This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Reload patch v4
- To: Bernd Schmidt <crux at pool dot informatik dot rwth-aachen dot de>
- Subject: Re: Reload patch v4
- From: Toon Moene <toon at moene dot indiv dot nluug dot nl>
- Date: Wed, 2 Sep 98 08:21:36 +0200
- Cc: egcs-patches at cygnus dot com
- Organization: Moene Computational Physics, Maartensdijk, The Netherlands
- References: <Pine.GSO.4.02A.9808311145340.24291-100000@colt.informatik.rwth-aachen.de>
> Here's an updated version of my reload patch.
>
> Fixes since version 3:
> - the problems with caller-save aborting should be fixed.
> - some attempts to speed reload up a bit.
> - more large functions broken up into manageable pieces.
> - deleted the hack in i386.h to prevent a compiler abort while
> compiling libgcc2.a, see my previous patch today for a
> suggestion on how to fix this properly.
I bootstrapped succesfully on my m68k-next-nextstep3 with this
patch; running c-torture-1.45 shows no regression w.r.t. to previous
egcs-1.1 prerelease snapshots.
There are three reasons why I think it's useful for _me_ to test
this patch (apart from the fact that I think it's a sound way of
improving reload):
1. The m68k, like the ix86, has few registers, so there will be an
impact.
2. The m68k, unlike the ix86, has postincrement and predecrement
addressing modes - this has to be tested also (see the REG_INC
parts of the patch).
3. Using the Fortran Frontend guarantees different "forcing" of
the backend and the reload pass.
Cheers,
Toon.