Morgan Deters [Tue, 8 Mar 2011 21:43:45 +0000 (21:43 +0000)]
Clean up Theory base class as per code review bug #60; also fixes to CodeTimer statistic, and adding a CodeTimer to TheoryEngine::EngineOutputChannel::newFact() for investigation into (possible) slow or redundant theory registration.
Tim King [Tue, 8 Mar 2011 01:46:31 +0000 (01:46 +0000)]
- Merges queue-interrogation branch into the trunk. This branch adds extra phases of looking for additional conflicts during and after the heuristic pivoting stage. (For the expected performance gain, comparing jobs 1676 and 1643 gives a rough idea.)
Tim King [Mon, 7 Mar 2011 19:10:16 +0000 (19:10 +0000)]
Merges branches/arithmetic/tableau-reset into the trunk. The tableau is now heuristically reset to its initial state during restarts.
Tim King [Sat, 5 Mar 2011 23:10:41 +0000 (23:10 +0000)]
- Adds PermissiveBackArithVarSet. This is very similar to ArithVarSet. The difference is that set.isMember(x) for an ArithVar x s.t. x > set.allocated() returns false for PermissiveBackArithVarSet and is an assertion failure for ArithVarSet. This cuts down on the memory usage of the ColumnMatrix slightly.
Tim King [Sat, 5 Mar 2011 19:30:06 +0000 (19:30 +0000)]
Enables the PreferenceFunction minBoundAndRowCount.
Tim King [Sat, 5 Mar 2011 17:33:01 +0000 (17:33 +0000)]
- Adds "PreferenceFunction" to SimplexDecisionProcedure. A PreferenceFunction allows for specifying how to choose between two nonbasic variables for which should become basic during the selectSlack(...) function. This partially addresses a point brought up by Dejan during the Code Review. (Unfortunately, function pointers are involved in the implementation. Because of this, I have had Morgan review this code before check-in.)
Tim King [Sat, 5 Mar 2011 16:59:13 +0000 (16:59 +0000)]
- Made Rational::sgn() function const.
Morgan Deters [Sat, 5 Mar 2011 00:03:08 +0000 (00:03 +0000)]
adding three features to CVC parser that drastically improve its support for the language: LET now supported (but not "type-lets" yet), OPTION now supported, and ^ operator (exponentiation) supported for nonnegative integer exponents. The latter simply expands to MULT, and of course fails assertions in arithmetic if a variable is raised to n >= 2. Also other CVC parser clean-up.
Morgan Deters [Thu, 3 Mar 2011 18:35:17 +0000 (18:35 +0000)]
fix for bug #244, "Segfault if file cannot be found and --stats is on"
Tim King [Thu, 3 Mar 2011 18:00:35 +0000 (18:00 +0000)]
- Creates a queue for lemmas discovered during the simplex procedure. Lemmas are sent to the sat solver during theory propagation. The lemmas currently come from additional conflicts that are discovered by findConflictOnTheQueue(...).
Morgan Deters [Thu, 3 Mar 2011 17:49:15 +0000 (17:49 +0000)]
resurrecting triple.h from r1023 (after which it was removed)
Tim King [Thu, 3 Mar 2011 16:34:48 +0000 (16:34 +0000)]
Merged the tableau-copy branch into trunk. This adds a copy constructor and operator=(...) to Tableau.
Dejan Jovanović [Thu, 3 Mar 2011 14:04:38 +0000 (14:04 +0000)]
fixing a type that caused the segfaults in the regressions
Dejan Jovanović [Wed, 2 Mar 2011 23:10:18 +0000 (23:10 +0000)]
fixing the big with lemma reallocation in minisat garbage collection
Morgan Deters [Mon, 28 Feb 2011 07:49:56 +0000 (07:49 +0000)]
CongruenceClosure module now should support nullary congruence operators (now that they are allowed for the datatypes theory, as in (APPLY_CONSTRUCTOR nil) or (APPLY_CONSTRUCTOR zero)). It does this by treating such terms with zero children as non-candidates for congruence, even though they have the congruence kind APPLY_CONSTRUCTOR.
Morgan Deters [Mon, 28 Feb 2011 07:03:33 +0000 (07:03 +0000)]
Review of mktheorytraits, mkrewriter, and recent changes to other mk* scripts. Minor changes only, correcting some documentation and fixing some warnings that were being issued about functions not existing.
Morgan Deters [Mon, 28 Feb 2011 06:27:12 +0000 (06:27 +0000)]
minor doxygen build target fixes
Morgan Deters [Mon, 28 Feb 2011 06:06:25 +0000 (06:06 +0000)]
Review of statistics code. Added lots of documentation, and fixed an issue (I think) that Tim found with TimerStat involving wild, sometimes negative, timer statistic values. (It was due to improper initialization.)
Tim King [Sun, 27 Feb 2011 19:59:52 +0000 (19:59 +0000)]
- Adds a path for Theory to be passed a reference to Options.
- Adds 3 choices of heuristic variable orders to use in ArithPriorityQueue.
- Adds the pivot-rule command line option.
Tim King [Sun, 27 Feb 2011 18:29:38 +0000 (18:29 +0000)]
- Makes VarCoeffPair a class instead of a typedef of pair<ArithVar, Rational>. This addresses a point Dejan brought up in the code review.
Tim King [Sun, 27 Feb 2011 17:34:37 +0000 (17:34 +0000)]
- Adds a buffer to the ReducedRowVector addRowTimesConstant operation to reduce the number of allocations made. Compare cluster jobs 1585 and 1584 for the expected performance increase.
Tim King [Sat, 26 Feb 2011 23:32:34 +0000 (23:32 +0000)]
- Merged RowVector and ReducedRowVector.
- Renamed NonZeroIterator to const_iterator.
- Both of these changes are in response to the code review.
Morgan Deters [Sat, 26 Feb 2011 22:19:47 +0000 (22:19 +0000)]
Commit to fix bug 241 (improper "using namespace std" in a header). This caused a number of latent errors in sources and headers to come up. Those are now fixed (by adding "using" or "std::" depending on the context). Took the opportunity to bring many rewriter sources in line with coding conventions.
Morgan Deters [Sat, 26 Feb 2011 21:24:18 +0000 (21:24 +0000)]
Merge from theory-break-dependences branch to break Theory and TheoryEngine dependences; now, if you touch theory_engine.h, only a few things in theory need be recompiled (TheoryEngine, SharedTermManager, .... but no theory implementations), along with the PropEngine and SmtEngine. If you touch a specific theory's .h file, only that theory must be recompiled (along with the TheoryEngine, since it uses traits, and SmtEngine, since it tells the TheoryEngine which theory implementations to use).
Morgan Deters [Sat, 26 Feb 2011 21:18:14 +0000 (21:18 +0000)]
fix serious regression breakage (segfaults) caused by an off-by-one error in initialization of the expression variable statistics in last commit
Dejan Jovanović [Sat, 26 Feb 2011 07:13:01 +0000 (07:13 +0000)]
adding the variables count to the statistics in the expr manager
Dejan Jovanović [Sat, 26 Feb 2011 05:40:55 +0000 (05:40 +0000)]
adding statistics about how many different kinds of expressions we have created in the expression manager.
this is useful, for example, with --parse-only, to figure out a bit of problem structure
Tim King [Fri, 25 Feb 2011 19:28:48 +0000 (19:28 +0000)]
- This commit adds some debugging information to ArithPriorityQueue.
- There was a missing break statement in ArithPriorityQueue. This addresses the bug discovered in the debug regression (http://goedel.cims.nyu.edu/regress-results/compare_jobs.php?job_id=1566&reference_id=1548&category=1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,29&p=10).
Dejan Jovanović [Fri, 25 Feb 2011 02:11:14 +0000 (02:11 +0000)]
slicing manager is not breaking the old regressions, time to sync
Tim King [Thu, 24 Feb 2011 23:00:08 +0000 (23:00 +0000)]
- Adds an additional round of checks for a conflict after the difference heuristic round has been completed. This happens immediately before switching to the variable order round.
Tim King [Thu, 24 Feb 2011 20:36:35 +0000 (20:36 +0000)]
- Adds an additional mode to ArithPriorityQueue, Collection. Collection is a mode where the heap structure is not maintained. There is just a list of variables that may be inconsistent. This is used up until the simplex procedure is invoked.
- Misc. cleanup and renaming in ArithPriorityQueue.
Tim King [Thu, 24 Feb 2011 16:52:15 +0000 (16:52 +0000)]
- Changes ArithPriorityQueue to use stl::vector<>'s plus stl's heap algorithms instead of stl's priority queue (which is really an stl vector plus the stl heap algorithms). This offers more control of the underlying data structure.
Tim King [Tue, 22 Feb 2011 01:13:56 +0000 (01:13 +0000)]
- Adds column based iterators.
Tim King [Mon, 21 Feb 2011 19:15:33 +0000 (19:15 +0000)]
- Adds the ArithPriorityQueue class. The ArithPriorityQueue class provides an abstraction to the previous priority queues representing the 2 different pivoting rules.
Tim King [Mon, 21 Feb 2011 00:22:18 +0000 (00:22 +0000)]
- Adds the statistic d_avgNumRowsNotContainingOnPivot.
- Removed a bug in row counting in row counting.
Tim King [Sat, 19 Feb 2011 00:22:34 +0000 (00:22 +0000)]
Changes:
- The Tableau is now in charge of managing what variables are basic in a unified manner. Specifically, TheoryArith::d_basicManager was merged into Tableau::d_basicVariables.
Tim King [Fri, 18 Feb 2011 23:28:19 +0000 (23:28 +0000)]
Changes:
- ArithVar is no longer an attribute
- RowVector's destructor reduces the row count of its variables upon.
- Tableau's destructor now free its rows instead of leaking memory.
- Added ability to convert ReducedRowVectors into equivalent Nodes.
- getValue() should work again.
Tim King [Thu, 17 Feb 2011 21:30:57 +0000 (21:30 +0000)]
This commit merges the branch branches/arithmetic/quick-row-has into trunk. quick-row-has has an optimization to make checking if a variable is in a row faster.
Tim King [Thu, 17 Feb 2011 18:22:16 +0000 (18:22 +0000)]
This commit is the promised clean up after removing row ejection.
Tim King [Thu, 17 Feb 2011 18:00:30 +0000 (18:00 +0000)]
Removed ActivityMonitor from arithmetic. This was only used for row ejection, and is now superfluous.
Tim King [Thu, 17 Feb 2011 17:46:31 +0000 (17:46 +0000)]
Row ejection is now completely disabled. Another commit cleaning this one up will follow shortly.
Tim King [Thu, 17 Feb 2011 17:41:15 +0000 (17:41 +0000)]
Removed vestigial normal form notes file from src/theory/arith/Makefile.am. Should have been removed earlier.
Dejan Jovanović [Thu, 17 Feb 2011 05:27:48 +0000 (05:27 +0000)]
some unit tests to work on slicing
Tim King [Thu, 17 Feb 2011 01:10:38 +0000 (01:10 +0000)]
I replaced the pattern "x = x + y;" with "x += y;" in a few places in DeltaRational. This addresses a point that came up in the code review.
Tim King [Thu, 17 Feb 2011 01:02:06 +0000 (01:02 +0000)]
Updates based on the group code review of arithmetic on 2011-02-15. The only substantive change is that UnatePropagator no longer uses attributes. The rest is comments and other beatification.
Tim King [Thu, 17 Feb 2011 00:55:11 +0000 (00:55 +0000)]
Deleting depricated files.
Dejan Jovanović [Thu, 17 Feb 2011 00:29:26 +0000 (00:29 +0000)]
getting ready for slicing bitvectors
Tim King [Wed, 16 Feb 2011 23:23:04 +0000 (23:23 +0000)]
Overview of the changes:
- 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
Dejan Jovanović [Wed, 16 Feb 2011 01:26:26 +0000 (01:26 +0000)]
updates for the rewriter, added some statistics
Tim King [Mon, 14 Feb 2011 02:47:01 +0000 (02:47 +0000)]
Reverses the order of the d_possiblyInconsistent queue. (It is that old termination bug again. *le sigh*)
Tim King [Sun, 13 Feb 2011 21:19:20 +0000 (21:19 +0000)]
3 heuristics were added to arithmetic. A heuristic for detecting an encoding of min added to static learning in LRA. A heuristic added for when the true branch and false branch are both constants (also in static learning). A heuristic for checking whether any variables begin in conflict before pivoting.
Dejan Jovanović [Wed, 5 Jan 2011 19:12:54 +0000 (19:12 +0000)]
fix for build errors
Dejan Jovanović [Wed, 5 Jan 2011 03:21:35 +0000 (03:21 +0000)]
Commit for the theory engine and rewriter changes. Changes are substantial and not yet finalized but I need to put it in to work further with the theory writers. Please check the files that you 'own'. Any comments or discussion is welcome. Further details will be coming in a follow up email later.
Morgan Deters [Fri, 17 Dec 2010 23:09:15 +0000 (23:09 +0000)]
tls.h, rational.h, and integer.h are only re-generated if changed. this obviates the need for a full rebuild just because you re-./configured.
Morgan Deters [Thu, 16 Dec 2010 00:36:53 +0000 (00:36 +0000)]
minor fixes for correct doxygen output
Morgan Deters [Tue, 14 Dec 2010 21:27:38 +0000 (21:27 +0000)]
make some CC module methods private that should not have been public
Morgan Deters [Tue, 14 Dec 2010 21:07:46 +0000 (21:07 +0000)]
congruence closure module now supports things other than APPLY_UF; ported from "arrays" branch to trunk
Morgan Deters [Tue, 14 Dec 2010 19:37:35 +0000 (19:37 +0000)]
permit PARAMETERIZED operators to be zero-ary
Morgan Deters [Tue, 14 Dec 2010 01:04:00 +0000 (01:04 +0000)]
fix to static learning application in UF, resolves bug# 239
Dejan Jovanović [Wed, 24 Nov 2010 18:50:54 +0000 (18:50 +0000)]
Changin the get() semantics to a CDQeue-sque semantics.
Morgan Deters [Fri, 19 Nov 2010 01:37:55 +0000 (01:37 +0000)]
Merge from ufprop branch, including:
* Theory::staticLearning() for statically adding new T-stuff before
normal preprocessing. UF's staticLearning() does transitivity of
equality/iff, solving the diamonds.
* more aggressive T-propagation for UF
* new KEEP_STATISTIC macro to hide Theories from having to
register/deregister statistics (and also has the advantage of
keeping the statistic type, field name, and the 'tag' used to output
the statistic in the same place---instead of scattered in the theory
definition and constructor initializer list. See documentation for
KEEP_STATISTIC in src/util/stats.h for more of an explanation).
* more statistics for UF
* restart notifications from SAT (through TheoryEngine) via
Theory::notifyRestart()
* StackingMap and UnionFind unit tests
* build fixes/adjustments
* code cleanup; minor other improvements
Morgan Deters [Fri, 19 Nov 2010 00:12:17 +0000 (00:12 +0000)]
add statistics support information to --show-config
Morgan Deters [Thu, 18 Nov 2010 03:57:52 +0000 (03:57 +0000)]
small changes to documentation; also, '\''make doc'\'' doesn't build dot graphs (but nightly build system will produce them)
Morgan Deters [Wed, 17 Nov 2010 07:22:35 +0000 (07:22 +0000)]
fix improper CongruenceClosureWhite test by merging from a uf branch; fixes the nightly test failure
Morgan Deters [Wed, 17 Nov 2010 02:45:13 +0000 (02:45 +0000)]
add some stats to UF/CC
Morgan Deters [Wed, 17 Nov 2010 01:39:37 +0000 (01:39 +0000)]
The "UF engineering issues" release, after much profiling.
* swap in backtracking data structures (that use and maintain their own
undo stack) in some places instead of the usual Context-aware
paradigm (MUCH faster).
* cosmetic changes to UF, CC modules.
Tim King [Tue, 16 Nov 2010 21:11:11 +0000 (21:11 +0000)]
Added Theory::presolve().
Morgan Deters [Tue, 16 Nov 2010 19:18:19 +0000 (19:18 +0000)]
SmtEngine now fails with a ModalException if --incremental is not enabled
but a push/pop or multiple query is attempted (previously it could give
incorrect answers)
Also, fix some multi-query and push-pop tests that had wrong answers, and
support a new "% COMMAND-LINE: " gesture in regression tests so that a
test can pass additional, specific command line flags it wants to run
with (here, --incremental).
Also fix mkbuilddir script for when it's called from contrib/switch-config.
Morgan Deters [Tue, 16 Nov 2010 00:20:30 +0000 (00:20 +0000)]
fix function signatures
Morgan Deters [Mon, 15 Nov 2010 23:58:41 +0000 (23:58 +0000)]
cleanup from today's commits: delegate as-yet-unimplemented prettyprinters in a better way; fix arith Makefile
Tim King [Mon, 15 Nov 2010 23:51:38 +0000 (23:51 +0000)]
Changes to Solver and PropEngine to support lemmasOnDemand during solve but not yet in d_checkSat.
Morgan Deters [Mon, 15 Nov 2010 22:57:14 +0000 (22:57 +0000)]
Pretty-printer infrastructure created (in src/printer) and SMT-LIBv2 printer
implemented. This new infrastructure removes support for pretty-printing
(even in the AST language) an Expr with reference count 0. Previously,
this was supported in a few places internally to the expr package, for
example in NodeBuilder. (Now, a NodeBuilder cannot be prettyprinted, you
must extract the Node before printing it.)
Tim King [Mon, 15 Nov 2010 21:15:45 +0000 (21:15 +0000)]
This commit merges the arith-prop-opt branch into the main trunk. This was done by way of the intermediate branch arith-prop-tmp. Both arith-prop-opt and arith-prop-tmp will now be phased out.
Morgan Deters [Mon, 15 Nov 2010 20:24:35 +0000 (20:24 +0000)]
minor tweaks to last commit, testing infrastructure
Morgan Deters [Mon, 15 Nov 2010 20:08:17 +0000 (20:08 +0000)]
fix some things with the build system (make dist, make install, make check)
Dejan Jovanović [Fri, 12 Nov 2010 04:39:54 +0000 (04:39 +0000)]
Some bug fixes in the SAT for lemmas, and an experiment with a more complete (wr propagation) splitter in arithmetic.
Morgan Deters [Thu, 11 Nov 2010 04:45:29 +0000 (04:45 +0000)]
make addsourcedir executable
Dejan Jovanović [Tue, 9 Nov 2010 21:57:06 +0000 (21:57 +0000)]
Lemmas on demand work, push-pop, some cleanup.
Morgan Deters [Mon, 8 Nov 2010 23:37:36 +0000 (23:37 +0000)]
command-line flag to disable theory registration, also SMT-LIBv2 compliance (per SMT-LIB mailing list this afternoon)
Morgan Deters [Mon, 8 Nov 2010 23:15:08 +0000 (23:15 +0000)]
cleanup, documentation, SMT-LIBv2 compliance
Morgan Deters [Mon, 8 Nov 2010 03:27:48 +0000 (03:27 +0000)]
fix out-of-date version/copyright for minisats
Christopher L. Conway [Fri, 5 Nov 2010 17:43:52 +0000 (17:43 +0000)]
Moving Options fiddling to options.h
Morgan Deters [Thu, 4 Nov 2010 21:01:31 +0000 (21:01 +0000)]
competition mode implies --no-checking
Tim King [Thu, 4 Nov 2010 16:30:25 +0000 (16:30 +0000)]
Moving the post_mortem.py script out of contrib and into the cvc4 scripts directory.
Tim King [Thu, 4 Nov 2010 16:28:02 +0000 (16:28 +0000)]
Updates post_mortem.py script to be able to handle certain kinds of crashes and new statistics.
Tim King [Thu, 4 Nov 2010 16:26:40 +0000 (16:26 +0000)]
This commit adds the ejected and un-ejected statistics.
Tim King [Wed, 3 Nov 2010 18:46:46 +0000 (18:46 +0000)]
Adds size() to RowVector.
Tim King [Wed, 3 Nov 2010 18:42:28 +0000 (18:42 +0000)]
Adds statistics for the number of Uservariables and Slack variables used by arithmetic.
Tim King [Wed, 3 Nov 2010 01:07:09 +0000 (01:07 +0000)]
Adds AverageStat to stats.h.
Morgan Deters [Sun, 31 Oct 2010 23:57:50 +0000 (23:57 +0000)]
small fix to debug segfaults
Morgan Deters [Sun, 31 Oct 2010 15:26:19 +0000 (15:26 +0000)]
enable dependence graphs in doxygen; fix lots of doxygen warnings, fix some documentation, and make it possible to "make doc" on a clean source tree (post-configure)
Morgan Deters [Sun, 31 Oct 2010 15:22:38 +0000 (15:22 +0000)]
maximize stack limit, handle SEGV signals on an alternate signal stack, and try to diagnose stack overflow
Tim King [Sat, 30 Oct 2010 19:18:36 +0000 (19:18 +0000)]
Adds a hueristic from Alberto's thesis. For a fixed window the row count is used to select which non-basic variable is a row in made basic.
Tim King [Fri, 29 Oct 2010 21:10:46 +0000 (21:10 +0000)]
Fix for a problem caused by using a != instead of == in generateConflictBelow(). Resolves a bug introduced in -r1063.
Morgan Deters [Fri, 29 Oct 2010 21:08:41 +0000 (21:08 +0000)]
portability updates to build system
Tim King [Fri, 29 Oct 2010 20:57:50 +0000 (20:57 +0000)]
Adds a very small test that triggers a bug. The bug is from the commit for -r1063.
Morgan Deters [Fri, 29 Oct 2010 19:44:03 +0000 (19:44 +0000)]
minor fixes as a result of review of Chris's getType() rewrite; also fix some macros to make various GCC versions happy
Tim King [Fri, 29 Oct 2010 15:51:25 +0000 (15:51 +0000)]
Fixes RowVector::has().
Tim King [Fri, 29 Oct 2010 00:45:10 +0000 (00:45 +0000)]
Factors out the QF_LRA decision procedure from TheoryArith and puts this into its own class SimplexDecisionProcedure. Implements about 1/2 of the pivoting rule from Alberto's thesis (section2.5.3).
Tim King [Thu, 28 Oct 2010 21:46:44 +0000 (21:46 +0000)]
The Row implementation has no been replaced by RowVector and 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.