tree 6009ab087899554718c2d46deaccb9d0e9dd527a
parent 40496f7d7bdc8cd58c09ce906babedeb592887d5
author Aniket Dhok <aniket.dhok@ittiam.com> 1540210912 +0530
committer Yunqing Wang <yunqingwang@google.com> 1540224622 +0000

Optimize highbd fwd_txfm modules

Added sse4_1 variant for identity tx_type of 32.

Enabled sse4_1 optimizations for tx_sizes 32x32 , 32x16 ,
16x32 , 32x8 , 8x32.

Module level gains:
Tx_size    Gain w.r.t. C
32x32      1.75x
32x16      2.68x
16x32      2.38x
32x8       2.88x
8x32       2.72x

When tested for 20 frames of crowd_run_360p_10 at 1 mbps
for speed=1 preset, observed ~0.12% reduction in encoder time.

Change-Id: I833c08ae30e84432b5258e1bf2c10307f3379d72