[PATCH, i386]: Option to select default 80387 precision control

[Jakub Jelinek]

On Tue, Apr 03, 2007 at 09:44:22AM +0200, Uros Bizjak wrote:
> On 4/2/07, Richard Guenther <richard.guenther@gmail.com> wrote:
> 
> >This lacks documentation in invoke.texi.  Please mention its effect on
> >using double with -mpc32 and long double with -mpc64 (if there is any).
> 
> I have reworded the documentation and committed attached patch:

"may affect the results returned by some of the mathematical functions"

IMHO that should be worded more strongly, the effect on results from various
libm functions will be totally disastrous.

More importantly, why not limit these options to -m32 code only?
For -m64, they are useless, float and double is evaluated in SSE, only
long double uses i?87 unit, so any setting other than -mpc80 doesn't make
sense.

	Jakub