Re: x86/MMX/SSE shift by immediate

[Stuart Hastings <stuart at apple dot com>]

On Jan 7, 2004, at 2:45 PM, Richard Henderson wrote:

> On Wed, Jan 07, 2004 at 02:36:13PM -0800, Stuart Hastings wrote:
> > Yes, both of the CPUID instructions here kill %eax, %ebx, %ecx, and
> > %edx.  However, if I admit as much in the __asm__, and I compile with
> > -fPIC (default on Darwin/x86), GCC stops here with the diagnostic
> >
> > 	/Volumes/sandbox/stuart/gcc.fsf/gcc/testsuite/gcc.dg/i386-cpuid.h:39:
> > error: PIC register `ebx' clobbered in `asm'
> >
> > Ergo, I chose to lie to GCC about the CPUID instruction.  :-(  I 
> > really
> > dislike doing this; sleazy tricks like this usually cause maintenance
> > headaches later on (e.g. if GCC gets smarter about the PIC register, 
> > we
> > could and should describe CPUID correctly here).  Do you know a better
> > way?
>
> 	__asm__ ("push %ebx; cpuid; pop %ebx"
> 		 : "=a" (fl1)
> 		 : "0" (0)
> 		 : "ecx" "edx", cc");

Right.  Wish I'd thought of that...

> Ok with that fix.

Thank-you.

> Which also removes the need for the nowhere call too, right?

Right again.

> r~

I've revised the CPUID stuff per above, and added some more NOINLINEs 
in i386-sse-5.c.  Everything still works on Darwin/x86(P4, -fPIC) and 
Linux/x86(P3, -fno-PIC).  If I hear no objections, I'll commit this 
later today.

stuart hastings
Apple Computer

Attachment: gcc.fsf.diffs.txt
Description: Text document

Attachment: i386-cpuid.h
Description: Text document

Attachment: i386-sse-5.c
Description: Text document