mirrors/z3
3
0
Fork 0
mirror of https://github.com/Z3Prover/z3 synced 2025-10-31 03:32:28 +00:00
Code Activity

Merge pull request #1700 from agurfinkel/deep_space

Browse source 
Deep space
This commit is contained in:
Nikolaj Bjorner 2018-06-24 19:13:34 -07:00 • committed by GitHub
commit a5f4980c92
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
7 changed files with 910 additions and 590 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/muz/spacer/spacer_iuc_proof.h
View file

@ -33,7 +33,7 @@ public:
bool is_h_marked(proof* p) {return m_h_mark.is_marked(p);}
bool is_b_pure (proof *p) {
return !is_h_marked (p) && is_core_pure(m.get_fact (p));
return !is_h_marked(p) && !this->is_a_marked(p) && is_core_pure(m.get_fact(p));
}
void display_dot(std::ostream &out);

34
src/muz/spacer/spacer_iuc_solver.cpp
View file

@ -322,6 +322,24 @@ void iuc_solver::get_iuc(expr_ref_vector &core)
// -- new hypothesis reducer
else
{
#if 0
static unsigned bcnt = 0;
{
bcnt++;
TRACE("spacer", tout << "Dumping pf bcnt: " << bcnt << "\n";);
if (bcnt == 123) {
std::ofstream ofs;
ofs.open("/tmp/bpf_" + std::to_string(bcnt) + ".dot");
iuc_proof iuc_pf_before(m, res.get(), core_lits);
iuc_pf_before.display_dot(ofs);
ofs.close();
proof_checker pc(m);
expr_ref_vector side(m);
ENSURE(pc.check(res, side));
}
}
#endif
scoped_watch _t_ (m_hyp_reduce2_sw);
// pre-process proof for better iuc extraction
@ -356,6 +374,22 @@ void iuc_solver::get_iuc(expr_ref_vector &core)
iuc_proof iuc_pf(m, res, core_lits);
#if 0
static unsigned cnt = 0;
{
cnt++;
TRACE("spacer", tout << "Dumping pf cnt: " << cnt << "\n";);
if (cnt == 123) {
std::ofstream ofs;
ofs.open("/tmp/pf_" + std::to_string(cnt) + ".dot");
iuc_pf.display_dot(ofs);
ofs.close();
proof_checker pc(m);
expr_ref_vector side(m);
ENSURE(pc.check(res, side));
}
}
#endif
unsat_core_learner learner(m, iuc_pf);
unsat_core_plugin* plugin;

1165
src/muz/spacer/spacer_proof_utils.cpp
View file

File diff suppressed because it is too large Load diff

36
src/muz/spacer/spacer_unsat_core_learner.cpp
View file

@ -36,28 +36,27 @@ void unsat_core_learner::register_plugin(unsat_core_plugin* plugin) {
}
void unsat_core_learner::compute_unsat_core(expr_ref_vector& unsat_core) {
// traverse proof
proof_post_order it(m_pr.get(), m);
while (it.hasNext()) {
proof* currentNode = it.next();
proof* curr = it.next();
if (m.get_num_parents(currentNode) > 0) {
bool need_to_mark_closed = true;
bool done = is_closed(curr);
if (done) continue;
for (proof* premise : m.get_parents(currentNode)) {
need_to_mark_closed &= (!m_pr.is_b_marked(premise) || m_closed.is_marked(premise));
}
// save result
m_closed.mark(currentNode, need_to_mark_closed);
if (m.get_num_parents(curr) > 0) {
done = true;
for (proof* p : m.get_parents(curr)) done &= !is_b_open(p);
set_closed(curr, done);
}
// we have now collected all necessary information, so we can visit the node
// if the node mixes A-reasoning and B-reasoning and contains non-closed premises
if (m_pr.is_a_marked(currentNode) &&
m_pr.is_b_marked(currentNode) &&
!m_closed.is_marked(currentNode)) {
compute_partial_core(currentNode); // then we need to compute a partial core
// we have now collected all necessary information,
// so we can visit the node
// if the node mixes A-reasoning and B-reasoning
// and contains non-closed premises
if (!done) {
if (is_a(curr) && is_b(curr)) {
compute_partial_core(curr);
}
}
}
@ -74,7 +73,7 @@ void unsat_core_learner::compute_unsat_core(expr_ref_vector& unsat_core) {
void unsat_core_learner::compute_partial_core(proof* step) {
for (unsat_core_plugin* plugin : m_plugins) {
if (m_closed.is_marked(step)) break;
if (is_closed(step)) break;
plugin->compute_partial_core(step);
}
}
@ -93,9 +92,6 @@ void unsat_core_learner::set_closed(proof* p, bool value) {
m_closed.mark(p, value);
}
bool unsat_core_learner::is_b_open(proof *p) {
return m_pr.is_b_marked(p) && !is_closed (p);
}
void unsat_core_learner::add_lemma_to_core(expr* lemma) {
m_unsat_core.push_back(lemma);

33
src/muz/spacer/spacer_unsat_core_learner.h
View file

@ -22,6 +22,7 @@ Revision History:
#include "ast/ast.h"
#include "muz/spacer/spacer_util.h"
#include "muz/spacer/spacer_proof_utils.h"
#include "muz/spacer/spacer_iuc_proof.h"
namespace spacer {
@ -31,13 +32,32 @@ namespace spacer {
class unsat_core_learner {
typedef obj_hashtable<expr> expr_set;
ast_manager& m;
iuc_proof& m_pr;
ptr_vector<unsat_core_plugin> m_plugins;
ast_mark m_closed;
expr_ref_vector m_unsat_core;
public:
unsat_core_learner(ast_manager& m, iuc_proof& pr) :
m(m), m_pr(pr), m_unsat_core(m) {};
virtual ~unsat_core_learner();
ast_manager& m;
iuc_proof& m_pr;
ast_manager& get_manager() {return m;}
bool is_a(proof *pr) {
// AG: treat hypotheses as A
// AG: assume that all B-hyp have been eliminated
// AG: this is not yet true in case of arithmetic eq_prop
return m_pr.is_a_marked(pr) || is_h(pr);
}
bool is_b(proof *pr) {return m_pr.is_b_marked(pr);}
bool is_h(proof *pr) {return m_pr.is_h_marked(pr);}
bool is_b_pure(proof *pr) { return m_pr.is_b_pure(pr);}
bool is_b_open(proof *p) {return m_pr.is_b_marked(p) && !is_closed (p);}
/*
* register a plugin for computation of partial unsat cores
@ -59,7 +79,6 @@ namespace spacer {
bool is_closed(proof* p);
void set_closed(proof* p, bool value);
bool is_b_open (proof *p);
/*
* adds a lemma to the unsat core
@ -67,14 +86,6 @@ namespace spacer {
void add_lemma_to_core(expr* lemma);
private:
ptr_vector<unsat_core_plugin> m_plugins;
ast_mark m_closed;
/*
* collects the lemmas of the unsat-core
* will at the end be inserted into unsat_core.
*/
expr_ref_vector m_unsat_core;
/*
* computes partial core for step by delegating computation to plugins

214
src/muz/spacer/spacer_unsat_core_plugin.cpp
View file

@ -34,82 +34,81 @@ Revision History:
namespace spacer {
unsat_core_plugin::unsat_core_plugin(unsat_core_learner& learner):
m(learner.m), m_learner(learner) {};
unsat_core_plugin::unsat_core_plugin(unsat_core_learner& ctx):
m(ctx.get_manager()), m_ctx(ctx) {};
void unsat_core_plugin_lemma::compute_partial_core(proof* step) {
SASSERT(m_learner.m_pr.is_a_marked(step));
SASSERT(m_learner.m_pr.is_b_marked(step));
for (proof* premise : m.get_parents(step)) {
if (m_learner.is_b_open (premise)) {
SASSERT(m_ctx.is_a(step));
SASSERT(m_ctx.is_b(step));
for (auto* p : m.get_parents(step)) {
if (m_ctx.is_b_open (p)) {
// by IH, premises that are AB marked are already closed
SASSERT(!m_learner.m_pr.is_a_marked(premise));
add_lowest_split_to_core(premise);
SASSERT(!m_ctx.is_a(p));
add_lowest_split_to_core(p);
}
}
m_learner.set_closed(step, true);
m_ctx.set_closed(step, true);
}
void unsat_core_plugin_lemma::add_lowest_split_to_core(proof* step) const
{
SASSERT(m_learner.is_b_open(step));
void unsat_core_plugin_lemma::add_lowest_split_to_core(proof* step) const {
SASSERT(m_ctx.is_b_open(step));
ptr_buffer<proof> todo;
todo.push_back(step);
while (!todo.empty()) {
proof* pf = todo.back();
todo.pop_back();
// if current step hasn't been processed,
if (!m_learner.is_closed(pf)) {
m_learner.set_closed(pf, true);
if (!m_ctx.is_closed(pf)) {
m_ctx.set_closed(pf, true);
// the step is b-marked and not closed.
// by I.H. the step must be already visited
// so if it is also a-marked, it must be closed
SASSERT(m_learner.m_pr.is_b_marked(pf));
SASSERT(!m_learner.m_pr.is_a_marked(pf));
SASSERT(m_ctx.is_b(pf));
SASSERT(!m_ctx.is_a(pf));
// the current step needs to be interpolated:
expr* fact = m.get_fact(pf);
// if we trust the current step and we are able to use it
if (m_learner.m_pr.is_b_pure (pf) &&
(m.is_asserted(pf) || is_literal(m, fact))) {
if (m_ctx.is_b_pure (pf) && (m.is_asserted(pf) || is_literal(m, fact))) {
// just add it to the core
m_learner.add_lemma_to_core(fact);
m_ctx.add_lemma_to_core(fact);
}
// otherwise recurse on premises
else {
for (proof* premise : m.get_parents(pf))
if (m_learner.is_b_open(premise))
for (proof* premise : m.get_parents(pf))
if (m_ctx.is_b_open(premise))
todo.push_back(premise);
}
}
}
}
void unsat_core_plugin_farkas_lemma::compute_partial_core(proof* step)
{
SASSERT(m_learner.m_pr.is_a_marked(step));
SASSERT(m_learner.m_pr.is_b_marked(step));
/***
* FARKAS
*/
void unsat_core_plugin_farkas_lemma::compute_partial_core(proof* step) {
SASSERT(m_ctx.is_a(step));
SASSERT(m_ctx.is_b(step));
// XXX this assertion should be true so there is no need to check for it
SASSERT (!m_learner.is_closed (step));
SASSERT (!m_ctx.is_closed (step));
func_decl* d = step->get_decl();
symbol sym;
if (!m_learner.is_closed(step) && // if step is not already interpolated
is_farkas_lemma(m, step)) {
// weaker check: d->get_num_parameters() >= m.get_num_parents(step) + 2
TRACE("spacer.farkas",
tout << "looking at: " << mk_pp(step, m) << "\n";);
if (!m_ctx.is_closed(step) && is_farkas_lemma(m, step)) {
// weaker check : d->get_num_parameters() >= m.get_num_parents(step) + 2
SASSERT(d->get_num_parameters() == m.get_num_parents(step) + 2);
SASSERT(m.has_fact(step));
coeff_lits_t coeff_lits;
expr_ref_vector pinned(m);
/* The farkas lemma represents a subproof starting from premise(-set)s A, BNP and BP(ure) and
* ending in a disjunction D. We need to compute the contribution of BP, i.e. a formula, which
* is entailed by BP and together with A and BNP entails D.
@ -134,34 +133,35 @@ namespace spacer {
* as workaround we take the absolute value of the provided coefficients.
*/
parameter const* params = d->get_parameters() + 2; // point to the first Farkas coefficient
STRACE("spacer.farkas",
verbose_stream() << "Farkas input: "<< "\n";
for (unsigned i = 0; i < m.get_num_parents(step); ++i) {
proof * prem = m.get_parent(step, i);
rational coef = params[i].get_rational();
bool b_pure = m_learner.m_pr.is_b_pure (prem);
verbose_stream() << (b_pure?"B":"A") << " " << coef << " " << mk_pp(m.get_fact(prem), m) << "\n";
}
);
bool can_be_closed = true;
TRACE("spacer.farkas",
tout << "Farkas input: "<< "\n";
for (unsigned i = 0; i < m.get_num_parents(step); ++i) {
proof * prem = m.get_parent(step, i);
rational coef = params[i].get_rational();
bool b_pure = m_ctx.is_b_pure (prem);
tout << (b_pure?"B":"A") << " " << coef << " " << mk_pp(m.get_fact(prem), m) << "\n";
}
);
bool done = true;
for (unsigned i = 0; i < m.get_num_parents(step); ++i) {
proof * premise = m.get_parent(step, i);
if (m_learner.is_b_open (premise)) {
SASSERT(!m_learner.m_pr.is_a_marked(premise));
if (m_learner.m_pr.is_b_pure (step)) {
if (m_ctx.is_b_open (premise)) {
SASSERT(!m_ctx.is_a(premise));
if (m_ctx.is_b_pure (premise)) {
if (!m_use_constant_from_a) {
rational coefficient = params[i].get_rational();
coeff_lits.push_back(std::make_pair(abs(coefficient), (app*)m.get_fact(premise)));
}
}
else {
can_be_closed = false;
// -- mixed premise, won't be able to close this proof step
done = false;
if (m_use_constant_from_a) {
rational coefficient = params[i].get_rational();
coeff_lits.push_back(std::make_pair(abs(coefficient), (app*)m.get_fact(premise)));
@ -175,10 +175,11 @@ namespace spacer {
}
}
}
// TBD: factor into another method
if (m_use_constant_from_a) {
params += m.get_num_parents(step); // point to the first Farkas coefficient, which corresponds to a formula in the conclusion
// the conclusion can either be a single formula or a disjunction of several formulas, we have to deal with both situations
if (m.get_num_parents(step) + 2 < d->get_num_parameters()) {
unsigned num_args = 1;
@ -190,7 +191,7 @@ namespace spacer {
args = _or->get_args();
}
SASSERT(m.get_num_parents(step) + 2 + num_args == d->get_num_parameters());
bool_rewriter brw(m);
for (unsigned i = 0; i < num_args; ++i) {
expr* premise = args[i];
@ -205,13 +206,12 @@ namespace spacer {
}
// only if all b-premises can be used directly, add the farkas core and close the step
if (can_be_closed) {
m_learner.set_closed(step, true);
expr_ref res = compute_linear_combination(coeff_lits);
m_learner.add_lemma_to_core(res);
}
// AG: this decision needs to be re-evaluated. If the proof cannot be closed, literals above
// AG: it will go into the core. However, it does not mean that this literal should/could not be added.
m_ctx.set_closed(step, done);
expr_ref res = compute_linear_combination(coeff_lits);
TRACE("spacer.farkas", tout << "Farkas core: " << res << "\n";);
m_ctx.add_lemma_to_core(res);
}
}
@ -235,12 +235,12 @@ namespace spacer {
void unsat_core_plugin_farkas_lemma_optimized::compute_partial_core(proof* step)
{
SASSERT(m_learner.m_pr.is_a_marked(step));
SASSERT(m_learner.m_pr.is_b_marked(step));
SASSERT(m_ctx.is_a(step));
SASSERT(m_ctx.is_b(step));
func_decl* d = step->get_decl();
symbol sym;
if (!m_learner.is_closed(step) && // if step is not already interpolated
if (!m_ctx.is_closed(step) && // if step is not already interpolated
is_farkas_lemma(m, step)) {
SASSERT(d->get_num_parameters() == m.get_num_parents(step) + 2);
SASSERT(m.has_fact(step));
@ -249,25 +249,25 @@ namespace spacer {
parameter const* params = d->get_parameters() + 2; // point to the first Farkas coefficient
STRACE("spacer.farkas",
verbose_stream() << "Farkas input: "<< "\n";
for (unsigned i = 0; i < m.get_num_parents(step); ++i) {
proof * prem = m.get_parent(step, i);
rational coef = params[i].get_rational();
bool b_pure = m_learner.m_pr.is_b_pure (prem);
verbose_stream() << (b_pure?"B":"A") << " " << coef << " " << mk_pp(m.get_fact(prem), m_learner.m) << "\n";
}
);
TRACE("spacer.farkas",
tout << "Farkas input: "<< "\n";
for (unsigned i = 0; i < m.get_num_parents(step); ++i) {
proof * prem = m.get_parent(step, i);
rational coef = params[i].get_rational();
bool b_pure = m_ctx.is_b_pure (prem);
tout << (b_pure?"B":"A") << " " << coef << " " << mk_pp(m.get_fact(prem), m) << "\n";
}
);
bool can_be_closed = true;
for (unsigned i = 0; i < m.get_num_parents(step); ++i) {
proof * premise = m.get_parent(step, i);
if (m_learner.m_pr.is_b_marked(premise) && !m_learner.is_closed(premise))
if (m_ctx.is_b(premise) && !m_ctx.is_closed(premise))
{
SASSERT(!m_learner.m_pr.is_a_marked(premise));
SASSERT(!m_ctx.is_a(premise));
if (m_learner.m_pr.is_b_pure(premise))
if (m_ctx.is_b_pure(premise))
{
rational coefficient = params[i].get_rational();
linear_combination.push_back
@ -283,7 +283,7 @@ namespace spacer {
// only if all b-premises can be used directly, close the step and add linear combinations for later processing
if (can_be_closed)
{
m_learner.set_closed(step, true);
m_ctx.set_closed(step, true);
if (!linear_combination.empty())
{
m_linear_combinations.push_back(linear_combination);
@ -340,7 +340,7 @@ namespace spacer {
// 4. extract linear combinations from matrix and add result to core
for (unsigned k = 0; k < i; ++k)// i points to the row after the last row which is non-zero
{
coeff_lits_t coeff_lits;
coeff_lits_t coeff_lits;
for (unsigned l = 0; l < matrix.num_cols(); ++l) {
if (!matrix.get(k,l).is_zero()) {
coeff_lits.push_back(std::make_pair(matrix.get(k, l), ordered_basis[l]));
@ -349,19 +349,19 @@ namespace spacer {
SASSERT(!coeff_lits.empty());
expr_ref linear_combination = compute_linear_combination(coeff_lits);
m_learner.add_lemma_to_core(linear_combination);
m_ctx.add_lemma_to_core(linear_combination);
}
}
expr_ref unsat_core_plugin_farkas_lemma_optimized::compute_linear_combination(const coeff_lits_t& coeff_lits) {
expr_ref unsat_core_plugin_farkas_lemma_optimized::compute_linear_combination(const coeff_lits_t& coeff_lits) {
smt::farkas_util util(m);
for (auto const & p : coeff_lits) {
util.add(p.first, p.second);
}
expr_ref negated_linear_combination = util.get();
SASSERT(m.is_not(negated_linear_combination));
return expr_ref(mk_not(m, negated_linear_combination), m);
return expr_ref(mk_not(m, negated_linear_combination), m);
//TODO: rewrite the get-method to return nonnegated stuff?
}
@ -449,7 +449,7 @@ namespace spacer {
for (unsigned j = 0; j < matrix.num_cols(); ++j) {
SASSERT(matrix.get(i, j).is_int());
app_ref a_ij(util.mk_numeral(matrix.get(i,j), true), m);
app_ref sum(util.mk_int(0), m);
for (unsigned k = 0; k < n; ++k) {
sum = util.mk_add(sum, util.mk_mul(coeffs[i][k].get(), bounded_vectors[j][k].get()));
@ -458,7 +458,7 @@ namespace spacer {
s->assert_expr(eq);
}
}
// check result
lbool res = s->check_sat(0, nullptr);
@ -480,7 +480,7 @@ namespace spacer {
SASSERT(!coeff_lits.empty()); // since then previous outer loop would have found solution already
expr_ref linear_combination = compute_linear_combination(coeff_lits);
m_learner.add_lemma_to_core(linear_combination);
m_ctx.add_lemma_to_core(linear_combination);
}
return;
}
@ -504,10 +504,10 @@ namespace spacer {
{
ptr_vector<proof> todo;
SASSERT(m_learner.m_pr.is_a_marked(step));
SASSERT(m_learner.m_pr.is_b_marked(step));
SASSERT(m_ctx.is_a(step));
SASSERT(m_ctx.is_b(step));
SASSERT(m.get_num_parents(step) > 0);
SASSERT(!m_learner.is_closed(step));
SASSERT(!m_ctx.is_closed(step));
todo.push_back(step);
while (!todo.empty())
@ -516,7 +516,7 @@ namespace spacer {
todo.pop_back();
// if we need to deal with the node and if we haven't added the corresponding edges so far
if (!m_learner.is_closed(current) && !m_visited.is_marked(current))
if (!m_ctx.is_closed(current) && !m_visited.is_marked(current))
{
// compute smallest subproof rooted in current, which has only good edges
// add an edge from current to each leaf of that subproof
@ -527,7 +527,7 @@ namespace spacer {
}
}
m_learner.set_closed(step, true);
m_ctx.set_closed(step, true);
}
@ -538,7 +538,7 @@ namespace spacer {
ptr_buffer<proof> todo_subproof;
for (proof* premise : m.get_parents(step)) {
if (m_learner.m_pr.is_b_marked(premise)) {
if (m_ctx.is_b(premise)) {
todo_subproof.push_back(premise);
}
}
@ -548,21 +548,21 @@ namespace spacer {
todo_subproof.pop_back();
// if we need to deal with the node
if (!m_learner.is_closed(current))
if (!m_ctx.is_closed(current))
{
SASSERT(!m_learner.m_pr.is_a_marked(current)); // by I.H. the step must be already visited
SASSERT(!m_ctx.is_a(current)); // by I.H. the step must be already visited
// and the current step needs to be interpolated:
if (m_learner.m_pr.is_b_marked(current))
if (m_ctx.is_b(current))
{
// if we trust the current step and we are able to use it
if (m_learner.m_pr.is_b_pure (current) &&
if (m_ctx.is_b_pure (current) &&
(m.is_asserted(current) ||
is_literal(m, m.get_fact(current))))
{
// we found a leaf of the subproof, so
// 1) we add corresponding edges
if (m_learner.m_pr.is_a_marked(step))
if (m_ctx.is_a(step))
{
add_edge(nullptr, current); // current is sink
}
@ -682,9 +682,9 @@ namespace spacer {
void unsat_core_plugin_min_cut::finalize() {
unsigned_vector cut_nodes;
m_min_cut.compute_min_cut(cut_nodes);
for (unsigned cut_node : cut_nodes) {
m_learner.add_lemma_to_core(m_node_to_formula[cut_node]);
m_ctx.add_lemma_to_core(m_node_to_formula[cut_node]);
}
}
}

16
src/muz/spacer/spacer_unsat_core_plugin.h
View file

@ -35,14 +35,14 @@ namespace spacer {
virtual ~unsat_core_plugin() {};
virtual void compute_partial_core(proof* step) = 0;
virtual void finalize(){};
unsat_core_learner& m_learner;
unsat_core_learner& m_ctx;
};
class unsat_core_plugin_lemma : public unsat_core_plugin {
class unsat_core_plugin_lemma : public unsat_core_plugin {
public:
unsat_core_plugin_lemma(unsat_core_learner& learner) : unsat_core_plugin(learner){};
void compute_partial_core(proof* step) override;
unsat_core_plugin_lemma(unsat_core_learner& learner) : unsat_core_plugin(learner){};
void compute_partial_core(proof* step) override;
private:
void add_lowest_split_to_core(proof* step) const;
};
@ -54,7 +54,7 @@ namespace spacer {
bool use_constant_from_a=true) :
unsat_core_plugin(learner),
m_split_literals(split_literals),
m_use_constant_from_a(use_constant_from_a) {};
m_use_constant_from_a(use_constant_from_a) {};
void compute_partial_core(proof* step) override;
private:
bool m_split_literals;
@ -64,8 +64,8 @@ namespace spacer {
*/
expr_ref compute_linear_combination(const coeff_lits_t& coeff_lits);
};
class unsat_core_plugin_farkas_lemma_optimized : public unsat_core_plugin {
class unsat_core_plugin_farkas_lemma_optimized : public unsat_core_plugin {
public:
unsat_core_plugin_farkas_lemma_optimized(unsat_core_learner& learner, ast_manager& m) : unsat_core_plugin(learner) {};
void compute_partial_core(proof* step) override;