tree 04d339eaa6322425f88df0506243c0e4216ce0ca
parent 7db820ef18b5b034b7850ee57718eca829be26e3
author Ravi Chaudhary <ravi.chaudhary@ittiam.com> 1521697376 +0530
committer Yunqing Wang <yunqingwang@google.com> 1522254611 +0000

SSE optimization of motion compensation functions

SSE4_1 optimizations of av1_highbd_jnt_convolve_2d_copy_c,
av1_highbd_jnt_convolve_y_c and av1_highbd_jnt_convolve_x_c
have been added.

av1_highbd_jnt_convolve_x_sse4_1: ~8.5 times faster than its
c implementation.

av1_highbd_jnt_convolve_y_sse4_1: ~9.4 times faster than its
c implementation.

av1_highbd_jnt_convolve_2d_copy_sse4_1: ~2.6 times faster than
its c implementation.

code cleanup for highbd_convolve_rounding_sse2 and
highbd_comp_avg_sse4_1.

Change-Id: I71290d1a04e0c0654c2008689ce85a2f1700685e