Newest loader from CVS not working
David O'Brien
obrien at freebsd.org
Tue May 24 07:44:50 PDT 2005
On Sun, May 22, 2005 at 08:47:36PM -0600, Scott Long wrote:
> David O'Brien wrote:
> >On Sun, May 22, 2005 at 04:21:12PM -0600, Scott Long wrote:
> >>>What I fixed was an amd64 build problem. The thread starter here was
> >>>talking about pentium-m builds, so I assume its i386 in this case.
> >>Yes, the threads jumped back and forth between people experiencing
> >>problems with non-default CFLAGS <..snip..>
> >
> >
> >I've heard those problems on and off for a year now - with no one
> >experiencing the problem spending sufficient effort to provide a decent
> >analysis of the issue.
>
> Re-read the threads. There is a lot of good analysis on how gcc was
> emitting SSE instructions.
I don't see how that could be the case:
RCS file: /home/ncvs/src/sys/boot/i386/Makefile.inc,v
..
----------------------------
revision 1.10
date: 2005/03/15 18:43:36; author: obrien; state: Exp; lines: +2 -1
Ensure GCC does not use FP registers in integer code.
I think all we really need is -fno-sse2.
I really don't like cluttering up the compiler invocation,
but this bigger hammer will fix reported problems for now.
----------------------------
diff -u -u -0 -r1.9 -r1.10
--- Makefile.inc 9 Feb 2004 14:11:55 -0000 1.9
+++ Makefile.inc 15 Mar 2005 18:43:36 -0000 1.10
@@ -8 +8,2 @@
-CFLAGS+= -ffreestanding -mpreferred-stack-boundary=2
+CFLAGS+= -ffreestanding -mpreferred-stack-boundary=2 \
+ -mno-mmx -mno-3dnow -mno-sse -mno-sse2
An _analysis_ would be someone experiencing a problem clearly showing
SSE[2] instructions in their .s files. Along with figuring out why the
above flags didn't handle the issue. Also telling the results of trying
CPUTYPE={<empty>,i486,pentium,pentium-pro,pentium3m,pentium4,pentium4m}
(or <empty>,i486,pentium,k6,k6-2,athlon,athlon-xp).
--
-- David (obrien at FreeBSD.org)
More information about the freebsd-current
mailing list