On Thu, 2003-09-11 at 22:14, Guido Guenther wrote: > On Thu, Sep 11, 2003 at 08:04:01PM +0200, Paul van Tilburg wrote: > > Thanks! I had trouble building mplayer 1.0pre1 with gcc-3.3 (got gcc > > internal > > error). You had no trouble with the AltiVec part of the build > > process? > I che ated, I hacked the Makefile of the libavcodec/ subdir to use a cvs > gcc and even that one ICEs if you pass '-g', so I had to disable that > too: > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11983 > > > As I understand when reading http://developer.apple.com/hardware/ve/ > > it's G4-only, so -g4 would be a better suffix IMO. > > Like Marillat's packages have a i386 (ppc/powerpc suffix here) version for > > ALL x86-based archs and i686,k7,k6 specific versions (g4 suffix) for CPU > > optimized mplayer/mencoder packages. > Well, I'm not very familiar with the PowerPC family of processors yet, > will the g5 have a different/improved altivec instruction set? Is it > still called altivec then? In this case -g4 would make more sense. I'm > also still unsure what will be the best -mcpu and -march flags for > G4s (although I doubt it will make much of a difference).
I used to use: -mcpu=7450 -mtune=7450 -maltivec -mabi=altivec but as already noted, gcc-3.3 segfaults on the libavcodec subdir Soeren.
