[FFmpeg-devel] [PATCH 6/6] avcodec/h264: add avx 8-bit 4:2:2 chroma h intra deblock/loop filter

[James Darnley]

~1.37x faster (147 vs. 108 cycles) compared to mmxext function
---
 libavcodec/x86/h264_deblock.asm | 18 ++++++++++++++++++
 libavcodec/x86/h264dsp_init.c   |  1 +
 2 files changed, 19 insertions(+)

diff --git a/libavcodec/x86/h264_deblock.asm b/libavcodec/x86/h264_deblock.asm
index 2197608..32aa3d3 100644
--- a/libavcodec/x86/h264_deblock.asm
+++ b/libavcodec/x86/h264_deblock.asm
@@ -1232,6 +1232,24 @@ cglobal deblock_h_chroma_intra_8, 4, 6, 8, pix_, stride_, alpha_, beta_
     STORE_8_ROWS PASS8ROWS(pix_q - 2, r4 - 2, stride_q, r5)
 RET
 
+cglobal deblock_h_chroma422_intra_8, 4, 6, 8, pix_, stride_, alpha_, beta_
+    CHROMA_H_START_XMM r4, r5
+    LOAD_8_ROWS PASS8ROWS(pix_q - 2, r4 - 2, stride_q, r5)
+    TRANSPOSE_8x4B_XMM
+    CHROMA_INTRA_BODY_XMM
+    TRANSPOSE_4x8B_XMM
+    STORE_8_ROWS PASS8ROWS(pix_q - 2, r4 - 2, stride_q, r5)
+
+    lea pix_q, [pix_q + 8*stride_q]
+    lea r4,    [r4    + 8*stride_q]
+
+    LOAD_8_ROWS PASS8ROWS(pix_q - 2, r4 - 2, stride_q, r5)
+    TRANSPOSE_8x4B_XMM
+    CHROMA_INTRA_BODY_XMM
+    TRANSPOSE_4x8B_XMM
+    STORE_8_ROWS PASS8ROWS(pix_q - 2, r4 - 2, stride_q, r5)
+RET
+
 %endmacro ; DEBLOCK_CHROMA_XMM
 
 DEBLOCK_CHROMA_XMM avx
diff --git a/libavcodec/x86/h264dsp_init.c b/libavcodec/x86/h264dsp_init.c
index b1246d0..51082e8 100644
--- a/libavcodec/x86/h264dsp_init.c
+++ b/libavcodec/x86/h264dsp_init.c
@@ -325,6 +325,7 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
                 c->h264_h_loop_filter_chroma_intra = ff_deblock_h_chroma_intra_8_avx;
             } else {
                 c->h264_h_loop_filter_chroma       = ff_deblock_h_chroma422_8_avx;
+                c->h264_h_loop_filter_chroma_intra = ff_deblock_h_chroma422_intra_8_avx;
             }
         }
     } else if (bit_depth == 10) {
-- 
2.8.3