arm: Implement luma intra deblocking
authorMartin Storsjö <martin@martin.st>
Wed, 2 Sep 2015 19:39:51 +0000 (22:39 +0300)
committerHenrik Gramner <henrik@gramner.com>
Sun, 11 Oct 2015 16:44:54 +0000 (18:44 +0200)
commit654901dfca73a21e2bb2366dda79eb413e9bfb66
tree6e213c9b50a258229b15a360204823eac93117d1
parente2696a60a3e58d92e88e149b63c0b06a066eea9e
arm: Implement luma intra deblocking

checkasm timing       Cortex-A7      A8     A9
deblock_luma_intra[0]_c      5988    4653   4316
deblock_luma_intra[0]_neon   3103    2170   2128
deblock_luma_intra[1]_c      7119    5905   5347
deblock_luma_intra[1]_neon   2068    1381   1412

This includes extra optimizations by Janne Grunau.

Timings from a separate build, on Exynos 5422:

                      Cortex-A7     A15
deblock_luma_intra[0]_c      6627   3300
deblock_luma_intra[0]_neon   3059   1128
deblock_luma_intra[1]_c      7314   4128
deblock_luma_intra[1]_neon   2038   720
common/arm/deblock-a.S
common/deblock.c