This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: alternative to -ffloat-store on x86?
- To: Ted dot Belding at umich dot edu, egcs at cygnus dot com
- Subject: Re: alternative to -ffloat-store on x86?
- From: N8TM at aol dot com
- Date: Tue, 24 Nov 1998 09:09:01 EST
In a message dated 11/24/98 5:14:32 AM Pacific Standard Time,
Ted.Belding@umich.edu writes:
> I need to force strict IEEE floating-point compliance in a program that I'm
> compiling with egcs 1.1 for x86, so that floating-point results are exactly
> the same as on other platforms. Is -ffloat-store still the best way to do
> this, or is there a better alternative?
If you wish to do this with reasonable execution speed, on p6-style CPU's you
may reset the rounding mode. If you don't do mixed precision, this should be
a viable solution. division and sqrt() will be noticeably faster than the
extended mode.