projects / platform / upstream / libvpx.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: c1c7dd3) | patch
Avoid vshr and vget_{low,high} in Neon d135 predictor impl
authorGeorge Steed <george.steed@arm.com>
Tue, 28 Mar 2023 14:49:37 +0000 (14:49 +0000)
committerGeorge Steed <george.steed@arm.com>
Thu, 30 Mar 2023 09:00:26 +0000 (09:00 +0000)
commita257b4d6be525c50aea1e9f33f791fd4b627e92b
treecf33b7291f6f335d2fa1ea20d7d8f1deeb14e38btree | snapshot
parentc1c7dd3138796c512c441fba901d1a48ef7d61d1commit | diff
Avoid vshr and vget_{low,high} in Neon d135 predictor impl

The shift instructions have marginally worse performance on some
micro-architectures, and the vget_{low,high} instructions are
unnecessary.

This commit improves performance of the d135 predictors by 1.5% geomean
averaged across a range of compilers and micro-architectures.

Change-Id: Ied4c3eecc12fc973841696459d868ce403ed4e6c
vpx_dsp/arm/intrapred_neon.c diff | blob | history
Domain: Multimedia / Codec;