diff options
| author | Loren Merritt <lorenm@u.washington.edu> | 2009-10-18 20:10:10 +0000 |
|---|---|---|
| committer | Loren Merritt <lorenm@u.washington.edu> | 2009-10-18 20:10:10 +0000 |
| commit | 2f77923d72c35f4a10b9bb1d1086d0edd7f43dde (patch) | |
| tree | 2c9e7c89285fb2e5117f3d0b4aad5da055aa0fc1 /libavcodec/x86/dsputil_mmx.c | |
| parent | 1303d62d8416fa315a0cc7bbbe35cfdab787ea92 (diff) | |
| download | ffmpeg-2f77923d72c35f4a10b9bb1d1086d0edd7f43dde.tar.gz | |
simd add_hfyu_left_prediction
2.2x faster than C on conroe, 3.6x on penryn.
4-6% faster huffyuv decoding if using left or plane mode and yuv
Originally committed as revision 20287 to svn://svn.ffmpeg.org/ffmpeg/trunk
Diffstat (limited to 'libavcodec/x86/dsputil_mmx.c')
| -rw-r--r-- | libavcodec/x86/dsputil_mmx.c | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c index 2e00aa2a24..3f6f1dae6b 100644 --- a/libavcodec/x86/dsputil_mmx.c +++ b/libavcodec/x86/dsputil_mmx.c @@ -2385,6 +2385,8 @@ void ff_float_to_int16_interleave6_sse(int16_t *dst, const float **src, int len) void ff_float_to_int16_interleave6_3dnow(int16_t *dst, const float **src, int len); void ff_float_to_int16_interleave6_3dn2(int16_t *dst, const float **src, int len); void ff_add_hfyu_median_prediction_mmx2(uint8_t *dst, uint8_t *top, uint8_t *diff, int w, int *left, int *left_top); +int ff_add_hfyu_left_prediction_ssse3(uint8_t *dst, uint8_t *src, int w, int left); +int ff_add_hfyu_left_prediction_sse4(uint8_t *dst, uint8_t *src, int w, int left); void ff_x264_deblock_v_luma_sse2(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0); void ff_x264_deblock_h_luma_sse2(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0); void ff_x264_deblock_v8_luma_intra_mmxext(uint8_t *pix, int stride, int alpha, int beta); @@ -2951,6 +2953,11 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_ssse3; c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_ssse3; c->add_png_paeth_prediction= add_png_paeth_prediction_ssse3; +#if HAVE_YASM + c->add_hfyu_left_prediction = ff_add_hfyu_left_prediction_ssse3; + if (mm_flags & FF_MM_SSE4) // not really sse4, just slow on Conroe + c->add_hfyu_left_prediction = ff_add_hfyu_left_prediction_sse4; +#endif } #endif |