1. 1 A theory-based decision heuristic for DPLL(T) Dan Goldwasser Ofer Strichman Shai Fine Haifa university TechnionIBM-HRL

2. 2 DPLL Decide BCP Analyze conflict Backtrack SAT UNSAT full assignment partial assignment conflict

3. 3 DPLL(T) Decide BCP DeductionAdd Clauses Analyze conflict Backtrack SAT UNSAT full assignment partial assignment conflict T-propagation / T-conflict

4. 4 Theory propagation Matters for efficiency, not correctness. Depending on the theory, the best strategy can be:  One T-implication at a time  All possible T-implications (“exhaustive theory-propagation”).  Cheap-to-compute T-implications

5. 5 In this work we are interested in Linear Arithmetic ( LA ) We will see:  The potential of theory propagation  Why doesn’t it work today  How can it be approximated efficiently Theory propagation for LA

6. 6 A geometric interpretation Let H be a finite set of hyperplanes in d dimensions. Let n = | H | An arrangement of H, denoted A ( H ), is a partition of R d. An arrangement in d =2: # cells · n d

7. 7 A geometric interpretation Consider a consistent partial assignment of size r.  e.g. assignment to ( l 1, l 2, l 3 ), hence r =3. How many such T-implications are there ? l1l1 l4l4 current partial assignment (1,0,0) n = 6 r = 3 l5l5 T-Implied

8. 8 A geometric interpretation Consider a consistent partial assignment of size r. Theorem 1: O(( n ¢ log r ) / r ) of the remaining constraints intersect the cell [HW87] with high probability (1 - 1/r c ). Some example numbers:  r = 3, ~47% of the remaining constraints are implied.  r = 12, ~70% of the remaining constraints are implied.  r = 60, ~90% of the remaining constraints are implied. [HW87] D. Haussler and E. Welzl. Epsilon-nets and simplex range queries. Comput. Geom., 2:127- 151, 1987.

9. 9 Assigned vs. implied in practice Two benchmarks. Measured averages at T-consistent points

10. 10 Theory propagation for LA Let l 1, l 2, l 3 be asserted. Is l 4 (or : l 4 ) T-implied ? Two techniques for finding T-implications. 1. “Plunging”: check satisfiability of ( l 1 Æ l 2 Æ l 3 Æ l 4 ) and of ( l 1 Æ l 2 Æ l 3 Æ : l 4 ) Requires solving a linear system. Too expensive in practice (see e.g. [DdM06]). [DdM06] Integrating simplex with DPLL(T), Dutertre and De Moura, SRI-CSL-06-01

11. 11 Theory propagation for LA Let l 1, l 2, l 3 be asserted. Is l 4 (or : l 4 ) T-implied ? Two techniques for finding T-implications. 2. Check if all vertices on the same side of l 4 There is an exponential number of vertices. Too expensive in practice.

12. 12 Approximating theory propagation Problem 1: How can we use conjectured information without losing soundness ? Problem 2: how can we find cheaply good conjectures  i.e., conjectured T-implications

13. 13 Problem 1: how to use conjectures ? We use conjectured implications just to bias decisions. SAT chooses a variable to decide, we conjecture its value. Might be better than the alternative: SAT’s heuristics are T-ignorant.

14. 14 Problem 2: conjecturing T-implications We examined two methods: 1. k - vertices  Find k -vertices.  If they are all on the same side of l 4 – conjecture that l 4 is implied. l4l4 In this case we conjecture : l 4

15. 15 Problem 2: conjecturing T-implications We examined two methods: 1. k - vertices  Find k -vertices.  If they are all on the same side of l 4 – conjecture that l 4 is implied. l4l4 In this case we conjecture nothing

16. 16 Problem 2: conjecturing T-implications We examined two methods: 1. k - vertices  Find k -vertices.  If they are all on the same side of l 4 – conjecture that l 4 is implied. l4l4 In this case we (falsely) conjecture l 4

17. 17 Problem 2: conjecturing T-implications We examined two methods: 1. k - vertices  Find k -vertices.  If they are all on the same side of l 4 – conjecture that l 4 is implied.  Too expensive in practice

18. 18 Problem 2: conjecturing T-implications We examined two methods: 2. One approximated point Here we always conjecture a T-implication. l4l4

19. 19 Problem 2: conjecturing T-implications We examined two methods: 2. One approximated point Here we always conjecture a T-implication. l4l4

20. 20 Problem 2: conjecturing T-implications We examined two methods: 2. One approximated point Here we always conjecture a T-implication. l4l4

21. 21 Problem 2: conjecturing T-implications We examined two methods: 2. One approximated point The idea: use the assignment maintained by Simplex. It’s for free. Competitive SMT solvers  Use general Simplex [DdM06], not classical Simplex  Do not activate Simplex after each assignment  They only update the assignment  according to the ‘simple’ constraints (e.g. “ x < c ”).

22. 22 Problem 2: conjecturing T-implications The assignment  maintained by general Simplex is updated after each partial (Boolean) assignment   Based on simple constraints only. Several possibilities:  is T-inconsistent  is T-consistent  doesn’t satisfy it  is T-consistent  satisfies it 22%

23. 23 Problem 2: conjecturing T-implications Our hope:  is ‘close’ to the polygon. Therefore it can be successful in guessing implications. Even if l 4 is not T-implied, it can guide the search. l4l4

24. 24 Results Some results for the 200 benchmarks from SMT-COMP’07 Implementation on top of ArgoLib Each column refers to a different strategy of choosing the value.

25. 25 0-pivot vs. Minisat MiniSat

26. 26 Back to the future # of cells is exponential in d rather than exponential in n  n d rather than 2 n   for n sufficiently larger than d, better worst-case complexity SMT-LIB + SRI’s GDP benchmarks Examples: n : d  QF_RDL_SCHEDULING 10.9: 1  QF_RDL_SAL6.7 : 1  QF_LRA_SC 3.9 : 1  QF_LRA_START_UP 6.9 : 1  QF_LRA_UART6.1 : 1  QF_LRA_CLOCK_SYNCH 3.3 : 1  QF_LRA_SPIDER_BENCHMARKS 3.2 : 1  QF_LRA_SAL6.1 : 1  MathSAT benchmarks (difference logic) 44.5: 1  SEP benchmarks (difference logic)17: 1

27. 27 P#2: a reversed lazy approach Current SAT-based ‘lazy’ approaches  Search the Boolean domain  check assignment in the theory domain A ‘reversed lazy approach’:  Search the theory domain  check assignment in the Boolean domain T-solver SAT

28. 28 How can we enumerate the cells ? There exists a data structure (“incidence graph”) that represents the linear arrangement Too large in practice…  Corresponds to an explicit representation of the search space.  Constructing a symbolic representation seems as hard as building the arrangement. For two years we worked on a random, incremental algorithm, each time adding a constraint and consulting SAT. The short summary: we were unable to beat Yices…

29. 29 Summary We showed how to use ‘free’ information computed by general simplex in order to improve SAT’s decision.  Somewhat compensates on the fact that there is no theory propagation. Future research:  How can we let the theory lead efficiently ?

30. 30 How many T-implications are there ? Let p be a polygon defined by a (consistent) assignment to r, ( r · n ), hyperplanes Theorem 1: O(( n ¢ log r ) / r ) of the remaining constraints intersect the cell [HW87] with high probability (1 - 1/r c ). In practice, less constraints are implied:  Due to the constant in O  Assignments and predicates are not random  Most decisions are made in low decision levels. [HW87] D. Haussler and E. Welzl. Epsilon-nets and simplex range queries. Comput. Geom., 2:127- 151, 1987.

31. 31 Let’s summarize our failed attempts… For two years we worked on a random algorithm:  Choose randomly r constraints.  Build the corresponding arrangement H ( r ).  Now each cell corresponds to a partial assignment.  Together with BCP may lead to a conflict.  Otherwise – need to refine.  … The short summary: we could never beat Yices.