This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Loop optimizer issues
- From: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>
- To: David Edelsohn <dje at watson dot ibm dot com>
- Cc: "S. Bosscher" <S dot Bosscher at student dot tudelft dot nl>,'Richard Henderson ' <rth at redhat dot com>,'Geoff Keating ' <geoffk at geoffk dot org>,'Diego Novillo ' <dnovillo at redhat dot com>,'Jason Merrill ' <jason at redhat dot com>, 'Jan Hubicka ' <jh at suse dot cz>,'Daniel Berlin ' <dberlin at dberlin dot org>,"'pop at gauvain dot u-strasbg dot fr '" <pop at gauvain dot u-strasbg dot fr>,"'gcc at gcc dot gnu dot org '" <gcc at gcc dot gnu dot org>
- Date: Thu, 3 Jul 2003 00:11:26 +0200
- Subject: Re: Loop optimizer issues
- References: <S.Bosscher@student.tudelft.nl> <4195D82C2DB1D211B9910008C7C9B06F01F37361@lr0nt3.lr.tudelft.nl> <200307022207.SAA31782@makai.watson.ibm.com>
Hello,
> Steven> Well, surely some loop optimizers will still be necessary on RTL even with
> Steven> tree-ssa, don't you think?
>
> Please read the full context of the thread:
>
> "but surely the code motion I'm concerned with will at least partially be
> done on tree-ssa (as loop invariant motion or pre or perhaps both)."
>
> "... ust do these high-level motions on ast-level when they are not
> represented in this ugly way (and either do only simpler pass on rtl, or
> don't do it there at all, depending on what will turn to be more
> practical)."
>
> Yes, RTL will have a simple loop optimizer, but not the
> optimizations for which Zdenek wants to assume the absence of LIBCALLs.
this is what I am not quite sure. Loop invariant motion is quite
dependent on some knowledge about register pressure, so it is not
unimaginable to me that it would mostly stay on rtl level (although
due to compile-time performance reasons it should be avoided).
Zdenek