PowerPC Floating Point compare bug?

[Geoff Keating]

Becky Gill <bgill@ibmoto.com> writes:

> I noticed today that the following code:
> 
>   static float f1 = 1.10000E+000;
>   static double d2;
>   static int a;
>   
>   d2 = (f1 == 2);
>   if (d2 != 0.0)
>     printf ("FAIL == %f\n", d2);
> 
> fails under the gcc 2.95.3 20010315 release (and on the gcc version 3.0
> 20010220 prerelease) whenever the -mcpu=750 option is specified, passing
> whenever no cpu option is given. 
> 
> Can someone take a look at this?  My entire compile line is included below.  
> 
> gcc -funroll-loops -funroll-all-loops -O3 -fno-inline-functions
> -fomit-frame-pointer -mcpu=750 -o test test.c 

Yup, it's a bug.  I think it's related to the generation of fsel.

-- 
- Geoffrey Keating <geoffk@geoffk.org>