Age | Commit message (Collapse) | Author |
|
- Splits the functionality of having a sparse matrix of Ts and a solved matrix of rationals in tableau.
- Splits ArithVarSet into DenseMap and CDDenseSet and simplifies the code.
- No performance loss!
|
|
Here are a summary of the changes:
- Adds CDMaybe and CDRaised in cdmaybe.h
- Add test for congruence over arithmetic terms and constants
- Renames DifferenceManager to CongruenceManager
- Changes a number of internal details for CongruenceManager
|
|
Below is a highlight of the changes:
- This introduces a new normal form to arithmetic.
-- Equalities and disequalities are in solved form.
Roughly speaking this means: (= x (+ y z)) is in normal form.
(See the comments in normal_form.h for what this formally requires.)
-- The normal form for inequality atoms always uses GEQ and GT instead of GEQ and LEQ.
Integer atoms always use GEQ.
- Constraint was added to TheoryArith.
-- A constraint is a triple of (k x v) where:
--- k is the type of the constraint (either LowerBound, UpperBound, Equality or Disequality),
--- x is an ArithVar, and
--- v is a DeltaRational value.
-- Constraints are always attached to a ConstraintDatabase.
-- A Constraint has its negation in the ConstraintDatabase [at least for now].
-- Every constraint belongs to a set of constraints for each ArithVar sorted by the delta rational values.
-- This set can be iterated over and provides efficient access to other constraints for this variable.
-- A literal may be attached to a constraint.
-- Constraints with attached literals may be marked as being asserted to the theory (sat context dependent).
-- Constraints can be propagated.
-- Every constraint has a proof (sat context dependent).
-- Proofs can be explained for either conflicts or propagations (if the node was propagated). (These proofs may be different.)
-- Equalities and disequalities can be marked as being split (user context dependent)
- This removes and replaces:
-- src/theory/arith/arith_prop_manager.*
-- src/theory/arith/atom_database.*
-- src/theory/arith/ordered_set.h
- Added isZero(), isOne() and isNegativeOne() to Rational and Integer.
- Added operator+ to CDList::const_iterator.
- Added const_iterator to CDQueue.
- Changes to regression tests.
|
|
- Improved the checks in AssertLower and AssertUpper so that redundant bounds cause less work.
- Because of the above change, d_constantIntegerVariables now cannot have duplicate elements enqueued. This allows removing d_varsInDioSolver.
- Fix to an assertion in CDQueue.
- Implements a CDArithVarSet using a vector of booleans and CDList.
- Refactored ArithVar out of arith_utilities.h. Miscellaneous cleanup of arithmetic.
|
|
This is a significant refactoring of code.
- r2820
-- Refactors Simplex so that it does significantly fewer functions.
-- Adds the LinearEqualityModule for handling update and pivotAndUpdate and other utility functions that require access to both the tableau and partial model.
-- Some of the code for propagation has moved to TheoryArith.
-r2826
-- Small changes to documentation and removes the no longer implemented deduceLowerBound() and deduceUpperBound().
- r2827
-- Adds isZero() to Rational. Adds cmp to DeltaRational.
- r2831
-- Refactored comparison to upper and lower in the partial model to use DeltaRational::cmp.
-- Refactored AssertUpper and AssertLower in TheoryArith to include functionality that has weaseled into TheoryArith::assertionCases.
|
|
now supports propagating equalities when a slack variable corresponding to a difference of shared terms must be 0. Similarly disequalities are propagated when these variables cannot be zero.
|
|
infrastructure, and takes care not to affect CVC4's performance on LRA
benchmarks.
|
|
theory of arithmetic.
* This code has been partially tested. (My testing situation is currently not so great.) The code for testing not preregistering equalities can be compile time enabled by setting the boolean turnOffEqualityPreRegister. Don't be shocked by slowdowns or failures. This does pass make regress as well as a fresh checkout does. (The Mac version has issues.)
* I need to disable the permanent row removal heuristic by default. We need to discuss why this needs to happen. We should probably detect pure QF_LRA/QF_RDL problems and enable this when this can safely be done.
* I have disabled the arithmetic rewrite equality flag. This code needs to be added to the parser.
* For all of the above changes, I have annotated the code with the key word BREADCRUMB.
* I have renamed ArithUnatePropagator to ArithAtomDatabase.
|
|
- This adds code for bounds refinement, and conflict weakening.
- This adds util/boolean_simplification.h.
- This adds a propagation manager to theory of arithmetic.
- Propagation is disabled by default.
- Propagation can be enabled by the command line flag "--enable-arithmetic-propagation"
- Propagation interacts *heavily* with rewriting equalities, and will work best if the command line flag "--rewrite-arithmetic-equalities" is enabled.
|
|
|
|
- Adds ArithStaticLearner. Consolidates and cleans up the code for static learning in arithmetic. Static learning is now associated with a small amount of state between calls. This is used to track the data for the miplib trick. The goal is to make this inference work without relying on the fact that all of the miplib problem is asserted under the same AND node.
- This commit contains miscellaneous other arithmetic cleanup.
|
|
an abstraction to the previous priority queues representing the 2 different pivoting rules.
|
|
Should have been removed earlier.
|
|
only substantive change is that UnatePropagator no longer uses attributes. The rest is comments and other beatification.
|
|
- Preparing to remove row ejection from the code base!
- Checks for conflicts immediately after a pivot to avoid potentially wasteful search.
- Added arithvar_set.h. This replaces ArithVarSet that was previously in the Tableau, and ArithVarDenseSet.
- Removes variables that have no preregistered bounds during presolve().
- Theory::isLeafOf() currently returns true for atoms. (I was unaware.) I modified Variable::isMember() to account for this exclude atoms.
- Added statistics all over the place.
This commit effects both boolean search and simplex search so expect running times to go all over the place. The time differences should be roughly as follows:
http://goedel.cims.nyu.edu/regress-results/compare_jobs.php?job_id=1486&reference_id=1447&p=10&category=1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,29
|
|
in a better way; fix arith Makefile
|
|
done by way of the intermediate branch arith-prop-tmp. Both arith-prop-opt and arith-prop-tmp will now be phased out.
|
|
into its own class SimplexDecisionProcedure. Implements about 1/2 of the pivoting rule from Alberto's thesis (section2.5.3).
|
|
ReducedRowVector. A RowVector is an array of ArithVar and Rational pairs. (This replaces a map based implementation in Row.) ReducedRowVector is a RowVector with a notion of having a basic variable. The Tableau is now a collection of ReduceRowVector's. A major difference between ReducedRowVectors and Rows is that the iterator now includes the basic variable and its coefficient (always -1). Before only nonbasic members were accessible by the iterator.
|
|
general ArithVarDenseSet. Renamed NextArithRewriter to ArithRewriter.
|
|
|
|
* src/theory/arith/normal_form.{h,cpp} contains the description for the new
normal form as well as utilities for dealing with the normal form.
* src/theory/arith/next_arith_rewriter.{h,cpp} contains the new rewriter.
The new rewriter implements preRewrite() and postRewrite() for arithmetic.
* src/theory/arith/arith_rewriter.{h,cpp} have been removed.
* TheoryArith::rewrite() has been removed.
* Arithmetic with the new normal form outperforms the trunk where the branch
occurred (-r797) on 46% of the examples in QF_LRA. (33% have no noticeable
difference.) Some important optimizations are stilling pending to the code
for handling the new normal form. (Bug 196.)
|
|
|
|
building with CLN or with GMP, the contrib/switch-config script
(enabling "fast switching" of different configurations in the same
builds/ directory), and also some minor changes.
./configure --with-gmp (or --without-cln) forces building with GMP
and doesn't even look for CLN. Configure fails if GMP isn't installed.
./configure --with-cln (or --without-gmp) forces building with CLN
and doesn't even look for GMP. Configure fails if CLN isn't installed.
./configure [no arguments] will detect what's installed. CLN is
default, if it isn't installed, or is too old, GMP is looked for (and
configure fails if neither is available).
It is an error to specify --with-gmp --with-cln (or --without-* for
both) at the same time.
Building with CLN (whether forced or detected) adds a note to the
configure output mentioning the fact that the build of CVC4 will be
linked against a GPLed library and notifying the user of the
--without-cln option.
Building with GMP (whether forced or detected) affects the build
directory, so CLN and GMP builds are kept separate.
./configure --with-cln debug builds in builds/$arch/debug
./configure --with-gmp debug builds in builds/$arch/debug-gmp
The final binaries are linked explicitly against either gmp or cln,
but not both. If linked against cln, cln pulls in gmp as a
dependency, so the result will be linked against both.
=== Details that you probably don't care about ===
The headers src/util/{integer,rational}.h are generated from the
corresponding .in versions. A user installing a CVC4-devel package
will get the headers for rational and integer that match the library
that s/he installs.
The preprocessor #defines CVC4_GMP_IMP and CVC4_CLN_IMP are added to
cvc4autoconfig.h. Only one is ever #defined. cvc4autoconfig.h
doesn't need to be #included directly; you get it through #including
cvc4_private.h (or the parser version).
AM_CONDITIONALs are also defined so that Makefiles get the cln/gmp
configuration. AC_SUBSTs are defined so that public headers (see
src/util/{integer,rational}.h.in) can use the setting.
*Public* headers that need to depend on the cln/gmp configuration
can't use cvc4autoconfig.h, because we're keeping that in the private,
internal-only space, never to be installed on users' machines. Here,
something special is required, like the configure-level generation of
headers that I used for src/util/{integer,rational}.h.in.
Tim's Integer and Rational wrappers are the only bits of code that
should care which library is used (and also src/util/configuration.h,
which gives the user of the library information about how CVC4 is
built), and possibly some unit tests (?).
|
|
|
|
expect a little turbulence. This commit will not compile. There will be a second commit that fixes this in a moment. I am delaying a change to avoid svn whining about a conflict.
|
|
|
|
|
|
fixes and organizational changes. The theory is now enabled to perform checking in the TheoryEngine. This draft can now solve 2 new regression tests test/regress/regress0/ineq_slack.smt and test/regress/regress0/ineq_basic.smt. There is also a small bug fix inside src/expr/attribute.h.
|
|
|
|
This should not affect other parts of the system. This code is untested, and is volatile. It is going to be improved in the future so don't pick on it too much. I am merging this code in its current state back into the main trunk because it was getting unruly to keep merging the updated trunk back into the branch.
|
|
instead of assuming it's atomic based on kind. Atomicity is
determined at node building time. Fixes bug #81. If this is
determined to make node building too slow, we can allocate another
attribute "AtomicHasBeenComputed" to lazily compute atomicity.
* TheoryImpl<> has gone away. Theory implementations now derive from
Theory directly and share a single RegisteredAttr attribute for term
registration (which shouldn't overlap: every term is "owned" by
exactly one Theory). Fixes bug #79.
* Additional atomicity tests in ExprBlack unit test.
* More appropriate whitebox testing for attribute ID assignment
(AttributeWhite unit test).
* Better (and more correct) assertion checking in NodeBuilderBlack.
* run-regression script now checks exit status against what's provided
in "% EXIT: " gesture in .cvc input files, and stderr against
"% EXPECT-ERROR: ". These can be used to support intended failures.
Fixes bug #84. Also add "% EXIT: " gestures to all .cvc regressions
in repository.
* Solved some "control reaches end of non-void function" warnings in
src/parser/bounded_token_buffer.cpp by replacing
"AlwaysAssert(false)" with "Unreachable()" (which is known
statically to never return normally).
* Regression tests now use the cvc4 binary under
builds/$(CURRENT_BUILD)/src/main instead of the one in bin/ which
may not be properly installed yet at that point of the build.
(Partially fixes bug #46.)
* -fvisibility=hidden is now included by configure.ac instead of each
Makefile.am, which will make it easier to support platforms
(e.g. cygwin) that do things a different way.
* TheoryUF code formatting. (re: my code review bug #64)
* CDMap<> is leaking memory again, pending a fix for bug #85 in the
context subsystem. (To avoid serious errors, can't free context
objects.)
* add ContextWhite unit test for bug #85 (though it's currently
"defanged," awaiting the bugfix)
* Minor documentation, other cleanup.
|
|
recognize an instantiation of the join conversion/copy ctor with
ref_count = ref_count_1 as a copy constructor. Problems with
reference counts ensue.
* src/theory/theory.h, src/theory/theory.cpp: Theory base
implementation work. Changed from continuation-style theory calls
to having an data member for the output channel. registerTerm() and
preRegisterTerm() work.
* src/theory/output_channel.h, src/theory/theory.h,
src/theory/theory.cpp, src/theory/uf/theory_uf.h,
src/theory/uf/theory_uf.cpp: merged ExplainOutputChannel into
OutputChannel.
* test/unit/expr/node_black.h: remove testPlusNode(),
testUMinusNode(), testMultNode().
* src/expr/attribute.h: new facilities ManagedAttribute<> and
CDAttribute<>, and add new template parameters to Attribute<>. Make
CDAttribute<>s work with context manager.
* src/expr/attribute.h, src/expr/node_manager.h: VarNameAttr and
TypeAttr are now "owned" (defined) by the NodeManager. The
AttributeManager knows nothing of specific attributes, it just as
all the code for dealing generically with attributes.
* test/unit/expr/node_white.h: test new attribute facilities.
* src/expr/soft_node.h: removed: We now have TNode, so SoftNode goes
away.
* src/theory/Makefile.am: fixed improper linking of theories
* src/theory/theory_engine.h: some implementation work (mainly stubs
for now, just to make sure TheoryUF can be instantiated properly,
etc.)
* src/expr/node_value.cpp, src/expr/node_value.h: move a number of
function implementations to the header and make them inline
* src/expr/node_manager.cpp, src/expr/node_manager.h: move a number of
function implementations to the header and make them inline
* src/theory/theoryof_table_prologue.h,
src/theory/theoryof_table_epilogue.h, src/theory/mktheoryof,
src/theory/Makefile.am: make the theoryOf() table from kinds and
implement TheoryEngine::theoryOf().
* src/theory/arith/Makefile, src/theory/bool/Makefile: generated these
stub Makefiles (with contrib/addsourcedir) as per policy
* src/theory/arith, src/theory/bool [directory properties]: add .deps
to svn:ignore.
* contrib/configure-in-place: permit configuring "in-place" in the
source directory.
* contrib/get-authors, contrib/dimacs_to_smt.pl,
contrib/update-copyright.pl, contrib/get-authors,
contrib/addsourcedir, src/expr/mkkind: copyright notice
* src/expr/node_manager.h, src/expr/node_builder.h,
src/prop/prop_engine.h, src/prop/prop_engine.cpp,
src/theory/theory_engine.h, src/smt/smt_engine.h, src/smt/smt_engine.cpp,
src/theory/output_channel.h: turn "const Node&"-typed formal
parameters into "TNode"
* src/theory/bool, src/theory/booleans: "bool" directory renamed "booleans"
to avoid keyword clash on containing namespace
* src/theory/booleans/theory_def.h, src/theory/uf/theory_def.h,
src/theory/arith/theory_def.h: "define" a theory simply (for automatic
theoryOf() generator).
* src/Makefile.am: build theory subdirectory before prop, smt, etc. so that
src/theory/theoryof_table.h header gets generated before it's needed
* src/expr/node_prologue.h, src/expr/node_middle.h: move "Kind" into a
separate CVC4::kind namespace to avoid its contents from cluttering
the CVC4 root namespace. Import the symbol "Kind" into the CVC4 namespace
but not the enum values.
* src/expr/node_manager.h, src/expr/node.h, src/expr/node_value.h,
src/expr/node_value.cpp, src/expr/expr.cpp, src/theory/uf/theory_uf.cpp,
src/prop/cnf_stream.cpp, src/parser/smt/smt_parser.g,
src/parser/cvc/cvc_parser.g, src/parser/antlr_parser.cpp,
test/unit/expr/node_white.h, test/unit/expr/node_black.h,
test/unit/expr/kind_black.h, test/unit/expr/node_builder_black.h:
update for having moved Kind into CVC4::kind.
* src/parser/parser.cpp: added file-does-not-exist check (was failing
silently).
|
|
Updated contrib/update-copyright to handle Antlr (.g) parser/lexer inputs.
Updated copyright year.
Added a new "bool" theory (right now just to hold a kind.h contribution).
Added "kinds" files to UF and the new "bool" theory.
|