diff options
| author | Daniel Kang <daniel.d.kang@gmail.com> | 2011-06-05 18:33:23 -0400 |
|---|---|---|
| committer | Ronald S. Bultje <rsbultje@gmail.com> | 2011-06-18 07:52:19 -0400 |
| commit | f188a1e0ca12822fd6c607924169d678c7254838 (patch) | |
| tree | 5e42d3b50ede4807a00e285d1bb32e96626f667a /libavcodec/x86/dsputil_mmx.c | |
| parent | 78440c007cd310bb27ac2af5fb7ea5b7555efc84 (diff) | |
| download | ffmpeg-f188a1e0ca12822fd6c607924169d678c7254838.tar.gz | |
H.264: Add x86 assembly for 10-bit MC Chroma H.264 functions.
Mainly ported from 8-bit H.264 MC Chroma.
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
Diffstat (limited to 'libavcodec/x86/dsputil_mmx.c')
| -rw-r--r-- | libavcodec/x86/dsputil_mmx.c | 32 |
1 files changed, 32 insertions, 0 deletions
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c index 214c6a3945..b174b8393f 100644 --- a/libavcodec/x86/dsputil_mmx.c +++ b/libavcodec/x86/dsputil_mmx.c @@ -1938,6 +1938,19 @@ void ff_avg_h264_chroma_mc8_ssse3_rnd (uint8_t *dst, uint8_t *src, void ff_avg_h264_chroma_mc4_ssse3 (uint8_t *dst, uint8_t *src, int stride, int h, int x, int y); +#define CHROMA_MC(OP, NUM, DEPTH, OPT) \ +void ff_ ## OP ## _h264_chroma_mc ## NUM ## _ ## DEPTH ## _ ## OPT \ + (uint8_t *dst, uint8_t *src,\ + int stride, int h, int x, int y); + +CHROMA_MC(put, 2, 10, mmxext) +CHROMA_MC(avg, 2, 10, mmxext) +CHROMA_MC(put, 4, 10, mmxext) +CHROMA_MC(avg, 4, 10, mmxext) +CHROMA_MC(put, 8, 10, sse2) +CHROMA_MC(avg, 8, 10, sse2) +CHROMA_MC(put, 8, 10, avx) +CHROMA_MC(avg, 8, 10, avx) /* CAVS specific */ void ff_put_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride) { @@ -2420,6 +2433,7 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) { int mm_flags = av_get_cpu_flags(); const int high_bit_depth = avctx->codec_id == CODEC_ID_H264 && avctx->bits_per_raw_sample > 8; + const int bit_depth = avctx->bits_per_raw_sample; if (avctx->dsp_mask) { if (avctx->dsp_mask & AV_CPU_FLAG_FORCE) @@ -2651,6 +2665,12 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) c->avg_h264_chroma_pixels_tab[2]= ff_avg_h264_chroma_mc2_mmx2; c->put_h264_chroma_pixels_tab[2]= ff_put_h264_chroma_mc2_mmx2; } + if (bit_depth == 10) { + c->put_h264_chroma_pixels_tab[2]= ff_put_h264_chroma_mc2_10_mmxext; + c->avg_h264_chroma_pixels_tab[2]= ff_avg_h264_chroma_mc2_10_mmxext; + c->put_h264_chroma_pixels_tab[1]= ff_put_h264_chroma_mc4_10_mmxext; + c->avg_h264_chroma_pixels_tab[1]= ff_avg_h264_chroma_mc4_10_mmxext; + } c->add_hfyu_median_prediction = ff_add_hfyu_median_prediction_mmx2; #endif @@ -2756,6 +2776,10 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) H264_QPEL_FUNCS(3, 2, sse2); H264_QPEL_FUNCS(3, 3, sse2); } + if (bit_depth == 10) { + c->put_h264_chroma_pixels_tab[0]= ff_put_h264_chroma_mc8_10_sse2; + c->avg_h264_chroma_pixels_tab[0]= ff_avg_h264_chroma_mc8_10_sse2; + } } #if HAVE_SSSE3 if(mm_flags & AV_CPU_FLAG_SSSE3){ @@ -2854,6 +2878,14 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) } #endif } +#if HAVE_AVX + if (mm_flags & AV_CPU_FLAG_AVX) { + if (bit_depth == 10) { + c->put_h264_chroma_pixels_tab[0]= ff_put_h264_chroma_mc8_10_avx; + c->avg_h264_chroma_pixels_tab[0]= ff_avg_h264_chroma_mc8_10_avx; + } + } +#endif } if (CONFIG_ENCODERS) |