projects / platform / upstream / libvpx.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: d40a860) | patch
Add AVX2 for convolve horizontal filter for block width 4
authorAnupam Pandey <anupam.pandey@ittiam.com>
Tue, 14 Mar 2023 11:20:31 +0000 (16:50 +0530)
committerAnupam Pandey <anupam.pandey@ittiam.com>
Thu, 16 Mar 2023 03:18:45 +0000 (08:48 +0530)
commit5c2cd048a05d8d06777cb9af5c0f4a261456023b
tree926db39d81dc9ab8d9a29ecd552e394566837f60tree | snapshot
parentd40a8608cceede09edec6e104b706a8823a7102ccommit | diff
Add AVX2 for convolve horizontal filter for block width 4

Introduced AVX2 intrinsic to compute convolve horizontal for
w = 4 case. This is a bit-exact change.

                 Instruction Count
cpu   Resolution   Reduction(%)
 0       LOWRES2      0.763
 0       MIDRES2      0.466
 0        HDRES2      0.317
 0       Average      0.516

Change-Id: I124f3f8e994c24461812f4963b113819466db44f
vpx_dsp/x86/vpx_subpixel_8t_intrin_avx2.c diff | blob | history
Domain: Multimedia / Codec;