ref: ea7e13e77efa4fc38597cc981cc755685143fb26
parent: 13fad75d96942a7c895d8f0f5e1577edbf7c0df4
author: Martin Storsjö <martin@martin.st>
date: Tue Sep 1 08:05:58 EDT 2020
arm32: mc: Use narrower vext.8 in 8tap_w4_h The previous form was a leftover from how it had to be written on aarch64.
--- a/src/arm/32/mc.S
+++ b/src/arm/32/mc.S
@@ -1403,12 +1403,12 @@
vld1.8 {d24}, [\sr2], \s_strdvmovl.u8 q8, d16
vmovl.u8 q12, d24
- vext.8 q9, q8, q8, #2
- vext.8 q10, q8, q8, #4
- vext.8 q11, q8, q8, #6
- vext.8 q13, q12, q12, #2
- vext.8 q14, q12, q12, #4
- vext.8 q15, q12, q12, #6
+ vext.8 d18, d16, d17, #2
+ vext.8 d20, d16, d17, #4
+ vext.8 d22, d16, d17, #6
+ vext.8 d26, d24, d25, #2
+ vext.8 d28, d24, d25, #4
+ vext.8 d30, d24, d25, #6
subs \h, \h, #2
vmul.s16 d4, d16, d0[0]
vmla.s16 d4, d18, d0[1]