Re: A question about bounded model checking

From: Fabio Somenzi (Fabio_at_colorado.edu)
Date: Mon Apr 18 2005 - 14:01:27 MDT


>>>>> "JD" == Junkie Dolphin <junkie.dolphin_at_libero.it> writes:

 JD> Hello everyone,

 JD> when bounded_model_check returns "formula undecided", but zchaff
 JD> has died trying to SAT-solving some CNF (for example, because it
 JD> tried to malloc() too much memory, and so on), the undecidability
 JD> of the formula is simply due to the abortion of zchaff, or one
 JD> can say it is a "true" undecidability?

LTL properties are decidable for finite models. When BMC says
"undecided" it's because it could not come to a conclusion within the
allotted limits. These limits may be explicitly specified (via the -k
option) or may arise from memory or CPU limitations. In all cases,
including when the SAT solver runs out of memory, an LTL property can
be proved, in theory, by checking for sufficiently long paths.
(Again, we are talking about finite models.)

Fabio



This archive was generated by hypermail 2.1.7 : Mon Apr 18 2005 - 14:10:28 MDT