Tutorial on SMT Solvers
Combinatorial Problem Solving (CPS)
Enric Rodrı́guez-Carbonell
May 27, 2016
SMT Solvers
■ SMT solvers take as input a (quantifier-free) first-order logic formula F
over a background theory T , and return:
◆ sat(+ model): if F is satisfiable
◆ unsat: if F is unsatisfiable
■ We will be using Z3: [Link]
(developed by L. de Moura and N. Bjorner at Microsoft Research)
■ Usage: z3 [ <options> ] <input>
■ Some options:
◆ -stm2: use parser for SMT-LIB 2 input format
◆ -st: display statistics
◆ -rs:<seed>: set random seed
◆ -h: help, shows all options
2 / 10
Input Format: SMT-LIB 2
■ We will be using a small subset of this language.
For going beyond:
■ Tutorial (standard version 2.0):
[Link]
■ Full standard (standard version 2.5):
[Link]
3 / 10
Input Format: SMT-LIB 2
■ First, directives. E.g., asking models to be reported:
( set - option : produce - models true )
■ Second, set background theory:
( set - logic QF_LIA )
■ Standard theories of interest to us:
◆ QF_LRA : quantifier-free linear real arithmetic
◆ QF_LIA : quantifier-free linear integer arithmetic
◆ QF_RDL : quantifier-free real difference logic
◆ QF_IDL : quantifier-free integer difference logic
■ SMT-LIB 2 does not allow to have mixed problems
(although some solvers support it outside the standard)
4 / 10
Input Format: SMT-LIB 2
■ Third, declare variables.
E.g., integer variable x:
( declare - fun x () Int )
E.g., real variable z 1 3 :
( declare - fun z_1_3 () Real )
5 / 10
Input Format: SMT-LIB 2
■ Fourth, assert formula.
■ Expressions should be written in prefix form:
( < operator > < arg1 > ... < argn > )
( assert
( and
( or
( <= (+ x 3) (* 2 u) )
( >= (+ v 4) y )
( >= (+ x y z ) 2)
)
(= 7
(+
( ite ( and ( <= x 2 ) ( <= 2 (+ x 3 ( - 1)))) 3 0)
( ite ( and ( <= u 2 ) ( <= 2 (+ u 3 ( - 1)))) 4 0)
)
)
)
) 6 / 10
Input Format: SMT-LIB 2
■ and, or, + have arbitrary arity
■ - is unary or binary
■ * is binary
■ ite is the if-then-else operator (like ? in C, C++, Java).
Let a be Boolean and b, c have the same sort S.
Then (ite a b c) is the expression of sort S equal to:
◆ b if a holds
◆ c if a does not hold
7 / 10
Input Format: SMT-LIB 2
■ Finally ask the SMT solver to check satisfiability ...
( check - sat )
■ ... and report the model
( get - model )
■ Anything following a ; up to an end-of-line is a comment
8 / 10
Input Format: SMT-LIB 2
( set - option : produce - models true )
( set - logic QF_LIA )
( declare - fun x () Int )
( declare - fun y () Int )
( declare - fun z () Int ) ; This is an example
( declare - fun u () Int )
( declare - fun v () Int )
( assert
( and
( or
( <= (+ x 3) (* 2 y ) )
( >= (+ x 4) z )
) ) )
( check - sat )
( get - model )
9 / 10
Output Format
■ 1st line is sat or unsat
■ If satisfiable, then comes a description of the solution in a model
expression, where the value of each variable is given by:
(define − fun < variable > () < sort > < value >)
■ Example:
sat
( model
( define - fun y () Int 0)
( define - fun x () Int ( - 3))
( define - fun z () Int 2)
)
10 / 10
