This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: What to do with new-ra for GCC 4.0
- From: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)
- To: dberlin at dberlin dot org
- Cc: gcc at gcc dot gnu dot org
- Date: Thu, 6 Jan 05 21:52:22 EST
- Subject: Re: What to do with new-ra for GCC 4.0
Nobody wants to go near reload or try an rtl with constraints better
amenable to register allocation with a 10 foot pole because they are
afraid of breaking every non-very-actively maintained port in
existence, when this is simply something that is likely to have to
happen to replace or significantly rewrite reload[1].
I disagree that one needs to do significant work to reload in order to get
better register allocation. As was pointed out a few times, if the
register allocator did all constraint matching and spll operations,
reload could remain unchanged, but do absolutely nothing. This is
an incremental path: as an improved ra did more and more spilling, reload
would do less and less.