This is the mail archive of the
mailing list for the GCC project.
Re: Reload patch version 2
- To: Bernd Schmidt <crux at pool dot informatik dot rwth-aachen dot de>
- Subject: Re: Reload patch version 2
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Mon, 07 Sep 1998 19:07:10 -0600
- cc: egcs-patches at cygnus dot com
- Reply-To: law at cygnus dot com
In message <Pine.GSO.4.02A.firstname.lastname@example.org>you write:
> > The caller-saves issues are important to deal with, particularly for
> > floating point performance on sparcs. If you want a way to quantify
> > the lossage for poor caller-saves code, that's the target to look at.
> It seems like I'm going to be able to test on some Sparcs soon. Do you
> have any special test cases to suggest? Otherwise I'm just going to throw
> things like POVray or XaoS at it.
C code with a fair amount of floating point.
The X server itself was greatly improved by having caller-save.c generate
efficient code, so that may be a reasonable benchmark. x11perf was good
at exposing improvements/regressions in the X server itself. There
may be better server benchmarks for the X server these days.
> > Or just run a hard reg life analysis pass after reload. Having this
> > info available would deal with the reorg problem mentioned above and
> > simplify its code. It would also be useful for the sched2 instruction
> > splitter.
> Do you want me to send a patch for this? That should be easy to do.
Yea. I've been sitting on the fence on this for a long time. Having
accurate live/dead info after reload makes so many things simpler and
less error prone that we ought to just bite the bullet and do it.