[x265] [arm64] improve luma_hps

[Pop, Sebastian]

Hi,
the attached patch improves the existing implementation of luma_hps as follows:

before the patch:
luma_hps[  4x4]         7.26x    155.66          1130.24
luma_hps[  8x8]         13.19x   229.34          3025.77
luma_hps[  8x4]         12.82x   170.88          2190.91
luma_hps[  4x8]         7.39x    209.01          1544.11
luma_hps[ 8x16]         13.33x   346.17          4614.47
luma_hps[12x16]         10.92x   628.93          6869.03
luma_hps[ 4x16]         7.42x    316.50          2348.13
luma_hps[24x32]         14.09x   1627.97         22935.64
luma_hps[ 8x32]         13.48x   579.82          7815.91

with the patch:
luma_hps[  4x4]         7.25x    154.91          1123.54
luma_hps[  8x8]         13.25x   229.03          3033.61
luma_hps[16x16]         13.90x   655.11          9105.27
luma_hps[32x32]         14.20x   2154.92         30597.96
luma_hps[64x64]         14.57x   7626.98         111121.77
luma_hps[  8x4]         12.98x   170.83          2218.02
luma_hps[  4x8]         7.38x    209.01          1543.31
luma_hps[ 16x8]         13.91x   427.22          5942.64
luma_hps[ 8x16]         13.31x   346.14          4607.25
luma_hps[32x16]         14.29x   1263.49         18052.49
luma_hps[16x32]         14.08x   1097.46         15447.23
luma_hps[64x32]         14.33x   4257.39         60995.71
luma_hps[32x64]         14.38x   3860.87         55516.82
luma_hps[16x12]         14.15x   533.02          7541.39
luma_hps[12x16]         10.92x   628.27          6861.97
luma_hps[ 16x4]         13.94x   312.73          4360.99
luma_hps[ 4x16]         7.44x    316.14          2352.54
luma_hps[32x24]         14.37x   1695.62         24370.08
luma_hps[24x32]         14.08x   1628.42         22924.21
luma_hps[ 32x8]         14.27x   821.42          11721.52
luma_hps[ 8x32]         13.49x   579.23          7815.46
luma_hps[64x48]         14.60x   5892.16         86022.95
luma_hps[48x64]         14.59x   5724.71         83546.84
luma_hps[64x16]         14.50x   2484.56         36024.81
luma_hps[16x64]         14.15x   1979.60         28009.11

Ok to commit?

Thanks,
Sebastian

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20210710/f61857d9/attachment.html>