On Wed, May 16, 2012 at 08:09:15AM -0700, Ronald S. Bultje wrote: > On Wed, May 16, 2012 at 5:34 AM, Diego Biurrun <[email protected]> wrote: > > On Mon, May 14, 2012 at 12:41:45PM +0200, Diego Biurrun wrote: > >> From: Michael Kostylev <[email protected]> > >> > >> --- > >> libavcodec/x86/h264_intrapred.asm | 4 ++++ > >> 1 files changed, 4 insertions(+), 0 deletions(-) > > > > ping > > If cmov/mmx is indendent, then this is too. We should use the same > solution here as we used in the other thread. There, we marked it as > mmx2. Here, we should likely do the same thing, or just make sure > these functions don't get assigned if HAVE_CMOV is not set.
Now that the cmov issue is fixed I would like to ping this again. The solution for the cmov issue was not, in the end, marking all relevant functions as mmx2/mmxext. Doing so here would disable quite a few optimizations and leave many many old CPUs behind. This patch is quite localized and non-intrusive IMO, so I'm in favor of applying it as-is. Diego _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
