http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46507
Martin Jambor <jamborm at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jamborm at gcc dot gnu.org
--- Comment #6 from Martin Jambor <jamborm at gcc dot gnu.org> ---
(In reply to Marc Glisse from comment #4)
> Martin, do you have an opinion on the testcase in comment 3? We get:
>
> _2 = &t_1(D)->first;
> _4 = MEM[(const struct type *)t_1(D)]._vptr.A;
> _5 = *_4;
> OBJ_TYPE_REF(_5;_2->0) (_2); [tail call]
Assuming that first is a non-artificial (as in DECL_ARTIFICIAL) field,
we should be able to devirtualize this but as you have already
noticed, our current type-based middle-end intraprocedural
devirtualization machinery only works on automatic variables, but this
example shows another interesting situation. Thanks for the testcase,
I will try to make it work after I cleanup the the code to better
interoperate with Honza's recent devirtualization work.
