Runtime for Octagon domain exceeds Polyhedral domain in programs with large number of variables

[DivyaRathore09]

For a program containing a large number of variables, let's say >240, I have observed that static analysis time under the octagon domain is higher than the polyhedral domain. Is it an issue with how octagon and polyhedral domains are implemented, or am I missing something

[antoinemine]

The octagon domain has a rather predicable polynomial cost, that does not depend on the complexity of the invariants.
The polyhedra domain uses a sparse representation, so, it can be very fast (faster than octagons) if you don't have many constraints in your system. This is not a very realistic setting, however, as in general programs, variables have bounds, which generate constraints. Another possible cause is the use of widening, that can require more iterations to converge if not configured well. Generally, the octagon domain is faster than the polyhedra one.

Could you please give details on the code you analyze and the analyzer you use?