[isabelle-dev] Metis: Unused Theorem Reduces Run-Time?

[Jasmin Christian Blanchette]

Hi Tjark,

Am 18.01.2013 um 16:44 schrieb Tjark Weber:

> The new AFP entry on Kleene Algebras contains a metis call
> (http://afp.hg.sourceforge.net/hgweb/afp/afp/file/082b389cb3f8/thys/Kleene_Algebra/Kleene_Algebra.thy#l652)
> that generates a warning about an unused theorem local.opp_mult_def.
> Not passing opp_mult_def as an argument to metis gets rid of the
> warning, but increases run-time of this metis call from ~ 1 s to ~ 15 s
> on my machine.

Sure, these things happen with pretty much any tool that performs a heuristic search. There are even (rare) cases where taking out the unused theorems will lead to an unprovable first-order problem (because of incompletenesses in the translation of HO constructs).

What's your concrete suggestion here?

Cheers,

Jasmin