https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66790
--- Comment #43 from Pierre-Marie de Rodat <derodat at adacore dot com> --- (In reply to Bernd Schmidt from comment #42) > I don't think that would actually help. Even if something is an actual > incoming argument register, it may still be uninitialized by the caller. Sure, but my understanding is that what matters in DF is what happens inside the function, so we can assume arguments are always initialized. Actually this makes sense in the context of REE as well: we have one (artificial) def for arguments so since this def has no corresponding instruction, REE will not be able to remove z-extensions reaching this def. That being said, I failed to get the “is it really used for incoming arguments” hard-reg predicate, so falling back to your proposal: I will followup on gcc-patches@. Thanks! :-)
