CAVS decoder by (Stefan Gehrer stefan.gehrer gmx.de)

Originally committed as revision 5590 to svn://svn.ffmpeg.org/ffmpeg/trunk

9 files changed, 161 insertions, 166 deletions

diff --git a/libavcodec/Makefile b/libavcodec/Makefile
index cd48f816e7..fd212f4aba 100644
--- a/libavcodec/Makefile
+++ b/libavcodec/Makefile
@@ -32,6 +32,9 @@ endif
 ifeq ($(CONFIG_AVS_DECODER),yes)
     OBJS+= avs.o
 endif
+ifeq ($(CONFIG_CAVS_DECODER),yes)
+    OBJS+= cavs.o cavsdsp.o
+endif
 ifeq ($(CONFIG_CINEPAK_DECODER),yes)
     OBJS+= cinepak.o
 endif
diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c
index f348617c05..ef97b871d3 100644
--- a/libavcodec/allcodecs.c
+++ b/libavcodec/allcodecs.c
@@ -529,6 +529,9 @@ void avcodec_register_all(void)
 #ifdef CONFIG_AVS_DECODER
     register_avcodec(&avs_decoder);
 #endif //CONFIG_AVS_DECODER
+#ifdef CONFIG_CAVS_DECODER
+    register_avcodec(&cavs_decoder);
+#endif //CONFIG_CAVS_DECODER
 #ifdef CONFIG_RAWVIDEO_DECODER
     register_avcodec(&rawvideo_decoder);
 #endif //CONFIG_RAWVIDEO_DECODER
@@ -637,6 +640,7 @@ PCM_CODEC(CODEC_ID_ADPCM_SBPRO_2, adpcm_sbpro_2);
     /* parsers */
     av_register_codec_parser(&mpegvideo_parser);
     av_register_codec_parser(&mpeg4video_parser);
+    av_register_codec_parser(&cavsvideo_parser);
 #if defined(CONFIG_H261_DECODER) || defined(CONFIG_H261_ENCODER)
     av_register_codec_parser(&h261_parser);
 #endif
diff --git a/libavcodec/cavs.c b/libavcodec/cavs.c
index 115c6ae5bf..0773045146 100644
--- a/libavcodec/cavs.c
+++ b/libavcodec/cavs.c
@@ -76,7 +76,7 @@ typedef struct {
     /* intra prediction is done with un-deblocked samples
      they are saved here before deblocking the MB  */
     uint8_t *top_border_y, *top_border_u, *top_border_v;
-    uint8_t left_border_y[16], left_border_u[8], left_border_v[8];
+    uint8_t left_border_y[16], left_border_u[10], left_border_v[10];
     uint8_t topleft_border_y, topleft_border_u, topleft_border_v;
 
     void (*intra_pred_l[8])(uint8_t *d,uint8_t *top,uint8_t *left,int stride);
@@ -136,22 +136,22 @@ static inline int get_bs_b(vector_t *mvP, vector_t *mvQ) {
     tc    =    tc_tab[clip(qp_avg + h->alpha_offset,0,63)];
 
 static void filter_mb(AVSContext *h, enum mb_t mb_type) {
-    uint8_t bs[8];
+    DECLARE_ALIGNED_8(uint8_t, bs[8]);
     int qp_avg, alpha, beta, tc;
     int i;
 
     /* save un-deblocked lines */
     h->topleft_border_y = h->top_border_y[h->mbx*16+15];
-    h->topleft_border_u = h->top_border_u[h->mbx*8+7];
-    h->topleft_border_v = h->top_border_v[h->mbx*8+7];
+    h->topleft_border_u = h->top_border_u[h->mbx*10+8];
+    h->topleft_border_v = h->top_border_v[h->mbx*10+8];
     memcpy(&h->top_border_y[h->mbx*16], h->cy + 15* h->l_stride,16);
-    memcpy(&h->top_border_u[h->mbx* 8], h->cu +  7* h->c_stride,8);
-    memcpy(&h->top_border_v[h->mbx* 8], h->cv +  7* h->c_stride,8);
+    memcpy(&h->top_border_u[h->mbx*10+1], h->cu +  7* h->c_stride,8);
+    memcpy(&h->top_border_v[h->mbx*10+1], h->cv +  7* h->c_stride,8);
     for(i=0;i<8;i++) {
         h->left_border_y[i*2+0] = *(h->cy + 15 + (i*2+0)*h->l_stride);
         h->left_border_y[i*2+1] = *(h->cy + 15 + (i*2+1)*h->l_stride);
-        h->left_border_u[i] = *(h->cu + 7 + i*h->c_stride);
-        h->left_border_v[i] = *(h->cv + 7 + i*h->c_stride);
+        h->left_border_u[i+1] = *(h->cu + 7 + i*h->c_stride);
+        h->left_border_v[i+1] = *(h->cv + 7 + i*h->c_stride);
     }
     if(!h->loop_filter_disable) {
         /* clear bs */
@@ -286,27 +286,6 @@ static inline void load_intra_pred_luma(AVSContext *h, uint8_t *top,
     }
 }
 
-static inline void load_intra_pred_chroma(uint8_t *stop, uint8_t *sleft,
-                                          uint8_t stopleft, uint8_t *dtop,
-                                          uint8_t *dleft, int stride, int flags) {
-    int i;
-
-    if(flags & A_AVAIL) {
-        for(i=0; i<8; i++)
-            dleft[i+1] = sleft[i];
-        dleft[0] = dleft[1];
-        dleft[9] = dleft[8];
-    }
-    if(flags & B_AVAIL) {
-        for(i=0; i<8; i++)
-            dtop[i+1] = stop[i];
-        dtop[0] = dtop[1];
-        dtop[9] = dtop[8];
-        if(flags & A_AVAIL)
-            dleft[0] = dtop[0] = stopleft;
-    }
-}
-
 static void intra_pred_vert(uint8_t *d,uint8_t *top,uint8_t *left,int stride) {
     int y;
     uint64_t a = *((uint64_t *)(&top[1]));
@@ -394,7 +373,7 @@ static void intra_pred_lp_top(uint8_t *d,uint8_t *top,uint8_t *left,int stride)
 
 #undef LOWPASS
 
-static inline void modify_pred(const int8_t *mod_table, int *mode) {
+static inline void modify_pred(const int_fast8_t *mod_table, int *mode) {
     int newmode = mod_table[*mode];
     if(newmode < 0) {
         av_log(NULL, AV_LOG_ERROR, "Illegal intra prediction mode\n");
@@ -688,8 +667,10 @@ static void mv_pred(AVSContext *h, enum mv_loc_t nP, enum mv_loc_t nC,
 
 /* kth-order exponential golomb code */
 static inline int get_ue_code(GetBitContext *gb, int order) {
-    if(order)
-        return (get_ue_golomb(gb) << order) + get_bits(gb,order);
+    if(order) {
+        int ret = get_ue_golomb(gb) << order;
+        return ret + get_bits(gb,order);
+    }
     return get_ue_golomb(gb);
 }
 
@@ -730,7 +711,7 @@ static int decode_residual_block(AVSContext *h, GetBitContext *gb,
         run_buf[i] = run;
     }
     /* inverse scan and dequantization */
-    for(i=i-1;i>=0;i--) {
+    while(--i >= 0){
         pos += 1 + run_buf[i];
         if(pos > 63) {
             av_log(h->s.avctx, AV_LOG_ERROR,
@@ -920,12 +901,24 @@ static void decode_mb_i(AVSContext *h, int is_i_pic) {
     }
 
     /* chroma intra prediction */
-    load_intra_pred_chroma(&h->top_border_u[h->mbx*8], h->left_border_u,
-                           h->topleft_border_u, top, left, h->c_stride, h->flags);
-    h->intra_pred_c[pred_mode_uv](h->cu, top, left, h->c_stride);
-    load_intra_pred_chroma(&h->top_border_v[h->mbx*8], h->left_border_v,
-                           h->topleft_border_v, top, left, h->c_stride, h->flags);
-    h->intra_pred_c[pred_mode_uv](h->cv, top, left, h->c_stride);
+    /* extend borders by one pixel */
+    h->left_border_u[9] = h->left_border_u[8];
+    h->left_border_v[9] = h->left_border_v[8];
+    h->top_border_u[h->mbx*10+9] = h->top_border_u[h->mbx*10+8];
+    h->top_border_v[h->mbx*10+9] = h->top_border_v[h->mbx*10+8];
+    if(h->mbx && h->mby) {
+        h->top_border_u[h->mbx*10] = h->left_border_u[0] = h->topleft_border_u;
+        h->top_border_v[h->mbx*10] = h->left_border_v[0] = h->topleft_border_v;
+    } else {
+        h->left_border_u[0] = h->left_border_u[1];
+        h->left_border_v[0] = h->left_border_v[1];
+        h->top_border_u[h->mbx*10] = h->top_border_u[h->mbx*10+1];
+        h->top_border_v[h->mbx*10] = h->top_border_v[h->mbx*10+1];
+    }
+    h->intra_pred_c[pred_mode_uv](h->cu, &h->top_border_u[h->mbx*10],
+                                  h->left_border_u, h->c_stride);
+    h->intra_pred_c[pred_mode_uv](h->cv, &h->top_border_v[h->mbx*10],
+                                  h->left_border_v, h->c_stride);
 
     decode_residual_chroma(h);
     filter_mb(h,I_8X8);
@@ -1324,10 +1317,10 @@ static void init_top_lines(AVSContext *h) {
     h->top_qp       = av_malloc( h->mb_width);
     h->top_mv[0]    = av_malloc((h->mb_width*2+1)*sizeof(vector_t));
     h->top_mv[1]    = av_malloc((h->mb_width*2+1)*sizeof(vector_t));
-    h->top_pred_Y   = av_malloc( h->mb_width*2*sizeof(int));
+    h->top_pred_Y   = av_malloc( h->mb_width*2*sizeof(*h->top_pred_Y));
     h->top_border_y = av_malloc((h->mb_width+1)*16);
-    h->top_border_u = av_malloc((h->mb_width+1)*8);
-    h->top_border_v = av_malloc((h->mb_width+1)*8);
+    h->top_border_u = av_malloc((h->mb_width)*10);
+    h->top_border_v = av_malloc((h->mb_width)*10);
 
     /* alloc space for co-located MVs and types */
     h->col_mv       = av_malloc( h->mb_width*h->mb_height*4*sizeof(vector_t));
@@ -1336,7 +1329,7 @@ static void init_top_lines(AVSContext *h) {
 
 static int decode_seq_header(AVSContext *h) {
     MpegEncContext *s = &h->s;
-    extern const AVRational frame_rate_tab[];
+    extern const AVRational ff_frame_rate_tab[];
     int frame_rate_code;
 
     h->profile =         get_bits(&s->gb,8);
@@ -1354,8 +1347,8 @@ static int decode_seq_header(AVSContext *h) {
     s->low_delay =       get_bits1(&s->gb);
     h->mb_width  = (s->width  + 15) >> 4;
     h->mb_height = (s->height + 15) >> 4;
-    h->s.avctx->time_base.den = frame_rate_tab[frame_rate_code].num;
-    h->s.avctx->time_base.num = frame_rate_tab[frame_rate_code].den;
+    h->s.avctx->time_base.den = ff_frame_rate_tab[frame_rate_code].num;
+    h->s.avctx->time_base.num = ff_frame_rate_tab[frame_rate_code].den;
     h->s.avctx->width  = s->width;
     h->s.avctx->height = s->height;
     if(!h->top_qp)
@@ -1541,6 +1534,6 @@ AVCodec cavs_decoder = {
     NULL,
     cavs_decode_end,
     cavs_decode_frame,
-    CODEC_CAP_TRUNCATED | CODEC_CAP_DELAY, //FIXME is this correct ?
+    CODEC_CAP_DR1 | CODEC_CAP_DELAY,
     .flush= ff_cavs_flush,
 };
diff --git a/libavcodec/cavsdata.h b/libavcodec/cavsdata.h
index a1334d554c..a73a3d60d7 100644
--- a/libavcodec/cavsdata.h
+++ b/libavcodec/cavsdata.h
@@ -187,7 +187,7 @@ static const uint16_t dequant_mul[64] = {
   32771,35734,38965,42497,46341,50535,55109,60099
 };
 
-typedef struct {
+DECLARE_ALIGNED_8(typedef, struct) {
     int16_t x;
     int16_t y;
     int16_t dist;
diff --git a/libavcodec/cavsdsp.c b/libavcodec/cavsdsp.c
index 71aaddba58..6a1a4b4a46 100644
--- a/libavcodec/cavsdsp.c
+++ b/libavcodec/cavsdsp.c
@@ -22,7 +22,6 @@
 
 #include <stdio.h>
 #include "dsputil.h"
-#include "cavsdsp.h"
 
 /*****************************************************************************
  *
@@ -110,7 +109,7 @@ static inline void loop_filter_c1(uint8_t *p0_p,int stride,int alpha, int beta,
 #undef Q1
 #undef Q2
 
-void cavs_filter_lv_c(uint8_t *d, int stride, int alpha, int beta, int tc,
+static void cavs_filter_lv_c(uint8_t *d, int stride, int alpha, int beta, int tc,
                            int bs1, int bs2) {
     int i;
     if(bs1==2)
@@ -126,7 +125,7 @@ void cavs_filter_lv_c(uint8_t *d, int stride, int alpha, int beta, int tc,
     }
 }
 
-void cavs_filter_lh_c(uint8_t *d, int stride, int alpha, int beta, int tc,
+static void cavs_filter_lh_c(uint8_t *d, int stride, int alpha, int beta, int tc,
                            int bs1, int bs2) {
     int i;
     if(bs1==2)
@@ -142,7 +141,7 @@ void cavs_filter_lh_c(uint8_t *d, int stride, int alpha, int beta, int tc,
     }
 }
 
-void cavs_filter_cv_c(uint8_t *d, int stride, int alpha, int beta, int tc,
+static void cavs_filter_cv_c(uint8_t *d, int stride, int alpha, int beta, int tc,
                            int bs1, int bs2) {
     int i;
     if(bs1==2)
@@ -158,7 +157,7 @@ void cavs_filter_cv_c(uint8_t *d, int stride, int alpha, int beta, int tc,
     }
 }
 
-void cavs_filter_ch_c(uint8_t *d, int stride, int alpha, int beta, int tc,
+static void cavs_filter_ch_c(uint8_t *d, int stride, int alpha, int beta, int tc,
                            int bs1, int bs2) {
     int i;
     if(bs1==2)
@@ -180,7 +179,7 @@ void cavs_filter_ch_c(uint8_t *d, int stride, int alpha, int beta, int tc,
  *
  ****************************************************************************/
 
-void cavs_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride) {
+static void cavs_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride) {
     int i;
     DCTELEM (*src)[8] = (DCTELEM(*)[8])block;
     uint8_t *cm = cropTbl + MAX_NEG_CROP;
@@ -416,66 +415,63 @@ static void OPNAME ## cavs_filt16_hv_ ## NAME(uint8_t *dst, uint8_t *src1, uint8
 }\
 
 #define CAVS_MC(OPNAME, SIZE) \
-void OPNAME ## cavs_qpel ## SIZE ## _mc00_c (uint8_t *dst, uint8_t *src, int stride){\
-    OPNAME ## pixels ## SIZE ## _c(dst, src, stride, SIZE);\
-}\
-void OPNAME ## cavs_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
     OPNAME ## cavs_filt ## SIZE ## _h_qpel_l(dst, src, stride, stride);\
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\
     OPNAME ## cavs_filt ## SIZE ## _h_hpel(dst, src, stride, stride);\
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
     OPNAME ## cavs_filt ## SIZE ## _h_qpel_r(dst, src, stride, stride);\
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
     OPNAME ## cavs_filt ## SIZE ## _v_qpel_l(dst, src, stride, stride);\
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\
     OPNAME ## cavs_filt ## SIZE ## _v_hpel(dst, src, stride, stride);\
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
     OPNAME ## cavs_filt ## SIZE ## _v_qpel_r(dst, src, stride, stride);\
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
   OPNAME ## cavs_filt ## SIZE ## _hv_jj(dst, src, NULL, stride, stride); \
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src, stride, stride); \
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+stride, stride, stride); \
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+1, stride, stride); \
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc33_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc33_c(uint8_t *dst, uint8_t *src, int stride){\
   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+stride+1,stride, stride); \
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
   OPNAME ## cavs_filt ## SIZE ## _hv_ff(dst, src, src+stride+1,stride, stride); \
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
   OPNAME ## cavs_filt ## SIZE ## _hv_ii(dst, src, src+stride+1,stride, stride); \
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
   OPNAME ## cavs_filt ## SIZE ## _hv_kk(dst, src, src+stride+1,stride, stride); \
 }\
 \
-void OPNAME ## cavs_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
+static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
   OPNAME ## cavs_filt ## SIZE ## _hv_qq(dst, src, src+stride+1,stride, stride); \
 }\
 
@@ -509,3 +505,37 @@ CAVS_MC(put_, 8)
 CAVS_MC(put_, 16)
 CAVS_MC(avg_, 8)
 CAVS_MC(avg_, 16)
+
+void ff_put_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
+void ff_avg_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
+void ff_put_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
+void ff_avg_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
+
+void ff_cavsdsp_init(DSPContext* c, AVCodecContext *avctx) {
+#define dspfunc(PFX, IDX, NUM) \
+    c->PFX ## _pixels_tab[IDX][ 0] = ff_ ## PFX ## NUM ## _mc00_c; \
+    c->PFX ## _pixels_tab[IDX][ 1] = ff_ ## PFX ## NUM ## _mc10_c; \
+    c->PFX ## _pixels_tab[IDX][ 2] = ff_ ## PFX ## NUM ## _mc20_c; \
+    c->PFX ## _pixels_tab[IDX][ 3] = ff_ ## PFX ## NUM ## _mc30_c; \
+    c->PFX ## _pixels_tab[IDX][ 4] = ff_ ## PFX ## NUM ## _mc01_c; \
+    c->PFX ## _pixels_tab[IDX][ 5] = ff_ ## PFX ## NUM ## _mc11_c; \
+    c->PFX ## _pixels_tab[IDX][ 6] = ff_ ## PFX ## NUM ## _mc21_c; \
+    c->PFX ## _pixels_tab[IDX][ 7] = ff_ ## PFX ## NUM ## _mc31_c; \
+    c->PFX ## _pixels_tab[IDX][ 8] = ff_ ## PFX ## NUM ## _mc02_c; \
+    c->PFX ## _pixels_tab[IDX][ 9] = ff_ ## PFX ## NUM ## _mc12_c; \
+    c->PFX ## _pixels_tab[IDX][10] = ff_ ## PFX ## NUM ## _mc22_c; \
+    c->PFX ## _pixels_tab[IDX][11] = ff_ ## PFX ## NUM ## _mc32_c; \
+    c->PFX ## _pixels_tab[IDX][12] = ff_ ## PFX ## NUM ## _mc03_c; \
+    c->PFX ## _pixels_tab[IDX][13] = ff_ ## PFX ## NUM ## _mc13_c; \
+    c->PFX ## _pixels_tab[IDX][14] = ff_ ## PFX ## NUM ## _mc23_c; \
+    c->PFX ## _pixels_tab[IDX][15] = ff_ ## PFX ## NUM ## _mc33_c
+    dspfunc(put_cavs_qpel, 0, 16);
+    dspfunc(put_cavs_qpel, 1, 8);
+    dspfunc(avg_cavs_qpel, 0, 16);
+    dspfunc(avg_cavs_qpel, 1, 8);
+    c->cavs_filter_lv = cavs_filter_lv_c;
+    c->cavs_filter_lh = cavs_filter_lh_c;
+    c->cavs_filter_cv = cavs_filter_cv_c;
+    c->cavs_filter_ch = cavs_filter_ch_c;
+    c->cavs_idct8_add = cavs_idct8_add_c;
+}
diff --git a/libavcodec/cavsdsp.h b/libavcodec/cavsdsp.h
deleted file mode 100644
index 9425b9e448..0000000000
--- a/libavcodec/cavsdsp.h
+++ /dev/null
@@ -1,95 +0,0 @@
-/*
- * Chinese AVS video (AVS1-P2, JiZhun profile) decoder.
- * Copyright (c) 2006  Stefan Gehrer <stefan.gehrer@gmx.de>
- *
- * DSP function prototypes
- *
- * This library is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2 of the License, or (at your option) any later version.
- *
- * This library is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with this library; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
- */
-
-void put_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc01_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc02_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc03_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc10_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc11_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc12_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc13_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc20_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc21_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc22_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc23_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc30_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc31_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc32_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc01_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc02_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc03_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc10_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc11_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc12_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc13_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc20_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc21_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc22_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc23_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc30_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc31_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc32_c(uint8_t *dst, uint8_t *src, int stride);
-void put_cavs_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc01_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc02_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc03_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc10_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc11_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc12_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc13_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc20_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc21_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc22_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc23_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc30_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc31_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc32_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc01_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc02_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc03_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc10_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc11_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc12_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc13_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc20_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc21_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc22_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc23_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc30_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc31_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc32_c(uint8_t *dst, uint8_t *src, int stride);
-void avg_cavs_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride);
-void cavs_filter_lv_c(uint8_t *d, int stride, int alpha, int beta, int tc, int bs1, int bs2);
-void cavs_filter_lh_c(uint8_t *d, int stride, int alpha, int beta, int tc, int bs1, int bs2);
-void cavs_filter_cv_c(uint8_t *d, int stride, int alpha, int beta, int tc, int bs1, int bs2);
-void cavs_filter_ch_c(uint8_t *d, int stride, int alpha, int beta, int tc, int bs1, int bs2);
-void cavs_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride);
-
-void put_pixels8_c(uint8_t *block, const uint8_t *pixels, int line_size, int h);
-void put_pixels16_c(uint8_t *block, const uint8_t *pixels, int line_size, int h);
-void avg_pixels8_c(uint8_t *block, const uint8_t *pixels, int line_size, int h);
-void avg_pixels16_c(uint8_t *block, const uint8_t *pixels, int line_size, int h);
diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index 453b38ab2a..f85794bd71 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -2570,6 +2570,22 @@ static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int
     }
 }
 
+/* AVS specific */
+void ff_cavsdsp_init(DSPContext* c, AVCodecContext *avctx);
+
+void ff_put_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride) {
+    put_pixels8_c(dst, src, stride, 8);
+}
+void ff_avg_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride) {
+    avg_pixels8_c(dst, src, stride, 8);
+}
+void ff_put_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride) {
+    put_pixels16_c(dst, src, stride, 16);
+}
+void ff_avg_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride) {
+    avg_pixels16_c(dst, src, stride, 16);
+}
+
 static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){
     uint8_t *cm = cropTbl + MAX_NEG_CROP;
     int i;
@@ -3989,6 +4005,8 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->biweight_h264_pixels_tab[8]= biweight_h264_pixels2x4_c;
     c->biweight_h264_pixels_tab[9]= biweight_h264_pixels2x2_c;
 
+    ff_cavsdsp_init(c,avctx);
+
     c->put_mspel_pixels_tab[0]= put_mspel8_mc00_c;
     c->put_mspel_pixels_tab[1]= put_mspel8_mc10_c;
     c->put_mspel_pixels_tab[2]= put_mspel8_mc20_c;
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index ce37cf46e5..fb84b20e00 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -274,6 +274,15 @@ typedef struct DSPContext {
     h264_weight_func weight_h264_pixels_tab[10];
     h264_biweight_func biweight_h264_pixels_tab[10];
 
+    /* AVS specific */
+    qpel_mc_func put_cavs_qpel_pixels_tab[2][16];
+    qpel_mc_func avg_cavs_qpel_pixels_tab[2][16];
+    void (*cavs_filter_lv)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
+    void (*cavs_filter_lh)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
+    void (*cavs_filter_cv)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
+    void (*cavs_filter_ch)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
+    void (*cavs_idct8_add)(uint8_t *dst, DCTELEM *block, int stride);
+
     me_cmp_func pix_abs[2][4];
 
     /* huffyuv specific */
diff --git a/libavcodec/parser.c b/libavcodec/parser.c
index 7281516628..1977dd4943 100644
--- a/libavcodec/parser.c
+++ b/libavcodec/parser.c
@@ -533,6 +533,30 @@ static int mpeg4video_parse(AVCodecParserContext *s,
     return next;
 }
 
+static int cavsvideo_parse(AVCodecParserContext *s,
+                           AVCodecContext *avctx,
+                           uint8_t **poutbuf, int *poutbuf_size,
+                           const uint8_t *buf, int buf_size)
+{
+    ParseContext *pc = s->priv_data;
+    int next;
+
+    if(s->flags & PARSER_FLAG_COMPLETE_FRAMES){
+        next= buf_size;
+    }else{
+        next= ff_cavs_find_frame_end(pc, buf, buf_size);
+
+        if (ff_combine_frame(pc, next, (uint8_t **)&buf, &buf_size) < 0) {
+            *poutbuf = NULL;
+            *poutbuf_size = 0;
+            return buf_size;
+        }
+    }
+    *poutbuf = (uint8_t *)buf;
+    *poutbuf_size = buf_size;
+    return next;
+}
+
 static int mpeg4video_split(AVCodecContext *avctx,
                            const uint8_t *buf, int buf_size)
 {
@@ -987,6 +1011,15 @@ AVCodecParser mpeg4video_parser = {
     mpeg4video_split,
 };
 
+AVCodecParser cavsvideo_parser = {
+    { CODEC_ID_CAVS },
+    sizeof(ParseContext1),
+    NULL,
+    cavsvideo_parse,
+    parse1_close,
+    mpeg4video_split,
+};
+
 AVCodecParser mpegaudio_parser = {
     { CODEC_ID_MP2, CODEC_ID_MP3 },
     sizeof(MpegAudioParseContext),