GCC 3.2.1 Optimizations and Via C3 Ezra CPU

Ken Moffat ken at kenmoffat.uklinux.net
Sun Feb 15 04:53:59 PST 2004

On Sat, 14 Feb 2004, Joel Miller wrote:

> On Sat, 14 Feb 2004 21:04:36 -0600, Bryan K. Walton <zxcvbnm at tds.net>
> wrote:
> > Hi,
> > 	I have a question regarding compiler optimizations for the Via
> > C3 Ezra processor. I did a little searching around on the web and found
> > this:
> >
> > CHOST="i586-pc-linux-gnu"
> > CFLAGS="-march=i586 -m3dnow -O3 -pipe -fomit-frame-pointer"
> > CXXFLAGS="-march=i586 -m3dnow -O3 -pipe -fomit-frame-pointer"
> >
> > Now, that makes sense to me.  However, when I issue a "uname -m" on
> > this box, it reports that it is a i686.  And sure enough, when I
> > compile gcc on this box, without any optimizations, it seems to be
> > identifying the host as i686-pc-linux-gnu.  Any ideas what might be
> > going on here.  Is gcc incorrectly identifying hardware on the box?
> > I'm using GCC 3.2.1.  Or can I specify i686 in those three
> > optimization lines?
> >
> > Thanks!!
> > Bryan Walton
> IIRC the C3 processor is a 686 processor but it lacks the CMov instruction
> that other 686 processors have. Thus default i686 instructions will call
> CMOV and you will get an error stating an illegal instruction has been
> called. I also seem to remember hearing that newer C3's didnt have this
> problem. Take what I say with a grain of salt, as I'm trying to recall all
> this from memory.

 Also from memory, there were postings on lkml suggesting it works
faster on i486 code (maybe only really significant in kernel space), and
I think I saw somewhere that gcc-3.3.2 will correctly identify these

Brighton tops UK Jedi league

More information about the lfs-support mailing list