Dejan Jovanović [Tue, 12 Jun 2012 03:21:47 +0000 (03:21 +0000)]
test cases for the
d_disequalityReasonsMap.find(pair) != d_disequalityReasonsMap.end()
assertion fail
Clark Barrett [Tue, 12 Jun 2012 03:01:41 +0000 (03:01 +0000)]
Adding constant propagation code - needs more testing - off by default
Tim King [Tue, 12 Jun 2012 00:37:27 +0000 (00:37 +0000)]
Adding incorrect qf_lia result.
Morgan Deters [Tue, 12 Jun 2012 00:32:02 +0000 (00:32 +0000)]
fix ordering issue of --dump-to and --default-dag-thresh. now can be specified in either order and the DAG threshold takes.
Tim King [Mon, 11 Jun 2012 23:40:47 +0000 (23:40 +0000)]
Fix to term normalization of integer equalities. Adds a regression test that triggered it.
Tim King [Mon, 11 Jun 2012 22:32:46 +0000 (22:32 +0000)]
Fixing type comparision assertion in getEqualityStatus().
Dejan Jovanović [Mon, 11 Jun 2012 21:53:00 +0000 (21:53 +0000)]
another benchmark that used to fail
Dejan Jovanović [Mon, 11 Jun 2012 18:54:38 +0000 (18:54 +0000)]
fixing bitvector bugs
* clauses shouldn't be erased when they could be a reason for outside propagation
* propagation of p and !p is ignored as this must lead to a conflict in the subtheory internally
Morgan Deters [Mon, 11 Jun 2012 18:27:58 +0000 (18:27 +0000)]
mark a quantifiers global var as "static" so we can find it easier later
Clark Barrett [Mon, 11 Jun 2012 17:31:13 +0000 (17:31 +0000)]
OK, now the rewrite issues are fixed
Morgan Deters [Mon, 11 Jun 2012 17:13:03 +0000 (17:13 +0000)]
distribute an .expect file. fixes a "make check" failure not for svn working dirs but for distributed tarballs.
Morgan Deters [Mon, 11 Jun 2012 16:28:23 +0000 (16:28 +0000)]
Merge from quantifiers2-trunkmerge branch.
Adds TheoryQuantifiers and TheoryRewriteRules, QuantifiersEngine, and other infrastructure.
Adds theory instantiators to many theories.
Adds the UF strong solver.
Clark Barrett [Mon, 11 Jun 2012 16:14:01 +0000 (16:14 +0000)]
Fix for array bug with decision heuristic
Also fixed one bv rewrite failure (more to come)
Morgan Deters [Mon, 11 Jun 2012 15:36:42 +0000 (15:36 +0000)]
fix issue referred to in bug 352 regarding infinite loop between SubstitutionMap (when debugging tag "substitutions" is on) and DagificationVisitor
Dejan Jovanović [Mon, 11 Jun 2012 13:54:42 +0000 (13:54 +0000)]
failing bv examples
Clark Barrett [Mon, 11 Jun 2012 11:43:55 +0000 (11:43 +0000)]
Fixed bug 352
Dejan Jovanović [Mon, 11 Jun 2012 02:32:04 +0000 (02:32 +0000)]
some failing examples
Clark Barrett [Sun, 10 Jun 2012 15:55:31 +0000 (15:55 +0000)]
Fixed one more bug in rewriter
Clark Barrett [Sun, 10 Jun 2012 14:43:58 +0000 (14:43 +0000)]
Added a very fruitful assertion to the rewriter: checks that rewriting after "REWRITE_DONE" is idempotent
Found several problems where this is not the case and fixed them
Dejan Jovanović [Sun, 10 Jun 2012 06:09:25 +0000 (06:09 +0000)]
adding an assertion to trigger the problem of bug349 and the testcase
bv rewriter apparently deosn't have a proper normal form for equalities
Dejan Jovanović [Sun, 10 Jun 2012 03:03:17 +0000 (03:03 +0000)]
fixes for bug347
it was an issue with constants being merged, which influenced explanations of disequalities. when constants are attempted to be merged, equality engine now enters conflict mode immediately
Clark Barrett [Sat, 9 Jun 2012 17:03:22 +0000 (17:03 +0000)]
Turning on unconstrained simp for QF_AUFBV
Morgan Deters [Sat, 9 Jun 2012 14:09:39 +0000 (14:09 +0000)]
Cleanup and comments for the dag-ifier. Also some unit testing for it.
Morgan Deters [Sat, 9 Jun 2012 00:35:38 +0000 (00:35 +0000)]
Dagification of output expressions.
By default, common subexpressions are dagified if they appear > 1 time and are not constants or variables.
This can be changed with --default-expr-dag=N --- N is a threshold such that if the subexpression occurs > N
times, it is dagified; a setting of 0 turns off dagification entirely.
If you notice strange dumping behavior (taking too long to print anything, e.g.), revert to the old behavior
with --default-expr-dag=0 and let me know of the problem.
Morgan Deters [Fri, 8 Jun 2012 23:35:21 +0000 (23:35 +0000)]
The option --arith-presolve-lemmas had previously been renamed --unate-lemmas.
This commit just renames it in --help documentation, etc.
Morgan Deters [Fri, 8 Jun 2012 21:21:42 +0000 (21:21 +0000)]
Extend Printer infrastructure also to the "Result" class, meaning that different output languages can write "sat", "unsat", etc., in different ways. No output is changed by this commit, but the flexibility is added that Francois wanted at today's meeting.
Morgan Deters [Fri, 8 Jun 2012 20:11:22 +0000 (20:11 +0000)]
minor fixes, for Mac OS
Dejan Jovanović [Fri, 8 Jun 2012 20:00:13 +0000 (20:00 +0000)]
very small fast example for the bv fail
Morgan Deters [Fri, 8 Jun 2012 19:00:44 +0000 (19:00 +0000)]
Fix SMT-LIBv2 ALL_SUPPORTED logic inference (by inferring it earlier).
Kshitij Bansal [Fri, 8 Jun 2012 10:18:01 +0000 (10:18 +0000)]
handle BitVectorSignExtend in pickler
Kshitij Bansal [Fri, 8 Jun 2012 09:34:20 +0000 (09:34 +0000)]
threadlocal
Kshitij Bansal [Fri, 8 Jun 2012 05:56:08 +0000 (05:56 +0000)]
Merge from decision branch (till r3663)
(no performace or search behavior changes expected)
Dejan Jovanović [Fri, 8 Jun 2012 05:48:19 +0000 (05:48 +0000)]
small fuzz examples where bv fails
Dejan Jovanović [Thu, 7 Jun 2012 22:19:53 +0000 (22:19 +0000)]
fixing the wrong results. arrays equality adaptor had a missing case when propagating disequalities between shared terms.
Morgan Deters [Thu, 7 Jun 2012 20:45:13 +0000 (20:45 +0000)]
LogicInfo locking implemented, and some initialization-order issues in SmtEngine resolved.
ALL_SUPPORTED and QF_ALL_SUPPORTED logics now supported by SMT-LIB parsers. In SMT-LIBv2, if a (set-logic..) command is missing, ALL_SUPPORTED is assumed, and a warning is issued, as discussed on the cvc4-devel mailing list.
Clark Barrett [Thu, 7 Jun 2012 19:34:21 +0000 (19:34 +0000)]
Fixed performance issue with ite_simplifier on some QF_AUFBV benchmarks
Morgan Deters [Thu, 7 Jun 2012 16:16:48 +0000 (16:16 +0000)]
Adding EchoCommand and associated printer and parser rules:
* SMT-LIBv2 parser now supports (echo...).
* Dump() gestures can now dump EchoCommands in CVC and SMT-LIB formats.
This can make it much easier to interpret output.
Dejan Jovanović [Thu, 7 Jun 2012 16:12:06 +0000 (16:12 +0000)]
cleaning up the expample for the future
Clark Barrett [Thu, 7 Jun 2012 15:31:08 +0000 (15:31 +0000)]
Added small test case for diseq propagation
Dejan Jovanović [Thu, 7 Jun 2012 07:11:24 +0000 (07:11 +0000)]
fixing some bugs in propagation of disequalities
still doesnt fix the wrong answers thought :(
Clark Barrett [Wed, 6 Jun 2012 21:11:37 +0000 (21:11 +0000)]
Don't ever call nonclausalSimplify if simplificationMode = NONE (even if
repeatSimp is true)
Clark Barrett [Wed, 6 Jun 2012 19:29:18 +0000 (19:29 +0000)]
Fixed assertion failures
Morgan Deters [Wed, 6 Jun 2012 18:08:28 +0000 (18:08 +0000)]
removing std::cout from trunk
Morgan Deters [Wed, 6 Jun 2012 17:52:47 +0000 (17:52 +0000)]
also remove now-incorrect comment from makefile
Clark Barrett [Wed, 6 Jun 2012 17:49:51 +0000 (17:49 +0000)]
Fixed broken test case, removed one that is a mistake
Morgan Deters [Wed, 6 Jun 2012 17:35:09 +0000 (17:35 +0000)]
unconstrained regressions are now run with "make check", but with --unconstrained-simp option
Morgan Deters [Wed, 6 Jun 2012 16:17:19 +0000 (16:17 +0000)]
Fixing numerous issues with tests and "make dist":
* test/regress/regress0/unconstrained wasn't being distributed. This caused the debian build failure last night. (It still doesn't run on "make check", but had to be distributed properly.)
* Fixing an issue where a test name was longer than 99 characters and couldn't be distributed in the "old" tar format. (Told automake to use a newer tar format.)
Dejan Jovanović [Wed, 6 Jun 2012 16:03:37 +0000 (16:03 +0000)]
disabling a super-expensive assertions to speed up debug runs
Dejan Jovanović [Wed, 6 Jun 2012 06:12:40 +0000 (06:12 +0000)]
Changes to the combination mechanism, lots of details. Not done yet, there are still the AUFBV wrong results, but it seems better.
http://church.cims.nyu.edu/regress-results/compare_jobs.php?job_id=4382&reference_id=4359&p=5
Clark Barrett [Wed, 6 Jun 2012 01:06:15 +0000 (01:06 +0000)]
Fixed problem causing crash at destruction time
Clark Barrett [Tue, 5 Jun 2012 20:40:22 +0000 (20:40 +0000)]
More clean-up
Clark Barrett [Tue, 5 Jun 2012 19:48:30 +0000 (19:48 +0000)]
Fixed a performance issue with unconstrained simplifier
Clark Barrett [Tue, 5 Jun 2012 14:43:11 +0000 (14:43 +0000)]
Adding missing files...
Clark Barrett [Mon, 4 Jun 2012 22:26:40 +0000 (22:26 +0000)]
Added preprocessing pass that propagates unconstrained values - solves all of
the unconstrained examples in QF_AUFBV/brummayerbiere3 - should also help
generally on at least BV and maybe others.
Off by default for now - results are mixed and it's hard to evaluate with so
many existing assertion failures and segfaults - will re-evaluate once those
are fixed
Morgan Deters [Fri, 1 Jun 2012 20:31:24 +0000 (20:31 +0000)]
add a global user-context push/pop in smt engine, just like clark's addition the other day of a push/pop of the sat context
Clark Barrett [Thu, 31 May 2012 17:15:02 +0000 (17:15 +0000)]
Fixed bug in bv: one more case where non-shared equality was getting propagated
Added a global push and pop around solving - fixes an assertion failure when
TNodes are still around in a CDHashMap at destruction time.
Morgan Deters [Thu, 31 May 2012 14:46:38 +0000 (14:46 +0000)]
svn:ignore a parallel-tests driver file that automake deposits in config/
Morgan Deters [Thu, 31 May 2012 14:45:09 +0000 (14:45 +0000)]
pass JAVA_CPPFLAGS properly
Clark Barrett [Wed, 30 May 2012 20:33:40 +0000 (20:33 +0000)]
Added BitwiseEq bitvector rewrite
Added option "--repeat-simp" which repeats nonclausal simplification: 4 times,
twice before ite removal and twice after
Enabled these options (plus ite-simp) on QF_AUFBV, obtaining significant
speedup on dwp examples
Clark Barrett [Wed, 30 May 2012 16:16:18 +0000 (16:16 +0000)]
Fixed problem with array queue growing too large
Morgan Deters [Wed, 30 May 2012 16:14:56 +0000 (16:14 +0000)]
remove unused/broken check build target
Morgan Deters [Tue, 29 May 2012 23:12:06 +0000 (23:12 +0000)]
removing now-unused TheoryEngine::setLogic() interface function
Clark Barrett [Tue, 29 May 2012 17:56:51 +0000 (17:56 +0000)]
Enabled SolveEq bv rewrite
Clark Barrett [Mon, 28 May 2012 16:27:50 +0000 (16:27 +0000)]
Added some BV rewrites, fixed bugs in array theory, made ite simp work with BV
Dejan Jovanović [Sun, 27 May 2012 16:59:01 +0000 (16:59 +0000)]
some reordering to keep invariants
Dejan Jovanović [Sun, 27 May 2012 05:44:13 +0000 (05:44 +0000)]
Committing the work on equality engine, I need to see how it does on the regressions. New additions:
* areDisequal(x, y) -> areDisequal(x, y, needProof): when asking for a disequality you must say needProof if you will ask for an explanation later.
* propagation of shared dis-equalities (not yet complete, once case missing)
* changes to the theories that use it, authors should check up on the changes
Clark Barrett [Sun, 27 May 2012 00:39:27 +0000 (00:39 +0000)]
Another expensive function call in a Debug trace
Clark Barrett [Sun, 27 May 2012 00:38:30 +0000 (00:38 +0000)]
Another expensive function call in a Debug line
Kshitij Bansal [Fri, 25 May 2012 16:45:21 +0000 (16:45 +0000)]
init bug fix
Clark Barrett [Fri, 25 May 2012 00:21:53 +0000 (00:21 +0000)]
Checking in fix for bug 340 - somehow didn't get checked in earlier
Dejan Jovanović [Thu, 24 May 2012 16:03:38 +0000 (16:03 +0000)]
Separating the subtheory implementations in the bitvector theory.
Dejan Jovanović [Thu, 24 May 2012 05:54:39 +0000 (05:54 +0000)]
Significant changes to the internals of the equality engine. Equality is not handled natively and not as a generic predicate. The changes also change the order of propagation, and can produce different conflicts. Since the engine is now used everywhere this means that so some crazy results are to be expected.
Tim King [Tue, 22 May 2012 15:09:03 +0000 (15:09 +0000)]
This commit merges in the branch arithmetic/cprop.
Dejan Jovanović [Mon, 21 May 2012 20:15:52 +0000 (20:15 +0000)]
Updating equality manager to handle tagged trigger terms. Notifications are pushed out for relationships between terms tagged with the same tag. No performance impact.
Morgan Deters [Mon, 21 May 2012 16:33:53 +0000 (16:33 +0000)]
main() no longer catches non-CVC4 exceptions. This means on memout and other C++-level exceptions, we'll exit the C++ way rather than our custom way (so we don't get statistics etc.)
Tim King [Sat, 19 May 2012 17:45:37 +0000 (17:45 +0000)]
Adding regress test for bug 341.
Tim King [Sat, 19 May 2012 17:24:52 +0000 (17:24 +0000)]
- The array type rules were fixed to use isSubtypeOf.
- The type of (kind::DIVISION x y) is RealType.
- A reference to util/pseudoboolean.i was removed.
- rec5.cvc is disabled for now. The type of 2 is Integer which is not a subtype of [0..11].
Tim King [Fri, 18 May 2012 23:48:38 +0000 (23:48 +0000)]
This commit adds TypeNode::leastCommonTypeNode(). The special case for arithmetic in TypeNode::operator==() has been removed. A number of faulty type checking checks were switched to use isSubtypeOf. The resolves bug #339
Tim King [Fri, 18 May 2012 20:20:58 +0000 (20:20 +0000)]
This commit removes the dead psuedoboolean code.
Tim King [Fri, 18 May 2012 16:02:50 +0000 (16:02 +0000)]
Removing long unsigned operator+ from CDList's const_iterator.
Dejan Jovanović [Fri, 18 May 2012 14:24:02 +0000 (14:24 +0000)]
removing failing regression
Tim King [Thu, 17 May 2012 23:00:43 +0000 (23:00 +0000)]
This fixes a fascinating segfault. This is the sequence of events:
1) A restart occurs
2) A shared term is registered to arithmetic.
3) Arithmetic sets this up.
4) No new linear relations are added to arithmetic.
5) Eventually a restart occurs.
6) Arithmetic resets the tableau as it has not had a row added since the last restart.
7) A new variable is added.
8) This exceeds the size of the column vector of the saved tableau by exactly one.
9) segfault
Morgan Deters [Thu, 17 May 2012 20:45:32 +0000 (20:45 +0000)]
Fixing an issue with LogicInfo::isPure() that turned off simplification in QF_UF and maybe others
Liana Hadarean [Thu, 17 May 2012 18:42:13 +0000 (18:42 +0000)]
Fixed bug 338:
- fixed buggy rewrite rules assuming certain nodes only had 2 children
- added support for bv-rewrite dump
- fixed smt2_printer for bv constants
Tim King [Thu, 17 May 2012 15:33:13 +0000 (15:33 +0000)]
Adding failing regression for ite type computation.
Dejan Jovanović [Thu, 17 May 2012 04:34:03 +0000 (04:34 +0000)]
Queueing up asserted literals in the proxy instead of sending them off to the theory engine immediately. The queue is discharged just before a check().
Dejan Jovanović [Wed, 16 May 2012 19:28:25 +0000 (19:28 +0000)]
adding simple-minded handling of (dis-)equalities where constants are involved
Morgan Deters [Wed, 16 May 2012 18:20:09 +0000 (18:20 +0000)]
Fixing C compatibility library (it still had a reference to CONST_INTEGER).
This hopefully fixes the Debian build.
Dejan Jovanović [Wed, 16 May 2012 17:54:16 +0000 (17:54 +0000)]
testcase for bug 337
Dejan Jovanović [Wed, 16 May 2012 17:53:41 +0000 (17:53 +0000)]
Changes to SAT solver:
* allowing propagation of false literals (handles conflict)
* allowing lemmas during BCP (bug 337)
* UF does direct propagation, without checking for literal value anymore
Dejan Jovanović [Wed, 16 May 2012 16:18:52 +0000 (16:18 +0000)]
equality status for bitvectors can now look into the sat solver to check for IN_MODEL status
Dejan Jovanović [Wed, 16 May 2012 16:01:48 +0000 (16:01 +0000)]
removing duplicate literals in explanations of propagations
Liana Hadarean [Wed, 16 May 2012 15:21:18 +0000 (15:21 +0000)]
refactored TheoryBV bitblaster and equality engine into subtheories (similar to TheoryEngine
Liana Hadarean [Tue, 15 May 2012 22:02:47 +0000 (22:02 +0000)]
fixed QF_BV performance problem due to using CDList instead of CDQueue
Dejan Jovanović [Tue, 15 May 2012 21:58:36 +0000 (21:58 +0000)]
Dejan Jovanović [Tue, 15 May 2012 21:58:09 +0000 (21:58 +0000)]
test cases
Tim King [Tue, 15 May 2012 21:20:56 +0000 (21:20 +0000)]
Fix to shared terms visitor.
Morgan Deters [Tue, 15 May 2012 19:28:18 +0000 (19:28 +0000)]
Implement TypeNode::isComparableTo() and add a unit test for it.
Clark Barrett [Tue, 15 May 2012 19:24:09 +0000 (19:24 +0000)]
Fixed several bugs in shared terms database
Tim King [Tue, 15 May 2012 19:01:19 +0000 (19:01 +0000)]
This commit removes the CONST_INTEGER kind from nodes. This code comes from the branch arithmetic/remove_const_int.