[FFmpeg-cvslog] avcodec/h264: add avx 8-bit chroma v intra deblock/loop filter

[James Darnley]

ffmpeg | branch: master | James Darnley <jdarnley at obe.tv> | Wed Feb 15 14:59:46 2017 +0100| [987ffe4b8dce88b8df12208eef788bd022a43598] | committer: James Darnley

avcodec/h264: add avx 8-bit chroma v intra deblock/loop filter

~1.14x faster (90 vs 78 cycles) compared with mmxext

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=987ffe4b8dce88b8df12208eef788bd022a43598
---

 libavcodec/x86/h264_deblock.asm | 33 +++++++++++++++++++++++++++++++++
 libavcodec/x86/h264dsp_init.c   |  1 +
 2 files changed, 34 insertions(+)

diff --git a/libavcodec/x86/h264_deblock.asm b/libavcodec/x86/h264_deblock.asm
index e2eb002..1e6d822 100644
--- a/libavcodec/x86/h264_deblock.asm
+++ b/libavcodec/x86/h264_deblock.asm
@@ -1117,6 +1117,28 @@ ff_chroma_intra_body_mmxext:
     DEBLOCK_P0_Q0
 %endmacro
 
+%macro CHROMA_INTRA_BODY_XMM 0
+    LOAD_MASK alpha_d, beta_d
+    mova    m5,  m1
+    mova    m6,  m2
+    pxor    m4,  m1, m3
+    pand    m4, [pb_1]
+    pavgb   m1,  m3
+    psubusb m1,  m4
+    pavgb   m1,  m0
+    pxor    m4,  m2, m0
+    pand    m4, [pb_1]
+    pavgb   m2,  m0
+    psubusb m2,  m4
+    pavgb   m2,  m3
+    psubb   m1,  m5
+    psubb   m2,  m6
+    pand    m1,  m7
+    pand    m2,  m7
+    paddb   m1,  m5
+    paddb   m2,  m6
+%endmacro
+
 %macro CHROMA_V_START_XMM 1
     movsxdifnidn stride_q, stride_d
     dec alpha_d
@@ -1190,6 +1212,17 @@ cglobal deblock_h_chroma422_8, 5, 7, 8, 0-16, pix_, stride_, alpha_, beta_, tc0_
     STORE_8_ROWS PASS8ROWS(pix_q - 2, r5 - 2, stride_q, r6)
 RET
 
+cglobal deblock_v_chroma_intra_8, 4, 5, 8, pix_, stride_, alpha_, beta_
+    CHROMA_V_START_XMM r4
+    movq m0, [r4]
+    movq m1, [r4 + stride_q]
+    movq m2, [pix_q]
+    movq m3, [pix_q + stride_q]
+    CHROMA_INTRA_BODY_XMM
+    movq [r4 + stride_q], m1
+    movq [pix_q], m2
+RET
+
 %endmacro ; DEBLOCK_CHROMA_XMM
 
 DEBLOCK_CHROMA_XMM avx
diff --git a/libavcodec/x86/h264dsp_init.c b/libavcodec/x86/h264dsp_init.c
index 6073932..cc86664 100644
--- a/libavcodec/x86/h264dsp_init.c
+++ b/libavcodec/x86/h264dsp_init.c
@@ -319,6 +319,7 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
 #endif
 
             c->h264_v_loop_filter_chroma       = ff_deblock_v_chroma_8_avx;
+            c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_8_avx;
             if (chroma_format_idc <= 1) {
                 c->h264_h_loop_filter_chroma       = ff_deblock_h_chroma_8_avx;
             } else {