This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: ping Dale's patches
- From: Roger Sayle <roger at eyesopen dot com>
- To: Andrew Pinski <pinskia at physics dot uc dot edu>
- Cc: Dale Johannesen <dalej at apple dot com>, <gcc-patches at gcc dot gnu dot org>
- Date: Sun, 30 May 2004 11:43:23 -0600 (MDT)
- Subject: Re: ping Dale's patches
On Sun, 30 May 2004, Andrew Pinski wrote:
> http://gcc.gnu.org/ml/gcc-patches/2004-04/msg01944.html
> Patch: RTL loop opt bug
This is OK for mainline. Clearly, it would be nice to also
apply this transformation to rotated loops, but obviously
we can't do that yet and this patch fixes the resulting
miscompilation.
> http://gcc.gnu.org/ml/gcc-patches/2004-04/msg01946.html
> Patch: disable invalid REG_EQUAL note
I think its best to explore the "very dusty twisty little
passages of darwin-specific address generation". I was
going to approve this patch (with Andrews suggested change
to use "== const0_rtx") when I realized that GCC's RTL
canonicalization rules mean that the RTX "(MINUS X CONST_INT)"
should never match! So Darwin (or the RTL optimizers) should
be generating addresses as (PLUS X -C) instead of (MINUS X C),
in addition to the fact that we should obviously never see
either (MINUS X 0) or (PLUS X 0).
Do you have a test case, so that others can help investigate?
Roger
--