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)

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)