Re: 2.1.128 / current EGCS
Andi Kleen (ak@muc.de)
Thu, 19 Nov 1998 18:32:28 +0100
On Thu, Nov 19, 1998 at 05:23:32PM +0100, Mike Galbraith wrote:
> Hi,
>
> On 14 Nov 1998, Andi Kleen wrote:
>
> > In article <Pine.GSO.4.02A.9811132159510.2863-100000@jennifer.informatik.rwth-aachen.de>,
> > Bernd Schmidt <crux@pool.informatik.rwth-aachen.de> writes:
> > >>
> > >> There's still a problem (last time I had forgotten to change the compiler
> > >> in the arch/i386/Makefile... Guess I should stop trying stuff after being
> > >> at the computer for 18 hours continually ;) ):
> > >>
> > >> gcc -D__KERNEL__ -I/usr/src/linux-AC/include -Wall -Wstrict-prototypes -O2
> > >> -fomit-frame-pointer -pipe -s -fno-exceptions -pipe -s
> > >> -fomit-frame-pointer -mk6 -march=k6 -DCPU=686 -c -o checksum.o
> > >> checksum.c
> > >> checksum.c: In function `csum_partial':
> > >> checksum.c:200: Invalid `asm' statement:
> > >> checksum.c:200: fixed or forbidden register 2 (cx) was spilled for class
> > >> CREG.
> > >> checksum.c: In function `csum_partial_copy_generic':
> > >> checksum.c:538: Invalid `asm' statement:
> > >> checksum.c:538: fixed or forbidden register 4 (si) was spilled for class
> > >> SIREG.
> >
> > > That's a known problem; I didn't fix it yet in the patch I sent to Alan. The
> > > asms in checksum.c are rather hard to convert as they'd need too many
> > > operands if they were done properly. IMO the whole file should be replaced
> > > with a new file checksum.S which is entirely in assembler - the only C code
> > > in that file has a comment saying "old compatibility stuff, will be removed
> > > soon".
> >
> > I did the conversion work, unfortunately it is not tested because the
> > resulting kernel crashes during boot. Tabs might be messed up a bit.
> >
>
> Thanks Andi!
>
> I had to tinker a little, but she works now. (I also reformatted a bit,
> and can only hope I didn't screw up the ppro part.. can't test that)
[...] Sorry I should have told, but I already fixed these bugs and some more
in my development sources.
Also one new feature:
CONFIG_CPU_GENERIC: If selected the checksum function is selected at run time,
this should make generic kernel images (as shipped by RedHat/Caldera/SuSE) much
faster on PIIs.
-Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/