ref: 7cd175d09729c676ed6d2b954e9d3216d40b81e4
parent: b13a399ab5c6062da4d094ce8c041954b305a216
author: Martin Storsjö <martin@martin.st>
date: Wed Apr 23 07:25:14 EDT 2014
Use the correct ext syntax in the gnu version of macros
--- a/codec/common/arm64/mc_aarch64_neon.S
+++ b/codec/common/arm64/mc_aarch64_neon.S
@@ -371,8 +371,8 @@
.macro UNPACK_FILTER_SINGLE_TAG_16BITS arg0, arg1, arg2, arg3, arg4, arg5
// { // each 16bits; input: d_dst, d_src[0:5], para, working, working, d(low part of d_dst)
- ext.16b \arg3, \arg1, \arg1, #14 // X[0][1][2][3][4][5]O
- ext.16b \arg4, \arg3, \arg3, #8 // [3][4][5]OX[0][1][2]
+ ext \arg3\().16b, \arg1\().16b, \arg1\().16b, #14 // X[0][1][2][3][4][5]O
+ ext \arg4\().16b, \arg3\().16b, \arg3\().16b, #8 // [3][4][5]OX[0][1][2]
rev64 \arg4\().8h, \arg4\().8h // X[5][4][3][2][1][0]O
add \arg3\().8h, \arg3\().8h, \arg4\().8h // each 16bits, *[50][41][32][23][14][05]*
smull \arg3\().4s, \arg3\().4h, \arg2\().4h // 0+1*[50]-5*[41]+20[32]