Re: PATCH: Support MIPS II

["H . J . Lu" <hjl at lucon dot org>]

On Wed, May 08, 2002 at 07:08:54AM -0700, Benjamin Kosnik wrote:
> 
> > I have been using this patch for a while.
> 
> ... it was rejected the first time, and will be rejected this time too,
> as it's the exact same patch.
> 
> Since it's the style a la mode, I'll also post what I posted six months
> ago. 
> 
> You'll need to work with Eric Christo or whoever the MIPS maintainer is
> to clean up the MIPS configure files. Instead of including one file for
> all MIPS targets, and then inside of atomicity.h switching if atomic ops
> are supported, I'd rather the mips targets that can hack atomic ops
> figure themselves out, so that configure.target has:
> 
> +   mips-that-can-hack-it*)
> +     cpu_include_dir=config/cpu/mips
> 
> else, default to generic. 
> 
> Any clearer the second time around?

Have you even looked at my new patch? It has

#ifndef _BITS_ATOMICITY_H

#ifdef _MIPS_ISA

#include <sgidefs.h>

#if _MIPS_ISA >= _MIPS_ISA_MIPS2

#define _BITS_ATOMICITY_H 1

asm for MIPS 2.

#endif /* _MIPS_ISA >= _MIPS_ISA_MIPS2 */

#endif /* !_MIPS_ISA */

#ifndef _BITS_ATOMICITY_H

#define _BITS_ATOMICITY_H 1

generic version

#endif

#endif

The idea is any mips targets which define _MIPS_ISA in CPP have <sgidefs.h>.
My new patch checks _MIPS_ISA before including <sgidefs.h>. One can add the
asm support to a mips target by adding <sgidefs.h> and defining _MIPS_ISA in
CPP. You can't do it at the gcc configure time since it is a runtime option
for gcc. Now, please tell me why it won't work for all mips targets?

BTW, Richard suggested adding a warning for generic version. But it causes so
many "make check" failures so that I had to remove the warning.


H.J.