Merge b_idx and edge variables, and optimize the ASM to directly load variables
authorRonald S. Bultje <rsbultje@gmail.com>
Wed, 29 Sep 2010 14:04:39 +0000 (14:04 +0000)
committerRonald S. Bultje <rsbultje@gmail.com>
Wed, 29 Sep 2010 14:04:39 +0000 (14:04 +0000)
commitcd17285e6cecefbeb286adbba881eb26dbabec29
tree6c89e8e8accadfd32a10a8d3466ff9efb01a0ff7
parent0cc8a5d0885396bfec7e02f1aea9e5eee386855a
Merge b_idx and edge variables, and optimize the ASM to directly load variables
from memory locations/offsets depending on b_idx plus constants, rather than
having gcc do this. This saves several lea calls and together saves about
10 cycles in h264_loop_filter_strength_mmx2().

Originally committed as revision 25256 to svn://svn.ffmpeg.org/ffmpeg/trunk
libavcodec/x86/h264dsp_mmx.c