This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix PR optimization/10157
- From: Eric Botcazou <ebotcazou at libertysurf dot fr>
- To: Dale Johannesen <dalej at apple dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Wed, 2 Apr 2003 23:38:08 +0200
- Subject: Re: [PATCH] Fix PR optimization/10157
- References: <56B0A5B8-654F-11D7-9A06-000393D76DAA@apple.com>
> Interesting interpretation. I'd say if it's not there at all it's
> as incomplete as you can get.
Of course. However, in French at least, "incomplet" clearly implies that
there is something; otherwise "inexistant" would be better suited. But I
certainly can't argue about the English semantics.
> > /* On the D30V, copying to/from CCmode is complete, but since there
> > are only
> > two CC registers usable for conditional tests, this helps gcse not
> > compound the reload problem. */
> > #define AVOID_CCMODE_COPIES
>
> This is clearly not doing what the doc says to do, although there seems
> to be a good reason for it.
Well, one could argue that the support is incomplete in that it doesn't
support the whole set of transformations (in that case, GCSE) of the
optimizer.
But let's get back to code: don't you think that the body of
compute_can_copy() indicates that AVOID_CCMODE_COPIES is a mean for ports
with the "movcc" pattern to prevent unwanted CCmode moves from being
generated?
--
Eric Botcazou