This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [please test] RTL sharing verifier
>
> >I was considering this too, but since the TODO way is way other
> >verifiers are hooked in, it seemed more consistent to do it that way.
> >We might wan to change gimple/flow verifier to be PROP controlled, that
> >is rahter easy and drop the TODO flags everywhere then - it don't seem
> >to be coolest idea to have some verifiers controlled one way and others
> >different...
> Well, verify_ssa is already PROP-driven.
>
> I think the reason for verify_flow_info to be TODO-driven is that for
> some passes you just know that it cannot perform a transformation that
> violates the verification (and could be much more fine grained if
> verify_flow_info was splitted in multiple parts).
Well, the same apply to sharing verifier too - the analysis passes are
harmless etc.... But I would not worry too much about the compilation
time implication. Except for SSA verifications, the whole verify_* stuff
seems to be resonably fast IMO.
Honza
>
> Paolo