diff options
author | Peter Ross <pross@xvid.org> | 2014-04-01 14:21:51 +0200 |
---|---|---|
committer | Vittorio Giovara <vittorio.giovara@gmail.com> | 2014-04-04 04:00:11 +0200 |
commit | ac4b32df71bd932838043a4838b86d11e169707f (patch) | |
tree | 50fbb1b0c61040f815d8f7bb22fe8c53987ed7a1 /libavcodec | |
parent | ce58ead40dd16f272ec1f127f04f69df67c7cd35 (diff) | |
download | ffmpeg-ac4b32df71bd932838043a4838b86d11e169707f.tar.gz |
On2 VP7 decoder
Further performance improvements and security fixes by
Vittorio Giovara, Luca Barbato and Diego Biurrun.
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
Signed-off-by: Diego Biurrun <diego@biurrun.de>
Diffstat (limited to 'libavcodec')
-rw-r--r-- | libavcodec/Makefile | 1 | ||||
-rw-r--r-- | libavcodec/allcodecs.c | 1 | ||||
-rw-r--r-- | libavcodec/arm/Makefile | 6 | ||||
-rw-r--r-- | libavcodec/arm/h264pred_init_arm.c | 8 | ||||
-rw-r--r-- | libavcodec/arm/vp8dsp.h | 2 | ||||
-rw-r--r-- | libavcodec/arm/vp8dsp_init_arm.c | 10 | ||||
-rw-r--r-- | libavcodec/arm/vp8dsp_init_armv6.c | 47 | ||||
-rw-r--r-- | libavcodec/arm/vp8dsp_init_neon.c | 45 | ||||
-rw-r--r-- | libavcodec/avcodec.h | 1 | ||||
-rw-r--r-- | libavcodec/codec_desc.c | 7 | ||||
-rw-r--r-- | libavcodec/h264pred.c | 17 | ||||
-rw-r--r-- | libavcodec/h264pred.h | 2 | ||||
-rw-r--r-- | libavcodec/ppc/Makefile | 1 | ||||
-rw-r--r-- | libavcodec/ppc/vp8dsp_altivec.c | 3 | ||||
-rw-r--r-- | libavcodec/version.h | 2 | ||||
-rw-r--r-- | libavcodec/vp8.c | 907 | ||||
-rw-r--r-- | libavcodec/vp8.h | 21 | ||||
-rw-r--r-- | libavcodec/vp8data.h | 151 | ||||
-rw-r--r-- | libavcodec/vp8dsp.c | 365 | ||||
-rw-r--r-- | libavcodec/vp8dsp.h | 13 | ||||
-rw-r--r-- | libavcodec/x86/Makefile | 3 | ||||
-rw-r--r-- | libavcodec/x86/h264_intrapred_init.c | 11 | ||||
-rw-r--r-- | libavcodec/x86/vp8dsp_init.c | 77 |
23 files changed, 1359 insertions, 342 deletions
diff --git a/libavcodec/Makefile b/libavcodec/Makefile index aa870d83ca..110e0dd74f 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -386,6 +386,7 @@ OBJS-$(CONFIG_VP5_DECODER) += vp5.o vp56.o vp56data.o vp56dsp.o \ vp56rac.o OBJS-$(CONFIG_VP6_DECODER) += vp6.o vp56.o vp56data.o vp56dsp.o \ vp6dsp.o vp56rac.o +OBJS-$(CONFIG_VP7_DECODER) += vp8.o vp8dsp.o vp56rac.o OBJS-$(CONFIG_VP8_DECODER) += vp8.o vp8dsp.o vp56rac.o OBJS-$(CONFIG_VP9_DECODER) += vp9.o vp9data.o vp9dsp.o \ vp9block.o vp9prob.o vp9mvs.o vp56rac.o diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c index aa5bdc0371..8066e6f7fd 100644 --- a/libavcodec/allcodecs.c +++ b/libavcodec/allcodecs.c @@ -261,6 +261,7 @@ void avcodec_register_all(void) REGISTER_DECODER(VP6, vp6); REGISTER_DECODER(VP6A, vp6a); REGISTER_DECODER(VP6F, vp6f); + REGISTER_DECODER(VP7, vp7); REGISTER_DECODER(VP8, vp8); REGISTER_DECODER(VP9, vp9); REGISTER_DECODER(VQA, vqa); diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile index 3d34c4a754..9d5b6aab5b 100644 --- a/libavcodec/arm/Makefile +++ b/libavcodec/arm/Makefile @@ -31,6 +31,7 @@ OBJS-$(CONFIG_MLP_DECODER) += arm/mlpdsp_init_arm.o OBJS-$(CONFIG_VC1_DECODER) += arm/vc1dsp_init_arm.o OBJS-$(CONFIG_VORBIS_DECODER) += arm/vorbisdsp_init_arm.o OBJS-$(CONFIG_VP6_DECODER) += arm/vp6dsp_init_arm.o +OBJS-$(CONFIG_VP7_DECODER) += arm/vp8dsp_init_arm.o OBJS-$(CONFIG_VP8_DECODER) += arm/vp8dsp_init_arm.o OBJS-$(CONFIG_RV30_DECODER) += arm/rv34dsp_init_arm.o OBJS-$(CONFIG_RV40_DECODER) += arm/rv34dsp_init_arm.o \ @@ -55,6 +56,9 @@ ARMV6-OBJS-$(CONFIG_HPELDSP) += arm/hpeldsp_init_armv6.o \ ARMV6-OBJS-$(CONFIG_MPEGAUDIODSP) += arm/mpegaudiodsp_fixed_armv6.o ARMV6-OBJS-$(CONFIG_MLP_DECODER) += arm/mlpdsp_armv6.o +ARMV6-OBJS-$(CONFIG_VP7_DECODER) += arm/vp8_armv6.o \ + arm/vp8dsp_init_armv6.o \ + arm/vp8dsp_armv6.o ARMV6-OBJS-$(CONFIG_VP8_DECODER) += arm/vp8_armv6.o \ arm/vp8dsp_init_armv6.o \ arm/vp8dsp_armv6.o @@ -102,5 +106,7 @@ NEON-OBJS-$(CONFIG_VC1_DECODER) += arm/vc1dsp_init_neon.o \ arm/vc1dsp_neon.o NEON-OBJS-$(CONFIG_VORBIS_DECODER) += arm/vorbisdsp_neon.o NEON-OBJS-$(CONFIG_VP6_DECODER) += arm/vp6dsp_neon.o +NEON-OBJS-$(CONFIG_VP7_DECODER) += arm/vp8dsp_init_neon.o \ + arm/vp8dsp_neon.o NEON-OBJS-$(CONFIG_VP8_DECODER) += arm/vp8dsp_init_neon.o \ arm/vp8dsp_neon.o diff --git a/libavcodec/arm/h264pred_init_arm.c b/libavcodec/arm/h264pred_init_arm.c index 2e25c2a271..bbfe63f2aa 100644 --- a/libavcodec/arm/h264pred_init_arm.c +++ b/libavcodec/arm/h264pred_init_arm.c @@ -56,10 +56,11 @@ static av_cold void h264_pred_init_neon(H264PredContext *h, int codec_id, h->pred8x8[VERT_PRED8x8 ] = ff_pred8x8_vert_neon; h->pred8x8[HOR_PRED8x8 ] = ff_pred8x8_hor_neon; - if (codec_id != AV_CODEC_ID_VP8) + if (codec_id != AV_CODEC_ID_VP7 && codec_id != AV_CODEC_ID_VP8) h->pred8x8[PLANE_PRED8x8] = ff_pred8x8_plane_neon; h->pred8x8[DC_128_PRED8x8 ] = ff_pred8x8_128_dc_neon; - if (codec_id != AV_CODEC_ID_RV40 && codec_id != AV_CODEC_ID_VP8) { + if (codec_id != AV_CODEC_ID_RV40 && codec_id != AV_CODEC_ID_VP7 && + codec_id != AV_CODEC_ID_VP8) { h->pred8x8[DC_PRED8x8 ] = ff_pred8x8_dc_neon; h->pred8x8[LEFT_DC_PRED8x8] = ff_pred8x8_left_dc_neon; h->pred8x8[TOP_DC_PRED8x8 ] = ff_pred8x8_top_dc_neon; @@ -75,7 +76,8 @@ static av_cold void h264_pred_init_neon(H264PredContext *h, int codec_id, h->pred16x16[LEFT_DC_PRED8x8] = ff_pred16x16_left_dc_neon; h->pred16x16[TOP_DC_PRED8x8 ] = ff_pred16x16_top_dc_neon; h->pred16x16[DC_128_PRED8x8 ] = ff_pred16x16_128_dc_neon; - if (codec_id != AV_CODEC_ID_SVQ3 && codec_id != AV_CODEC_ID_RV40 && codec_id != AV_CODEC_ID_VP8) + if (codec_id != AV_CODEC_ID_SVQ3 && codec_id != AV_CODEC_ID_RV40 && + codec_id != AV_CODEC_ID_VP7 && codec_id != AV_CODEC_ID_VP8) h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_plane_neon; } diff --git a/libavcodec/arm/vp8dsp.h b/libavcodec/arm/vp8dsp.h index ce00e4abf5..0d55e0ffc0 100644 --- a/libavcodec/arm/vp8dsp.h +++ b/libavcodec/arm/vp8dsp.h @@ -21,6 +21,8 @@ #include "libavcodec/vp8dsp.h" +void ff_vp78dsp_init_armv6(VP8DSPContext *dsp); +void ff_vp78dsp_init_neon(VP8DSPContext *dsp); void ff_vp8dsp_init_armv6(VP8DSPContext *dsp); void ff_vp8dsp_init_neon(VP8DSPContext *dsp); diff --git a/libavcodec/arm/vp8dsp_init_arm.c b/libavcodec/arm/vp8dsp_init_arm.c index b7897cd9fd..aa77dbab98 100644 --- a/libavcodec/arm/vp8dsp_init_arm.c +++ b/libavcodec/arm/vp8dsp_init_arm.c @@ -23,6 +23,16 @@ #include "libavcodec/vp8dsp.h" #include "vp8dsp.h" +av_cold void ff_vp78dsp_init_arm(VP8DSPContext *dsp) +{ + int cpu_flags = av_get_cpu_flags(); + + if (have_armv6(cpu_flags)) + ff_vp78dsp_init_armv6(dsp); + if (have_neon(cpu_flags)) + ff_vp78dsp_init_neon(dsp); +} + av_cold void ff_vp8dsp_init_arm(VP8DSPContext *dsp) { int cpu_flags = av_get_cpu_flags(); diff --git a/libavcodec/arm/vp8dsp_init_armv6.c b/libavcodec/arm/vp8dsp_init_armv6.c index e15e191e30..febe4e71a2 100644 --- a/libavcodec/arm/vp8dsp_init_armv6.c +++ b/libavcodec/arm/vp8dsp_init_armv6.c @@ -40,29 +40,8 @@ VP8_BILIN(16, armv6); VP8_BILIN(8, armv6); VP8_BILIN(4, armv6); -av_cold void ff_vp8dsp_init_armv6(VP8DSPContext *dsp) +av_cold void ff_vp78dsp_init_armv6(VP8DSPContext *dsp) { - dsp->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_armv6; - dsp->vp8_luma_dc_wht_dc = ff_vp8_luma_dc_wht_dc_armv6; - - dsp->vp8_idct_add = ff_vp8_idct_add_armv6; - dsp->vp8_idct_dc_add = ff_vp8_idct_dc_add_armv6; - dsp->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_armv6; - dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_armv6; - - dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_armv6; - dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_armv6; - dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_armv6; - dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_armv6; - - dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_armv6; - dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_armv6; - dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_armv6; - dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_armv6; - - dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_armv6; - dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_armv6; - dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_armv6; dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_armv6; dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_armv6; @@ -118,3 +97,27 @@ av_cold void ff_vp8dsp_init_armv6(VP8DSPContext *dsp) dsp->put_vp8_bilinear_pixels_tab[2][2][1] = ff_put_vp8_bilin4_hv_armv6; dsp->put_vp8_bilinear_pixels_tab[2][2][2] = ff_put_vp8_bilin4_hv_armv6; } + +av_cold void ff_vp8dsp_init_armv6(VP8DSPContext *dsp) +{ + dsp->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_armv6; + dsp->vp8_luma_dc_wht_dc = ff_vp8_luma_dc_wht_dc_armv6; + + dsp->vp8_idct_add = ff_vp8_idct_add_armv6; + dsp->vp8_idct_dc_add = ff_vp8_idct_dc_add_armv6; + dsp->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_armv6; + dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_armv6; + + dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_armv6; + dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_armv6; + dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_armv6; + dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_armv6; + + dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_armv6; + dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_armv6; + dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_armv6; + dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_armv6; + + dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_armv6; + dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_armv6; +} diff --git a/libavcodec/arm/vp8dsp_init_neon.c b/libavcodec/arm/vp8dsp_init_neon.c index 0468181f70..2b6c7750d3 100644 --- a/libavcodec/arm/vp8dsp_init_neon.c +++ b/libavcodec/arm/vp8dsp_init_neon.c @@ -39,28 +39,8 @@ VP8_BILIN(16, neon); VP8_BILIN(8, neon); VP8_BILIN(4, neon); -av_cold void ff_vp8dsp_init_neon(VP8DSPContext *dsp) +av_cold void ff_vp78dsp_init_neon(VP8DSPContext *dsp) { - dsp->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_neon; - - dsp->vp8_idct_add = ff_vp8_idct_add_neon; - dsp->vp8_idct_dc_add = ff_vp8_idct_dc_add_neon; - dsp->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_neon; - dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_neon; - - dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_neon; - dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_neon; - dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_neon; - dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_neon; - - dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_neon; - dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_neon; - dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_neon; - dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_neon; - - dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_neon; - dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_neon; - dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon; dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_neon; dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_neon; @@ -114,3 +94,26 @@ av_cold void ff_vp8dsp_init_neon(VP8DSPContext *dsp) dsp->put_vp8_bilinear_pixels_tab[2][2][1] = ff_put_vp8_bilin4_hv_neon; dsp->put_vp8_bilinear_pixels_tab[2][2][2] = ff_put_vp8_bilin4_hv_neon; } + +av_cold void ff_vp8dsp_init_neon(VP8DSPContext *dsp) +{ + dsp->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_neon; + + dsp->vp8_idct_add = ff_vp8_idct_add_neon; + dsp->vp8_idct_dc_add = ff_vp8_idct_dc_add_neon; + dsp->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_neon; + dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_neon; + + dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_neon; + dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_neon; + dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_neon; + dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_neon; + + dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_neon; + dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_neon; + dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_neon; + dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_neon; + + dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_neon; + dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_neon; +} diff --git a/libavcodec/avcodec.h b/libavcodec/avcodec.h index a95e735125..439c426863 100644 --- a/libavcodec/avcodec.h +++ b/libavcodec/avcodec.h @@ -288,6 +288,7 @@ enum AVCodecID { AV_CODEC_ID_BRENDER_PIX, AV_CODEC_ID_PAF_VIDEO, AV_CODEC_ID_EXR, + AV_CODEC_ID_VP7, /* various PCM "codecs" */ AV_CODEC_ID_FIRST_AUDIO = 0x10000, ///< A dummy id pointing at the start of audio codecs diff --git a/libavcodec/codec_desc.c b/libavcodec/codec_desc.c index f46e082926..11671bf68c 100644 --- a/libavcodec/codec_desc.c +++ b/libavcodec/codec_desc.c @@ -1099,6 +1099,13 @@ static const AVCodecDescriptor codec_descriptors[] = { .long_name = NULL_IF_CONFIG_SMALL("Amazing Studio Packed Animation File Video"), .props = AV_CODEC_PROP_LOSSY, }, + { + .id = AV_CODEC_ID_VP7, + .type = AVMEDIA_TYPE_VIDEO, + .name = "vp7", + .long_name = NULL_IF_CONFIG_SMALL("On2 VP7"), + .props = AV_CODEC_PROP_LOSSY, + }, /* image codecs */ { diff --git a/libavcodec/h264pred.c b/libavcodec/h264pred.c index 531760c145..8b515fd126 100644 --- a/libavcodec/h264pred.c +++ b/libavcodec/h264pred.c @@ -410,7 +410,7 @@ av_cold void ff_h264_pred_init(H264PredContext *h, int codec_id, #define H264_PRED(depth) \ if(codec_id != AV_CODEC_ID_RV40){\ - if(codec_id == AV_CODEC_ID_VP8) {\ + if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) {\ h->pred4x4[VERT_PRED ]= FUNCD(pred4x4_vertical_vp8);\ h->pred4x4[HOR_PRED ]= FUNCD(pred4x4_horizontal_vp8);\ } else {\ @@ -425,15 +425,14 @@ av_cold void ff_h264_pred_init(H264PredContext *h, int codec_id, h->pred4x4[DIAG_DOWN_RIGHT_PRED]= FUNCC(pred4x4_down_right , depth);\ h->pred4x4[VERT_RIGHT_PRED ]= FUNCC(pred4x4_vertical_right , depth);\ h->pred4x4[HOR_DOWN_PRED ]= FUNCC(pred4x4_horizontal_down , depth);\ - if (codec_id == AV_CODEC_ID_VP8) {\ + if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) {\ h->pred4x4[VERT_LEFT_PRED ]= FUNCD(pred4x4_vertical_left_vp8);\ } else\ h->pred4x4[VERT_LEFT_PRED ]= FUNCC(pred4x4_vertical_left , depth);\ h->pred4x4[HOR_UP_PRED ]= FUNCC(pred4x4_horizontal_up , depth);\ - if(codec_id != AV_CODEC_ID_VP8) {\ + if (codec_id != AV_CODEC_ID_VP7 && codec_id != AV_CODEC_ID_VP8) {\ h->pred4x4[LEFT_DC_PRED ]= FUNCC(pred4x4_left_dc , depth);\ h->pred4x4[TOP_DC_PRED ]= FUNCC(pred4x4_top_dc , depth);\ - h->pred4x4[DC_128_PRED ]= FUNCC(pred4x4_128_dc , depth);\ } else {\ h->pred4x4[TM_VP8_PRED ]= FUNCD(pred4x4_tm_vp8);\ h->pred4x4[DC_127_PRED ]= FUNCC(pred4x4_127_dc , depth);\ @@ -441,6 +440,8 @@ av_cold void ff_h264_pred_init(H264PredContext *h, int codec_id, h->pred4x4[VERT_VP8_PRED ]= FUNCC(pred4x4_vertical , depth);\ h->pred4x4[HOR_VP8_PRED ]= FUNCC(pred4x4_horizontal , depth);\ }\ + if (codec_id != AV_CODEC_ID_VP8)\ + h->pred4x4[DC_128_PRED ]= FUNCC(pred4x4_128_dc , depth);\ }else{\ h->pred4x4[VERT_PRED ]= FUNCC(pred4x4_vertical , depth);\ h->pred4x4[HOR_PRED ]= FUNCC(pred4x4_horizontal , depth);\ @@ -479,7 +480,7 @@ av_cold void ff_h264_pred_init(H264PredContext *h, int codec_id, h->pred8x8[VERT_PRED8x8 ]= FUNCC(pred8x16_vertical , depth);\ h->pred8x8[HOR_PRED8x8 ]= FUNCC(pred8x16_horizontal , depth);\ }\ - if (codec_id != AV_CODEC_ID_VP8) {\ + if (codec_id != AV_CODEC_ID_VP7 && codec_id != AV_CODEC_ID_VP8) {\ if (chroma_format_idc <= 1) {\ h->pred8x8[PLANE_PRED8x8]= FUNCC(pred8x8_plane , depth);\ } else {\ @@ -487,7 +488,8 @@ av_cold void ff_h264_pred_init(H264PredContext *h, int codec_id, }\ } else\ h->pred8x8[PLANE_PRED8x8]= FUNCD(pred8x8_tm_vp8);\ - if(codec_id != AV_CODEC_ID_RV40 && codec_id != AV_CODEC_ID_VP8){\ + if (codec_id != AV_CODEC_ID_RV40 && codec_id != AV_CODEC_ID_VP7 && \ + codec_id != AV_CODEC_ID_VP8) {\ if (chroma_format_idc <= 1) {\ h->pred8x8[DC_PRED8x8 ]= FUNCC(pred8x8_dc , depth);\ h->pred8x8[LEFT_DC_PRED8x8]= FUNCC(pred8x8_left_dc , depth);\ @@ -509,7 +511,7 @@ av_cold void ff_h264_pred_init(H264PredContext *h, int codec_id, h->pred8x8[DC_PRED8x8 ]= FUNCD(pred8x8_dc_rv40);\ h->pred8x8[LEFT_DC_PRED8x8]= FUNCD(pred8x8_left_dc_rv40);\ h->pred8x8[TOP_DC_PRED8x8 ]= FUNCD(pred8x8_top_dc_rv40);\ - if (codec_id == AV_CODEC_ID_VP8) {\ + if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) {\ h->pred8x8[DC_127_PRED8x8]= FUNCC(pred8x8_127_dc , depth);\ h->pred8x8[DC_129_PRED8x8]= FUNCC(pred8x8_129_dc , depth);\ }\ @@ -530,6 +532,7 @@ av_cold void ff_h264_pred_init(H264PredContext *h, int codec_id, case AV_CODEC_ID_RV40:\ h->pred16x16[PLANE_PRED8x8 ]= FUNCD(pred16x16_plane_rv40);\ break;\ + case AV_CODEC_ID_VP7:\ case AV_CODEC_ID_VP8:\ h->pred16x16[PLANE_PRED8x8 ]= FUNCD(pred16x16_tm_vp8);\ h->pred16x16[DC_127_PRED8x8]= FUNCC(pred16x16_127_dc , depth);\ diff --git a/libavcodec/h264pred.h b/libavcodec/h264pred.h index 62e5b79dc8..5c4ef17c2a 100644 --- a/libavcodec/h264pred.h +++ b/libavcodec/h264pred.h @@ -60,7 +60,7 @@ #define VERT_VP8_PRED 10 ///< for VP8, #VERT_PRED is the average of ///< (left col+cur col x2+right col) / 4; ///< this is the "unaveraged" one -#define HOR_VP8_PRED 11 ///< unaveraged version of #HOR_PRED, see +#define HOR_VP8_PRED 14 ///< unaveraged version of #HOR_PRED, see ///< #VERT_VP8_PRED for details #define DC_127_PRED 12 #define DC_129_PRED 13 diff --git a/libavcodec/ppc/Makefile b/libavcodec/ppc/Makefile index 80580b5d9e..80ed14bb19 100644 --- a/libavcodec/ppc/Makefile +++ b/libavcodec/ppc/Makefile @@ -13,6 +13,7 @@ OBJS-$(CONFIG_VP3DSP) += ppc/vp3dsp_altivec.o OBJS-$(CONFIG_VC1_DECODER) += ppc/vc1dsp_altivec.o OBJS-$(CONFIG_VORBIS_DECODER) += ppc/vorbisdsp_altivec.o +OBJS-$(CONFIG_VP7_DECODER) += ppc/vp8dsp_altivec.o OBJS-$(CONFIG_VP8_DECODER) += ppc/vp8dsp_altivec.o ALTIVEC-OBJS-$(CONFIG_DSPUTIL) += ppc/dsputil_altivec.o \ diff --git a/libavcodec/ppc/vp8dsp_altivec.c b/libavcodec/ppc/vp8dsp_altivec.c index 366dc3ccbb..653ffb2786 100644 --- a/libavcodec/ppc/vp8dsp_altivec.c +++ b/libavcodec/ppc/vp8dsp_altivec.c @@ -311,7 +311,8 @@ static void put_vp8_pixels16_altivec(uint8_t *dst, ptrdiff_t dstride, uint8_t *s #endif /* HAVE_ALTIVEC */ -av_cold void ff_vp8dsp_init_ppc(VP8DSPContext *c) + +av_cold void ff_vp78dsp_init_ppc(VP8DSPContext *c) { #if HAVE_ALTIVEC if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC)) diff --git a/libavcodec/version.h b/libavcodec/version.h index 6555dbf6a4..58c4aa619f 100644 --- a/libavcodec/version.h +++ b/libavcodec/version.h @@ -29,7 +29,7 @@ #include "libavutil/version.h" #define LIBAVCODEC_VERSION_MAJOR 55 -#define LIBAVCODEC_VERSION_MINOR 42 +#define LIBAVCODEC_VERSION_MINOR 43 #define LIBAVCODEC_VERSION_MICRO 0 #define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \ diff --git a/libavcodec/vp8.c b/libavcodec/vp8.c index a1d24cdc1f..e73ee822e5 100644 --- a/libavcodec/vp8.c +++ b/libavcodec/vp8.c @@ -1,10 +1,11 @@ /* - * VP8 compatible video decoder + * VP7/VP8 compatible video decoder * * Copyright (C) 2010 David Conrad * Copyright (C) 2010 Ronald S. Bultje * Copyright (C) 2010 Jason Garrett-Glaser * Copyright (C) 2012 Daniel Kang + * Copyright (C) 2014 Peter Ross * * This file is part of Libav. * @@ -75,6 +76,7 @@ static void vp8_release_frame(VP8Context *s, VP8Frame *f) ff_thread_release_buffer(s->avctx, &f->tf); } +#if CONFIG_VP8_DECODER static int vp8_ref_frame(VP8Context *s, VP8Frame *dst, VP8Frame *src) { int ret; @@ -91,6 +93,7 @@ static int vp8_ref_frame(VP8Context *s, VP8Frame *dst, VP8Frame *src) return 0; } +#endif /* CONFIG_VP8_DECODER */ static void vp8_decode_flush_impl(AVCodecContext *avctx, int free_mem) { @@ -110,7 +113,32 @@ static void vp8_decode_flush(AVCodecContext *avctx) vp8_decode_flush_impl(avctx, 0); } -static int update_dimensions(VP8Context *s, int width, int height) +static VP8Frame *vp8_find_free_buffer(VP8Context *s) +{ + VP8Frame *frame = NULL; + int i; + + // find a free buffer + for (i = 0; i < 5; i++) + if (&s->frames[i] != s->framep[VP56_FRAME_CURRENT] && + &s->frames[i] != s->framep[VP56_FRAME_PREVIOUS] && + &s->frames[i] != s->framep[VP56_FRAME_GOLDEN] && + &s->frames[i] != s->framep[VP56_FRAME_GOLDEN2]) { + frame = &s->frames[i]; + break; + } + if (i == 5) { + av_log(s->avctx, AV_LOG_FATAL, "Ran out of free frames!\n"); + abort(); + } + if (frame->tf.f->data[0]) + vp8_release_frame(s, frame); + + return frame; +} + +static av_always_inline +int update_dimensions(VP8Context *s, int width, int height, int is_vp7) { AVCodecContext *avctx = s->avctx; int i, ret; @@ -127,8 +155,8 @@ static int update_dimensions(VP8Context *s, int width, int height) s->mb_width = (s->avctx->coded_width + 15) / 16; s->mb_height = (s->avctx->coded_height + 15) / 16; - s->mb_layout = (avctx->active_thread_type == FF_THREAD_SLICE) && - (FFMIN(s->num_coeff_partitions, avctx->thread_count) > 1); + s->mb_layout = is_vp7 || avctx->active_thread_type == FF_THREAD_SLICE && + FFMIN(s->num_coeff_partitions, avctx->thread_count) > 1; if (!s->mb_layout) { // Frame threading and one thread s->macroblocks_base = av_mallocz((s->mb_width + s->mb_height * 2 + 1) * sizeof(*s->macroblocks)); @@ -158,6 +186,16 @@ static int update_dimensions(VP8Context *s, int width, int height) return 0; } +static int vp7_update_dimensions(VP8Context *s, int width, int height) +{ + return update_dimensions(s, width, height, IS_VP7); +} + +static int vp8_update_dimensions(VP8Context *s, int width, int height) +{ + return update_dimensions(s, width, height, IS_VP8); +} + static void parse_segment_info(VP8Context *s) { VP56RangeCoder *c = &s->c; @@ -229,6 +267,25 @@ static int setup_partitions(VP8Context *s, const uint8_t *buf, int buf_size) return 0; } +static void vp7_get_quants(VP8Context *s) +{ + VP56RangeCoder *c = &s->c; + + int yac_qi = vp8_rac_get_uint(c, 7); + int ydc_qi = vp8_rac_get(c) ? vp8_rac_get_uint(c, 7) : yac_qi; + int y2dc_qi = vp8_rac_get(c) ? vp8_rac_get_uint(c, 7) : yac_qi; + int y2ac_qi = vp8_rac_get(c) ? vp8_rac_get_uint(c, 7) : yac_qi; + int uvdc_qi = vp8_rac_get(c) ? vp8_rac_get_uint(c, 7) : yac_qi; + int uvac_qi = vp8_rac_get(c) ? vp8_rac_get_uint(c, 7) : yac_qi; + + s->qmat[0].luma_qmul[0] = vp7_ydc_qlookup[ydc_qi]; + s->qmat[0].luma_qmul[1] = vp7_yac_qlookup[yac_qi]; + s->qmat[0].luma_dc_qmul[0] = vp7_y2dc_qlookup[y2dc_qi]; + s->qmat[0].luma_dc_qmul[1] = vp7_y2ac_qlookup[y2ac_qi]; + s->qmat[0].chroma_qmul[0] = FFMIN(vp7_ydc_qlookup[uvdc_qi], 132); + s->qmat[0].chroma_qmul[1] = vp7_yac_qlookup[uvac_qi]; +} + static void get_quants(VP8Context *s) { VP56RangeCoder *c = &s->c; @@ -291,6 +348,54 @@ static VP56Frame ref_to_update(VP8Context *s, int update, VP56Frame ref) return VP56_FRAME_NONE; } +static void vp78_reset_probability_tables(VP8Context *s) +{ + int i, j; + for (i = 0; i < 4; i++) + for (j = 0; j < 16; j++) + memcpy(s->prob->token[i][j], vp8_token_default_probs[i][vp8_coeff_band[j]], + sizeof(s->prob->token[i][j])); +} + +static void vp78_update_probability_tables(VP8Context *s) +{ + VP56RangeCoder *c = &s->c; + int i, j, k, l, m; + + for (i = 0; i < 4; i++) + for (j = 0; j < 8; j++) + for (k = 0; k < 3; k++) + for (l = 0; l < NUM_DCT_TOKENS-1; l++) + if (vp56_rac_get_prob_branchy(c, vp8_token_update_probs[i][j][k][l])) { + int prob = vp8_rac_get_uint(c, 8); + for (m = 0; vp8_coeff_band_indexes[j][m] >= 0; m++) + s->prob->token[i][vp8_coeff_band_indexes[j][m]][k][l] = prob; + } +} + +#define VP7_MVC_SIZE 17 +#define VP8_MVC_SIZE 19 + +static void vp78_update_pred16x16_pred8x8_mvc_probabilities(VP8Context *s, + int mvc_size) +{ + VP56RangeCoder *c = &s->c; + int i, j; + + if (vp8_rac_get(c)) + for (i = 0; i < 4; i++) + s->prob->pred16x16[i] = vp8_rac_get_uint(c, 8); + if (vp8_rac_get(c)) + for (i = 0; i < 3; i++) + s->prob->pred8x8c[i] = vp8_rac_get_uint(c, 8); + + // 17.2 MV probability update + for (i = 0; i < 2; i++) + for (j = 0; j < mvc_size; j++) + if (vp56_rac_get_prob_branchy(c, vp8_mv_update_prob[i][j])) + s->prob->mvc[i][j] = vp8_rac_get_nn(c); +} + static void update_refs(VP8Context *s) { VP56RangeCoder *c = &s->c; @@ -302,10 +407,211 @@ static void update_refs(VP8Context *s) s->update_altref = ref_to_update(s, update_altref, VP56_FRAME_GOLDEN2); } -static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size) +static void copy_luma(AVFrame *dst, AVFrame *src, int width, int height) +{ + int i, j; + + for (j = 1; j < 3; j++) { + for (i = 0; i < height / 2; i++) + memcpy(dst->data[j] + i * dst->linesize[j], + src->data[j] + i * src->linesize[j], width / 2); + } +} + +static void fade(uint8_t *dst, uint8_t *src, + int width, int height, int linesize, + int alpha, int beta) +{ + int i, j; + + for (j = 0; j < height; j++) { + for (i = 0; i < width; i++) { + uint8_t y = src[j * linesize + i]; + dst[j * linesize + i] = av_clip_uint8(y + ((y * beta) >> 8) + alpha); + } + } +} + +static int vp7_fade_frame(VP8Context *s, VP56RangeCoder *c) +{ + int alpha = (int8_t) vp8_rac_get_uint(c, 8); + int beta = (int8_t) vp8_rac_get_uint(c, 8); + int ret; + + if (!s->keyframe && (alpha || beta)) { + int width = s->mb_width * 16; + int height = s->mb_height * 16; + AVFrame *src, *dst; + + if (!s->framep[VP56_FRAME_PREVIOUS]) + return AVERROR_INVALIDDATA; + + dst = + src = s->framep[VP56_FRAME_PREVIOUS]->tf.f; + + /* preserve the golden frame, write a new previous frame */ + if (s->framep[VP56_FRAME_GOLDEN] == s->framep[VP56_FRAME_PREVIOUS]) { + s->framep[VP56_FRAME_PREVIOUS] = vp8_find_free_buffer(s); + if ((ret = vp8_alloc_frame(s, s->framep[VP56_FRAME_PREVIOUS], 1)) < 0) + return ret; + + dst = s->framep[VP56_FRAME_PREVIOUS]->tf.f; + + copy_luma(dst, src, width, height); + } + + fade(dst->data[0], src->data[0], + width, height, dst->linesize[0], alpha, beta); + } + + return 0; +} + +static int vp7_decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size) { VP56RangeCoder *c = &s->c; - int header_size, hscale, vscale, i, j, k, l, m, ret; + int part1_size, hscale, vscale, i, j, ret; + int width = s->avctx->width; + int height = s->avctx->height; + + s->profile = (buf[0] >> 1) & 7; + if (s->profile > 1) { + avpriv_request_sample(s->avctx, "Unknown profile %d", s->profile); + return AVERROR_INVALIDDATA; + } + + s->keyframe = !(buf[0] & 1); + s->invisible = 0; + part1_size = AV_RL24(buf) >> 4; + + buf += 4 - s->profile; + buf_size -= 4 - s->profile; + + memcpy(s->put_pixels_tab, s->vp8dsp.put_vp8_epel_pixels_tab, sizeof(s->put_pixels_tab)); + + ff_vp56_init_range_decoder(c, buf, part1_size); + buf += part1_size; + buf_size -= part1_size; + + /* A. Dimension information (keyframes only) */ + if (s->keyframe) { + width = vp8_rac_get_uint(c, 12); + height = vp8_rac_get_uint(c, 12); + hscale = vp8_rac_get_uint(c, 2); + vscale = vp8_rac_get_uint(c, 2); + if (hscale || vscale) + avpriv_request_sample(s->avctx, "Upscaling"); + + s->update_golden = s->update_altref = VP56_FRAME_CURRENT; + vp78_reset_probability_tables(s); + memcpy(s->prob->pred16x16, vp8_pred16x16_prob_inter, + sizeof(s->prob->pred16x16)); + memcpy(s->prob->pred8x8c, vp8_pred8x8c_prob_inter, + sizeof(s->prob->pred8x8c)); + for (i = 0; i < 2; i++) + memcpy(s->prob->mvc[i], vp7_mv_default_prob[i], + sizeof(vp7_mv_default_prob[i])); + memset(&s->segmentation, 0, sizeof(s->segmentation)); + memset(&s->lf_delta, 0, sizeof(s->lf_delta)); + memcpy(s->prob[0].scan, zigzag_scan, sizeof(s->prob[0].scan)); + } + + if (s->keyframe || s->profile > 0) + memset(s->inter_dc_pred, 0 , sizeof(s->inter_dc_pred)); + + /* B. Decoding information for all four macroblock-level features */ + for (i = 0; i < 4; i++) { + s->feature_enabled[i] = vp8_rac_get(c); + if (s->feature_enabled[i]) { + s->feature_present_prob[i] = vp8_rac_get_uint(c, 8); + + for (j = 0; j < 3; j++) + s->feature_index_prob[i][j] = + vp8_rac_get(c) ? vp8_rac_get_uint(c, 8) : 255; + + if (vp7_feature_value_size[i]) + for (j = 0; j < 4; j++) + s->feature_value[i][j] = + vp8_rac_get(c) ? vp8_rac_get_uint(c, vp7_feature_value_size[s->profile][i]) : 0; + } + } + + s->segmentation.enabled = 0; + s->segmentation.update_map = 0; + s->lf_delta.enabled = 0; + + s->num_coeff_partitions = 1; + ff_vp56_init_range_decoder(&s->coeff_partition[0], buf, buf_size); + + if (!s->macroblocks_base || /* first frame */ + width != s->avctx->width || height != s->avctx->height || + (width + 15) / 16 != s->mb_width || (height + 15) / 16 != s->mb_height) { + if ((ret = vp7_update_dimensions(s, width, height)) < 0) + return ret; + } + + /* C. Dequantization indices */ + vp7_get_quants(s); + + /* D. Golden frame update flag (a Flag) for interframes only */ + if (!s->keyframe) { + s->update_golden = vp8_rac_get(c) ? VP56_FRAME_CURRENT : VP56_FRAME_NONE; + s->sign_bias[VP56_FRAME_GOLDEN] = 0; + } + + s->update_last = 1; + s->update_probabilities = 1; + s->fade_present = 1; + + if (s->profile > 0) { + s->update_probabilities = vp8_rac_get(c); + if (!s->update_probabilities) + s->prob[1] = s->prob[0]; + + if (!s->keyframe) + s->fade_present = vp8_rac_get(c); + } + + /* E. Fading information for previous frame */ + if (s->fade_present && vp8_rac_get(c)) { + if ((ret = vp7_fade_frame(s ,c)) < 0) + return ret; + } + + /* F. Loop filter type */ + if (!s->profile) + s->filter.simple = vp8_rac_get(c); + + /* G. DCT coefficient ordering specification */ + if (vp8_rac_get(c)) + for (i = 1; i < 16; i++) + s->prob[0].scan[i] = zigzag_scan[vp8_rac_get_uint(c, 4)]; + + /* H. Loop filter levels */ + if (s->profile > 0) + s->filter.simple = vp8_rac_get(c); + s->filter.level = vp8_rac_get_uint(c, 6); + s->filter.sharpness = vp8_rac_get_uint(c, 3); + + /* I. DCT coefficient probability update; 13.3 Token Probability Updates */ + vp78_update_probability_tables(s); + + s->mbskip_enabled = 0; + + /* J. The remaining frame header data occurs ONLY FOR INTERFRAMES */ + if (!s->keyframe) { + s->prob->intra = vp8_rac_get_uint(c, 8); + s->prob->last = vp8_rac_get_uint(c, 8); + vp78_update_pred16x16_pred8x8_mvc_probabilities(s, VP7_MVC_SIZE); + } + + return 0; +} + +static int vp8_decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size) +{ + VP56RangeCoder *c = &s->c; + int header_size, hscale, vscale, ret; int width = s->avctx->width; int height = s->avctx->height; @@ -348,11 +654,7 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size) avpriv_request_sample(s->avctx, "Upscaling"); s->update_golden = s->update_altref = VP56_FRAME_CURRENT; - for (i = 0; i < 4; i++) - for (j = 0; j < 16; j++) - memcpy(s->prob->token[i][j], - vp8_token_default_probs[i][vp8_coeff_band[j]], - sizeof(s->prob->token[i][j])); + vp78_reset_probability_tables(s); memcpy(s->prob->pred16x16, vp8_pred16x16_prob_inter, sizeof(s->prob->pred16x16)); memcpy(s->prob->pred8x8c, vp8_pred8x8c_prob_inter, @@ -393,7 +695,7 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size) if (!s->macroblocks_base || /* first frame */ width != s->avctx->width || height != s->avctx->height) - if ((ret = update_dimensions(s, width, height)) < 0) + if ((ret = vp8_update_dimensions(s, width, height)) < 0) return ret; get_quants(s); @@ -411,15 +713,7 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size) s->update_last = s->keyframe || vp8_rac_get(c); - for (i = 0; i < 4; i++) - for (j = 0; j < 8; j++) - for (k = 0; k < 3; k++) - for (l = 0; l < NUM_DCT_TOKENS - 1; l++) - if (vp56_rac_get_prob_branchy(c, vp8_token_update_probs[i][j][k][l])) { - int prob = vp8_rac_get_uint(c, 8); - for (m = 0; vp8_coeff_band_indexes[j][m] >= 0; m++) - s->prob->token[i][vp8_coeff_band_indexes[j][m]][k][l] = prob; - } + vp78_update_probability_tables(s); if ((s->mbskip_enabled = vp8_rac_get(c))) s->prob->mbskip = vp8_rac_get_uint(c, 8); @@ -428,19 +722,7 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size) s->prob->intra = vp8_rac_get_uint(c, 8); s->prob->last = vp8_rac_get_uint(c, 8); s->prob->golden = vp8_rac_get_uint(c, 8); - - if (vp8_rac_get(c)) - for (i = 0; i < 4; i++) - s->prob->pred16x16[i] = vp8_rac_get_uint(c, 8); - if (vp8_rac_get(c)) - for (i = 0; i < 3; i++) - s->prob->pred8x8c[i] = vp8_rac_get_uint(c, 8); - - // 17.2 MV probability update - for (i = 0; i < 2; i++) - for (j = 0; j < 19; j++) - if (vp56_rac_get_prob_branchy(c, vp8_mv_update_prob[i][j])) - s->prob->mvc[i][j] = vp8_rac_get_nn(c); + vp78_update_pred16x16_pred8x8_mvc_probabilities(s, VP8_MVC_SIZE); } return 0; @@ -456,7 +738,7 @@ void clamp_mv(VP8Context *s, VP56mv *dst, const VP56mv *src) /** * Motion vector coding, 17.1. */ -static int read_mv_component(VP56RangeCoder *c, const uint8_t *p) +static int read_mv_component(VP56RangeCoder *c, const uint8_t *p, int vp7) { int bit, x = 0; @@ -465,9 +747,9 @@ static int read_mv_component(VP56RangeCoder *c, const uint8_t *p) for (i = 0; i < 3; i++) x += vp56_rac_get_prob(c, p[9 + i]) << i; - for (i = 9; i > 3; i--) + for (i = (vp7 ? 7 : 9); i > 3; i--) x += vp56_rac_get_prob(c, p[9 + i]) << i; - if (!(x & 0xFFF0) || vp56_rac_get_prob(c, p[12])) + if (!(x & (vp7 ? 0xF0 : 0xFFF0)) || vp56_rac_get_prob(c, p[12])) x += 8; } else { // small_mvtree @@ -485,8 +767,11 @@ static int read_mv_component(VP56RangeCoder *c, const uint8_t *p) } static av_always_inline -const uint8_t *get_submv_prob(uint32_t left, uint32_t top) +const uint8_t *get_submv_prob(uint32_t left, uint32_t top, int is_vp7) { + if (is_vp7) + return vp7_submv_prob; + if (left == top) return vp8_submv_prob[4 - !!left]; if (!top) @@ -499,7 +784,8 @@ const uint8_t *get_submv_prob(uint32_t left, uint32_t top) * @returns the number of motion vectors parsed (2, 4 or 16) */ static av_always_inline -int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, int layout) +int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, + int layout, int is_vp7) { int part_idx; int n, num; @@ -546,13 +832,15 @@ int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, int lay else above = AV_RN32A(&cur_mv[mbsplits_cur[k - 4]]); - submv_prob = get_submv_prob(left, above); + submv_prob = get_submv_prob(left, above, is_vp7); if (vp56_rac_get_prob_branchy(c, submv_prob[0])) { if (vp56_rac_get_prob_branchy(c, submv_prob[1])) { if (vp56_rac_get_prob_branchy(c, submv_prob[2])) { - mb->bmv[n].y = mb->mv.y + read_mv_component(c, s->prob->mvc[0]); - mb->bmv[n].x = mb->mv.x + read_mv_component(c, s->prob->mvc[1]); + mb->bmv[n].y = mb->mv.y + + read_mv_component(c, s->prob->mvc[0], is_vp7); + mb->bmv[n].x = mb->mv.x + + read_mv_component(c, s->prob->mvc[1], is_vp7); } else { AV_ZERO32(&mb->bmv[n]); } @@ -567,9 +855,129 @@ int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, int lay return num; } +/** + * The vp7 reference decoder uses a padding macroblock column (added to right + * edge of the frame) to guard against illegal macroblock offsets. The + * algorithm has bugs that permit offsets to straddle the padding column. + * This function replicates those bugs. + * + * @param[out] edge_x macroblock x address + * @param[out] edge_y macroblock y address + * + * @return macroblock offset legal (boolean) + */ +static int vp7_calculate_mb_offset(int mb_x, int mb_y, int mb_width, + int xoffset, int yoffset, int boundary, + int *edge_x, int *edge_y) +{ + int vwidth = mb_width + 1; + int new = (mb_y + yoffset) * vwidth + mb_x + xoffset; + if (new < boundary || new % vwidth == vwidth - 1) + return 0; + *edge_y = new / vwidth; + *edge_x = new % vwidth; + return 1; +} + +static const VP56mv *get_bmv_ptr(const VP8Macroblock *mb, int subblock) +{ + return &mb->bmv[mb->mode == VP8_MVMODE_SPLIT ? vp8_mbsplits[mb->partitioning][subblock] : 0]; +} + static av_always_inline -void decode_mvs(VP8Context *s, VP8Macroblock *mb, - int mb_x, int mb_y, int layout) +void vp7_decode_mvs(VP8Context *s, VP8Macroblock *mb, + int mb_x, int mb_y, int layout) +{ + VP8Macroblock *mb_edge[12]; + enum { CNT_ZERO, CNT_NEAREST, CNT_NEAR }; + enum { VP8_EDGE_TOP, VP8_EDGE_LEFT, VP8_EDGE_TOPLEFT }; + int idx = CNT_ZERO; + VP56mv near_mv[3]; + uint8_t cnt[3] = { 0 }; + VP56RangeCoder *c = &s->c; + int i; + + AV_ZERO32(&near_mv[0]); + AV_ZERO32(&near_mv[1]); + AV_ZERO32(&near_mv[2]); + + for (i = 0; i < VP7_MV_PRED_COUNT; i++) { + const VP7MVPred * pred = &vp7_mv_pred[i]; + int edge_x, edge_y; + + if (vp7_calculate_mb_offset(mb_x, mb_y, s->mb_width, pred->xoffset, + pred->yoffset, !s->profile, &edge_x, &edge_y)) { + VP8Macroblock *edge = mb_edge[i] = (s->mb_layout == 1) + ? s->macroblocks_base + 1 + edge_x + + (s->mb_width + 1) * (edge_y + 1) + : s->macroblocks + edge_x + + (s->mb_height - edge_y - 1) * 2; + uint32_t mv = AV_RN32A(get_bmv_ptr(edge, vp7_mv_pred[i].subblock)); + if (mv) { + if (AV_RN32A(&near_mv[CNT_NEAREST])) { + if (mv == AV_RN32A(&near_mv[CNT_NEAREST])) { + idx = CNT_NEAREST; + } else if (AV_RN32A(&near_mv[CNT_NEAR])) { + if (mv != AV_RN32A(&near_mv[CNT_NEAR])) + continue; + idx = CNT_NEAR; + } else { + AV_WN32A(&near_mv[CNT_NEAR], mv); + idx = CNT_NEAR; + } + } else { + AV_WN32A(&near_mv[CNT_NEAREST], mv); + idx = CNT_NEAREST; + } + } else { + idx = CNT_ZERO; + } + } else { + idx = CNT_ZERO; + } + cnt[idx] += vp7_mv_pred[i].score; + } + + mb->partitioning = VP8_SPLITMVMODE_NONE; + + if (vp56_rac_get_prob_branchy(c, vp7_mode_contexts[cnt[CNT_ZERO]][0])) { + mb->mode = VP8_MVMODE_MV; + + if (vp56_rac_get_prob_branchy(c, vp7_mode_contexts[cnt[CNT_NEAREST]][1])) { + + if (vp56_rac_get_prob_branchy(c, vp7_mode_contexts[cnt[CNT_NEAR]][2])) { + + if (cnt[CNT_NEAREST] > cnt[CNT_NEAR]) + AV_WN32A(&mb->mv, cnt[CNT_ZERO] > cnt[CNT_NEAREST] ? 0 : AV_RN32A(&near_mv[CNT_NEAREST])); + else + AV_WN32A(&mb->mv, cnt[CNT_ZERO] > cnt[CNT_NEAR] ? 0 : AV_RN32A(&near_mv[CNT_NEAR])); + + if (vp56_rac_get_prob_branchy(c, vp7_mode_contexts[cnt[CNT_NEAR]][3])) { + mb->mode = VP8_MVMODE_SPLIT; + mb->mv = mb->bmv[decode_splitmvs(s, c, mb, layout, IS_VP7) - 1]; + } else { + mb->mv.y += read_mv_component(c, s->prob->mvc[0], IS_VP7); + mb->mv.x += read_mv_component(c, s->prob->mvc[1], IS_VP7); + mb->bmv[0] = mb->mv; + } + } else { + mb->mv = near_mv[CNT_NEAR]; + mb->bmv[0] = mb->mv; + } + } else { + mb->mv = near_mv[CNT_NEAREST]; + mb->bmv[0] = mb->mv; + } + } else { + mb->mode = VP8_MVMODE_ZERO; + AV_ZERO32(&mb->mv); + mb->bmv[0] = mb->mv; + } +} + +static av_always_inline +void vp8_decode_mvs(VP8Context *s, VP8Macroblock *mb, + int mb_x, int mb_y, int layout) { VP8Macroblock *mb_edge[3] = { 0 /* top */, mb - 1 /* left */, @@ -646,10 +1054,10 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, if (vp56_rac_get_prob_branchy(c, vp8_mode_contexts[cnt[CNT_SPLITMV]][3])) { mb->mode = VP8_MVMODE_SPLIT; - mb->mv = mb->bmv[decode_splitmvs(s, c, mb, layout) - 1]; + mb->mv = mb->bmv[decode_splitmvs(s, c, mb, layout, IS_VP8) - 1]; } else { - mb->mv.y += read_mv_component(c, s->prob->mvc[0]); - mb->mv.x += read_mv_component(c, s->prob->mvc[1]); + mb->mv.y += read_mv_component(c, s->prob->mvc[0], IS_VP8); + mb->mv.x += read_mv_component(c, s->prob->mvc[1], IS_VP8); mb->bmv[0] = mb->mv; } } else { @@ -704,11 +1112,28 @@ void decode_intra4x4_modes(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, static av_always_inline void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, - uint8_t *segment, uint8_t *ref, int layout) + uint8_t *segment, uint8_t *ref, int layout, int is_vp7) { VP56RangeCoder *c = &s->c; - - if (s->segmentation.update_map) + const char *vp7_feature_name[] = { "q-index", + "lf-delta", + "partial-golden-update", + "blit-pitch" }; + if (is_vp7) { + int i; + *segment = 0; + for (i = 0; i < 4; i++) { + if (s->feature_enabled[i]) { + if (vp56_rac_get_prob(c, s->feature_present_prob[i])) { + int index = vp8_rac_get_tree(c, vp7_feature_index_tree, + s->feature_index_prob[i]); + av_log(s->avctx, AV_LOG_WARNING, + "Feature %s present in macroblock (value 0x%x)\n", + vp7_feature_name[i], s->feature_value[i][index]); + } + } + } + } else if (s->segmentation.update_map) *segment = vp8_rac_get_tree(c, vp8_segmentid_tree, s->prob->segmentid); else if (s->segmentation.enabled) *segment = ref ? *ref : *segment; @@ -723,7 +1148,8 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, if (mb->mode == MODE_I4x4) { decode_intra4x4_modes(s, c, mb, mb_x, 1, layout); } else { - const uint32_t modes = vp8_pred4x4_mode[mb->mode] * 0x01010101u; + const uint32_t modes = (is_vp7 ? vp7_pred4x4_mode + : vp8_pred4x4_mode)[mb->mode] * 0x01010101u; if (s->mb_layout == 1) AV_WN32A(mb->intra4x4_pred_mode_top, modes); else @@ -738,14 +1164,17 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, // inter MB, 16.2 if (vp56_rac_get_prob_branchy(c, s->prob->last)) mb->ref_frame = - vp56_rac_get_prob(c, s->prob->golden) ? VP56_FRAME_GOLDEN2 /* altref */ - : VP56_FRAME_GOLDEN; + (!is_vp7 && vp56_rac_get_prob(c, s->prob->golden)) ? VP56_FRAME_GOLDEN2 /* altref */ + : VP56_FRAME_GOLDEN; else mb->ref_frame = VP56_FRAME_PREVIOUS; s->ref_count[mb->ref_frame - 1]++; // motion vectors, 16.3 - decode_mvs(s, mb, mb_x, mb_y, layout); + if (is_vp7) + vp7_decode_mvs(s, mb, mb_x, mb_y, layout); + else + vp8_decode_mvs(s, mb, mb_x, mb_y, layout); } else { // intra MB, 16.1 mb->mode = vp8_rac_get_tree(c, vp8_pred16x16_tree_inter, s->prob->pred16x16); @@ -761,7 +1190,6 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, } } -#ifndef decode_block_coeffs_internal /** * @param r arithmetic bitstream reader context * @param block destination for block coefficients @@ -772,15 +1200,17 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, * @return 0 if no coeffs were decoded * otherwise, the index of the last coeff decoded plus one */ -static int decode_block_coeffs_internal(VP56RangeCoder *r, int16_t block[16], - uint8_t probs[16][3][NUM_DCT_TOKENS - 1], - int i, uint8_t *token_prob, - int16_t qmul[2]) +static av_always_inline +int decode_block_coeffs_internal(VP56RangeCoder *r, int16_t block[16], + uint8_t probs[16][3][NUM_DCT_TOKENS - 1], + int i, uint8_t *token_prob, int16_t qmul[2], + const uint8_t scan[16], int vp7) { VP56RangeCoder c = *r; goto skip_eob; do { int coeff; +restart: if (!vp56_rac_get_prob_branchy(&c, token_prob[0])) // DCT_EOB break; @@ -789,6 +1219,8 @@ skip_eob: if (++i == 16) break; // invalid input; blocks should end with EOB token_prob = probs[i][0]; + if (vp7) + goto restart; goto skip_eob; } @@ -821,12 +1253,57 @@ skip_eob: } token_prob = probs[i + 1][2]; } - block[zigzag_scan[i]] = (vp8_rac_get(&c) ? -coeff : coeff) * qmul[!!i]; + block[scan[i]] = (vp8_rac_get(&c) ? -coeff : coeff) * qmul[!!i]; } while (++i < 16); *r = c; return i; } + +static av_always_inline +int inter_predict_dc(int16_t block[16], int16_t pred[2]) +{ + int16_t dc = block[0]; + int ret = 0; + + if (pred[1] > 3) { + dc += pred[0]; + ret = 1; + } + + if (!pred[0] | !dc | ((int32_t)pred[0] ^ (int32_t)dc) >> 31) { + block[0] = pred[0] = dc; + pred[1] = 0; + } else { + if (pred[0] == dc) + pred[1]++; + block[0] = pred[0] = dc; + } + + return ret; +} + +static int vp7_decode_block_coeffs_internal(VP56RangeCoder *r, + int16_t block[16], + uint8_t probs[16][3][NUM_DCT_TOKENS - 1], + int i, uint8_t *token_prob, + int16_t qmul[2], + const uint8_t scan[16]) +{ + return decode_block_coeffs_internal(r, block, probs, i, + token_prob, qmul, scan, IS_VP7); +} + +#ifndef vp8_decode_block_coeffs_internal +static int vp8_decode_block_coeffs_internal(VP56RangeCoder *r, + int16_t block[16], + uint8_t probs[16][3][NUM_DCT_TOKENS - 1], + int i, uint8_t *token_prob, + int16_t qmul[2]) +{ + return decode_block_coeffs_internal(r, block, probs, i, + token_prob, qmul, zigzag_scan, IS_VP8); +} #endif /** @@ -844,30 +1321,42 @@ skip_eob: static av_always_inline int decode_block_coeffs(VP56RangeCoder *c, int16_t block[16], uint8_t probs[16][3][NUM_DCT_TOKENS - 1], - int i, int zero_nhood, int16_t qmul[2]) + int i, int zero_nhood, int16_t qmul[2], + const uint8_t scan[16], int vp7) { uint8_t *token_prob = probs[i][zero_nhood]; if (!vp56_rac_get_prob_branchy(c, token_prob[0])) // DCT_EOB return 0; - return decode_block_coeffs_internal(c, block, probs, i, token_prob, qmul); + return vp7 ? vp7_decode_block_coeffs_internal(c, block, probs, i, + token_prob, qmul, scan) + : vp8_decode_block_coeffs_internal(c, block, probs, i, + token_prob, qmul); } static av_always_inline void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, - VP8Macroblock *mb, uint8_t t_nnz[9], uint8_t l_nnz[9]) + VP8Macroblock *mb, uint8_t t_nnz[9], uint8_t l_nnz[9], + int is_vp7) { int i, x, y, luma_start = 0, luma_ctx = 3; int nnz_pred, nnz, nnz_total = 0; int segment = mb->segment; int block_dc = 0; - if (mb->mode != MODE_I4x4 && mb->mode != VP8_MVMODE_SPLIT) { + if (mb->mode != MODE_I4x4 && (is_vp7 || mb->mode != VP8_MVMODE_SPLIT)) { nnz_pred = t_nnz[8] + l_nnz[8]; // decode DC values and do hadamard nnz = decode_block_coeffs(c, td->block_dc, s->prob->token[1], 0, - nnz_pred, s->qmat[segment].luma_dc_qmul); + nnz_pred, s->qmat[segment].luma_dc_qmul, + zigzag_scan, is_vp7); l_nnz[8] = t_nnz[8] = !!nnz; + + if (is_vp7 && mb->mode > MODE_I4x4) { + nnz |= inter_predict_dc(td->block_dc, + s->inter_dc_pred[mb->ref_frame - 1]); + } + if (nnz) { nnz_total += nnz; block_dc = 1; @@ -887,7 +1376,8 @@ void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, nnz = decode_block_coeffs(c, td->block[y][x], s->prob->token[luma_ctx], luma_start, nnz_pred, - s->qmat[segment].luma_qmul); + s->qmat[segment].luma_qmul, + s->prob[0].scan, is_vp7); /* nnz+block_dc may be one more than the actual last index, * but we don't care */ td->non_zero_count_cache[y][x] = nnz + block_dc; @@ -902,13 +1392,13 @@ void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, for (y = 0; y < 2; y++) for (x = 0; x < 2; x++) { nnz_pred = l_nnz[i + 2 * y] + t_nnz[i + 2 * x]; - nnz = decode_block_coeffs(c, td->block[i][(y << 1) + x], - s->prob->token[2], - 0, nnz_pred, - s->qmat[segment].chroma_qmul); + nnz = decode_block_coeffs(c, td->block[i][(y << 1) + x], + s->prob->token[2], 0, nnz_pred, + s->qmat[segment].chroma_qmul, + s->prob[0].scan, is_vp7); td->non_zero_count_cache[i][(y << 1) + x] = nnz; t_nnz[i + 2 * x] = l_nnz[i + 2 * y] = !!nnz; - nnz_total += nnz; + nnz_total += nnz; } // if there were no coded coeffs despite the macroblock not being marked skip, @@ -974,42 +1464,43 @@ int check_dc_pred8x8_mode(int mode, int mb_x, int mb_y) } static av_always_inline -int check_tm_pred8x8_mode(int mode, int mb_x, int mb_y) +int check_tm_pred8x8_mode(int mode, int mb_x, int mb_y, int vp7) { if (!mb_x) - return mb_y ? VERT_PRED8x8 : DC_129_PRED8x8; + return mb_y ? VERT_PRED8x8 : (vp7 ? DC_128_PRED8x8 : DC_129_PRED8x8); else return mb_y ? mode : HOR_PRED8x8; } static av_always_inline -int check_intra_pred8x8_mode_emuedge(int mode, int mb_x, int mb_y) +int check_intra_pred8x8_mode_emuedge(int mode, int mb_x, int mb_y, int vp7) { switch (mode) { case DC_PRED8x8: return check_dc_pred8x8_mode(mode, mb_x, mb_y); case VERT_PRED8x8: - return !mb_y ? DC_127_PRED8x8 : mode; + return !mb_y ? (vp7 ? DC_128_PRED8x8 : DC_127_PRED8x8) : mode; case HOR_PRED8x8: - return !mb_x ? DC_129_PRED8x8 : mode; + return !mb_x ? (vp7 ? DC_128_PRED8x8 : DC_129_PRED8x8) : mode; case PLANE_PRED8x8: /* TM */ - return check_tm_pred8x8_mode(mode, mb_x, mb_y); + return check_tm_pred8x8_mode(mode, mb_x, mb_y, vp7); } return mode; } static av_always_inline -int check_tm_pred4x4_mode(int mode, int mb_x, int mb_y) +int check_tm_pred4x4_mode(int mode, int mb_x, int mb_y, int vp7) { if (!mb_x) { - return mb_y ? VERT_VP8_PRED : DC_129_PRED; + return mb_y ? VERT_VP8_PRED : (vp7 ? DC_128_PRED : DC_129_PRED); } else { return mb_y ? mode : HOR_VP8_PRED; } } static av_always_inline -int check_intra_pred4x4_mode_emuedge(int mode, int mb_x, int mb_y, int *copy_buf) +int check_intra_pred4x4_mode_emuedge(int mode, int mb_x, int mb_y, + int *copy_buf, int vp7) { switch (mode) { case VERT_PRED: @@ -1020,7 +1511,7 @@ int check_intra_pred4x4_mode_emuedge(int mode, int mb_x, int mb_y, int *copy_buf /* fall-through */ case DIAG_DOWN_LEFT_PRED: case VERT_LEFT_PRED: - return !mb_y ? DC_127_PRED : mode; + return !mb_y ? (vp7 ? DC_128_PRED : DC_127_PRED) : mode; case HOR_PRED: if (!mb_y) { *copy_buf = 1; @@ -1028,9 +1519,9 @@ int check_intra_pred4x4_mode_emuedge(int mode, int mb_x, int mb_y, int *copy_buf } /* fall-through */ case HOR_UP_PRED: - return !mb_x ? DC_129_PRED : mode; + return !mb_x ? (vp7 ? DC_128_PRED : DC_129_PRED) : mode; case TM_VP8_PRED: - return check_tm_pred4x4_mode(mode, mb_x, mb_y); + return check_tm_pred4x4_mode(mode, mb_x, mb_y, vp7); case DC_PRED: /* 4x4 DC doesn't use the same "H.264-style" exceptions * as 16x16/8x8 DC */ case DIAG_DOWN_RIGHT_PRED: @@ -1045,7 +1536,7 @@ int check_intra_pred4x4_mode_emuedge(int mode, int mb_x, int mb_y, int *copy_buf static av_always_inline void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3], - VP8Macroblock *mb, int mb_x, int mb_y) + VP8Macroblock *mb, int mb_x, int mb_y, int is_vp7) { int x, y, mode, nnz; uint32_t tr; @@ -1058,12 +1549,14 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3], s->filter.simple, 1); if (mb->mode < MODE_I4x4) { - mode = check_intra_pred8x8_mode_emuedge(mb->mode, mb_x, mb_y); + mode = check_intra_pred8x8_mode_emuedge(mb->mode, mb_x, mb_y, is_vp7); s->hpc.pred16x16[mode](dst[0], s->linesize); } else { uint8_t *ptr = dst[0]; uint8_t *intra4x4 = mb->intra4x4_pred_mode_mb; - uint8_t tr_top[4] = { 127, 127, 127, 127 }; + const uint8_t lo = is_vp7 ? 128 : 127; + const uint8_t hi = is_vp7 ? 128 : 129; + uint8_t tr_top[4] = { lo, lo, lo, lo }; // all blocks on the right edge of the macroblock use bottom edge // the top macroblock for their topright edge @@ -1091,19 +1584,18 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3], } else if (x == 3) topright = tr_right; - mode = check_intra_pred4x4_mode_emuedge(intra4x4[x], - mb_x + x, mb_y + y, - ©); + mode = check_intra_pred4x4_mode_emuedge(intra4x4[x], mb_x + x, + mb_y + y, ©, is_vp7); if (copy) { dst = copy_dst + 12; linesize = 8; if (!(mb_y + y)) { - copy_dst[3] = 127U; - AV_WN32A(copy_dst + 4, 127U * 0x01010101U); + copy_dst[3] = lo; + AV_WN32A(copy_dst + 4, lo * 0x01010101U); } else { AV_COPY32(copy_dst + 4, ptr + 4 * x - s->linesize); if (!(mb_x + x)) { - copy_dst[3] = 129U; + copy_dst[3] = hi; } else { copy_dst[3] = ptr[4 * x - s->linesize - 1]; } @@ -1112,7 +1604,7 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3], copy_dst[11] = copy_dst[19] = copy_dst[27] = - copy_dst[35] = 129U; + copy_dst[35] = hi; } else { copy_dst[11] = ptr[4 * x - 1]; copy_dst[19] = ptr[4 * x + s->linesize - 1]; @@ -1145,7 +1637,8 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3], } } - mode = check_intra_pred8x8_mode_emuedge(mb->chroma_pred_mode, mb_x, mb_y); + mode = check_intra_pred8x8_mode_emuedge(mb->chroma_pred_mode, + mb_x, mb_y, is_vp7); s->hpc.pred8x8[mode](dst[1], s->uvlinesize); s->hpc.pred8x8[mode](dst[2], s->uvlinesize); @@ -1300,6 +1793,8 @@ void vp8_mc_part(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3], /* U/V */ if (s->profile == 3) { + /* this block only applies VP8; it is safe to check + * only the profile, as VP7 profile <= 1 */ uvmv.x &= ~7; uvmv.y &= ~7; } @@ -1490,7 +1985,8 @@ chroma_idct_end: } static av_always_inline -void filter_level_for_mb(VP8Context *s, VP8Macroblock *mb, VP8FilterStrength *f) +void filter_level_for_mb(VP8Context *s, VP8Macroblock *mb, + VP8FilterStrength *f, int is_vp7) { int interior_limit, filter_level; @@ -1517,15 +2013,15 @@ void filter_level_for_mb(VP8Context *s, VP8Macroblock *mb, VP8FilterStrength *f) f->filter_level = filter_level; f->inner_limit = interior_limit; - f->inner_filter = !mb->skip || mb->mode == MODE_I4x4 || + f->inner_filter = is_vp7 || !mb->skip || mb->mode == MODE_I4x4 || mb->mode == VP8_MVMODE_SPLIT; } static av_always_inline void filter_mb(VP8Context *s, uint8_t *dst[3], VP8FilterStrength *f, - int mb_x, int mb_y) + int mb_x, int mb_y, int is_vp7) { - int mbedge_lim, bedge_lim, hev_thresh; + int mbedge_lim, bedge_lim_y, bedge_lim_uv, hev_thresh; int filter_level = f->filter_level; int inner_limit = f->inner_limit; int inner_filter = f->inner_filter; @@ -1545,8 +2041,15 @@ void filter_mb(VP8Context *s, uint8_t *dst[3], VP8FilterStrength *f, if (!filter_level) return; - bedge_lim = 2 * filter_level + inner_limit; - mbedge_lim = bedge_lim + 4; + if (is_vp7) { + bedge_lim_y = filter_level; + bedge_lim_uv = filter_level * 2; + mbedge_lim = filter_level + 2; + } else { + bedge_lim_y = + bedge_lim_uv = filter_level * 2 + inner_limit; + mbedge_lim = bedge_lim_y + 4; + } hev_thresh = hev_thresh_lut[s->keyframe][filter_level]; @@ -1557,18 +2060,24 @@ void filter_mb(VP8Context *s, uint8_t *dst[3], VP8FilterStrength *f, mbedge_lim, inner_limit, hev_thresh); } - if (inner_filter) { - s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0] + 4, linesize, bedge_lim, - inner_limit, hev_thresh); - s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0] + 8, linesize, bedge_lim, - inner_limit, hev_thresh); - s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0] + 12, linesize, bedge_lim, - inner_limit, hev_thresh); - s->vp8dsp.vp8_h_loop_filter8uv_inner(dst[1] + 4, dst[2] + 4, - uvlinesize, bedge_lim, - inner_limit, hev_thresh); +#define H_LOOP_FILTER_16Y_INNER(cond) \ + if (cond && inner_filter) { \ + s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0] + 4, linesize, \ + bedge_lim_y, inner_limit, \ + hev_thresh); \ + s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0] + 8, linesize, \ + bedge_lim_y, inner_limit, \ + hev_thresh); \ + s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0] + 12, linesize, \ + bedge_lim_y, inner_limit, \ + hev_thresh); \ + s->vp8dsp.vp8_h_loop_filter8uv_inner(dst[1] + 4, dst[2] + 4, \ + uvlinesize, bedge_lim_uv, \ + inner_limit, hev_thresh); \ } + H_LOOP_FILTER_16Y_INNER(!is_vp7) + if (mb_y) { s->vp8dsp.vp8_v_loop_filter16y(dst[0], linesize, mbedge_lim, inner_limit, hev_thresh); @@ -1578,19 +2087,21 @@ void filter_mb(VP8Context *s, uint8_t *dst[3], VP8FilterStrength *f, if (inner_filter) { s->vp8dsp.vp8_v_loop_filter16y_inner(dst[0] + 4 * linesize, - linesize, bedge_lim, + linesize, bedge_lim_y, inner_limit, hev_thresh); s->vp8dsp.vp8_v_loop_filter16y_inner(dst[0] + 8 * linesize, - linesize, bedge_lim, + linesize, bedge_lim_y, inner_limit, hev_thresh); s->vp8dsp.vp8_v_loop_filter16y_inner(dst[0] + 12 * linesize, - linesize, bedge_lim, + linesize, bedge_lim_y, inner_limit, hev_thresh); s->vp8dsp.vp8_v_loop_filter8uv_inner(dst[1] + 4 * uvlinesize, dst[2] + 4 * uvlinesize, - uvlinesize, bedge_lim, + uvlinesize, bedge_lim_uv, inner_limit, hev_thresh); } + + H_LOOP_FILTER_16Y_INNER(is_vp7) } static av_always_inline @@ -1627,8 +2138,9 @@ void filter_mb_simple(VP8Context *s, uint8_t *dst, VP8FilterStrength *f, } #define MARGIN (16 << 2) -static void vp8_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *curframe, - VP8Frame *prev_frame) +static av_always_inline +void vp78_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *curframe, + VP8Frame *prev_frame, int is_vp7) { VP8Context *s = avctx->priv_data; int mb_x, mb_y; @@ -1650,7 +2162,7 @@ static void vp8_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *curframe, DC_PRED * 0x01010101); decode_mb_mode(s, mb, mb_x, mb_y, curframe->seg_map->data + mb_xy, prev_frame && prev_frame->seg_map ? - prev_frame->seg_map->data + mb_xy : NULL, 1); + prev_frame->seg_map->data + mb_xy : NULL, 1, is_vp7); s->mv_min.x -= 64; s->mv_max.x -= 64; } @@ -1659,6 +2171,18 @@ static void vp8_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *curframe, } } +static void vp7_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *cur_frame, + VP8Frame *prev_frame) +{ + vp78_decode_mv_mb_modes(avctx, cur_frame, prev_frame, IS_VP7); +} + +static void vp8_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *cur_frame, + VP8Frame *prev_frame) +{ + vp78_decode_mv_mb_modes(avctx, cur_frame, prev_frame, IS_VP8); +} + #if HAVE_THREADS #define check_thread_pos(td, otd, mb_x_check, mb_y_check) \ do { \ @@ -1700,7 +2224,7 @@ static void vp8_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *curframe, #endif static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata, - int jobnr, int threadnr) + int jobnr, int threadnr, int is_vp7) { VP8Context *s = avctx->priv_data; VP8ThreadData *prev_td, *next_td, *td = &s->thread_data[threadnr]; @@ -1736,7 +2260,8 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata, AV_WN32A(s->intra4x4_pred_mode_left, DC_PRED * 0x01010101); } - memset(td->left_nnz, 0, sizeof(td->left_nnz)); + if (!is_vp7 || mb_y == 0) + memset(td->left_nnz, 0, sizeof(td->left_nnz)); s->mv_min.x = -MARGIN; s->mv_max.x = ((s->mb_width - 1) << 6) + MARGIN; @@ -1745,10 +2270,13 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata, // Wait for previous thread to read mb_x+2, and reach mb_y-1. if (prev_td != td) { if (threadnr != 0) { - check_thread_pos(td, prev_td, mb_x + 1, mb_y - 1); + check_thread_pos(td, prev_td, + mb_x + (is_vp7 ? 2 : 1), + mb_y - (is_vp7 ? 2 : 1)); } else { check_thread_pos(td, prev_td, - (s->mb_width + 3) + (mb_x + 1), mb_y - 1); + mb_x + (is_vp7 ? 2 : 1) + s->mb_width + 3, + mb_y - (is_vp7 ? 2 : 1)); } } @@ -1760,15 +2288,15 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata, if (!s->mb_layout) decode_mb_mode(s, mb, mb_x, mb_y, curframe->seg_map->data + mb_xy, prev_frame && prev_frame->seg_map ? - prev_frame->seg_map->data + mb_xy : NULL, 0); + prev_frame->seg_map->data + mb_xy : NULL, 0, is_vp7); prefetch_motion(s, mb, mb_x, mb_y, mb_xy, VP56_FRAME_PREVIOUS); if (!mb->skip) - decode_mb_coeffs(s, td, c, mb, s->top_nnz[mb_x], td->left_nnz); + decode_mb_coeffs(s, td, c, mb, s->top_nnz[mb_x], td->left_nnz, is_vp7); if (mb->mode <= MODE_I4x4) - intra_predict(s, td, dst, mb, mb_x, mb_y); + intra_predict(s, td, dst, mb, mb_x, mb_y, is_vp7); else inter_predict(s, td, dst, mb, mb_x, mb_y); @@ -1789,7 +2317,7 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata, } if (s->deblock_filter) - filter_level_for_mb(s, mb, &td->filter_strength[mb_x]); + filter_level_for_mb(s, mb, &td->filter_strength[mb_x], is_vp7); if (s->deblock_filter && num_jobs != 1 && threadnr == num_jobs - 1) { if (s->filter.simple) @@ -1817,7 +2345,7 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata, } static void vp8_filter_mb_row(AVCodecContext *avctx, void *tdata, - int jobnr, int threadnr) + int jobnr, int threadnr, int is_vp7) { VP8Context *s = avctx->priv_data; VP8ThreadData *td = &s->thread_data[threadnr]; @@ -1866,7 +2394,7 @@ static void vp8_filter_mb_row(AVCodecContext *avctx, void *tdata, if (s->filter.simple) filter_mb_simple(s, dst[0], f, mb_x, mb_y); else - filter_mb(s, dst, f, mb_x, mb_y); + filter_mb(s, dst, f, mb_x, mb_y, is_vp7); dst[0] += 16; dst[1] += 8; dst[2] += 8; @@ -1875,8 +2403,9 @@ static void vp8_filter_mb_row(AVCodecContext *avctx, void *tdata, } } -static int vp8_decode_mb_row_sliced(AVCodecContext *avctx, void *tdata, - int jobnr, int threadnr) +static av_always_inline +int vp78_decode_mb_row_sliced(AVCodecContext *avctx, void *tdata, int jobnr, + int threadnr, int is_vp7) { VP8Context *s = avctx->priv_data; VP8ThreadData *td = &s->thread_data[jobnr]; @@ -1889,9 +2418,9 @@ static int vp8_decode_mb_row_sliced(AVCodecContext *avctx, void *tdata, if (mb_y >= s->mb_height) break; td->thread_mb_pos = mb_y << 16; - vp8_decode_mb_row_no_filter(avctx, tdata, jobnr, threadnr); + vp8_decode_mb_row_no_filter(avctx, tdata, jobnr, threadnr, is_vp7); if (s->deblock_filter) - vp8_filter_mb_row(avctx, tdata, jobnr, threadnr); + vp8_filter_mb_row(avctx, tdata, jobnr, threadnr, is_vp7); update_pos(td, mb_y, INT_MAX & 0xFFFF); s->mv_min.y -= 64; @@ -1904,15 +2433,34 @@ static int vp8_decode_mb_row_sliced(AVCodecContext *avctx, void *tdata, return 0; } -int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame, - AVPacket *avpkt) +static int vp7_decode_mb_row_sliced(AVCodecContext *avctx, void *tdata, + int jobnr, int threadnr) +{ + return vp78_decode_mb_row_sliced(avctx, tdata, jobnr, threadnr, IS_VP7); +} + +static int vp8_decode_mb_row_sliced(AVCodecContext *avctx, void *tdata, + int jobnr, int threadnr) +{ + return vp78_decode_mb_row_sliced(avctx, tdata, jobnr, threadnr, IS_VP8); +} + + +static av_always_inline +int vp78_decode_frame(AVCodecContext *avctx, void *data, int *got_frame, + AVPacket *avpkt, int is_vp7) { VP8Context *s = avctx->priv_data; int ret, i, referenced, num_jobs; enum AVDiscard skip_thresh; VP8Frame *av_uninit(curframe), *prev_frame; - if ((ret = decode_frame_header(s, avpkt->data, avpkt->size)) < 0) + if (is_vp7) + ret = vp7_decode_frame_header(s, avpkt->data, avpkt->size); + else + ret = vp8_decode_frame_header(s, avpkt->data, avpkt->size); + + if (ret < 0) goto err; prev_frame = s->framep[VP56_FRAME_CURRENT]; @@ -1940,21 +2488,7 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame, &s->frames[i] != s->framep[VP56_FRAME_GOLDEN2]) vp8_release_frame(s, &s->frames[i]); - // find a free buffer - for (i = 0; i < 5; i++) - if (&s->frames[i] != prev_frame && - &s->frames[i] != s->framep[VP56_FRAME_PREVIOUS] && - &s->frames[i] != s->framep[VP56_FRAME_GOLDEN] && - &s->frames[i] != s->framep[VP56_FRAME_GOLDEN2]) { - curframe = s->framep[VP56_FRAME_CURRENT] = &s->frames[i]; - break; - } - if (i == 5) { - av_log(avctx, AV_LOG_FATAL, "Ran out of free frames!\n"); - abort(); - } - if (curframe->tf.f->data[0]) - vp8_release_frame(s, curframe); + curframe = s->framep[VP56_FRAME_CURRENT] = vp8_find_free_buffer(s); /* Given that arithmetic probabilities are updated every frame, it's quite * likely that the values we have on a random interframe are complete @@ -2017,7 +2551,10 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame, if (prev_frame && s->segmentation.enabled && !s->segmentation.update_map) ff_thread_await_progress(&prev_frame->tf, 1, 0); - vp8_decode_mv_mb_modes(avctx, curframe, prev_frame); + if (is_vp7) + vp7_decode_mv_mb_modes(avctx, curframe, prev_frame); + else + vp8_decode_mv_mb_modes(avctx, curframe, prev_frame); } if (avctx->active_thread_type == FF_THREAD_FRAME) @@ -2033,8 +2570,12 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame, s->thread_data[i].thread_mb_pos = 0; s->thread_data[i].wait_mb_pos = INT_MAX; } - avctx->execute2(avctx, vp8_decode_mb_row_sliced, - s->thread_data, NULL, num_jobs); + if (is_vp7) + avctx->execute2(avctx, vp7_decode_mb_row_sliced, s->thread_data, NULL, + num_jobs); + else + avctx->execute2(avctx, vp8_decode_mb_row_sliced, s->thread_data, NULL, + num_jobs); ff_thread_report_progress(&curframe->tf, INT_MAX, 0); memcpy(&s->framep[0], &s->next_framep[0], sizeof(s->framep[0]) * 4); @@ -2057,6 +2598,20 @@ err: return ret; } +int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame, + AVPacket *avpkt) +{ + return vp78_decode_frame(avctx, data, got_frame, avpkt, IS_VP8); +} + +#if CONFIG_VP7_DECODER +static int vp7_decode_frame(AVCodecContext *avctx, void *data, int *got_frame, + AVPacket *avpkt) +{ + return vp78_decode_frame(avctx, data, got_frame, avpkt, IS_VP7); +} +#endif /* CONFIG_VP7_DECODER */ + av_cold int ff_vp8_decode_free(AVCodecContext *avctx) { VP8Context *s = avctx->priv_data; @@ -2080,7 +2635,8 @@ static av_cold int vp8_init_frames(VP8Context *s) return 0; } -av_cold int ff_vp8_decode_init(AVCodecContext *avctx) +static av_always_inline +int vp78_decode_init(AVCodecContext *avctx, int is_vp7) { VP8Context *s = avctx->priv_data; int ret; @@ -2090,8 +2646,18 @@ av_cold int ff_vp8_decode_init(AVCodecContext *avctx) avctx->internal->allocate_progress = 1; ff_videodsp_init(&s->vdsp, 8); - ff_h264_pred_init(&s->hpc, AV_CODEC_ID_VP8, 8, 1); - ff_vp8dsp_init(&s->vp8dsp); + + ff_vp78dsp_init(&s->vp8dsp); + if (CONFIG_VP7_DECODER && is_vp7) { + ff_h264_pred_init(&s->hpc, AV_CODEC_ID_VP7, 8, 1); + ff_vp7dsp_init(&s->vp8dsp); + } else if (CONFIG_VP8_DECODER && !is_vp7) { + ff_h264_pred_init(&s->hpc, AV_CODEC_ID_VP8, 8, 1); + ff_vp8dsp_init(&s->vp8dsp); + } + + /* does not change for VP8 */ + memcpy(s->prob[0].scan, zigzag_scan, sizeof(s->prob[0].scan)); if ((ret = vp8_init_frames(s)) < 0) { ff_vp8_decode_free(avctx); @@ -2101,6 +2667,19 @@ av_cold int ff_vp8_decode_init(AVCodecContext *avctx) return 0; } +#if CONFIG_VP7_DECODER +static int vp7_decode_init(AVCodecContext *avctx) +{ + return vp78_decode_init(avctx, IS_VP7); +} +#endif /* CONFIG_VP7_DECODER */ + +av_cold int ff_vp8_decode_init(AVCodecContext *avctx) +{ + return vp78_decode_init(avctx, IS_VP8); +} + +#if CONFIG_VP8_DECODER static av_cold int vp8_decode_init_thread_copy(AVCodecContext *avctx) { VP8Context *s = avctx->priv_data; @@ -2151,7 +2730,24 @@ static int vp8_decode_update_thread_context(AVCodecContext *dst, return 0; } +#endif /* CONFIG_VP8_DECODER */ + +#if CONFIG_VP7_DECODER +AVCodec ff_vp7_decoder = { + .name = "vp7", + .long_name = NULL_IF_CONFIG_SMALL("On2 VP7"), + .type = AVMEDIA_TYPE_VIDEO, + .id = AV_CODEC_ID_VP7, + .priv_data_size = sizeof(VP8Context), + .init = vp7_decode_init, + .close = ff_vp8_decode_free, + .decode = vp7_decode_frame, + .capabilities = CODEC_CAP_DR1, + .flush = vp8_decode_flush, +}; +#endif /* CONFIG_VP7_DECODER */ +#if CONFIG_VP8_DECODER AVCodec ff_vp8_decoder = { .name = "vp8", .long_name = NULL_IF_CONFIG_SMALL("On2 VP8"), @@ -2166,3 +2762,4 @@ AVCodec ff_vp8_decoder = { .init_thread_copy = ONLY_IF_THREADS_ENABLED(vp8_decode_init_thread_copy), .update_thread_context = ONLY_IF_THREADS_ENABLED(vp8_decode_update_thread_context), }; +#endif /* CONFIG_VP7_DECODER */ diff --git a/libavcodec/vp8.h b/libavcodec/vp8.h index 1eb0512626..99389051eb 100644 --- a/libavcodec/vp8.h +++ b/libavcodec/vp8.h @@ -237,6 +237,7 @@ typedef struct VP8Context { uint8_t pred8x8c[3]; uint8_t token[4][16][3][NUM_DCT_TOKENS - 1]; uint8_t mvc[2][19]; + uint8_t scan[16]; } prob[2]; VP8Macroblock *macroblocks_base; @@ -270,6 +271,26 @@ typedef struct VP8Context { * 1 -> Macroblocks for entire frame alloced (sliced thread). */ int mb_layout; + + /** + * Fade bit present in bitstream (VP7) + */ + int fade_present; + + /** + * Interframe DC prediction (VP7) + * [0] VP56_FRAME_PREVIOUS + * [1] VP56_FRAME_GOLDEN + */ + uint16_t inter_dc_pred[2][2]; + + /** + * Macroblock features (VP7) + */ + uint8_t feature_enabled[4]; + uint8_t feature_present_prob[4]; + uint8_t feature_index_prob[4][3]; + uint8_t feature_value[4][4]; } VP8Context; int ff_vp8_decode_init(AVCodecContext *avctx); diff --git a/libavcodec/vp8data.h b/libavcodec/vp8data.h index 2a17cc2024..b49dea98fa 100644 --- a/libavcodec/vp8data.h +++ b/libavcodec/vp8data.h @@ -30,6 +30,13 @@ #include "vp8.h" #include "h264pred.h" +static const uint8_t vp7_pred4x4_mode[] = { + [DC_PRED8x8] = DC_PRED, + [VERT_PRED8x8] = TM_VP8_PRED, + [HOR_PRED8x8] = TM_VP8_PRED, + [PLANE_PRED8x8] = TM_VP8_PRED, +}; + static const uint8_t vp8_pred4x4_mode[] = { [DC_PRED8x8] = DC_PRED, [VERT_PRED8x8] = VERT_PRED, @@ -51,6 +58,63 @@ static const int8_t vp8_pred16x16_tree_inter[4][2] = { { -PLANE_PRED8x8, -MODE_I4x4 }, // '110', '111' }; +typedef struct VP7MVPred { + int8_t yoffset; + int8_t xoffset; + uint8_t subblock; + uint8_t score; +} VP7MVPred; + +#define VP7_MV_PRED_COUNT 12 +static const VP7MVPred vp7_mv_pred[VP7_MV_PRED_COUNT] = { + { -1, 0, 12, 8 }, + { 0, -1, 3, 8 }, + { -1, -1, 15, 2 }, + { -1, 1, 12, 2 }, + { -2, 0, 12, 2 }, + { 0, -2, 3, 2 }, + { -1, -2, 15, 1 }, + { -2, -1, 15, 1 }, + { -2, 1, 12, 1 }, + { -1, 2, 12, 1 }, + { -2, -2, 15, 1 }, + { -2, 2, 12, 1 }, +}; + +static const int vp7_mode_contexts[31][4] = { + { 3, 3, 1, 246 }, + { 7, 89, 66, 239 }, + { 10, 90, 78, 238 }, + { 14, 118, 95, 241 }, + { 14, 123, 106, 238 }, + { 20, 140, 109, 240 }, + { 13, 155, 103, 238 }, + { 21, 158, 99, 240 }, + { 27, 82, 108, 232 }, + { 19, 99, 123, 217 }, + { 45, 139, 148, 236 }, + { 50, 117, 144, 235 }, + { 57, 128, 164, 238 }, + { 69, 139, 171, 239 }, + { 74, 154, 179, 238 }, + { 112, 165, 186, 242 }, + { 98, 143, 185, 245 }, + { 105, 153, 190, 250 }, + { 124, 167, 192, 245 }, + { 131, 186, 203, 246 }, + { 59, 184, 222, 224 }, + { 148, 215, 214, 213 }, + { 137, 211, 210, 219 }, + { 190, 227, 128, 228 }, + { 183, 228, 128, 228 }, + { 194, 234, 128, 228 }, + { 202, 236, 128, 228 }, + { 205, 240, 128, 228 }, + { 205, 244, 128, 228 }, + { 225, 246, 128, 228 }, + { 233, 251, 128, 228 }, +}; + static const int vp8_mode_contexts[6][4] = { { 7, 1, 1, 143 }, { 14, 18, 14, 107 }, @@ -82,6 +146,10 @@ static const uint8_t vp8_mbsplit_prob[3] = { 110, 111, 150 }; +static const uint8_t vp7_submv_prob[3] = { + 180, 162, 25 +}; + static const uint8_t vp8_submv_prob[5][3] = { { 147, 136, 18 }, { 106, 145, 1 }, @@ -94,7 +162,7 @@ static const uint8_t vp8_pred16x16_prob_intra[4] = { 145, 156, 163, 128 }; static const uint8_t vp8_pred16x16_prob_inter[4] = { - 112, 86, 140, 37 + 112, 86, 140, 37 }; static const int8_t vp8_pred4x4_tree[9][2] = { @@ -641,7 +709,7 @@ static const uint8_t vp8_token_update_probs[4][8][3][NUM_DCT_TOKENS - 1] = { }; // fixme: copied from h264data.h -static const uint8_t zigzag_scan[16] = { +static const uint8_t zigzag_scan[16]={ 0 + 0 * 4, 1 + 0 * 4, 0 + 1 * 4, 0 + 2 * 4, 1 + 1 * 4, 2 + 0 * 4, 3 + 0 * 4, 2 + 1 * 4, 1 + 2 * 4, 0 + 3 * 4, 1 + 3 * 4, 2 + 2 * 4, @@ -674,11 +742,22 @@ static const uint8_t vp8_mv_update_prob[2][19] = { { 237, 246, 253, 253, 254, 254, 254, 254, 254, - 254, 254, 254, 254, 254, 250, 250, 252, 254, 254 }, + 254, 254, 254, 254, 254, 250, 250, 252, /* VP8 only: */ 254, 254 }, { 231, 243, 245, 253, 254, 254, 254, 254, 254, - 254, 254, 254, 254, 254, 251, 251, 254, 254, 254 } + 254, 254, 254, 254, 254, 251, 251, 254, /* VP8 only: */ 254, 254 } +}; + +static const uint8_t vp7_mv_default_prob[2][17] = { + { 162, + 128, + 225, 146, 172, 147, 214, 39, 156, + 247, 210, 135, 68, 138, 220, 239, 246 }, + { 164, + 128, + 204, 170, 119, 235, 140, 230, 228, + 244, 184, 201, 44, 173, 221, 239, 253 } }; static const uint8_t vp8_mv_default_prob[2][19] = { @@ -692,4 +771,68 @@ static const uint8_t vp8_mv_default_prob[2][19] = { 128, 130, 130, 74, 148, 180, 203, 236, 254, 254 } }; +static const uint8_t vp7_feature_value_size[2][4] = { + { 7, 6, 0, 8 }, + { 7, 6, 0, 5 }, +}; + +static const int8_t vp7_feature_index_tree[4][2] = +{ + { 1, 2 }, + { -0, -1 }, // '00', '01' + { -2, -3 }, // '10', '11' +}; + +static const uint16_t vp7_ydc_qlookup[] = { + 4, 4, 5, 6, 6, 7, 8, 8, 9, 10, 11, 12, 13, 14, 15, + 16, 17, 18, 19, 20, 21, 22, 23, 23, 24, 25, 26, 27, 28, 29, + 30, 31, 32, 33, 33, 34, 35, 36, 36, 37, 38, 39, 39, 40, 41, + 41, 42, 43, 43, 44, 45, 45, 46, 47, 48, 48, 49, 50, 51, 52, + 53, 53, 54, 56, 57, 58, 59, 60, 62, 63, 65, 66, 68, 70, 72, + 74, 76, 79, 81, 84, 87, 90, 93, 96, 100, 104, 108, 112, 116, 121, + 126, 131, 136, 142, 148, 154, 160, 167, 174, 182, 189, 198, 206, 215, 224, + 234, 244, 254, 265, 277, 288, 301, 313, 327, 340, 355, 370, 385, 401, 417, + 434, 452, 470, 489, 509, 529, 550, 572, +}; + +static const uint16_t vp7_yac_qlookup[] = { + 4, 4, 5, 5, 6, 6, 7, 8, 9, 10, 11, 12, 13, 15, + 16, 17, 19, 20, 22, 23, 25, 26, 28, 29, 31, 32, 34, 35, + 37, 38, 40, 41, 42, 44, 45, 46, 48, 49, 50, 51, 53, 54, + 55, 56, 57, 58, 59, 61, 62, 63, 64, 65, 67, 68, 69, 70, + 72, 73, 75, 76, 78, 80, 82, 84, 86, 88, 91, 93, 96, 99, + 102, 105, 109, 112, 116, 121, 125, 130, 135, 140, 146, 152, 158, 165, + 172, 180, 188, 196, 205, 214, 224, 234, 245, 256, 268, 281, 294, 308, + 322, 337, 353, 369, 386, 404, 423, 443, 463, 484, 506, 529, 553, 578, + 604, 631, 659, 688, 718, 749, 781, 814, 849, 885, 922, 960, 1000, 1041, + 1083, 1127, +}; + +static const uint16_t vp7_y2dc_qlookup[] = { + 7, 9, 11, 13, 15, 17, 19, 21, 23, 26, 28, 30, 33, 35, + 37, 39, 42, 44, 46, 48, 51, 53, 55, 57, 59, 61, 63, 65, + 67, 69, 70, 72, 74, 75, 77, 78, 80, 81, 83, 84, 85, 87, + 88, 89, 90, 92, 93, 94, 95, 96, 97, 99, 100, 101, 102, 104, + 105, 106, 108, 109, 111, 113, 114, 116, 118, 120, 123, 125, 128, 131, + 134, 137, 140, 144, 148, 152, 156, 161, 166, 171, 176, 182, 188, 195, + 202, 209, 217, 225, 234, 243, 253, 263, 274, 285, 297, 309, 322, 336, + 350, 365, 381, 397, 414, 432, 450, 470, 490, 511, 533, 556, 579, 604, + 630, 656, 684, 713, 742, 773, 805, 838, 873, 908, 945, 983, 1022, 1063, + 1105, 1148, +}; + +static const uint16_t vp7_y2ac_qlookup[] = { + 7, 9, 11, 13, 16, 18, 21, 24, 26, 29, 32, 35, + 38, 41, 43, 46, 49, 52, 55, 58, 61, 64, 66, 69, + 72, 74, 77, 79, 82, 84, 86, 88, 91, 93, 95, 97, + 98, 100, 102, 104, 105, 107, 109, 110, 112, 113, 115, 116, + 117, 119, 120, 122, 123, 125, 127, 128, 130, 132, 134, 136, + 138, 141, 143, 146, 149, 152, 155, 158, 162, 166, 171, 175, + 180, 185, 191, 197, 204, 210, 218, 226, 234, 243, 252, 262, + 273, 284, 295, 308, 321, 335, 350, 365, 381, 398, 416, 435, + 455, 476, 497, 520, 544, 569, 595, 622, 650, 680, 711, 743, + 776, 811, 848, 885, 925, 965, 1008, 1052, 1097, 1144, 1193, 1244, + 1297, 1351, 1407, 1466, 1526, 1588, 1652, 1719, +}; + #endif /* AVCODEC_VP8DATA_H */ diff --git a/libavcodec/vp8dsp.c b/libavcodec/vp8dsp.c index 431c96ea77..4e4012f88e 100644 --- a/libavcodec/vp8dsp.c +++ b/libavcodec/vp8dsp.c @@ -1,6 +1,7 @@ /* * Copyright (C) 2010 David Conrad * Copyright (C) 2010 Ronald S. Bultje + * Copyright (C) 2014 Peter Ross * * This file is part of Libav. * @@ -29,7 +30,126 @@ #include "mathops.h" #include "vp8dsp.h" +#define MK_IDCT_DC_ADD4_C(name) \ +static void name ## _idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16], \ + ptrdiff_t stride) \ +{ \ + name ## _idct_dc_add_c(dst + stride * 0 + 0, block[0], stride); \ + name ## _idct_dc_add_c(dst + stride * 0 + 4, block[1], stride); \ + name ## _idct_dc_add_c(dst + stride * 4 + 0, block[2], stride); \ + name ## _idct_dc_add_c(dst + stride * 4 + 4, block[3], stride); \ +} \ + \ +static void name ## _idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], \ + ptrdiff_t stride) \ +{ \ + name ## _idct_dc_add_c(dst + 0, block[0], stride); \ + name ## _idct_dc_add_c(dst + 4, block[1], stride); \ + name ## _idct_dc_add_c(dst + 8, block[2], stride); \ + name ## _idct_dc_add_c(dst + 12, block[3], stride); \ +} + +#if CONFIG_VP7_DECODER +static void vp7_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16]) +{ + int i, a1, b1, c1, d1; + int16_t tmp[16]; + + for (i = 0; i < 4; i++) { + a1 = (dc[i * 4 + 0] + dc[i * 4 + 2]) * 23170; + b1 = (dc[i * 4 + 0] - dc[i * 4 + 2]) * 23170; + c1 = dc[i * 4 + 1] * 12540 - dc[i * 4 + 3] * 30274; + d1 = dc[i * 4 + 1] * 30274 + dc[i * 4 + 3] * 12540; + tmp[i * 4 + 0] = (a1 + d1) >> 14; + tmp[i * 4 + 3] = (a1 - d1) >> 14; + tmp[i * 4 + 1] = (b1 + c1) >> 14; + tmp[i * 4 + 2] = (b1 - c1) >> 14; + } + + for (i = 0; i < 4; i++) { + a1 = (tmp[i + 0] + tmp[i + 8]) * 23170; + b1 = (tmp[i + 0] - tmp[i + 8]) * 23170; + c1 = tmp[i + 4] * 12540 - tmp[i + 12] * 30274; + d1 = tmp[i + 4] * 30274 + tmp[i + 12] * 12540; + dc[i * 4 + 0] = 0; + dc[i * 4 + 1] = 0; + dc[i * 4 + 2] = 0; + dc[i * 4 + 3] = 0; + block[0][i][0] = (a1 + d1 + 0x20000) >> 18; + block[3][i][0] = (a1 - d1 + 0x20000) >> 18; + block[1][i][0] = (b1 + c1 + 0x20000) >> 18; + block[2][i][0] = (b1 - c1 + 0x20000) >> 18; + } +} + +static void vp7_luma_dc_wht_dc_c(int16_t block[4][4][16], int16_t dc[16]) +{ + int i, val = (23170 * (23170 * dc[0] >> 14) + 0x20000) >> 18; + dc[0] = 0; + + for (i = 0; i < 4; i++) { + block[i][0][0] = val; + block[i][1][0] = val; + block[i][2][0] = val; + block[i][3][0] = val; + } +} + +static void vp7_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride) +{ + int i, a1, b1, c1, d1; + int16_t tmp[16]; + + for (i = 0; i < 4; i++) { + a1 = (block[i * 4 + 0] + block[i * 4 + 2]) * 23170; + b1 = (block[i * 4 + 0] - block[i * 4 + 2]) * 23170; + c1 = block[i * 4 + 1] * 12540 - block[i * 4 + 3] * 30274; + d1 = block[i * 4 + 1] * 30274 + block[i * 4 + 3] * 12540; + block[i * 4 + 0] = 0; + block[i * 4 + 1] = 0; + block[i * 4 + 2] = 0; + block[i * 4 + 3] = 0; + tmp[i * 4 + 0] = (a1 + d1) >> 14; + tmp[i * 4 + 3] = (a1 - d1) >> 14; + tmp[i * 4 + 1] = (b1 + c1) >> 14; + tmp[i * 4 + 2] = (b1 - c1) >> 14; + } + + for (i = 0; i < 4; i++) { + a1 = (tmp[i + 0] + tmp[i + 8]) * 23170; + b1 = (tmp[i + 0] - tmp[i + 8]) * 23170; + c1 = tmp[i + 4] * 12540 - tmp[i + 12] * 30274; + d1 = tmp[i + 4] * 30274 + tmp[i + 12] * 12540; + dst[0 * stride + i] = av_clip_uint8(dst[0 * stride + i] + + ((a1 + d1 + 0x20000) >> 18)); + dst[3 * stride + i] = av_clip_uint8(dst[3 * stride + i] + + ((a1 - d1 + 0x20000) >> 18)); + dst[1 * stride + i] = av_clip_uint8(dst[1 * stride + i] + + ((b1 + c1 + 0x20000) >> 18)); + dst[2 * stride + i] = av_clip_uint8(dst[2 * stride + i] + + ((b1 - c1 + 0x20000) >> 18)); + } +} + +static void vp7_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride) +{ + int i, dc = (23170 * (23170 * block[0] >> 14) + 0x20000) >> 18; + block[0] = 0; + + for (i = 0; i < 4; i++) { + dst[0] = av_clip_uint8(dst[0] + dc); + dst[1] = av_clip_uint8(dst[1] + dc); + dst[2] = av_clip_uint8(dst[2] + dc); + dst[3] = av_clip_uint8(dst[3] + dc); + dst += stride; + } +} + +MK_IDCT_DC_ADD4_C(vp7) +#endif /* CONFIG_VP7_DECODER */ + // TODO: Maybe add dequant +#if CONFIG_VP8_DECODER static void vp8_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16]) { int i, t0, t1, t2, t3; @@ -128,23 +248,8 @@ static void vp8_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride) } } -static void vp8_idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16], - ptrdiff_t stride) -{ - vp8_idct_dc_add_c(dst + stride * 0 + 0, block[0], stride); - vp8_idct_dc_add_c(dst + stride * 0 + 4, block[1], stride); - vp8_idct_dc_add_c(dst + stride * 4 + 0, block[2], stride); - vp8_idct_dc_add_c(dst + stride * 4 + 4, block[3], stride); -} - -static void vp8_idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], - ptrdiff_t stride) -{ - vp8_idct_dc_add_c(dst + 0, block[0], stride); - vp8_idct_dc_add_c(dst + 4, block[1], stride); - vp8_idct_dc_add_c(dst + 8, block[2], stride); - vp8_idct_dc_add_c(dst + 12, block[3], stride); -} +MK_IDCT_DC_ADD4_C(vp8) +#endif /* CONFIG_VP8_DECODER */ // because I like only having two parameters to pass functions... #define LOAD_PIXELS \ @@ -160,7 +265,7 @@ static void vp8_idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], #define clip_int8(n) (cm[n + 0x80] - 0x80) static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, - int is4tap) + int is4tap, int is_vp7) { LOAD_PIXELS int a, f1, f2; @@ -176,7 +281,11 @@ static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, // We deviate from the spec here with c(a+3) >> 3 // since that's what libvpx does. f1 = FFMIN(a + 4, 127) >> 3; - f2 = FFMIN(a + 3, 127) >> 3; + + if (is_vp7) + f2 = f1 - ((a & 7) == 4); + else + f2 = FFMIN(a + 3, 127) >> 3; // Despite what the spec says, we do need to clamp here to // be bitexact with libvpx. @@ -185,13 +294,33 @@ static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, // only used for _inner on blocks without high edge variance if (!is4tap) { - a = (f1 + 1) >> 1; + a = (f1 + 1) >> 1; p[-2 * stride] = cm[p1 + a]; p[ 1 * stride] = cm[q1 - a]; } } -static av_always_inline int simple_limit(uint8_t *p, ptrdiff_t stride, int flim) +static av_always_inline void vp7_filter_common(uint8_t *p, ptrdiff_t stride, + int is4tap) +{ + filter_common(p, stride, is4tap, IS_VP7); +} + +static av_always_inline void vp8_filter_common(uint8_t *p, ptrdiff_t stride, + int is4tap) +{ + filter_common(p, stride, is4tap, IS_VP8); +} + +static av_always_inline int vp7_simple_limit(uint8_t *p, ptrdiff_t stride, + int flim) +{ + LOAD_PIXELS + return FFABS(p0 - q0) <= flim; +} + +static av_always_inline int vp8_simple_limit(uint8_t *p, ptrdiff_t stride, + int flim) { LOAD_PIXELS return 2 * FFABS(p0 - q0) + (FFABS(p1 - q1) >> 1) <= flim; @@ -201,19 +330,21 @@ static av_always_inline int simple_limit(uint8_t *p, ptrdiff_t stride, int flim) * E - limit at the macroblock edge * I - limit for interior difference */ -static av_always_inline int normal_limit(uint8_t *p, ptrdiff_t stride, - int E, int I) -{ - LOAD_PIXELS - return simple_limit(p, stride, E) && - FFABS(p3 - p2) <= I && - FFABS(p2 - p1) <= I && - FFABS(p1 - p0) <= I && - FFABS(q3 - q2) <= I && - FFABS(q2 - q1) <= I && - FFABS(q1 - q0) <= I; +#define NORMAL_LIMIT(vpn) \ +static av_always_inline int vp ## vpn ## _normal_limit(uint8_t *p, \ + ptrdiff_t stride, \ + int E, int I) \ +{ \ + LOAD_PIXELS \ + return vp ## vpn ## _simple_limit(p, stride, E) && \ + FFABS(p3 - p2) <= I && FFABS(p2 - p1) <= I && \ + FFABS(p1 - p0) <= I && FFABS(q3 - q2) <= I && \ + FFABS(q2 - q1) <= I && FFABS(q1 - q0) <= I; \ } +NORMAL_LIMIT(7) +NORMAL_LIMIT(8) + // high edge variance static av_always_inline int hev(uint8_t *p, ptrdiff_t stride, int thresh) { @@ -243,82 +374,91 @@ static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride) p[ 2 * stride] = cm[q2 - a2]; } -#define LOOP_FILTER(dir, size, stridea, strideb, maybe_inline) \ +#define LOOP_FILTER(vpn, dir, size, stridea, strideb, maybe_inline) \ static maybe_inline \ -void vp8_ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, \ - ptrdiff_t stride, \ - int flim_E, int flim_I, \ - int hev_thresh) \ +void vpn ## _ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, \ + ptrdiff_t stride, \ + int flim_E, int flim_I, \ + int hev_thresh) \ { \ int i; \ for (i = 0; i < size; i++) \ - if (normal_limit(dst + i * stridea, strideb, flim_E, flim_I)) { \ + if (vpn ## _normal_limit(dst + i * stridea, strideb, \ + flim_E, flim_I)) { \ if (hev(dst + i * stridea, strideb, hev_thresh)) \ - filter_common(dst + i * stridea, strideb, 1); \ + vpn ## _filter_common(dst + i * stridea, strideb, 1); \ else \ filter_mbedge(dst + i * stridea, strideb); \ } \ } \ \ static maybe_inline \ -void vp8_ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, \ - ptrdiff_t stride, \ - int flim_E, int flim_I, \ - int hev_thresh) \ +void vpn ## _ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, \ + ptrdiff_t stride, \ + int flim_E, \ + int flim_I, \ + int hev_thresh) \ { \ int i; \ for (i = 0; i < size; i++) \ - if (normal_limit(dst + i * stridea, strideb, flim_E, flim_I)) { \ + if (vpn ## _normal_limit(dst + i * stridea, strideb, \ + flim_E, flim_I)) { \ int hv = hev(dst + i * stridea, strideb, hev_thresh); \ if (hv) \ - filter_common(dst + i * stridea, strideb, 1); \ + vpn ## _filter_common(dst + i * stridea, strideb, 1); \ else \ - filter_common(dst + i * stridea, strideb, 0); \ + vpn ## _filter_common(dst + i * stridea, strideb, 0); \ } \ } -LOOP_FILTER(v, 16, 1, stride, ) -LOOP_FILTER(h, 16, stride, 1, ) - -#define UV_LOOP_FILTER(dir, stridea, strideb) \ -LOOP_FILTER(dir, 8, stridea, strideb, av_always_inline) \ -static void vp8_ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV, \ - ptrdiff_t stride, int fE, \ - int fI, int hev_thresh) \ +#define UV_LOOP_FILTER(vpn, dir, stridea, strideb) \ +LOOP_FILTER(vpn, dir, 8, stridea, strideb, av_always_inline) \ +static void vpn ## _ ## dir ## _loop_filter8uv_c(uint8_t *dstU, \ + uint8_t *dstV, \ + ptrdiff_t stride, int fE, \ + int fI, int hev_thresh) \ { \ - vp8_ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh); \ - vp8_ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh); \ + vpn ## _ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh); \ + vpn ## _ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh); \ } \ \ -static void vp8_ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, \ - uint8_t *dstV, \ - ptrdiff_t stride, int fE, \ - int fI, int hev_thresh) \ +static void vpn ## _ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, \ + uint8_t *dstV, \ + ptrdiff_t stride, \ + int fE, int fI, \ + int hev_thresh) \ { \ - vp8_ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh); \ - vp8_ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh); \ + vpn ## _ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, \ + hev_thresh); \ + vpn ## _ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, \ + hev_thresh); \ } -UV_LOOP_FILTER(v, 1, stride) -UV_LOOP_FILTER(h, stride, 1) - -static void vp8_v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim) -{ - int i; - - for (i = 0; i < 16; i++) - if (simple_limit(dst + i, stride, flim)) - filter_common(dst + i, stride, 1); +#define LOOP_FILTER_SIMPLE(vpn) \ +static void vpn ## _v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, \ + int flim) \ +{ \ + int i; \ + for (i = 0; i < 16; i++) \ + if (vpn ## _simple_limit(dst + i, stride, flim)) \ + vpn ## _filter_common(dst + i, stride, 1); \ +} \ + \ +static void vpn ## _h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, \ + int flim) \ +{ \ + int i; \ + for (i = 0; i < 16; i++) \ + if (vpn ## _simple_limit(dst + i * stride, 1, flim)) \ + vpn ## _filter_common(dst + i * stride, 1, 1); \ } -static void vp8_h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim) -{ - int i; - - for (i = 0; i < 16; i++) - if (simple_limit(dst + i * stride, 1, flim)) - filter_common(dst + i * stride, 1, 1); -} +#define LOOP_FILTERS(vpn) \ + LOOP_FILTER(vpn, v, 16, 1, stride, ) \ + LOOP_FILTER(vpn, h, 16, stride, 1, ) \ + UV_LOOP_FILTER(vpn, v, 1, stride) \ + UV_LOOP_FILTER(vpn, h, stride, 1) \ + LOOP_FILTER_SIMPLE(vpn) \ static const uint8_t subpel_filters[7][6] = { { 0, 6, 123, 12, 1, 0 }, @@ -507,7 +647,7 @@ VP8_BILINEAR(16) VP8_BILINEAR(8) VP8_BILINEAR(4) -#define VP8_MC_FUNC(IDX, SIZE) \ +#define VP78_MC_FUNC(IDX, SIZE) \ dsp->put_vp8_epel_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \ dsp->put_vp8_epel_pixels_tab[IDX][0][1] = put_vp8_epel ## SIZE ## _h4_c; \ dsp->put_vp8_epel_pixels_tab[IDX][0][2] = put_vp8_epel ## SIZE ## _h6_c; \ @@ -518,8 +658,8 @@ VP8_BILINEAR(4) dsp->put_vp8_epel_pixels_tab[IDX][2][1] = put_vp8_epel ## SIZE ## _h4v6_c; \ dsp->put_vp8_epel_pixels_tab[IDX][2][2] = put_vp8_epel ## SIZE ## _h6v6_c -#define VP8_BILINEAR_MC_FUNC(IDX, SIZE) \ - dsp->put_vp8_bilinear_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \ +#define VP78_BILINEAR_MC_FUNC(IDX, SIZE) \ + dsp->put_vp8_bilinear_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \ dsp->put_vp8_bilinear_pixels_tab[IDX][0][1] = put_vp8_bilinear ## SIZE ## _h_c; \ dsp->put_vp8_bilinear_pixels_tab[IDX][0][2] = put_vp8_bilinear ## SIZE ## _h_c; \ dsp->put_vp8_bilinear_pixels_tab[IDX][1][0] = put_vp8_bilinear ## SIZE ## _v_c; \ @@ -529,6 +669,54 @@ VP8_BILINEAR(4) dsp->put_vp8_bilinear_pixels_tab[IDX][2][1] = put_vp8_bilinear ## SIZE ## _hv_c; \ dsp->put_vp8_bilinear_pixels_tab[IDX][2][2] = put_vp8_bilinear ## SIZE ## _hv_c +av_cold void ff_vp78dsp_init(VP8DSPContext *dsp) +{ + VP78_MC_FUNC(0, 16); + VP78_MC_FUNC(1, 8); + VP78_MC_FUNC(2, 4); + + VP78_BILINEAR_MC_FUNC(0, 16); + VP78_BILINEAR_MC_FUNC(1, 8); + VP78_BILINEAR_MC_FUNC(2, 4); + + if (ARCH_ARM) + ff_vp78dsp_init_arm(dsp); + if (ARCH_PPC) + ff_vp78dsp_init_ppc(dsp); + if (ARCH_X86) + ff_vp78dsp_init_x86(dsp); +} + +#if CONFIG_VP7_DECODER +LOOP_FILTERS(vp7) + +av_cold void ff_vp7dsp_init(VP8DSPContext *dsp) +{ + dsp->vp8_luma_dc_wht = vp7_luma_dc_wht_c; + dsp->vp8_luma_dc_wht_dc = vp7_luma_dc_wht_dc_c; + dsp->vp8_idct_add = vp7_idct_add_c; + dsp->vp8_idct_dc_add = vp7_idct_dc_add_c; + dsp->vp8_idct_dc_add4y = vp7_idct_dc_add4y_c; + dsp->vp8_idct_dc_add4uv = vp7_idct_dc_add4uv_c; + + dsp->vp8_v_loop_filter16y = vp7_v_loop_filter16_c; + dsp->vp8_h_loop_filter16y = vp7_h_loop_filter16_c; + dsp->vp8_v_loop_filter8uv = vp7_v_loop_filter8uv_c; + dsp->vp8_h_loop_filter8uv = vp7_h_loop_filter8uv_c; + + dsp->vp8_v_loop_filter16y_inner = vp7_v_loop_filter16_inner_c; + dsp->vp8_h_loop_filter16y_inner = vp7_h_loop_filter16_inner_c; + dsp->vp8_v_loop_filter8uv_inner = vp7_v_loop_filter8uv_inner_c; + dsp->vp8_h_loop_filter8uv_inner = vp7_h_loop_filter8uv_inner_c; + + dsp->vp8_v_loop_filter_simple = vp7_v_loop_filter_simple_c; + dsp->vp8_h_loop_filter_simple = vp7_h_loop_filter_simple_c; +} +#endif /* CONFIG_VP7_DECODER */ + +#if CONFIG_VP8_DECODER +LOOP_FILTERS(vp8) + av_cold void ff_vp8dsp_init(VP8DSPContext *dsp) { dsp->vp8_luma_dc_wht = vp8_luma_dc_wht_c; @@ -551,18 +739,9 @@ av_cold void ff_vp8dsp_init(VP8DSPContext *dsp) dsp->vp8_v_loop_filter_simple = vp8_v_loop_filter_simple_c; dsp->vp8_h_loop_filter_simple = vp8_h_loop_filter_simple_c; - VP8_MC_FUNC(0, 16); - VP8_MC_FUNC(1, 8); - VP8_MC_FUNC(2, 4); - - VP8_BILINEAR_MC_FUNC(0, 16); - VP8_BILINEAR_MC_FUNC(1, 8); - VP8_BILINEAR_MC_FUNC(2, 4); - if (ARCH_ARM) ff_vp8dsp_init_arm(dsp); - if (ARCH_PPC) - ff_vp8dsp_init_ppc(dsp); if (ARCH_X86) ff_vp8dsp_init_x86(dsp); } +#endif /* CONFIG_VP8_DECODER */ diff --git a/libavcodec/vp8dsp.h b/libavcodec/vp8dsp.h index 9e313a7067..4864cf7969 100644 --- a/libavcodec/vp8dsp.h +++ b/libavcodec/vp8dsp.h @@ -88,9 +88,18 @@ void ff_put_vp8_pixels8_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride, void ff_put_vp8_pixels4_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int h, int x, int y); +void ff_vp7dsp_init(VP8DSPContext *c); + +void ff_vp78dsp_init(VP8DSPContext *c); +void ff_vp78dsp_init_arm(VP8DSPContext *c); +void ff_vp78dsp_init_ppc(VP8DSPContext *c); +void ff_vp78dsp_init_x86(VP8DSPContext *c); + void ff_vp8dsp_init(VP8DSPContext *c); -void ff_vp8dsp_init_x86(VP8DSPContext *c); void ff_vp8dsp_init_arm(VP8DSPContext *c); -void ff_vp8dsp_init_ppc(VP8DSPContext *c); +void ff_vp8dsp_init_x86(VP8DSPContext *c); + +#define IS_VP7 1 +#define IS_VP8 0 #endif /* AVCODEC_VP8DSP_H */ diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile index eb312a2dd1..910471aeb5 100644 --- a/libavcodec/x86/Makefile +++ b/libavcodec/x86/Makefile @@ -37,6 +37,7 @@ OBJS-$(CONFIG_TRUEHD_DECODER) += x86/mlpdsp.o OBJS-$(CONFIG_VC1_DECODER) += x86/vc1dsp_init.o OBJS-$(CONFIG_VORBIS_DECODER) += x86/vorbisdsp_init.o OBJS-$(CONFIG_VP6_DECODER) += x86/vp6dsp_init.o +OBJS-$(CONFIG_VP7_DECODER) += x86/vp8dsp_init.o OBJS-$(CONFIG_VP8_DECODER) += x86/vp8dsp_init.o OBJS-$(CONFIG_VP9_DECODER) += x86/vp9dsp_init.o @@ -92,6 +93,8 @@ YASM-OBJS-$(CONFIG_RV40_DECODER) += x86/rv34dsp.o \ YASM-OBJS-$(CONFIG_VC1_DECODER) += x86/vc1dsp.o YASM-OBJS-$(CONFIG_VORBIS_DECODER) += x86/vorbisdsp.o YASM-OBJS-$(CONFIG_VP6_DECODER) += x86/vp6dsp.o +YASM-OBJS-$(CONFIG_VP7_DECODER) += x86/vp8dsp.o \ + x86/vp8dsp_loopfilter.o YASM-OBJS-$(CONFIG_VP8_DECODER) += x86/vp8dsp.o \ x86/vp8dsp_loopfilter.o YASM-OBJS-$(CONFIG_VP9_DECODER) += x86/vp9dsp.o diff --git a/libavcodec/x86/h264_intrapred_init.c b/libavcodec/x86/h264_intrapred_init.c index f934256706..6bd33ed388 100644 --- a/libavcodec/x86/h264_intrapred_init.c +++ b/libavcodec/x86/h264_intrapred_init.c @@ -195,7 +195,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, h->pred8x8 [VERT_PRED8x8 ] = ff_pred8x8_vertical_8_mmx; h->pred8x8 [HOR_PRED8x8 ] = ff_pred8x8_horizontal_8_mmx; } - if (codec_id == AV_CODEC_ID_VP8) { + if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) { h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_tm_vp8_8_mmx; h->pred8x8 [PLANE_PRED8x8 ] = ff_pred8x8_tm_vp8_8_mmx; h->pred4x4 [TM_VP8_PRED ] = ff_pred4x4_tm_vp8_8_mmx; @@ -231,7 +231,8 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, h->pred4x4 [VERT_RIGHT_PRED ] = ff_pred4x4_vertical_right_8_mmxext; h->pred4x4 [HOR_DOWN_PRED ] = ff_pred4x4_horizontal_down_8_mmxext; h->pred4x4 [DC_PRED ] = ff_pred4x4_dc_8_mmxext; - if (codec_id == AV_CODEC_ID_VP8 || codec_id == AV_CODEC_ID_H264) { + if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8 || + codec_id == AV_CODEC_ID_H264) { h->pred4x4 [DIAG_DOWN_LEFT_PRED] = ff_pred4x4_down_left_8_mmxext; } if (codec_id == AV_CODEC_ID_SVQ3 || codec_id == AV_CODEC_ID_H264) { @@ -246,7 +247,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, h->pred8x8[DC_PRED8x8 ] = ff_pred8x8_dc_8_mmxext; } } - if (codec_id == AV_CODEC_ID_VP8) { + if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) { h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_tm_vp8_8_mmxext; h->pred8x8 [DC_PRED8x8 ] = ff_pred8x8_dc_rv40_8_mmxext; h->pred8x8 [PLANE_PRED8x8 ] = ff_pred8x8_tm_vp8_8_mmxext; @@ -276,7 +277,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, h->pred8x8l [VERT_RIGHT_PRED ] = ff_pred8x8l_vertical_right_8_sse2; h->pred8x8l [VERT_LEFT_PRED ] = ff_pred8x8l_vertical_left_8_sse2; h->pred8x8l [HOR_DOWN_PRED ] = ff_pred8x8l_horizontal_down_8_sse2; - if (codec_id == AV_CODEC_ID_VP8) { + if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) { h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_tm_vp8_8_sse2; h->pred8x8 [PLANE_PRED8x8 ] = ff_pred8x8_tm_vp8_8_sse2; } else { @@ -307,7 +308,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, h->pred8x8l [VERT_LEFT_PRED ] = ff_pred8x8l_vertical_left_8_ssse3; h->pred8x8l [HOR_UP_PRED ] = ff_pred8x8l_horizontal_up_8_ssse3; h->pred8x8l [HOR_DOWN_PRED ] = ff_pred8x8l_horizontal_down_8_ssse3; - if (codec_id == AV_CODEC_ID_VP8) { + if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) { h->pred8x8 [PLANE_PRED8x8 ] = ff_pred8x8_tm_vp8_8_ssse3; h->pred4x4 [TM_VP8_PRED ] = ff_pred4x4_tm_vp8_8_ssse3; } else { diff --git a/libavcodec/x86/vp8dsp_init.c b/libavcodec/x86/vp8dsp_init.c index 69460aa73f..b3b32b7314 100644 --- a/libavcodec/x86/vp8dsp_init.c +++ b/libavcodec/x86/vp8dsp_init.c @@ -315,25 +315,69 @@ DECLARE_LOOP_FILTER(sse4) c->put_vp8_bilinear_pixels_tab[IDX][2][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT -av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c) +av_cold void ff_vp78dsp_init_x86(VP8DSPContext *c) { #if HAVE_YASM int cpu_flags = av_get_cpu_flags(); if (EXTERNAL_MMX(cpu_flags)) { - c->vp8_idct_dc_add = ff_vp8_idct_dc_add_mmx; - c->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_mmx; #if ARCH_X86_32 - c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_mmx; - c->vp8_idct_add = ff_vp8_idct_add_mmx; - c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_mmx; c->put_vp8_epel_pixels_tab[0][0][0] = c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_mmx; #endif c->put_vp8_epel_pixels_tab[1][0][0] = c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_mmx; + } + + /* note that 4-tap width=16 functions are missing because w=16 + * is only used for luma, and luma is always a copy or sixtap. */ + if (EXTERNAL_MMXEXT(cpu_flags)) { + VP8_MC_FUNC(2, 4, mmxext); + VP8_BILINEAR_MC_FUNC(2, 4, mmxext); +#if ARCH_X86_32 + VP8_LUMA_MC_FUNC(0, 16, mmxext); + VP8_MC_FUNC(1, 8, mmxext); + VP8_BILINEAR_MC_FUNC(0, 16, mmxext); + VP8_BILINEAR_MC_FUNC(1, 8, mmxext); +#endif + } + + if (EXTERNAL_SSE(cpu_flags)) { + c->put_vp8_epel_pixels_tab[0][0][0] = + c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse; + } + + if (EXTERNAL_SSE2(cpu_flags) && (cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { + VP8_LUMA_MC_FUNC(0, 16, sse2); + VP8_MC_FUNC(1, 8, sse2); + VP8_BILINEAR_MC_FUNC(0, 16, sse2); + VP8_BILINEAR_MC_FUNC(1, 8, sse2); + } + + if (EXTERNAL_SSSE3(cpu_flags)) { + VP8_LUMA_MC_FUNC(0, 16, ssse3); + VP8_MC_FUNC(1, 8, ssse3); + VP8_MC_FUNC(2, 4, ssse3); + VP8_BILINEAR_MC_FUNC(0, 16, ssse3); + VP8_BILINEAR_MC_FUNC(1, 8, ssse3); + VP8_BILINEAR_MC_FUNC(2, 4, ssse3); + } +#endif /* HAVE_YASM */ +} +av_cold void ff_vp8dsp_init_x86(VP8DSPContext *c) +{ +#if HAVE_YASM + int cpu_flags = av_get_cpu_flags(); + + if (EXTERNAL_MMX(cpu_flags)) { + c->vp8_idct_dc_add = ff_vp8_idct_dc_add_mmx; + c->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_mmx; #if ARCH_X86_32 + c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_mmx; + c->vp8_idct_add = ff_vp8_idct_add_mmx; + c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_mmx; + c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmx; c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmx; @@ -352,14 +396,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c) /* note that 4-tap width=16 functions are missing because w=16 * is only used for luma, and luma is always a copy or sixtap. */ if (EXTERNAL_MMXEXT(cpu_flags)) { - VP8_MC_FUNC(2, 4, mmxext); - VP8_BILINEAR_MC_FUNC(2, 4, mmxext); #if ARCH_X86_32 - VP8_LUMA_MC_FUNC(0, 16, mmxext); - VP8_MC_FUNC(1, 8, mmxext); - VP8_BILINEAR_MC_FUNC(0, 16, mmxext); - VP8_BILINEAR_MC_FUNC(1, 8, mmxext); - c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmxext; c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmxext; @@ -378,16 +415,9 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c) if (EXTERNAL_SSE(cpu_flags)) { c->vp8_idct_add = ff_vp8_idct_add_sse; c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_sse; - c->put_vp8_epel_pixels_tab[0][0][0] = - c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse; } if (EXTERNAL_SSE2(cpu_flags) && (cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { - VP8_LUMA_MC_FUNC(0, 16, sse2); - VP8_MC_FUNC(1, 8, sse2); - VP8_BILINEAR_MC_FUNC(0, 16, sse2); - VP8_BILINEAR_MC_FUNC(1, 8, sse2); - c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2; c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2; @@ -410,13 +440,6 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c) } if (EXTERNAL_SSSE3(cpu_flags)) { - VP8_LUMA_MC_FUNC(0, 16, ssse3); - VP8_MC_FUNC(1, 8, ssse3); - VP8_MC_FUNC(2, 4, ssse3); - VP8_BILINEAR_MC_FUNC(0, 16, ssse3); - VP8_BILINEAR_MC_FUNC(1, 8, ssse3); - VP8_BILINEAR_MC_FUNC(2, 4, ssse3); - c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_ssse3; c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_ssse3; |