https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114087
Jeffrey A. Law <law at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Ever confirmed|0 |1
Last reconfirmed| |2024-12-09
--- Comment #2 from Jeffrey A. Law <law at gcc dot gnu.org> ---
True that there's a tradeoff here. But in the spot where we're most likely to
be able to do something with these cases the loop optimizer has already been
run and any profitable LICM has already been realized.
Unfortunately I don't see good ways to handle #2 and #3. The basic problem is
combine is limited in how many insns it processes at a time and it's not enough
to expose both constants in the equality comparison step.
Oliver has a change for the first case though. I'm still evaluating it.
