tree de7c5eef6fc33bdb4f74ff27218777869e2be3bb
parent c199fb47ed1d301461f05e3ed1b1050241a7b9c8
author Remya <remya.prakasan@ittiam.com> 1537799340 +0530
committer Yunqing Wang <yunqingwang@google.com> 1537816968 +0000

Optimize highbd 4x16 and 16x4 inv_txfm

Enabled sse4_1 optimizations for tx_sizes 4x16 and 16x4.

Module level gains:
Tx_size    Gain w.r.t. C
4x16         6.01x
16x4         7.36x

When tested for 20 frames of crowd_run_360p_10 at 1 mbps
for speed=1 preset, observed ~0.5% reduction in encoder time.

Change-Id: I0c5d6530c666150be0de062a7084c7a2bf61410f