projects / platform / upstream / libvpx.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 6d0e5e5) | patch
Optimize Neon paths of high bitdepth SAD and SAD4d for 8xh blocks
authorSalome Thirot <salome.thirot@arm.com>
Mon, 27 Mar 2023 13:31:40 +0000 (14:31 +0100)
committerSalome Thirot <salome.thirot@arm.com>
Wed, 29 Mar 2023 15:50:34 +0000 (16:50 +0100)
commitcf1efecebf0ed2e01bafea6804c98f80ab7e12e0
treec51ad537cc018d08d519ff8aa13c37a18a1661b9tree | snapshot
parent6d0e5e56aede37f1047ba795c02acd886c89fe11commit | diff
Optimize Neon paths of high bitdepth SAD and SAD4d for 8xh blocks

For these block sizes there is no need to widen to 32-bits until the
final reduction, so use a single vabaq instead of vabd + vpadalq.

Change-Id: I9c19d620f7bb8b3a6b0bedd37789c03bb628b563
vpx_dsp/arm/highbd_sad4d_neon.c diff | blob | history
vpx_dsp/arm/highbd_sad_neon.c diff | blob | history
Domain: Multimedia / Codec;