tree d9c93c9e31c320425fd419ab1af2c784a61bc1f2
parent 699fa47ee8a7e06c20d4da268395f22f62f93069
author Jonathan Wright <jonathan.wright@arm.com> 1657529668 +0100
committer James Zern <jzern@google.com> 1662507837 +0000

Implement speciality variance functions using UDOT instruction

Accelerate the following Neon speciality variance (helper) functions
by calculating the sum of squares using a single Armv8.4-A UDOT
instruction instead of four MLAs:
 * mse8xh_neon
 * mse16xh_neon
 * aom_get4x4sse_cs_neon

The previous implementation is retained for use on systems that do
not support the Armv8.4-A dot-product instructions.

Change-Id: I20b083dc36d7f7bb5550ad82b384d7b96667e5ea