diff options
| author | James Darnley <jdarnley@obe.tv> | 2017-02-10 23:17:57 +0100 |
|---|---|---|
| committer | James Darnley <jdarnley@obe.tv> | 2017-02-18 20:26:52 +0100 |
| commit | 533688786799b22d0711eedcfe8f84deea014f30 (patch) | |
| tree | d4a4655fd9916238a27b5054bd8d92db8e6482bd /libavutil/x86 | |
| parent | 7627df15d411a69f236b4650e88b1ab911f38efc (diff) | |
| download | ffmpeg-533688786799b22d0711eedcfe8f84deea014f30.tar.gz | |
avcodec/h264: sse2, avx h luma mbaff deblock/loop filter
x86-64 only
Yorkfield:
- sse2: ~2.17x (434 vs. 200 cycles)
Nehalem:
- sse2: ~2.94x (409 vs. 139 cycles)
Skylake:
- sse2: ~3.10x (370 vs. 119 cycles)
- avx: ~3.29x (370 vs. 112 cycles)
Diffstat (limited to 'libavutil/x86')
| -rw-r--r-- | libavutil/x86/x86util.asm | 15 |
1 files changed, 15 insertions, 0 deletions
diff --git a/libavutil/x86/x86util.asm b/libavutil/x86/x86util.asm index 1408f0a176..c50ddc6c54 100644 --- a/libavutil/x86/x86util.asm +++ b/libavutil/x86/x86util.asm @@ -265,6 +265,21 @@ SWAP %12, %15 %endmacro +%macro TRANSPOSE_8X8B 8 + %if mmsize == 8 + %error "This macro does not support mmsize == 8" + %endif + punpcklbw m%1, m%2 + punpcklbw m%3, m%4 + punpcklbw m%5, m%6 + punpcklbw m%7, m%8 + TRANSPOSE4x4W %1, %3, %5, %7, %2 + MOVHL m%2, m%1 + MOVHL m%4, m%3 + MOVHL m%6, m%5 + MOVHL m%8, m%7 +%endmacro + ; PABSW macro assumes %1 != %2, while ABS1/2 macros work in-place %macro PABSW 2 %if cpuflag(ssse3) |