Date: Thu, 02 Jun 2005 22:24:49 +0200 From: Pawel Worach <pawel.worach@gmail.com> To: obrien@freebsd.org Cc: freebsd-current@freebsd.org, David Gurvich <david.freebsd@verizon.net> Subject: Re: Newest loader from CVS not working Message-ID: <429F6B11.2070407@gmail.com> In-Reply-To: <20050531181121.GA85431@dragon.NUXI.org> References: <20050518051111.GA33262@Athena.infor.org> <20050520164349.GD6982@dragon.NUXI.org> <428E1815.8080500@samsco.org> <200505221453.44007.peter@wemm.org> <429105D8.6000106@samsco.org> <20050523021527.GA62693@dragon.NUXI.org> <42914446.4000203@gmail.com> <42992034.3080403@gmail.com> <20050531181121.GA85431@dragon.NUXI.org>
next in thread | previous in thread | raw e-mail | index | archive | help
David O'Brien wrote: > On Sun, May 29, 2005 at 03:51:48AM +0200, Pawel Worach wrote: > >>I did some more investigation into this and found that MMX registers are >>used in the FICL code in the pentium-m case. Here is a full diff of >>sys/boot built with CPUTYPE?=pentium2 and pentium-m. > > > This is great! > I know what needs to be done now. > Hi David, I found two more problems, after beeing really puzzled why the p-m loader works fine after a build where the installed world was built with p2 and the next p-m build breaks the loader when installed world was built with p-m. It seems like libstand uses mmx registers and the sys/boot build uses the installed version of libstand instead of the one in obj/. libstand p2 vs. p-m save-temps diff: http://62.119.221.30/boot/libstand.diff wfm fix: --- lib/libstand/Makefile.orig Thu Jun 2 22:13:31 2005 +++ lib/libstand/Makefile Thu Jun 2 22:13:24 2005 @@ -20,6 +20,7 @@ .endif .if ${MACHINE_ARCH} == "i386" CFLAGS+= -mpreferred-stack-boundary=2 +CFLAGS+= -mno-mmx -mno-3dnow -mno-sse -mno-sse2 .endif .if ${MACHINE_ARCH} == "powerpc" CFLAGS+= -msoft-float -D_STANDALONE -- Pawel
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?429F6B11.2070407>