Jason Garrett-Glaser <ja...@x264.com> writes: > On Wed, Mar 7, 2012 at 2:13 PM, Ronald S. Bultje <rsbul...@gmail.com> wrote: >> Hi, >> >> 2012/3/7 Måns Rullgård <m...@mansr.com>: >>> "Ronald S. Bultje" <rsbul...@gmail.com> writes: >>>> On Wed, Mar 7, 2012 at 11:14 AM, Kostya Shishkov >>>> <kostya.shish...@gmail.com> wrote: >>>>> On Wed, Mar 07, 2012 at 10:52:33AM -0800, Ronald S. Bultje wrote: >>>>>> From: "Ronald S. Bultje" <rsbul...@gmail.com> >>>>>> >>>>>> This reverts commit ba36f14e5d34bcca6ecf77ba1c7512d1b51a9740. It >>>>>> broke decoding on x86-32 on some systems. >>>>>> --- >>>>>> libavcodec/aacsbr.c | 18 ++++++++++++------ >>>>> >>>>> Maybe it's better to fix alignment of those variables instead? >>>> >>>> x86-32 should support unaligned writes, so I'm not sure if that's the >>>> issue... >>> >>> The issue is that those macros use mmx instructions which don't get >>> along very well with x87 floating-point. >> >> Then I wonder why it works on x86-64? > > It uses 64-bit GPRs instead of MMX.
No, but x86_64 uses SSE for floating-point so there's no conflict. -- Måns Rullgård m...@mansr.com _______________________________________________ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel