ref: 06ca5744a1169ff2d106f601e3f618d813efe486
parent: 8b3985fde557ec7e3feeddd217827b52e33f00f4
author: Martin Storsjö <martin@martin.st>
date: Thu Oct 10 11:33:06 EDT 2019
arm64: looprestoration: Minimal scheduling improvements
--- a/src/arm/64/looprestoration.S
+++ b/src/arm/64/looprestoration.S
@@ -1688,14 +1688,14 @@
2:
subs x5, x5, #8
- ext v22.16b, v0.16b, v1.16b, #2 // -stride
- ext v23.16b, v2.16b, v3.16b, #2 // +stride
ext v24.16b, v0.16b, v1.16b, #4 // +1-stride
ext v25.16b, v2.16b, v3.16b, #4 // +1+stride
+ ext v22.16b, v0.16b, v1.16b, #2 // -stride
+ ext v23.16b, v2.16b, v3.16b, #2 // +stride
add v0.8h, v0.8h, v24.8h // -1-stride, +1-stride
add v25.8h, v2.8h, v25.8h // -1+stride, +1+stride
- add v0.8h, v0.8h, v25.8h
add v2.8h, v22.8h, v23.8h // -stride, +stride
+ add v0.8h, v0.8h, v25.8h
ext v22.16b, v16.16b, v17.16b, #4 // -stride
ext v23.16b, v17.16b, v18.16b, #4
@@ -1760,8 +1760,8 @@
4:
subs x5, x5, #8
- ext v22.16b, v0.16b, v1.16b, #2 // 0
ext v23.16b, v0.16b, v1.16b, #4 // +1
+ ext v22.16b, v0.16b, v1.16b, #2 // 0
add v0.8h, v0.8h, v23.8h // -1, +1
ext v24.16b, v16.16b, v17.16b, #4 // 0