gcc33 creates a zImage which doesnt boot

[Olaf Hering]

 On Tue, Nov 11, Alan Modra wrote:

> On Tue, Nov 11, 2003 at 09:05:16AM +1030, Alan Modra wrote:
> >
> > On Mon, Nov 10, 2003 at 10:54:28PM +0100, Olaf Hering wrote:
> > > Good morning,
> > >
> > > the attached .config creates a zImage which boots on a p610 when
> > > compiled with SLES8 gcc3.2.2, but not with gcc 3.3.2.
> >
> > Hi Olaf,
> >
> > > CONFIG_64BIT=y
> >
> > This means you use -mcpu=power4 doesn't it?
>
> OK, so it doesn't mean that at all.  I was confusing this one with
> Anton's new option to choose power4 compilation.  However, if you've
> hacked the makefiles and are somehow getting -mpower4 being passed to
> gas, the following comment is still true.
>
> >  Probably a bad idea on
> > power3 processors.  Likely gcc-3.3 is generating some mtcrf insns with a
> > single cr field destination, and gas is then using the power4 form of
> > the instruction.
>
> Incidentally, I think gcc-3.3 is more likely to generate single field
> mtcrf insns than gcc-3.2.

It is exaxtly the same source, ameslab 2.5 from yesterday. I had to
tweak compiler.h for 3.2.2.
(people/akpm/patches/2.6/2.6.0-test9/2.6.0-test9-mm2/broken-out/ppc64-reloc_hide.patch)

I tried to use -mcpu=power3 instead of -mcpu=power4 with the gcc33
build, but that did not help. I will copy the 3.2 .o files into the 3.3
tree and see where it breaks.

--
USB is for mice, FireWire is for men!

sUse lINUX ag, nÜRNBERG

** Sent via the linuxppc64-dev mail list. See http://lists.linuxppc.org/