2012/12/1 Jason Garrett-Glaser <ja...@x264.com>: > movu doesn't look like an aligned load to me...
Ah ah, no idea why I kept understanding unaligned. And indeed the address is aligned. So I ended up doing: mova m4, [s_mq + max_mq] movq m3, m4 unpcklps m3, m3 unpckhps m4, m4 which is *1* cycle faster for the whole function... The loading of q_filtq can be modified similarly, but it didn't help. -- Christophe _______________________________________________ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel