Re: optimization/9736: same fp comparison can lead to different results

[Tim Prince <timothyprince@sbcglobal.net>]

The following reply was made to PR optimization/9736; it has been noted by GNATS.

From: Tim Prince <timothyprince@sbcglobal.net>
To: Eric Botcazou <ebotcazou@libertysurf.fr>,
   Richard Addison-Wood <richard@wetafx.co.nz>
Cc: gcc-bugs@gcc.gnu.org, nobody@gcc.gnu.org, gcc-gnats@gcc.gnu.org
Subject: Re: optimization/9736: same fp comparison can lead to different results
Date: Wed, 19 Feb 2003 05:56:06 -0800

 On Wednesday 19 February 2003 01:02, Eric Botcazou wrote:
 > > If the compiler produces valid assembly code that does not
 > > correctly execute the input source code, that is a compiler bug.
 >
 > Sure. But why do you keep using a compilation mode that you know will
 > generate a faulty executable for your program? I can't say much more than
 > the GCC manual here: if your program relies on exact IEEE floating-point
 > semantics to properly work, compile it with -ffloat-store. Otherwise
 > eliminate the dependencies on the IEEE format.
 You might consider the options -march=pentium[34] -mfpmath=sse, or changing 
 the precision mode you run in, since you seem to want an alternative to 
 -ffloat-store to achieve those effects.
 -- 
 Tim Prince