Re: [tree-ssa] life analysis improvement

[Andrew Pinski <pinskia at physics dot uc dot edu>]

But that is because of the x86 floating point registers being 80bits 
instead of 64 bit.
It passes on ppc still.

From http://www.suse.de/~aj/SPEC/:

On Linux/ix86 we use long double precision for evaluation of floating 
point variables.  This can lead to different values than the expected 
values.  I'm experimenting on the tests for cfg-branch with a small 
shared library that is preloaded.

The complete code is below, I compiled it with  gcc -Wall -O2 -shared 
-o libchange-precission.so change-precission.c and set LD_PRELOAD to 
this library.
extern void change_precission (void) __attribute__ ((constructor));

void
change_precission (void)
{
  unsigned int cw = 0x27f;

  __asm__ volatile ("fldcw %0" :: "m" (*&cw));
}

Thanks,
Andrew Pinski

On Wednesday, Apr 23, 2003, at 15:33 US/Eastern, Diego Novillo wrote:

> On Wed, Apr 23, 2003 at 12:53:09PM -0600, law at redhat dot com wrote:
>
> > ps.  It looks like my C++ changes to enable the tree-ssa optimizers
> > may be causing eon to fail to compile from spec2000.  I'll be taking
> > a look at that shortly.
> If it's any consolation, not even mainline can handle eon.  The
> closest we got, was compile it but fail at runtime :)
>
>
> Diego.