[MPlayer-users] K 6-2 Compilation problem ..

[D Richard Felker III]

On Tue, Jun 01, 2004 at 11:26:56PM +1000, Daniel Kasak wrote:
> flanker wrote:
> 
> >I CAN'T succesfully compile mplayer 1.0xx and CVS on K6-2 CPU. 
> >This is short clip from log:
> >
> >"cc -c -I../libvo -I../../libvo -I/usr/X11/include -O4 -march=k6-2 -
> >mcpu=k6-2
> >-pipe -ffast-math -fomit-frame-pointer -D_REENTRANT -
> >D_LARGEFILE_SOURCE
> >-D_FILE_OFFSET_BITS=64  -Ilibmpdemux -Iloader -Ilibvo
> >-I/usr/include/freetype2    -I/usr/X11/include      -o subreader.o
> >subreader.c
> >{standard input}: Assembler messages:
> >{standard input}:5280: Error: value of 134 too large for field of 1 
> >bytes at
> >0000000000003919
> >make: *** [subreader.o] Error 1"
> > 
> >
> I'm pretty sure that's a compiler bug.
> I've have compile problems that look like that ( Assembler messages ... 
> ) when I try to compile some things for a k6-2 when doing Gentoo 
> installs - luckily none of them were *my* computer :)
> I got around it by switching my CFLAGS down from:
> 
> -march=k6-2 to:
> -march=i586
> 
> It probably wouldn'y hurt to also use:
> -mmmx -m3dnow
> as well. I've never bothered to investigate the problem ( partly because 
> I'm not sure whether anyone would be interested tracking down an obscure 
> k6-2 compilation bug now that the k6-2 is ... well ... not exactly in 
> it's prime any more ). But maybe you should report it to a gcc mailing list?

Another good option, if you have a K6-2, is use a compiler that
actually generates good code for K6-2. gcc generates notoriously slow
code, and takes a LOT longer to compile. You should "upgrade" to gcc
2.95.3 and all your problems will go away. :)

Rich