PPL semantic objects provided by the C++ interface
are also supported by all the non-C++ language interfaces. A few
domains (in particular, many of the possible Box instantiations)
are only available via the C++ interface.
o Almost all the public methods for the PPL semantic objects are
provided as methods/functions/predicates in the non-C++ language
interfaces with a uniform naming policy. In particular:
* in the C interface, the methods named
ppl_Polyhedron_{constraints,generators,congruences}
ppl_Polyhedron_minimized_{constraints,generators,congruences}
have been renamed
ppl_Polyhedron_get_{constraints,generators,congruences}
ppl_Polyhedron_get_minimized_{constraints,generators,congruences},
respectively;
* in the Prolog interfaces, the predicates
ppl_Grid_generalized_image_lhs_rhs/5 and
ppl_Grid_generalized_preimage_lhs_rhs/5
ppl_Grid_generalized_image/6 and
ppl_Grid_generalized_preimage/6
have been renamed as
ppl_Grid_generalized_image_lhs_rhs_with_congruence/5
ppl_Grid_generalized_preimage_lhs_rhs_with_congruence/5
ppl_Grid_generalized_image_with_congruence/6
ppl_Grid_generalized_preimage_with_congruence/6
respectively, so as to allow for /4 and /5, resp., versions.
o As already reported for the C++ interface, in the C interface,
all identifiers containing the strings `less_than_or_equal' or
`greater_than_or_equal', any case, have been renamed so as to contain
`less_or_equal' or `greater_or_equal', respectively.
o In the C interface it is no longer an error to call ppl_initialize()
or ppl_finalize() multiple times (this matches the behavior of the
other non-C++ language interfaces).
Documentation changes
---------------------
o The documentation for the library has been deeply reorganized and
split into several documents: besides the user and developer manuals
for the core library and its C++ interface, we now provide separate
user and developer manuals for each one of the other available
language interfaces (namely, C, Java, OCaml, and Prolog). It is
also possible to produce "configuration dependent" variants of the
non-C++ language interface manuals, where the contents of the
manual take into account the value of configuration option
`--enable-instantiations'.
All the manuals are provided in HTML, PDF and PostScript formats.
o New man pages libppl(3) and libppl_c(3) have been added. These
give short overviews on how to use the PPL in C++ and C programs,
respectively, on Unix-like operating systems.
Configuration changes
---------------------
o Several options have been added to the configuration script. These
allow to control the generated language interfaces, the floating
point instruction set to be used, the use of Valgrind during `make
check', the exclusion of some PPL-based programs from the build.
The README.configure file has been updated consequently.
Bugfixes
========
o Fixed bugs that prevented building the library on systems not supported
by the Parma Watchdog Library or when the `--disable-watchdog' configure
was used.
o Fixed a bug in Grid::constraints() and Grid::minimized_constraints()
that caused an internal assertion to fail when the grid had 0 space
dimensions.
o Fixed a bug in Linear_System::insert(const Linear_Row&) whereby a
wrong result could have been obtained when inserting a not necessarily
closed constraint/generator in an empty system having a higher space
dimension.
--------------------------------------------------------------------------
NEWS for version 0.9 (released on March 12, 2006)
--------------------------------------------------------------------------
New and Changed Features
========================
o The class Grid provides a complete implementation of the relational
domain of rational grids. This can represent all sets that can
be expressed by the conjunction of a finite number of congruence
equations. Operations provided include everything that is needed
in the field of static analysis and verification, including affine
images, preimages and their generalizations, grid-difference and
widening operators. This is the first time such a complete domain
is made freely available to the community.
o Several important portability improvements. Among other things,
it is now possible to build only the static libraries or only
the shared libraries. (Notice that some interfaces depend on
the availability of the shared libraries: these will not be built
when shared libraries are disabled.)
Bugfixes
========
o Fixed a bug whereby the SICStus Prolog interface could not be built
on x86_64.
o Fixed a bug in an internal method that, under some circumstances,
could cause a wrong result to be computed.
--------------------------------------------------------------------------
NEWS for version 0.8 (released on January 20, 2006)
--------------------------------------------------------------------------
New and Changed Features
========================
o The class template BD_Shape provides an implementation of the
abstract domain of bounded difference shapes. The template type
parameter T specifies the basic type used for the inhomogeneous term
of bounded difference constraints; it can be instantiated to either
GMP's unbounded precision types (mpq_class, mpz_class), native
floating point types (float, double), or native integer types
(8, 16, 32 and 64 bits wide).
o New class LP_Problem provides an implementation of the
primal simplex algorithm using exact arithmetic.
o The new program `ppl-config' returns information about the
configuration and the installed components of the PPL.
This greatly simplifies the task of writing makefiles and
automatic configuration scripts. A manual page for `ppl-config'
has also been added.
o New Autoconf function
AM_PATH_PPL([MINIMUM-VERSION, [ACTION-IF-FOUND [, ACTION-IF-NOT-FOUND]]])
allows to test the existence and usability of particular versions of the
PPL, defining macros containing the required paths. The simple addition
of, e.g.,
AM_PATH_PPL(0.8)
to an application's configure.ac file is all that is needed in most
cases. Paths to the installed version of the PPL will be detected,
the version number will be checked to be at least the one
indicated, the variables PPL_CPPFLAGS and PPL_LDFLAGS will be set
accordingly and a quick sanity check of the PPL installation will
be performed. For more complex tasks, the AM_PATH_PPL function
defines the `--with-ppl-prefix' and `--with-ppl-exec-prefix'
configure options (useful when the PPL is installed into a
non-standard place or when multiple versions of the PPL are
installed). AM_PATH_PPL also defines the `--disable-ppltest'
configure option to disable the quick sanity check. When something
fails, AM_PATH_PPL provides accurate indications about what went
wrong and how to fix it. See the sources in m4/ppl.m4 for more
information.
o The browse and print versions of the PS and PDF formats of the user
manual have been merged into single documents: ppl-user-0.8.pdf and
ppl-user-0.8.ps.gz. The equivalent developer reference documents
have also been merged.
o One of the possible values for the configuration option
`--enable-coefficients' has been renamed from `gmp' to `mpz'.
o New configuration option `--enable-interfaces' allows some or all of
the Prolog and C interfaces to be selectively enabled.
o Portability has been further improved.
o Added to C_Polyhedron (resp., NNC_Polyhedron) new method
bool poly_hull_assign_if_exact(const C_Polyhedron&)
(resp. bool poly_hull_assign_if_exact(const NNC_Polyhedron&))
and its synonym
bool upper_bound_assign_if_exact(const C_Polyhedron&)
(resp. bool upper_bound_assign_if_exact(const NNC_Polyhedron&)).
o Added new typedef `element_type' to template Polyhedra_Powerset,
which corresponds to the type of the underlying numeric domain.
o Output operators have been added for Generator::Type and
Constraint::Type.
o Class Bounding_Box has new method
Constraint_System Bounding_Box::constraints() const,
which returns the system of constraints.
o Class Bounding_Box has new widening methods
Bounding_Box::CC76_widening_assign(const Bounding_Box& y)
and
template
Bounding_Box::CC76_widening_assign(const Bounding_Box& y,
Iterator first,
Iterator last).
o All methods in class Determinate that are specific to the Polyhedra
template parameter have been dropped. If needed, they can still be
invoked through element().
o Method
bool Constraint_System::has_strict_inequalities() const
is now publicly accessible.
o Added Polyhedron methods difference_assign() and join_assign(),
behaving as poly_difference_assign() and poly_hull_assign(),
so as to have more uniform interfaces.
o The helper function widen_fun_ref() building a limited widening
function is now templatic even on the second argument (i.e., the
limiting constraint system). The template widening method
Polyhedra_Powerset::BHZ03_widening_assign()
no longer has a default value for the certificate parameter.
o The signatures of Polyhedron methods maximize() and minimize()
have been greatly simplified.
o The function template
template
bool check_containment(const PH&, const Polyhedra_Powerset&)
now works whenever there exists a lossless conversion mapping an
object of type PH into an NNC_Polyhedron (e.g., when PH = BD_Shape).
The same holds for methods
bool Polyhedra_Powerset::geometrically_covers()
and
bool Polyhedra_Powerset::geometrically_equals().
o Disjuncts can be added to an instance of Polyhedra_Powerset with
the new method
void add_disjunct(const PH& ph).
o The two generalized_affine_image() methods of class Polyhedron
are now matched by corresponding methods for computing preimages
of affine relations.
o Added to class Polyhedron the method
void bounded_affine_image(Variable v,
const Linear_Expression& lb,
const Linear_Expression& ub,
Coefficient_traits::const_reference d
= Coefficient_one())
computing the image of the polyhedron according to the
transfer relation lb/d <= v' <= ub/d.
Also added the corresponding method for computing preimages.
o The enumeration
Polyhedron::Degenerate_Kind
has been placed outside of class Polyhedron and renamed as
Degenerate_Element.
o New output operators in namespace IO_Operators:
std::ostream& operator<<(std::ostream&, const Constraint::Type&)
and
std::ostream& operator<<(std::ostream&, const Generator::Type&).
o Added to class Constraint the methods
bool is_tautological() const
and
bool is_inconsistent() const
returning true when the constraint is always or never satisfied,
respectively.
o Added to classes Constraint (resp., Generator) the method
bool is_equivalent_to(const Constraint& y) const
(resp., bool is_equivalent_to(const Generator& y) const)
which check for semantic equivalence of corresponding class
instances. Also made available the (semantic) comparison operators
`==' and `!='.
o The swap() methods of classes Linear_Expression, Constraint, Generator,
Constraint_System and Generator_System are now publicly accessible.
o Added to classes Constraint and Generator the methods
void ascii_dump(std::ostream& s) const
and
void ascii_load(std::istream& s) const.
o In classes Poly_Con_Relation and Poly_Gen_Relation the methods
void ascii_dump(std::ostream& s) const
and
void ascii_load(std::istream& s) const
are now publicly accessible.
o All classes which provide the method
void ascii_dump(std::ostream& s) const
now also provide the methods
void ascii_dump() const
and
void print() const.
These methods print to std::cerr the textual and user-level
representation (resp.) of the given object. This enables the
output of such object representations in GDB.
o New functions added to the C interface:
int ppl_Coefficient_is_bounded(void),
int ppl_Coefficient_min(mpz_t min),
int ppl_Coefficient_max(mpz_t max)
allow C applications to obtain information about the Coefficient
integer numerical type.
The new Prolog interface predicates ppl_Coefficient_is_bounded/0,
ppl_Coefficient_max/1 and ppl_Coefficient_min/1 provide the same
functionality.
o All predicates in the Prolog interface that require an input list
as an argument will now throw an exception if that argument is not
a list. Before, some predicates, such as
ppl_Polyhedron_remove_space_dimensions/2, would fail.
o In the Prolog interface, the names and arities of the "with_token"
widening and extrapolation predicates have been revised to
"with_tokens" with an extra argument and the functionality has been
revised to match more closely the corresponding C++ interface
operators.
o In the Prolog interface, the names and arities of the predicates
that create handles for new polyhedra have been revised to match
more closely the corresponding C and C++ interface operators. That
is, instead of having "c" and/or "nnc" as arguments to indicate the
topology of the polyhedron, the topologies are now part of the
names of the predicates.
o The SWI-Prolog interface allows now the exchange of unbounded numbers
between the PPL and Prolog applications. This requires SWI-Prolog
version 5.6.0 or later version. Previous versions of SWI-Prolog
are no longer supported.
o The YAP interface allows now the exchange of unbounded numbers
between the PPL and Prolog applications. This requires YAP
version 5.1.0 or later version. Previous versions of YAP
are no longer supported.
o The `ppl_lpsol' demo has now two more options: with `--enumerate' it
solves the given linear programming problem by vertex/point
enumeration; with `--simplex' (the default) it uses our simplex
implementation with exact arithmetic. The `ppl_lpsol' program,
which is only built if a suitable version of GLPK is available, is
installed into the directory (selectable at configuration time) for
user executables.
o Manual pages have been added for the ppl_lpsol and ppl_lcdd
programs.
o The new class BD_Shape as well as the "checked" native
coefficients selectable with the `--enable-coefficients' configure
options, are based on a very general and powerful notion of "number
family with a policy". This is made available to the users of the
PPL via the wrapper template class Checked_Number, where T is
the underlying numeric type (native integer or float of any width,
unbounded integer or rational) and `P' is a policy specifying
things such as whether to check for overflows and other
"exceptional" conditions and what to do in such cases. The policy
also specifies whether T should be extended with representations
for infinities or NAN (Not A Number) and default rounding modes.
A complete set of arithmetic functions and operators are provided:
they either use the default rounding mode or accept a rounding mode
as an extra parameter and, depending on the policy, may return a result
that indicates the relation that exists between the true mathematical
result and the (possibly approximate) computed result. Input/output
functions with the same properties (controlled rounding and indications
of the approximations) are also provided.
Bugfixes
========
o Fixed a bug in Polyhedra_Powerset::concatenate_assign() whereby
a temporary Polyhedra_Powerset object was created with the wrong
dimension.
o Corrected a memory leak bug in the demo ppl_lpsol.
o Corrected a bug in method
NNC_Polyhedron::minimized_constraints()
whereby an internal assertion might have been violated.
o Fixed a bug whereby calling the methods
Polyhedron::generalized_affine_image()
on an empty polyhedron could have resulted in an exception thrown.
o Fixed a bug whereby the occurrence of an `out of memory' error during
the allocation of a row of integer coefficients could have resulted
in a memory leak.
o Fixed a bug affecting the specialized constructors
Polyhedra_Powerset::
Polyhedra_Powerset(const Polyhedra_Powerset& y)
and
Polyhedra_Powerset::
Polyhedra_Powerset(const Polyhedra_Powerset& y)
whereby the newly built Polyhedra_Powerset object could have been
flagged as non-redundant even though it was containing redundant
disjuncts. Fixed a similar bug in generic constructor
Polyhedra_Powerset(const Constraint_System& cs)
that manifests when `cs' is denoting an empty polyhedron.
--------------------------------------------------------------------------
NEWS for version 0.7 (released on December 24, 2004)
--------------------------------------------------------------------------
New and Changed Features
========================
o The new configuration option `--enable-coefficients' allows for the
use of alternative (integral) coefficient types. Besides GMP
integers, the user can now use checked native integers (8, 16, 32
or 64 bits wide). The use of such native coefficients is
completely safe, since systematic (yet efficient) overflow
detection is performed and, in case of overflow, an exception is
raised. GMP coefficients are used by default.
o Significant efficiency improvements have been achieved everywhere.
o We now require GMP 4.1.3 or higher.
o The following classes have been renamed as indicated:
AskTell -> Ask_Tell
BoundingBox -> Bounding_Box
ConSys -> Constraint_System
GenSys -> Generator_System
Integer -> Coefficient
LinExpression -> Linear_Expression
Polyhedra_PowerSet -> Polyhedra_Powerset
PowerSet -> Powerset
SatMatrix -> Saturation_Matrix
SatRow -> Saturation_Row
o The helper function `widen_fun' has been renamed `widen_fun_ref'.
o New assignment operators allowing to obtain an NNC_Polyhedron from a
C_Polyhedron and the other way around. In the latter case, the
topological closure of the argument polyhedron is computed.
o New explicit constructors and assignment operators allowing to obtain a
Polyhedra_Powerset from a Polyhedra_Powerset
and the other way around. In the latter case, the topological closure
of the element polyhedra is computed.
o New explicit constructor Powerset::Powerset(const CS& d): if `d'
is not bottom, builds a powerset containing only `d'; builds the empty
powerset otherwise.
o New explicit constructor
Polyhedra_Powerset::Polyhedra_Powerset(const PH& ph): if `ph' is
not empty, builds a powerset containing only `ph'; builds the empty
powerset otherwise.
o New method
Polyhedra_Powerset::poly_difference_assign(const Polyhedra_Powerset& y)
assigns to `*this' the poly-difference of `*this' and `y'.
o All the public classes of the library have been endowed with methods
memory_size_type total_memory_in_bytes() const
and
memory_size_type external_memory_in_bytes() const
returning (lower bounds for) the total size in bytes of the memory
occupied by *this and of the memory managed by *this, respectively.
The type `memory_size_type' is a newly added unsigned integral type
suitable to the representation of such information.
o New method dimension_type Polyhedron::affine_dimension() returns
the affine dimension of *this (not to be confused with the dimension
of its enclosing vector space) or 0, if *this is empty.
o All the methods changing (i.e., adding, removing, mapping, etc.)
the dimensions of the vector space have been renamed so as to avoid
any possible ambiguity with the affine dimension of the modified object.
For instance,
Polyhedron::add_dimensions_and_embed(dimension_type m);
has been renamed as
Polyhedron::add_space_dimensions_and_embed(dimension_type m);
o The constructor C_Polyhedron(const NNC_Polyhedron& y) no longer
throws an exception if `y' is not topologically closed. Rather,
it constructs a C_Polyhedron representing the topological closure
of `y'.
o The following constructors have been made explicit:
Constraint_System::Constraint_System(const Constraint& c),
Generator_System::Generator_System(const Generator& g),
C_Polyhedron::C_Polyhedron(const Constraint_System& cs),
C_Polyhedron::C_Polyhedron(const Generator_System& cs),
C_Polyhedron::C_Polyhedron(Constraint_System& cs),
C_Polyhedron::C_Polyhedron(Generator_System& cs),
NNC_Polyhedron::NNC_Polyhedron(const Constraint_System& cs),
NNC_Polyhedron::NNC_Polyhedron(const Generator_System& cs),
NNC_Polyhedron::NNC_Polyhedron(Constraint_System& cs),
NNC_Polyhedron::NNC_Polyhedron(Generator_System& cs).
Polyhedra_Powerset::Polyhedra_Powerset(const Constraint_System& cs).
o Functions in the C interface that compute (space) dimensions
no longer return their result. The caller is now required to pass,
as an extra argument, a pointer to a memory area where the result
will be written. All the C interface functions now use the return
value to signal the success or failure of the requested operation.
o Now `make check' runs a number of tests with `ppl_lcdd', comparing
the results to expected ones.
o The `ppl_lcdd' demo is now able to parse problems produced by lrs,
i.e., where the number of rows of the matrix is omitted and replaced
by "*****" (five asterisks).
o The enumeration values of enum Complexity_Class have been renamed
POLYNOMIAL_COMPLEXITY, SIMPLEX_COMPLEXITY and ANY_COMPLEXITY.
o In the Prolog interface, the predicates
ppl_new_polyhedron_from_dimension/3 and
ppl_new_polyhedron_empty_from_dimension/3 have been replaced by a
single predicate ppl_new_polyhedron_from_space_dimension/4 where
the (extra) third argument indicates whether the polyhedron to be
created should be the universe or the empty polyhedron.
o As the unary plus operator is not in standard Prolog, '+'/1 in linear
expressions is no longer supported by the Prolog interface.
Bugfixes
========
o Fixed a bug that was causing an unwanted exception to be thrown
when adding to a C_Polyhedron some generators obtained from an
NNC_Polyhedron (even though no closure point was being added).
o Fixed a bug in the handling of empty generator systems having
a strictly positive space dimension.
o Fixed a bug that was affecting Polyhedra_PowerSet::geometrically_covers()
and Polyhedra_PowerSet::geometrically_equals().
o Method C_Polyhedron::H79_widening_assign() now widens the polyhedron
itself instead of the homogenized polyhedral cone representing it.
o Fixed a bug that was affecting Polyhedron::H79_widening_assign()
as well as all the limited and bounded extrapolation operators.
o Fixed a bug in Polyhedron::map_space_dimensions() that could manifest
itself when used with a partial function encoding permutation.
o Fixed a bug in the C interface function
ppl_new_Linear_Expression_with_dimension().
o Fixed a bug in the `ppl_lpsol' demo.
o Fixed a bug in Polyhedron::is_universe() that could manifest itself
when the polyhedron is described by a generator system that is
not in minimal form.
--------------------------------------------------------------------------
NEWS for version 0.6.1 (released on August 20, 2004)
--------------------------------------------------------------------------
New and Changed Features
========================
o Some packaging issues have been fixed.
o The documentation has been completed and improved.
o The methods
Polyhedra_PowerSet::semantically_contains(const Polyhedra_PowerSet&) and
Polyhedra_PowerSet::semantically_equals(const Polyhedra_PowerSet&)
have been renamed
Polyhedra_PowerSet::geometrically_covers(const Polyhedra_PowerSet&) and
Polyhedra_PowerSet::geometrically_equals(const Polyhedra_PowerSet& y),
respectively.
--------------------------------------------------------------------------
NEWS for version 0.6 (released on August 18, 2004)
--------------------------------------------------------------------------
New and Changed Features
========================
o New template classes Determinate, PowerSet, and Polyhedra_PowerSet.
The first two classes realize, in a completely generic way, the
determinate and powerset constructions described by Roberto Bagnara
in his 1998, Science of Computer Programming paper. The third class
is a specialization of the powerset construction on polyhedra.
The powerset construction comes with the generic widening technique
proposed by Roberto Bagnara, Patricia Hill and Enea Zaffanella
in their VMCAI 2004 paper. Actually, the Polyhedra_PowerSet class
provides the first genuine, non-trivial widening ever proposed
(let alone made available) on a domain of sets of convex polyhedra.
o New methods
void Polyhedron::expand_dimension(Variable, dimension_type) and
void Polyhedron::fold_dimensions(const Variables_Set&, Variable)
allow the easy realization of summary dimensions as proposed
by Denis Gopan and colleagues in their TACAS 2004 paper.
o A new `demos' directory has been added. In the `ppl_lcdd'
subdirectory, this contains a sort of clone of the cddlib test
program `lcdd', written using the PPL's C++ interface, together
with several example polyhedra in the formats that the program can
handle. The `ppl_lpsol' subdirectory contains another demo program
that solves linear programming problems by vertex/point
enumeration. This is written using the PPL's C interface and comes
with several example problems in the Mathematical Programming
System (MPS) format. In order to read MPS files, `ppl_lpsol' uses
the GNU Linear Programming Kit (GLPK); thus `ppl_lpsol' is only compiled
if a suitable version of GLPK is available.
o New macro PPL_VERSION expands to the version string of the library.
o New file README.configure provides additional information about
the configuration and compilation of the library.
o The documentation has been improved in various ways.
o The documentation for users, in PostScript, PDF and HTML formats,
is now installed in a standard place by `make install'.
o The variable `abandon_exponential_computations' has been renamed
`abandon_expensive_computations'.
o The methods
void Polyhedron::add_constraints(ConSys& cs),
void Polyhedron::add_generators(GenSys& gs),
bool Polyhedron::add_constraints_and_minimize(ConSys& cs), and
bool Polyhedron::add_generators_and_minimize(GenSys& gs)
have been renamed
void Polyhedron::add_recycled_constraints(ConSys& cs),
void Polyhedron::add_recycled_generators(GenSys& gs),
bool Polyhedron::add_recycled_constraints_and_minimize(ConSys& cs), and
bool Polyhedron::add_recycled_generators_and_minimize(GenSys& gs),
respectively. At the same time, the methods
void Polyhedron::add_constraints(const ConSys& cs),
void Polyhedron::add_generators(const GenSys& gs),
bool Polyhedron::add_constraints_and_minimize(const ConSys& cs), and
bool Polyhedron::add_generators_and_minimize(const GenSys& gs)
have been added. Corresponding changes have been made to the C and
Prolog interfaces.
o New methods Polyhedron::maximize() and Polyhedron::minimize()
for maximizing and minimizing a linear expression subject to the
polyhedron.
o New output operator in namespace IO_Operators:
std::ostream& operator<<(std::ostream&, const LinExpression&).
o The method Polyhedron::map_dimensions(const PartialFunction& pfunc)
has been significantly optimized for the case when `pfunc' is a
permutation. A simple "renaming" of the dimensions is now
extremely cheap.
o The function Parma_Polyhedra_Library::max_space_dimension() has been
given a new semantics and destiny: it returns the maximum space
dimension that _all_ the abstractions provided by the library can
handle. Each individual abstraction provides its versions of this
function. Thus, e.g., NNC_Polyhedron::max_space_dimension()
gives the maximum space dimensions an NNC_Polyhedron can handle.
o The type of output functions for the class Variable,
`Variable::Output_Function_Type', has been renamed
`Variable::output_function_type' and is now defined as
void output_function_type(std::ostream& s, const Variable& v).
In other words, `v' is now passed by const reference and not by value.
o Thanks to Bruno Haible, it is now possible to use versions of the
GMP library installed into nonstandard places. The new configure
options `--with-libgmp-prefix[=DIR]' and `--with-libgmpxx-prefix[=DIR]'
substitute the old (and not really working) options
`--with-gmp-includes=DIR' and `--with-gmp-dir=DIR'.
Bugfixes
========
o Fixed a bug in the C interface function ppl_Polyhedron_map_dimensions()
whereby a wrong result was computed.
o Fixed a bug in Polyhedron::poly_difference_assign(const Polyhedron&)
whereby a wrong result was computed.
o Fixed a bug in the Prolog interface predicate
ppl_Polyhedron_get_bounding_box/3 whereby a wrong result was
sometimes computed in the case of an empty polyhedron.
o Fixed a bug in the Prolog interface predicate
ppl_new_Polyhedron_from_bounding_box/3 whereby the predicate was
failing when given an empty bounding box.
--------------------------------------------------------------------------
NEWS for version 0.5 (released on April 28, 2003)
--------------------------------------------------------------------------
New and Changed Features
========================
o New methods Polyhedron::BHRZ03_widening_assign() and
Polyhedron::BHRZ03_limited_extrapolation_assign(). The BHRZ03 widening
is a novel widening that is always more precise than the one, by now
standard, we call H79.
o The novel "widening with tokens" technique improves on the good old
widening delay technique by refraining from widening only when
necessary. Precision is thus increased still guaranteeing
convergence. All widening operators can now be supplied with an
optional argument, recording the number of available tokens, which
is decremented when tokens are used.
o Two new methods have been defined that compute the image of
a polyhedron under an affine relation. The first method,
Polyhedron::generalized_affine_image(var, relsym, expr, denom),
generalizes the classical Polyhedron::affine_image() method by allowing
`relsym' to denote any of the relations <, <=, =, >=, >.
The second method, Polyhedron::generalized_affine_image(lhs, relsym, rhs),
is a variant where an arbitrary linear expression `lhs' is allowed to
occur on the left-hand side of the affine relation.
o New constructors to build polyhedra from read-only constraint and
generator systems: C_Polyhedron(const ConSys&),
C_Polyhedron(const GenSys&), NNC_Polyhedron(const ConSys&), and
NNC_Polyhedron(const GenSys&). In the C interface the functions
taking non-const arguments named ppl_new__from_~~ have been
renamed ppl_new__recycle_~~~~, where is either "C" or "NNC",
and ~~~~ is either "ConSys" or "GenSys". The old names have been
given to the new const functions.
o New function LinExpression& operator*=(LinExpression&, const Integer&)
to multiply (in place) an expression by a scalar.
o The methods Polyhedron::check_empty() and Polyhedron::check_universe()
have been renamed is_empty() and is_universe(), respectively.
o New method bool Polyhedron::is_disjoint_from(const Polyhedron& y)
returning true if and only `*this' and `y' are disjoint.
o New methods bool Polyhedron::add_constraint_and_minimize(const
Constraint&) and bool Polyhedron::add_generator_and_minimize(const
Generator&) to add a constraint or a generator and minimizing the
result at the same time.
o New method: template
void Polyhedron::map_dimensions(const PartialFunction&).
This allows to rename the dimensions of a polyhedron according
to a partial function mapping dimensions to dimensions.
o New function LinExpression operator+(const LinExpression&): previously
an expressions like `+x2-x3-x4 <= 0' could not constitute valid syntax
for a constraint.
o New type `dimension_type': an unsigned integral type for representing
space dimensions.
o New function dimension_type max_space_dimension():
returns the maximum space dimension this library can handle.
o New function dimension_type not_a_dimension():
returns a value that does not designate a valid dimension.
o The method Polyhedron::add_dimensions_and_constraints(ConSys&)
has gone. A similar functionality is provided by the new method
Polyhedron::concatenate_assign(const Polyhedron&). The same change
has, of course, been performed on all the PPL interfaces.
o New macros PPL_VERSION_MAJOR, PPL_VERSION_MINOR, PPL_VERSION_REVISION,
and PPL_VERSION_BETA allow the client application to adapt to different
versions of the library.
o New function const char* version() returns a character string
containing the PPL version.
o New function const char* banner() returns a character string
containing information about the PPL version, the licensing,
the lack of any warranty whatsoever, the C++ compiler used
to build the library, where to report bugs and where to look
for further information.
o The Prolog interface now supports also Ciao Prolog and XSB.
o The C and Prolog interfaces have been extended so as to make more of
the library's functionality available to Prolog and C users.
o Timeout computation facilities have been added to the Prolog interfaces:
new predicates ppl_set_timeout_exception_atom/1,
ppl_timeout_exception_atom/1, ppl_set_timeout/1, ppl_reset_timeout/0.
o Many efficiency improvements have been achieved. Part of these have
been obtained by increasing the degree of "laziness" of the library.
o Many portability and standard-conformance improvements: the library
can now be compiled with GNU g++, Intel C++ Compiler 7.0 for Linux,
and Comeau C/C++ 4.3.0.1 Compiler Front-End; the library has also
been tested on a variety of platforms.
o The functions
Polyhedron::operator<=(const Polyhedron&, const Polyhedron&),
Polyhedron::operator>=(const Polyhedron&, const Polyhedron&),
Polyhedron::operator<(const Polyhedron&, const Polyhedron&), and
Polyhedron::operator>(const Polyhedron&, const Polyhedron&)
have been removed. The methods
Polyhedron::contains(const Polyhedron&) and
Polyhedron::strictly_contains(const Polyhedron&)
provide the same functionality.
o The method Polyhedron::limited_H79_widening_assign() has been renamed
Polyhedron::limited_H79_extrapolation_assign(). From now on, the name
`widening' is reserved for operators that come with a convergence
guarantee (i.e., with the ability of turning infinite chains to finite
ones). Upper bound operators without such a guarantee contain the word
`extrapolation' in their name.
o The renamed method Polyhedron::limited_H79_extrapolation_assign()
takes the constraint system argument by const reference (in the
old Polyhedron::limited_H79_widening_assign() that argument was
passed by non-const reference).
o We now require GMP 4.1.2 or higher.
o In conformance with the C++ standard [17.4.3.1.2], in all the
identifiers exported by the C interface, any occurrence
of "__" (double underscore) has been replaced by "_" (underscore).
o Added a parameter to Polyhedron::shrink_bounding_box(): this specifies
the complexity class of the algorithm to be used.
o All the input/output operators have been confined into namespace
Parma_Polyhedra_Library::IO_Operators. This way they do not conflict
with the operators the user might want to define.
o The operator Constraint operator>>(const Constraint&, unsigned int)
has been removed.
o The method
Polyhedron::poly_difference_assign_and_minimize(const Polyhedron&)
has been removed.
Bugfixes
========
o Fixed a bug in operator-=(LinExpression&, const LinExpression&)
whereby we computed a wrong result in some circumstances.
o Fixed a bug in method Polyhedron::minimized_constraints() that,
under some circumstances, could cause a wrong result or a program
crash.
--------------------------------------------------------------------------
NEWS for version 0.4.2 (released on October 4, 2002)
--------------------------------------------------------------------------
Bugfixes
========
o Fixed a bug in method Polyhedron::add_generator(const Generator&)
whereby we were not adding the corresponding closure point when adding
a point to an empty NNC polyhedron.
o Fixed a bug in method GenSys::insert(const Generator&) whereby the
insertion of a generator into an empty generator system might be
mishandled.
o Fixed a bug in method Polyhedron::operator<=(const Polyhedron&)
whereby the lines of the polyhedron were handled improperly.
o Fixed a bug in a private method used to implement public method
Polyhedron::relation_with(const Generator& g),
whereby a wrong result was obtained when `g' was a line.
o Fixed a bug in methods Polyhedron::affine_image() and
Polyhedron::affine_preimage(), whereby a wrong result could be
obtained when using a negative denominator for the affine expression.
o Fixed a bug in methods Polyhedron::minimized_constraints() and
Polyhedron::minimized_generators(), whereby a library invariant
was violated when calling these methods on a zero-dimensional space
universe NNC polyhedron.
--------------------------------------------------------------------------
NEWS for version 0.4.1 (released on July 30, 2002)
--------------------------------------------------------------------------
Bugfixes
========
o Fixed a bug in Polyhedron::poly_difference_assign(const Polyhedron& y)
whereby the equality constraints of `y' were ignored (the bug was
affecting both C and NNC computations).
o Fixed a bug in Polyhedron::operator=(const Polyhedron& y). This bug,
which is triggered in some cases when `y' is empty, should only affect
versions of the library obtained with the `--enable-assertions'
configuration flag.
o Fixed a bug in Polyhedron::check_universe(), which was returning
the wrong result when called on a zero-dim universe polyhedron.
o Fixed a bug in NNC_Polyhedron::NNC_Polyhedron(ConSys& cs) whereby
an invariant was violated in case `cs' was empty. This bug
should only affect versions of the library obtained with the
`--enable-assertions' configuration flag.
--------------------------------------------------------------------------
NEWS for version 0.4 (released on July 1, 2002)
--------------------------------------------------------------------------
New and Changed Features
========================
o Added full support for Not Necessarily Closed (NNC) polyhedra:
- creation of strict inequality constraints and mixed constraint
systems;
- creation of closure points and extended generator systems;
- added classes C_Polyhedron (for the representation of Closed
polyhedra) and NNC_Polyhedron (the user no longer can create
Polyhedron objects);
- added topology compatibility checks to avoid mixing objects of
the two kinds;
- added explicit constructors to create a polyhedron of a given
topology kind starting from a polyhedron of the other kind;
- added methods Polyhedron::is_topologically_closed() and
Polyhedron::topological_closure_assign();
- implemented methods Polyhedron::minimized_constraints() and
Polyhedron::minimized_generators() to obtain minimal
descriptions, both for closed and for NNC polyhedra.
o New method Polyhedron::time_elapse_assign(const Polyhedron&):
it computes the time-elapse operation defined in
N. Halbwachs, Y.-E. Proy, and P. Roumanoff.
Verification of real-time systems using linear relation analysis.
Formal Methods in System Design, 11(2):157-185, 1997.
o New method Polyhedron::is_bounded(): it returns true if and only
if the polyhedron is bounded, i.e., finite.
o New methods Polyhedron::bounds_from_above(const LinExpression& e)
and Polyhedron::bounds_from_below(const LinExpression& e): they
return true if and only if the linear expression `e' is bounded
from above/below in the polyhedron.
o New, complete C interface. As a demo, a toy solver for pure linear
programming problems has been implemented using this interface.
Notice that solving linear programming problems is completely
outside the scope of the library. As a consequence the toy provided
as a demo is only a toy provided as a demo.
o Revised and completed Prolog interface:
- now supporting GNU Prolog, SICStus Prolog, SWI-Prolog and YAP.
- all predicates have been renamed to match their intended
semantics;
- arguments have been reordered where necessary so as to follow the
rule "input arguments before output arguments";
- predicates added so that all the public methods for Polyhedra in
the C++ library are available as Prolog predicates;
- the interface has been extended to allow for closed and not
necessarily closed polyhedra.
o Added support for timeout-guarded operations. It is now possible
for client applications to safely interrupt any exponential
computation paths in the library and get control back in a time
that is a linear function of the space dimension of the object
(polyhedron, system of constraints or generators) of highest
dimension on which the library is operating upon.
o The methods named convex_hull_* and convex_difference_*
have been renamed poly_hull_* and poly_difference_*.
o All methods named *_and_minimize() now return a Boolean
flag that is false if the result is empty.
o All method and variable names containing the word "vertex"
have been replaced by names containing the word "point"
(some previous uses of the word "vertex" were not formally correct).
o The methods Polyhedron::includes(const Generator&) and
Polyhedron::satisfies(const Constraint&) have been removed,
as well as the enumeration GenSys_Con_Rel.
They have been replaced by the new methods
Polyhedron::relation_with(const Generator&) and
Polyhedron::relation_with(const Constraint&),
which return values of the new enumeration types
Relation_Poly_Gen and Relation_Poly_Con, respectively.
o The method Constraint::coefficient(void) has been renamed
to Constraint::inhomogeneous_term(void).
o The methods Polyhedron::widening_assign() and
Polyhedron::limited_widening_assign() have been renamed
Polyhedron::H79_widening_assign() and
Polyhedron::limited_H79_widening_assign(), respectively.
This emphasizes the fact that they implement extensions
of the widenings introduced in
N. Halbwachs.
Determination Automatique de Relations Lineaires
Verifiees par les Variables d'un Programme.
These de 3eme cicle d'informatique,
Universite scientifique et medicale de Grenoble,
Grenoble, France, March 1979.
and described in
N. Halbwachs, Y.-E. Proy, and P. Roumanoff.
Verification of real-time systems using linear relation analysis.
Formal Methods in System Design, 11(2):157-185, 1997.
o The library no longer calls abort(): appropriate exceptions
are always thrown instead.
Bugfixes
========
o Fixed a bug whereby creating a point with a negative denominator
caused the library to misbehave.
o Fixed a bug in Polyhedron::add_constraints(ConSys&) whereby
we could have created constraint systems having rows with
different capacities.
o Several other bugs have been fixed.
--------------------------------------------------------------------------
NEWS for version 0.3 (released on February 26, 2002)
--------------------------------------------------------------------------
New Features
============
o The library has been ported under Libtool: it is now possible
to build dynamic libraries as well.
o We now use the integer C++ class of GMP to represent the
coefficients of constraints and generators, instead of our own
(very much inferior) Integer class. The drawback is that we
now require GMP 4.0.1 or higher.
o New methods Polyhedron::convex_difference_assign(const Polyhedron&) and
Polyhedron::convex_difference_assign_and_minimize(const Polyhedron&).
They assign to `*this' the convex hull of the set-theoretic difference
of `*this' and the argument (possibly non minimized or minimized,
respectively).
o The method Polyhedron::add_generators(GenSys&) is now lazy,
i.e., no minimization is performed. Adding generators and
minimizing at the same time is provided by the method
Polyhedron::add_generators_and_minimize(GenSys&).
These methods now throw an exception if the resulting
polyhedron would be illegal.
o Some performance improvements.
o Added more test programs.
Bugfixes
========
o Fixed Polyhedron::satisfies(const Constraint&) and
Polyhedron::affine_image().
o Polyhedron::limited_widening_assign(const Polyhedron&, ConSys&)
was erroneously returning a (random) Boolean: it is now a void
method.
--------------------------------------------------------------------------
NEWS for version 0.2 (released on November 14, 2001)
--------------------------------------------------------------------------
New Features
============
o Massive API changes. This would not normally be called "a feature",
but the old API was very wrong in a number of ways. More API changes
are to be expected for the next few releases.
o All user-accessible library methods are now guarded by suitable
sanity checks. Exception are thrown whenever the library is not
called in the intended way.
o A SICStus Prolog interface is now available. It comes with a somewhat
interesting demo: a toy CLP(Q) interpreter.
o Greatly improved documentation.
Bugfixes
========
o Many, many more than we would like to admit.
--------------------------------------------------------------------------
NEWS for version 0.1 (released on October 24, 2001)
--------------------------------------------------------------------------
New Features
============
o The library has been released under the GNU General Public License.
~~