tree 8076dc06e7a2285211cd56fdc38ce3163bf55c1c
parent 4e95d5f17e6068ad4999dda3eb193142839bebe6
author Konstantinos Margaritis <konma@vectorcamp.gr> 1684493857 +0000
committer James Zern <jzern@google.com> 1684949429 +0000

[NEON] Optimize highbd weighted convolution functions

These were submitted in one patchset as they are similar.
Differences in CPU profiling against scalar:

Scalar:
0.71%  aomenc  av1_highbd_dist_wtd_convolve_y_c
0.37%  aomenc  av1_highbd_dist_wtd_convolve_2d_c
0.19%  aomenc  av1_highbd_dist_wtd_convolve_2d_copy_c
0.12%  aomenc  av1_highbd_dist_wtd_convolve_x_c

NEON:
0.14%  aomenc  av1_highbd_dist_wtd_convolve_y_neon
0.07%  aomenc  av1_highbd_dist_wtd_convolve_2d_copy_neon
0.04%  aomenc  av1_highbd_dist_wtd_convolve_2d_neon
0.04%  aomenc  av1_highbd_dist_wtd_convolve_x_neon

Change-Id: I653dfb07eb399e8cf9e72b80357267c204e828ba