This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: For after egcs-2.0, when everything works and we get bored
- To: Toon Moene <toon at moene dot indiv dot nluug dot nl>
- Subject: Re: For after egcs-2.0, when everything works and we get bored
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Sat, 01 Aug 1998 23:40:42 -0600
- cc: Richard Henderson <rth at cygnus dot com>, egcs at cygnus dot com
- Reply-To: law at cygnus dot com
In message <9808011438.AA02114@moene.indiv.nluug.nl>you write:
> > I think this pretty much has to be done in a higher-level
> > intermediate language than our current RTL. Anything
> > else has too much hair.
>
> Hmm, I'm glad you reached the same conclusion as I did: It's
> simply too hard at this point in time (note the reference to
> egcs-2.0 in the subject :-)
Most of the interesting loop transformations need to happen at a
higher level than RTL.
Having entire functions as trees and optimizing on tree structures
before converting to RTL would be one approach to getting the
info we need to perform high level loop transformations.
This is (of course) a lot of work.
However, it occurs to me that if we do something like represent reducible
loops as a tree, optimize them, then hand them off for RTL conversion
then we would have a chance to implement some of cool loop transformations.
So the idea would be to try and build maximal trees to hand off to
rtl conversion, but when we encounter something "weird" we fall back
to the existing statement by statement conversion.
And (of course) the definition of "weird" changes as we extend our
tree mechanisms to handle more cases.
jeff