https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88626
--- Comment #6 from Jan Hubicka <hubicka at ucw dot cz> --- > Maybe a heuristic that gives a "bonus" for inlining a function that contains > __builtin_constant_p (only if the argument looks like it depends on the > function parameters?) would be easier? I don't know if we have something like > that yet. Yes, we have inline bonuses (called inline_hint) and I guess adding bonus for builtin_constant_p calls which depends on function parameter should be doable and indeed makes sense to me. Still for serious builtin_constant_p performance I think we need to do something about the double accounting problem I explained above that is bit harder to fix (the time/size vector themselves do not have any if-then-else semantics and predicates have no NOT operation, but merging the outcomes is possible, just bit hard to implement) Honza
