Set timeout on the fly, using command line interface about boolector

[mpreiner]

There is currently no such feature implemented in the SMT2 front-end. However, we can add this feature in the future via the terminate function callbacks in the SMT2 front-end.

[Joool]

Thank you very much for your quick response, much appreciated!

While you are at it, I'm also missing a feature to reset the solver, i.e., z3's (reset). My workflow is as follows:

• I spawn multiple sover processes in queue
• then alternating between them I send them queries
• then reset them and add them back to queue

However, I kind of want to avoid resetting them every time.

I hope I am not asking to much.

[mpreiner]

Can you provide a small SMT2 example for this workflow? That would be really helpful!

[Joool]

Sure:

(declare-const sub_1 (_ BitVec 256))
(declare-const sub_3 (_ BitVec 256))
(declare-const sub_2 (_ BitVec 256))
(assert (= sub_1 (_ bv1 256) ))
(assert (= sub_2 (_ bv0 256) ))
(assert (= sub_3 (ite (= sub_2 sub_1) (_ bv1 256) (_ bv0 256)) ))
(assert (= sub_3 (_ bv1 256) ))
(check-sat)
(reset)
(declare-const sub_1 (_ BitVec 256))
(declare-const sub_3 (_ BitVec 256))
(declare-const sub_2 (_ BitVec 256))
(assert (= sub_1 (_ bv0 256) ))
(assert (= sub_2 (_ bv0 256) ))
(assert (= sub_3 (ite (= sub_2 sub_1) (_ bv1 256) (_ bv0 256)) ))
(assert (= sub_3 (_ bv1 256) ))
(check-sat)

Maybe as context, I'm doing symbolic execution. However, in a multi threaded environment, i.e., I explore multiple paths simultaneously. I run multiple solver instances and tracking which solver has which symbols defined is too much of an overhead. Thus, I just cycle through the solvers, reseting them every time.
Note (reset), is defined by z3 not by SMT2.

[mpreiner]

Have you tried using push/pop instead of a reset? Or are the queries not related at all so that you always want to query a "fresh" instance of the solver?

Here would be the version with push/pop:

(push 1)
(declare-const sub_1 (_ BitVec 256))
(declare-const sub_3 (_ BitVec 256))
(declare-const sub_2 (_ BitVec 256))
(assert (= sub_1 (_ bv1 256) ))
(assert (= sub_2 (_ bv0 256) ))
(assert (= sub_3 (ite (= sub_2 sub_1) (_ bv1 256) (_ bv0 256)) ))
(assert (= sub_3 (_ bv1 256) ))
(check-sat)
(pop 1)
(push 1)
(declare-const sub_1 (_ BitVec 256))
(declare-const sub_3 (_ BitVec 256))
(declare-const sub_2 (_ BitVec 256))
(assert (= sub_1 (_ bv0 256) ))
(assert (= sub_2 (_ bv0 256) ))
(assert (= sub_3 (ite (= sub_2 sub_1) (_ bv1 256) (_ bv0 256)) ))
(assert (= sub_3 (_ bv1 256) ))
(check-sat)
(pop 1)

[Joool]

Ohh yeah forgot about that mode, I tried it with z3, but performance really suffers (https://stackoverflow.com/questions/26416814/why-does-z3-check-slow-when-it-is-immediately-preceded-by-z3-push). Thanks! I will give it a try with boolector.

[pointhi]

Any update in the meantime?

According to the help message there is the possibility of a time limit, but its an absolute timeout starting with the program execution:

  -t <seconds>,
    --time=<seconds>                  set time limit

It would be nice to have the ability to set a timeout which applies to commands like (check-sat), and thus is meaningful when using boolector for iterative solving. A global time limit does not make sense in those cases.

[aytey]

Hi @pointhi,

This is definitely not an official answer, but I had a need for something like this, which ended-up with this branch:

https://github.com/Boolector/boolector/compare/smtcomp19...andrewvaughanj:timeouts_for_term?expand=1

This adds:

• a -to <VALUE> argument to the main command-line application

• support for (set-option :timeout <VALUE>) in the SMT2 front-end

• support for self.btor.Set_opt(boolector.BTOR_OPT_TIMEOUT, <VALUE>) via the Python API

(all values are in millis).

Some important notes:

• this is based off of the smtcomp19 branch, which I am not sure is officially "ready for consumption" (so please be wary)

• this will only work (currently) with Lingeling; CaDiCaL does not adhere to the call-back mechanism (yet, there's another PR for that), and CMS/MiniSAT do not have a call-back mechanism at all

Let me know how you get on and, if it is successful, I can see about wrapping it up into an official PR.

Cheers,

Andrew