Re: x86/MMX/SSE shift by immediate

[Stuart Hastings <stuart at apple dot com>]

On Jan 14, 2004, at 5:12 PM, Richard Henderson wrote:

> On Wed, Jan 14, 2004 at 05:03:36PM -0800, Stuart Hastings wrote:
> > Does this look plausible (IANA x86 expert ;-) ?
> >
> > { target pentium-mmx-*-* pentium2-*-* winchip*-*-* c3*-*-* athlon*-*-*
> > k[68]*-*-* x86_64-*-* } for MMX
> > { target pentium3-*-* k[68]*-*-* x86_64-*-* } for SSE
> > { target pentium4-*-* k8*-*-* x86_64-*-* } for SSE2
>
> No.  Most folks just type configure, which won't come up with all
> these exotic names.  Indeed, config.guess is going to choose i686
> (not listed) virtually all of the time with hardware produced in
> the last 4 years.
>
> If we're generating cmov just because -msse is set, then that is a bug.

O.K.

> We shouldn't be generating those insns without the appropriate -march 
> argument.

This fragment from i386.h seems to explain it:

/* For sane SSE instruction set generation we need fcomi instruction.  
It is
   safe to enable all CMOVE instructions.  */
#define TARGET_CMOVE ((x86_cmove & (1 << ix86_arch)) || TARGET_SSE)

The patch apparently came from here:

	http://gcc.gnu.org/ml/gcc-patches/2001-02/msg00719.html

stuart hastings
Apple Computer