mirrors/z3
3
0
Fork 0
mirror of https://github.com/Z3Prover/z3 synced 2025-11-04 21:39:13 +00:00
Code Activity

fix param evaluation non-determinism

Browse source 
Signed-off-by: Lev Nachmanson <levnach@hotmail.com>
This commit is contained in:
Lev Nachmanson 2025-10-30 07:12:06 -07:00
parent 4b2a4a902d
commit 3c5cb7c044
5 changed files with 94 additions and 59 deletions
Download patch file Download diff file Expand all files Collapse all files

6
src/nlsat/tactic/goal2nlsat.cpp
View file

@ -125,8 +125,9 @@ struct goal2nlsat::imp {
m_qm.div(lcm, d2, d2);
m_qm.neg(d2);
polynomial_ref p(m_pm);
// TODO: non-deterministic parameter evaluation
p = m_pm.addmul(d1, m_pm.mk_unit(), p1, d2, m_pm.mk_unit(), p2);
polynomial::monomial * unit1 = m_pm.mk_unit();
polynomial::monomial * unit2 = m_pm.mk_unit();
p = m_pm.addmul(d1, unit1, p1, d2, unit2, p2);
TRACE(goal2nlsat_bug, tout << mk_pp(f, m) << " p: " << p << "\nk: " << k << "\n";);
if (is_const(p)) {
int sign;
@ -444,4 +445,3 @@ expr_ref nlsat2goal::operator()(nlsat::solver& s, u_map<expr*> const& b2a, u_map
return (*m_imp)(s, b2a, x2t, l);
}

69
src/nlsat/tactic/qfnra_nlsat_tactic.cpp
View file

@ -42,30 +42,47 @@ tactic * mk_qfnra_nlsat_tactic(ast_manager & m, params_ref const & p) {
else
factor = mk_skip_tactic();
return and_then(
mk_report_verbose_tactic("(qfnra-nlsat-tactic)", 10),
// TODO: non-deterministic parameter evaluation
and_then(using_params(mk_simplify_tactic(m, p),
main_p),
using_params(mk_purify_arith_tactic(m, p),
purify_p),
mk_propagate_values_tactic(m, p),
mk_solve_eqs_tactic(m, p),
mk_elim_uncnstr_tactic(m, p),
mk_elim_term_ite_tactic(m, p),
using_params(mk_purify_arith_tactic(m, p),
purify_p)),
// TODO: non-deterministic parameter evaluation
and_then(/* mk_degree_shift_tactic(m, p), */ // may affect full dimensionality detection
factor,
mk_solve_eqs_tactic(m, p),
using_params(mk_purify_arith_tactic(m, p),
purify_p),
using_params(mk_simplify_tactic(m, p),
main_p),
mk_tseitin_cnf_core_tactic(m, p),
using_params(mk_simplify_tactic(m, p),
main_p),
mk_nlsat_tactic(m, p)));
}
tactic * report = mk_report_verbose_tactic("(qfnra-nlsat-tactic)", 10);
tactic * simplify_base1 = mk_simplify_tactic(m, p);
tactic * simplify_main1 = using_params(simplify_base1, main_p);
tactic * purify_base1 = mk_purify_arith_tactic(m, p);
tactic * purify_main1 = using_params(purify_base1, purify_p);
tactic * propagate_values = mk_propagate_values_tactic(m, p);
tactic * solve_eqs1 = mk_solve_eqs_tactic(m, p);
tactic * elim_unc = mk_elim_uncnstr_tactic(m, p);
tactic * elim_term_ite = mk_elim_term_ite_tactic(m, p);
tactic * purify_base2 = mk_purify_arith_tactic(m, p);
tactic * purify_main2 = using_params(purify_base2, purify_p);
tactic * preprocessing = and_then(
simplify_main1,
purify_main1,
propagate_values,
solve_eqs1,
elim_unc,
elim_term_ite,
purify_main2);
tactic * solve_eqs2 = mk_solve_eqs_tactic(m, p);
tactic * purify_base3 = mk_purify_arith_tactic(m, p);
tactic * purify_main3 = using_params(purify_base3, purify_p);
tactic * simplify_base2 = mk_simplify_tactic(m, p);
tactic * simplify_main2 = using_params(simplify_base2, main_p);
tactic * tseitin = mk_tseitin_cnf_core_tactic(m, p);
tactic * simplify_base3 = mk_simplify_tactic(m, p);
tactic * simplify_main3 = using_params(simplify_base3, main_p);
tactic * nlsat = mk_nlsat_tactic(m, p);
tactic * post = and_then(
/* mk_degree_shift_tactic(m, p), */ // may affect full dimensionality detection
factor,
solve_eqs2,
purify_main3,
simplify_main2,
tseitin,
simplify_main3,
nlsat);
return and_then(report, preprocessing, post);
}

20
src/tactic/arith/bv2real_rewriter.cpp
View file

@ -585,8 +585,9 @@ br_status bv2real_rewriter::mk_ite(expr* c, expr* s, expr* t, expr_ref& result)
u().align_divisors(s1, s2, t1, t2, d1, d2);
u().align_sizes(s1, t1);
u().align_sizes(s2, t2);
// TODO: non-deterministic parameter evaluation
if (u().mk_bv2real(m().mk_ite(c, s1, t1), m().mk_ite(c, s2, t2), d1, r1, result)) {
expr_ref ite_s1(m().mk_ite(c, s1, t1), m());
expr_ref ite_s2(m().mk_ite(c, s2, t2), m());
if (u().mk_bv2real(ite_s1.get(), ite_s2.get(), d1, r1, result)) {
return BR_DONE;
}
}
@ -615,8 +616,9 @@ br_status bv2real_rewriter::mk_uminus(expr * s, expr_ref & result) {
if (u().is_bv2real(s, s1, s2, d1, r1)) {
s1 = u().mk_extend(1, s1);
s2 = u().mk_extend(1, s2);
// TODO: non-deterministic parameter evaluation
if (u().mk_bv2real(m_bv.mk_bv_neg(s1), m_bv.mk_bv_neg(s2), d1, r1, result)) {
expr_ref neg_s1(m_bv.mk_bv_neg(s1), m());
expr_ref neg_s2(m_bv.mk_bv_neg(s2), m());
if (u().mk_bv2real(neg_s1.get(), neg_s2.get(), d1, r1, result)) {
return BR_DONE;
}
}
@ -638,8 +640,9 @@ br_status bv2real_rewriter::mk_add(expr* s, expr* t, expr_ref& result) {
rational d1, d2, r1, r2;
if (u().is_bv2real(s, s1, s2, d1, r1) && u().is_bv2real(t, t1, t2, d2, r2) && r1 == r2) {
u().align_divisors(s1, s2, t1, t2, d1, d2);
// TODO: non-deterministic parameter evaluation
if (u().mk_bv2real(u().mk_bv_add(s1, t1), u().mk_bv_add(t2, s2), d1, r1, result)) {
expr_ref sum1(u().mk_bv_add(s1, t1), m());
expr_ref sum2(u().mk_bv_add(t2, s2), m());
if (u().mk_bv2real(sum1.get(), sum2.get(), d1, r1, result)) {
return BR_DONE;
}
}
@ -699,8 +702,9 @@ br_status bv2real_rewriter::mk_sub(expr* s, expr* t, expr_ref& result) {
rational d1, d2, r1, r2;
if (u().is_bv2real(s, s1, s2, d1, r1) && u().is_bv2real(t, t1, t2, d2, r2) && r1 == r2) {
u().align_divisors(s1, s2, t1, t2, d1, d2);
// TODO: non-deterministic parameter evaluation
if (u().mk_bv2real(u().mk_bv_sub(s1, t1), u().mk_bv_sub(s2, t2), d1, r1, result)) {
expr_ref diff1(u().mk_bv_sub(s1, t1), m());
expr_ref diff2(u().mk_bv_sub(s2, t2), m());
if (u().mk_bv2real(diff1.get(), diff2.get(), d1, r1, result)) {
return BR_DONE;
}
}

18
src/tactic/arith/factor_tactic.cpp
View file

@ -157,8 +157,10 @@ class factor_tactic : public tactic {
}
}
else {
// TODO: non-deterministic parameter evaluation
args.push_back(m.mk_app(m_util.get_family_id(), k, mk_mul(odd_factors.size(), odd_factors.data()), mk_zero_for(odd_factors[0])));
expr_ref odd_product(mk_mul(odd_factors.size(), odd_factors.data()), m);
expr* zero = mk_zero_for(odd_factors[0]);
expr_ref comparison(m.mk_app(m_util.get_family_id(), k, odd_product, zero), m);
args.push_back(comparison);
}
SASSERT(!args.empty());
if (args.size() == 1)
@ -186,11 +188,12 @@ class factor_tactic : public tactic {
scoped_mpz lcm(m_qm);
m_qm.lcm(d1, d2, lcm);
m_qm.div(lcm, d1, d1);
m_qm.div(lcm, d2, d2);
m_qm.neg(d2);
polynomial_ref p(m_pm);
// TODO: non-deterministic parameter evaluation
p = m_pm.addmul(d1, m_pm.mk_unit(), p1, d2, m_pm.mk_unit(), p2);
m_qm.div(lcm, d2, d2);
m_qm.neg(d2);
polynomial_ref p(m_pm);
polynomial::monomial* unit1 = m_pm.mk_unit();
polynomial::monomial* unit2 = m_pm.mk_unit();
p = m_pm.addmul(d1, unit1, p1, d2, unit2, p2);
if (is_const(p))
return BR_FAILED;
polynomial::factors fs(m_pm);
@ -337,4 +340,3 @@ tactic * mk_factor_tactic(ast_manager & m, params_ref const & p) {
return clean(alloc(factor_tactic, m, p));
}

40
src/tactic/sls/sls_tactic.cpp
View file

@ -244,23 +244,35 @@ static tactic * mk_preamble(ast_manager & m, params_ref const & p, bool add_nnf)
// conservative gaussian elimination.
gaussian_p.set_uint("gaussian_max_occs", 2);
return and_then(
// TODO: non-deterministic parameter evaluation
and_then(mk_simplify_tactic(m, p),
mk_propagate_values_tactic(m),
using_params(mk_solve_eqs_tactic(m), gaussian_p),
mk_elim_uncnstr_tactic(m),
mk_bv_size_reduction_tactic(m),
using_params(mk_simplify_tactic(m), simp2_p)),
using_params(mk_simplify_tactic(m), hoist_p),
mk_max_bv_sharing_tactic(m),
add_nnf ? mk_nnf_tactic(m, p) : mk_skip_tactic()
);
tactic* simplify_primary = mk_simplify_tactic(m, p);
tactic* propagate_values = mk_propagate_values_tactic(m);
tactic* solve_eqs_base = mk_solve_eqs_tactic(m);
tactic* solve_eqs_gaussian = using_params(solve_eqs_base, gaussian_p);
tactic* elim_unc = mk_elim_uncnstr_tactic(m);
tactic* size_reduction = mk_bv_size_reduction_tactic(m);
tactic* simplify_secondary_base = mk_simplify_tactic(m);
tactic* simplify_secondary = using_params(simplify_secondary_base, simp2_p);
tactic* inner = and_then(
simplify_primary,
propagate_values,
solve_eqs_gaussian,
elim_unc,
size_reduction,
simplify_secondary);
tactic* simplify_hoist_base = mk_simplify_tactic(m);
tactic* simplify_hoist = using_params(simplify_hoist_base, hoist_p);
tactic* max_sharing = mk_max_bv_sharing_tactic(m);
tactic* nnf = add_nnf ? mk_nnf_tactic(m, p) : mk_skip_tactic();
return and_then(inner, simplify_hoist, max_sharing, nnf);
}
tactic * mk_qfbv_sls_tactic(ast_manager & m, params_ref const & p) {
// TODO: non-deterministic parameter evaluation
tactic * t = and_then(mk_preamble(m, p, true), mk_sls_tactic(m, p));
tactic * preamble = mk_preamble(m, p, true);
tactic * sls = mk_sls_tactic(m, p);
tactic * t = and_then(preamble, sls);
t->updt_params(p);
return t;
}