diff options
| author | James Darnley <jdarnley@obe.tv> | 2017-02-15 15:05:42 +0100 |
|---|---|---|
| committer | James Darnley <jdarnley@obe.tv> | 2017-02-27 13:22:06 +0100 |
| commit | cd893b9307b8c871fec067928dc613fa987e38c3 (patch) | |
| tree | 9bfaa2cf5e7f2776a2e2d3fb87c69801c01d0f3f | |
| parent | 0e16b3e2be3a3e20a5da94a0c48049ba6a5baacd (diff) | |
| download | ffmpeg-cd893b9307b8c871fec067928dc613fa987e38c3.tar.gz | |
avcodec/h264: add avx 8-bit 4:2:2 chroma h intra deblock/loop filter
~1.37x faster (147 vs. 108 cycles) compared to mmxext function
| -rw-r--r-- | libavcodec/x86/h264_deblock.asm | 18 | ||||
| -rw-r--r-- | libavcodec/x86/h264dsp_init.c | 1 |
2 files changed, 19 insertions, 0 deletions
diff --git a/libavcodec/x86/h264_deblock.asm b/libavcodec/x86/h264_deblock.asm index 2197608d73..32aa3d376b 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 b1246d029e..51082e88e2 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) { |