[PATCH, i386]: Review FLOAT_MODE_P macro for i386 target
Uros Bizjak
ubizjak@gmail.com
Thu Apr 12 07:05:00 GMT 2007
On 4/12/07, Eric Christopher <echristo@apple.com> wrote:
> > Patch is bootstrapped on i686-pc-linux-gnu and regression tested for
> > all defaul languages. If there are no comments, I plan to commit this
> > patch to mainline in a couple of days.
> Anyhow, we also have a lot of !SSE_FLOAT_MODE_P (operands[1]) in some
> of these patterns which could also probably be removed. Future patch
> though. :)
No, these are needed to selectively disable pattern when SSE can
provide better alternative (in some cases). Usually, SSE_FLOAT_MODE is
coupled to TARGET_SSE_MATH. SSE_FLOAT_MODE_P has no effect on x86_64
(if you don't use -mfpmath=387 ;), but different x86_32 targets rely
on it.
Also, SSE can be SSE1 or SSE2 and SSE_FLOAT_MODE resolves this.
Uros.
More information about the Gcc-patches
mailing list