[vlc-devel] commit: Contribs: Use mtune=pentium2 for Windows (Jean-Baptiste Kempf )

[Rémi Denis-Courmont]

Le samedi 9 janvier 2010 11:26:38 Jean-Baptiste Kempf, vous avez écrit :
> On Sat, Jan 09, 2010 at 11:14:12AM +0200, Rémi Denis-Courmont wrote :
> > Le samedi 9 janvier 2010 01:25:27 git version control, vous avez écrit :
> > > vlc | branch: master | Jean-Baptiste Kempf <jb at videolan.org> | Sat Jan 
> > > 9 00:17:44 2010 +0100| [bcd6f556b7a41613074491553a765e5ed6510e0c] |
> > > committer: Jean-Baptiste Kempf
> > >
> > > Contribs: Use mtune=pentium2 for Windows
> >
> > That will schedule instructions as if running on a Pentium II, regardless
> > of the actual processor. I guess it is in fact more likely a P3 or P4. I
> > don't know the arcanes of IA-32, but are you sure this is really a good
> > idea?
> 
> No, it isn't a good idea.
> But well:
>  - default on gcc 4.4.2, the windows compiler we will use from now on, is
>    mtune=generic, which is i686, which is even lower than pentium2

'generic' is a compromise of the most common IA32 processors (whatever GCC 
developpers believe those are). 'i686' is the same as 'generic' as far as no 
386/486/586 are common anymore.

>  - we could try mtune=pentium3, but wouldn't that kill some stupid
>    (read: AMD) processors,

It's just as bad as pentium2 unless the instruction scheduling of Pentium 4 
and later processors is similar to that of Pentium 3.

>  - I believe that we could even --march=pentium2, but that would imply
>    --mtune=pentium2.

It's not the same. If we want to drop PPro support/enforce MMX, I would use
'-march=pentium2 -mtune=i686'.

-- 
Rémi Denis-Courmont
http://www.remlab.net/
http://fi.linkedin.com/in/remidenis