Age | Commit message (Collapse) | Author |
|
|
|
|
|
support incrementality.
Some clean-up work will likely follow, but the CNF/Minisat stuff should be
left pretty much untouched.
Expected performance change negligible; slightly better on memory:
http://church.cims.nyu.edu/regress-results/compare_jobs.php?job_id=3705&reference_id=3697&mode=&category=&p=5
Note that there are crashes, but that these are exhibited in the nightly
regression run too!
|
|
* separated SatSolverInput interface class into two classes:
- TheoryProxy for the sat solver to communicate with the theories
- SatSolverInterface abstract class to communicate with the sat solver
* instead of using #ifdef typedef for SatClauses and SatLiterals, now there are CVC4 SatLiteral/SatClause types and mappings between them and the internal sat solver clause/literal representation
* added abstract classes for DPLLSatSolver and BVSatSolver different interfaces
Replaced TheoryBV with bitblasting implementation:
* all operators bitblasted
* only operator elimination rewrite rules so far
|
|
to request a decision on a literal. All these theory requests are kept in a
context-dependent queue and serviced in order when the SAT solver goes to make a
decision. Requests that don't have a SAT literal give an assert-fail. Requests
for literals that already have an assignment are silently ignored.
Since the queue is CD, requests can actually be serviced more than once (e.g., if
a request is made at DL 5, but not serviced until DL 10, and later, a conflict
backtracks to level 7, the request may be serviced again).
Performance impact: none to negligible for theories that don't use it
See http://church.cims.nyu.edu/regress-results/compare_jobs.php?job_id=3620&reference_id=3614&mode=&category=&p=0
|
|
|
|
|
|
for bug #275. will finally close #275.
|
|
Details here: http://goedel.cims.nyu.edu/wiki/Meeting_Minutes_-_October_14,_2011#Resource.2Ftime_limiting_API
This will need more work, but it's a start.
Also implemented TheoryEngine::properPropagation().
Other minor things.
|
|
|
|
SAT solver on every backtrack
* Updated UF to handle the context dependent pre-registration
* Additionally some small changes in order to satisfy warnings of the eclipse code analysis tool
|
|
comparison <http://goedel.cims.nyu.edu/regress-results/compare_jobs.php?job_id=2673&&p=5&reference_id=2637>
|
|
|
|
the sat solver. Also added command line options for setting both.
|
|
|
|
* Makes Options an "omnipresent thread-local global" (like the notion
of the "current NodeManager" was already). Options::current() accesses
this structure.
* Removes Options from constructors and data structures everywhere
(this cleans up a lot of things).
* No longer uses StatisticsRegistry statically. An instance of the
registry is created and linked to a NodeManager.
* StatisticsRegistry::current() is similar to Options::current(), but
the pointer is stowed in the NodeManager (rather than stored)
* The static functions of StatisticsRegistry have been left, for backward
compatibility; they now use the "current" statistics registry.
* SmtEngine::getStatisticsRegistry() is a public accessor for the
registry; this is needed by main() to reach in and get the registry,
for flushing statistics at the end.
|
|
|
|
are somewhat disparate but belonged on the same branch because they were
held back from trunk all for the same reason (to keep the trunk stable
for furious bitvector development). Dejan has now given me the go-ahead
for a merge.
=========================================
THIS COMMIT CHANGES THE THEORY INTERFACE!
=========================================
Theory constructors are expected to take an additional "Valuation*"
parameter that each Theory should send along to the base class
constructor. The base class Theory keeps the Valuation* in a
d_valuation field for use by it and by its derived classes.
Theory::getValue() no longer takes a Valuation* (it is expected
to use d_valuation instead). This allows other theory functions
to take advantage of getValue() for debugging or heuristic
purposes.
TODO BEFORE MERGE TO TRUNK:
****implement BitIterator find() in CDAttrHash<bool>.
Specifically:
* Added QF_BV support for SMT-LIB v2.
* Two adjustments to the theory interface as requested by Tim King:
1. As described above.
2. Theories now have const access to the fact queue through base
class functions facts_begin() and facts_end(); useful for
debugging.
* Added an "Asserted" attribute so that theories can check if something
has been asserted or not (and therefore not propagate it). However, this
has been disabled for now, pending more data on the overhead of it, and
pending discussion at the 3/25/2011 meeting.
* Do not define NDEBUG in MiniSat in assertion-enabled builds (so
that MiniSat asserts are evaluated).
* As a result of the new MiniSat assertions, some --incremental
regressions had to be disabled; also, some bitvectors ?!!
* Bug 71 is resolved by adding a specialization for CDAttrHash<> in the
attribute package.
* Fixes for some warnings flagged by clang.
* System tests have arrived! So far mainly infrastructure for having
system tests, but there is a system test aimed at improving code
coverage of the printer package.
* Minor other adjustments to documentation and coding to be more
conformant to CVC4 policy.
Tests have been performed to demonstrate that these changes have no or
negligible effect on performance. In particular, changing the
CDAttrHash<> doesn't have any real effect on performance or memory right
now, since there is only one context-dependent boolean flag (as soon
as another is added, the effect is noticeable but probably still slight).
|
|
* 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
|
|
|
|
util/options.h,cpp
|
|
* Adding options --lazy-type-checking and --eager-type-checking
to control type checking in NodeBuilder, which can now be enabled
in production mode and disabled in debug mode
* Option --no-checking implies --no-type-checking
* Adding constructor SmtEngine(ExprManager* em) that uses default options
|
|
supported; work on Result type (biggest noticeable change is that CVC4 now outputs lowercase "sat" and "unsat"), Options class moved to src/smt, to allow for future work on runtime configuration via (set-option) command
|
|
(get-value ...) SMT-LIBv2 command. As per SMT-LIBv2 spec,
you must pass --interactive --produce-models on the command
line (although they don't currently make us do any extra
work). Closes bug #213.
|
|
modified in this commit
|
|
|
|
|
|
|
|
working. Still need to implement theory-specific shared term propagation.
|
|
should have no impact on production performance and speed up debug
performance considerably, while making the code much cleaner. On some
benchmarks, debug builds now run _much_ faster.
We no longer have to sprinkle our code with things like:
if(debugTagIsOn("context")) {
Debug("context") << theContext << std::endl;
}
which we had to do to get around performance problems previously.
Now just writing:
Debug("context") << theContext << std::endl;
does the same in production and debug builds. That is, if "context"
debugging is off, theContext isn't converted to a string, nor is it
output to a "/dev/null" ostream. I've confirmed this. In production
builds, the whole statement inlines to nothing. I've confirmed this
too.
Also, "Debug" is no longer a #definition, so you can use it directly
in production builds where you couldn't previously, e.g.
if(Debug.isOn("paranoid:check_tableau")) {
checkTableau();
}
I'm leaving debugTagIsOn() for now, but marking it as deprecated.
|
|
|
|
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.
|
|
this Tuesday during the meeting. You'll need to run autogen and receonfigure after updating.
|
|
** file-level documentation at the top of the sources. **
This is the "make bugzilla stop bugging me" bugfix commit.
* Remove BackedNodeBuilder<> and collapse NodeBuilder<> hierarchy.
Updated documentation in the file. Resolves bug #99.
* Convenience NodeBuilders (PlusNodeBuilder, OrNodeBuilder, etc.)
moved into a separate file. Partially resolves bug #100.
* Moved isAssociative(Kind) into kind.h (and into the CVC4::kind
namespace) instead of metakind.h (where it was in CVC4::metakind).
This clears up a warning (private #inclusion) from the SMT and SMT2
parsers, and maybe makes more sense anyways, since this is based on
the kind (and not the metakind) of an operator.
* Documentation improvement; doxygen top-level \file gestures, \brief
gestures for files, etc. Changed contrib/update-copyright.pl for
this change, and post-processed to add \brief. Resolves bug #98.
* Removed ExprManager::mkExpr(Kind) and NodeManager::mkNode(Kind).
They no longer made sense. Resolves bug #91.
|
|
To be done:
* Add erasable map Clause* to bool to minisat (backtracks with the solver)
* Add map from Clause* to Node (clauses that came from a node)
* Add reference counting Literal -> Node to CNFManager
* If Literal -> Node refcount goes to zero, the clauses of Node can be erased (if eresable)
* If clause is erased for each L in clause L -> Node refcount goes down
|
|
|
|
|
|
|
|
comments to files without them
|
|
* Other minor changes to the new parser to match coding guidelines,
add documentation, ....
* Add CFLAGS stuff to configure.ac parser Makefile.ams. This ensures
that profiling, coverage, optimization, debugging, and warning
level options will apply to the new parser as well (which is in C,
not C++). This fixes the deprecated warning we were seeing this
evening.
* Now, if you have ANTLR_HOME set in your environment, you don't need
to specify --with-antlr-dir to ./configure or have libantlr3c
installed in standard places. --with-antlr-dir still overrides
$ANTLR_HOME, and if the installation in $ANTLR_HOME is missing or
doesn't work, the standard places are still tried.
* Extend "silent make" to new parser stuff.
* Added src/parser/bounded_token_buffer.{h,cpp} to the list of
exclusions in contrib/update-copyright.pl and mention them as
excluded from CVC4 copyright in COPYING. They are antlr3-derived
works, covered under a BSD license.
OTHER STUFF:
* expr_manager.h, expr.h, expr_manager.cpp, and expr.cpp are now
auto-generated by a "mkexpr" script. This provides the correct
instantiations of mkConst() for public use, e.g., by the parser.
* Fix doxygen documentation in expr, expr_manager.. closes bug #35
* Node::isAtomic() implemented in a better way, based on theory kinds
files. Fixes bug #40. To support this, a "nonatomic_operator"
command has been added. All other "parameterized" or "operator"
kinds are atomic.
* Added expr_black test
* Remove kind::TRUE and kind::FALSE and make a new CONST_BOOLEAN kind
that takes a "bool" payload; for example, to make "true" you now do
nodeManager->mkConst(true).
* Make new "cvc4_public.h" and "cvc4parser_public.h" headers. Private
headers should include "cvc4_private.h"
(resp. "cvc4parser_private.h"), which existed previously. Public
headers should include the others. **No one** should include the
autoheader #include (which has been renamed "cvc4autoconfig.h")
directly, and public CVC4 headers can't access its #defines. This
is to avoid us having the same distribution problem as libantlr3c.
* Preliminary fixes based on Tim's code review of attributes (bug #61).
This includes splitting hairy template internals into
attribute_internals.h, for which another code review ticket will be
opened. Bug is still outstanding, but pending further
refactoring/documentation.
* Some *HashFcns renamed to *HashStrategy to match refactoring done
elsewhere (done by Chris?) earlier this week.
* Simplified creation of make rules for generated files (expr.cpp,
expr.h, expr_manager.cpp, expr_manager.h, theoryof_table.h, kind.h,
metakind.h).
* CVC4::Configuration interface and implementation split (so private
stuff doesn't leak into public headers).
* Some documentation/code formatting fixes.
* Add required versions of autotools to autogen.sh.
* src/expr/mkmetakind: fix a nonportable thing in invocation of "expr"
that was causing warnings on Red Hat.
* src/context/cdmap.h: add workaround to what appears to be a g++ 4.1
parsing bug.
|
|
Minor name changes for cleanup and hash function templates
|
|
clauses and:
* adding the smallest test case (eq_diamond23.smt) that memouts in 50s
* adding the initial attributes black box test
|
|
DecisionEngine, and Theory. These are triggered from the SmtEngine
dtor before the other engines are deleted. This is important
because of potential issues with outstanding TNodes in Theory
implementations (which fail if the destruction is done in the wrong
order, in certain cases).
* Favor "FactQueueResetter" instead of clearAssertionQueues() for
resetting facts queue in Theory implementations.
* Better theory-rewriting code.
* Minor cleanups.
|
|
that fail on "x != x"
|
|
the theories didn't get cleared on SatSolver backtracking so there were unasserted literals being returned as part of some conflicts. Sat solver now explicitely calls in the theory engine after it backtracks in order to clear the queues (clearAssertionQueues).
Also, changed the let.smt as it used to exibit "single literal conflict" problem. The sat solve can not except conflicts similar to (x != x), these should be rewritten to false during pre-processing.
Adding 3 more small problems from the library that we can solve now to the regressions.
|
|
|
|
|
|
(A Lightweight Component Caching Scheme for Satisfiability Solvers <http://www.springerlink.com/content/y802q03263x84159/>)
|
|
|
|
test behavior of grow(), which was previously very broken, fixed by
Tim earlier this afternoon.
* add the notion of a "private header". Private header files (those
not intended for distribution) should now #include "cvc4_private.h"
(or "cvc4parser_private.h" for the parser code). When not actually
building libcvc4 (resp. libcvc4parser), or associated unit tests, a
warning is emitted by the preprocessor. This should make it easier
to notice (and disentangle early) any unwanted public/private
mixing. Currently the warning identifies a couple places where we
need to fix things.
* added directory infrastructure for arrays and BV theories.
* the Theory inheritance hierarchy makes some assumptions about the
way inheritance is done. These are checked at runtime when
CVC4_ASSERTIONS is on. See src/theory/theory.h's TheoryImpl<>
definition for details.
* src/theory/booleans/theory_bool.h, src/theory/booleans/theory_def.h,
src/theory/arith/theory_arith.h, src/theory/arith/theory_def.h,
src/theory/uf/theory_uf.h, src/theory/uf/theory_def.h,
src/parser/antlr_parser.h: minor code formatting fixes as per
policy.
* src/theory/uf/theory_uf.cpp: fix for non-debug builds.
* src/util/options.h, src/util/model.h, src/util/result.h,
src/expr/type.h: make CVC4_PUBLIC.
* src/util/decision_engine.h: no longer CVC4_PUBLIC.
* src/expr/expr_manager.cpp: ExprManager::booleanType() and
ExprManager::kindType() weren't returning a value ?! Fixed.
* src/expr/expr_manager.h, src/expr/node_manager.h: ExprManager no
longer depends on NodeManager (public/private interface mixing).
ExprManagerScope is an internal implementation detail, and is moved
to node_manager.h.
* src/expr/node.h: mark gdb debug routines as "used" so that GCC
always emits code for them (even though its static analysis shows
they're unused).
|