diff options
| author | Michael Niedermayer <michaelni@gmx.at> | 2014-01-15 15:27:07 +0100 |
|---|---|---|
| committer | Michael Niedermayer <michaelni@gmx.at> | 2014-01-15 15:27:26 +0100 |
| commit | bf0470a5be1c06ac48e97a9e6aa082154bab7656 (patch) | |
| tree | 02dbd6ca89332dc17b4f321961f97bbd9d754e79 /libavcodec/aarch64/neon.S | |
| parent | 28207bfafb905ce02fcc549a27921ac085714623 (diff) | |
| parent | 36e3b1f2fd262028834a9d7b1eb533c1218ee6c2 (diff) | |
| download | ffmpeg-bf0470a5be1c06ac48e97a9e6aa082154bab7656.tar.gz | |
Merge commit '36e3b1f2fd262028834a9d7b1eb533c1218ee6c2'
* commit '36e3b1f2fd262028834a9d7b1eb533c1218ee6c2':
aarch64: h264 loop filter NEON optimizations
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavcodec/aarch64/neon.S')
| -rw-r--r-- | libavcodec/aarch64/neon.S | 24 |
1 files changed, 24 insertions, 0 deletions
diff --git a/libavcodec/aarch64/neon.S b/libavcodec/aarch64/neon.S index 19ab6cb0bd..619aec6426 100644 --- a/libavcodec/aarch64/neon.S +++ b/libavcodec/aarch64/neon.S @@ -80,6 +80,30 @@ trn2 \r7\().4S, \t1\().4S, \r7\().4S .endm +.macro transpose_4x16B r0, r1, r2, r3, t4, t5, t6, t7 + trn1 \t4\().16B, \r0\().16B, \r1\().16B + trn2 \t5\().16B, \r0\().16B, \r1\().16B + trn1 \t6\().16B, \r2\().16B, \r3\().16B + trn2 \t7\().16B, \r2\().16B, \r3\().16B + + trn1 \r0\().8H, \t4\().8H, \t6\().8H + trn2 \r2\().8H, \t4\().8H, \t6\().8H + trn1 \r1\().8H, \t5\().8H, \t7\().8H + trn2 \r3\().8H, \t5\().8H, \t7\().8H +.endm + +.macro transpose_4x8B r0, r1, r2, r3, t4, t5, t6, t7 + trn1 \t4\().8B, \r0\().8B, \r1\().8B + trn2 \t5\().8B, \r0\().8B, \r1\().8B + trn1 \t6\().8B, \r2\().8B, \r3\().8B + trn2 \t7\().8B, \r2\().8B, \r3\().8B + + trn1 \r0\().4H, \t4\().4H, \t6\().4H + trn2 \r2\().4H, \t4\().4H, \t6\().4H + trn1 \r1\().4H, \t5\().4H, \t7\().4H + trn2 \r3\().4H, \t5\().4H, \t7\().4H +.endm + .macro transpose_4x4H r0, r1, r2, r3, r4, r5, r6, r7 trn1 \r4\().4H, \r0\().4H, \r1\().4H trn2 \r5\().4H, \r0\().4H, \r1\().4H |