using atomic builtins

[Richard Guenther]

On Mon, 26 Jun 2006, Benjamin Kosnik wrote:

> 
> > > ... and if you search a bit the mailing lists, certainly we discussed a bit
> > > the usefulness of adding a builtin macro providing exactly that info. if we
> > > cannot change the default i?86 target, that would be definitely very usuful.
> > > At some point, I think Richard Guenther even posted a patch essentially
> > > complete, then I don't know what happened...
> >
> > I will re-surrect it (I don't remember what happened to it either) with
> > gcc-math work once we get to stage1 again.  
> 
> Great.
> 
> Thanks for the pointer Paolo, I have found the thread you are talking about:
> http://gcc.gnu.org/ml/libstdc++/2005-05/msg00277.html
> 
> Ugh. This gets considerably more complicated if -march=x changes
> things. We would need some way of detecting that (say) i486 code was
> trying to be linked with i386 code....

I think we can do runtime detection instead.  But we'll see.

> hmmm...
> 
> > First I want to do a BOF on the gcc-math subject in general.
> 
> A ha! Yes, well, I'm interested in this (separate) issue as well. I
> found the FSF decision most unfortunate.
> 
> If you do a BOF at the gcc summit, can you or somebody take notes and
> post minutes on the gcc list?

Sure.

Richard.

--
Richard Guenther <rguenther@suse.de>
Novell / SUSE Labs