Split direct mode (macro)block decoding off h264.c.
authorMichael Niedermayer <michaelni@gmx.at>
Tue, 12 Jan 2010 14:32:58 +0000 (14:32 +0000)
committerMichael Niedermayer <michaelni@gmx.at>
Tue, 12 Jan 2010 14:32:58 +0000 (14:32 +0000)
No speedloss meassured (its slightly faster here but that may be random fluctuations)

Originally committed as revision 21165 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/Makefile
libavcodec/h264.c
libavcodec/h264.h
libavcodec/h264_direct.c [new file with mode: 0644]

index bb00904..8fef51a 100644 (file)
@@ -131,7 +131,7 @@ OBJS-$(CONFIG_H263_VAAPI_HWACCEL)      += vaapi_mpeg4.o
 OBJS-$(CONFIG_H263_ENCODER)            += mpegvideo_enc.o mpeg4video.o mpeg4videoenc.o motion_est.o      \
                                           ratecontrol.o h263.o ituh263enc.o flvenc.o mpeg12data.o \
                                           mpegvideo.o error_resilience.o
-OBJS-$(CONFIG_H264_DECODER)            += h264.o h264idct.o h264pred.o h264_loopfilter.o cabac.o \
+OBJS-$(CONFIG_H264_DECODER)            += h264.o h264idct.o h264pred.o h264_loopfilter.o h264_direct.o cabac.o \
                                           mpegvideo.o error_resilience.o
 OBJS-$(CONFIG_H264_ENCODER)            += h264enc.o h264dspenc.o
 OBJS-$(CONFIG_H264_VAAPI_HWACCEL)      += vaapi_h264.o
@@ -281,7 +281,7 @@ OBJS-$(CONFIG_SVQ1_DECODER)            += svq1dec.o svq1.o h263.o \
 OBJS-$(CONFIG_SVQ1_ENCODER)            += svq1enc.o svq1.o    \
                                           motion_est.o h263.o \
                                           mpegvideo.o error_resilience.o
-OBJS-$(CONFIG_SVQ3_DECODER)            += h264.o svq3.o h264idct.o h264pred.o h264_loopfilter.o  \
+OBJS-$(CONFIG_SVQ3_DECODER)            += h264.o svq3.o h264idct.o h264pred.o h264_loopfilter.o h264_direct.o  \
                                           cabac.o          \
                                           mpegvideo.o error_resilience.o \
                                           svq1dec.o svq1.o h263.o
index 9dfd912..bf45c7c 100644 (file)
@@ -726,433 +726,6 @@ static inline void pred_pskip_motion(H264Context * const h, int * const mx, int
     return;
 }
 
-static int get_scale_factor(H264Context * const h, int poc, int poc1, int i){
-    int poc0 = h->ref_list[0][i].poc;
-    int td = av_clip(poc1 - poc0, -128, 127);
-    if(td == 0 || h->ref_list[0][i].long_ref){
-        return 256;
-    }else{
-        int tb = av_clip(poc - poc0, -128, 127);
-        int tx = (16384 + (FFABS(td) >> 1)) / td;
-        return av_clip((tb*tx + 32) >> 6, -1024, 1023);
-    }
-}
-
-static inline void direct_dist_scale_factor(H264Context * const h){
-    MpegEncContext * const s = &h->s;
-    const int poc = h->s.current_picture_ptr->field_poc[ s->picture_structure == PICT_BOTTOM_FIELD ];
-    const int poc1 = h->ref_list[1][0].poc;
-    int i, field;
-    for(field=0; field<2; field++){
-        const int poc  = h->s.current_picture_ptr->field_poc[field];
-        const int poc1 = h->ref_list[1][0].field_poc[field];
-        for(i=0; i < 2*h->ref_count[0]; i++)
-            h->dist_scale_factor_field[field][i^field] = get_scale_factor(h, poc, poc1, i+16);
-    }
-
-    for(i=0; i<h->ref_count[0]; i++){
-        h->dist_scale_factor[i] = get_scale_factor(h, poc, poc1, i);
-    }
-}
-
-static void fill_colmap(H264Context *h, int map[2][16+32], int list, int field, int colfield, int mbafi){
-    MpegEncContext * const s = &h->s;
-    Picture * const ref1 = &h->ref_list[1][0];
-    int j, old_ref, rfield;
-    int start= mbafi ? 16                      : 0;
-    int end  = mbafi ? 16+2*h->ref_count[list] : h->ref_count[list];
-    int interl= mbafi || s->picture_structure != PICT_FRAME;
-
-    /* bogus; fills in for missing frames */
-    memset(map[list], 0, sizeof(map[list]));
-
-    for(rfield=0; rfield<2; rfield++){
-        for(old_ref=0; old_ref<ref1->ref_count[colfield][list]; old_ref++){
-            int poc = ref1->ref_poc[colfield][list][old_ref];
-
-            if     (!interl)
-                poc |= 3;
-            else if( interl && (poc&3) == 3) //FIXME store all MBAFF references so this isnt needed
-                poc= (poc&~3) + rfield + 1;
-
-            for(j=start; j<end; j++){
-                if(4*h->ref_list[list][j].frame_num + (h->ref_list[list][j].reference&3) == poc){
-                    int cur_ref= mbafi ? (j-16)^field : j;
-                    map[list][2*old_ref + (rfield^field) + 16] = cur_ref;
-                    if(rfield == field)
-                        map[list][old_ref] = cur_ref;
-                    break;
-                }
-            }
-        }
-    }
-}
-
-static inline void direct_ref_list_init(H264Context * const h){
-    MpegEncContext * const s = &h->s;
-    Picture * const ref1 = &h->ref_list[1][0];
-    Picture * const cur = s->current_picture_ptr;
-    int list, j, field;
-    int sidx= (s->picture_structure&1)^1;
-    int ref1sidx= (ref1->reference&1)^1;
-
-    for(list=0; list<2; list++){
-        cur->ref_count[sidx][list] = h->ref_count[list];
-        for(j=0; j<h->ref_count[list]; j++)
-            cur->ref_poc[sidx][list][j] = 4*h->ref_list[list][j].frame_num + (h->ref_list[list][j].reference&3);
-    }
-
-    if(s->picture_structure == PICT_FRAME){
-        memcpy(cur->ref_count[1], cur->ref_count[0], sizeof(cur->ref_count[0]));
-        memcpy(cur->ref_poc  [1], cur->ref_poc  [0], sizeof(cur->ref_poc  [0]));
-    }
-
-    cur->mbaff= FRAME_MBAFF;
-
-    if(cur->pict_type != FF_B_TYPE || h->direct_spatial_mv_pred)
-        return;
-
-    for(list=0; list<2; list++){
-        fill_colmap(h, h->map_col_to_list0, list, sidx, ref1sidx, 0);
-        for(field=0; field<2; field++)
-            fill_colmap(h, h->map_col_to_list0_field[field], list, field, field, 1);
-    }
-}
-
-static inline void pred_direct_motion(H264Context * const h, int *mb_type){
-    MpegEncContext * const s = &h->s;
-    int b8_stride = h->b8_stride;
-    int b4_stride = h->b_stride;
-    int mb_xy = h->mb_xy;
-    int mb_type_col[2];
-    const int16_t (*l1mv0)[2], (*l1mv1)[2];
-    const int8_t *l1ref0, *l1ref1;
-    const int is_b8x8 = IS_8X8(*mb_type);
-    unsigned int sub_mb_type;
-    int i8, i4;
-
-    assert(h->ref_list[1][0].reference&3);
-
-#define MB_TYPE_16x16_OR_INTRA (MB_TYPE_16x16|MB_TYPE_INTRA4x4|MB_TYPE_INTRA16x16|MB_TYPE_INTRA_PCM)
-
-    if(IS_INTERLACED(h->ref_list[1][0].mb_type[mb_xy])){ // AFL/AFR/FR/FL -> AFL/FL
-        if(!IS_INTERLACED(*mb_type)){                    //     AFR/FR    -> AFL/FL
-            int cur_poc = s->current_picture_ptr->poc;
-            int *col_poc = h->ref_list[1]->field_poc;
-            int col_parity = FFABS(col_poc[0] - cur_poc) >= FFABS(col_poc[1] - cur_poc);
-            mb_xy= s->mb_x + ((s->mb_y&~1) + col_parity)*s->mb_stride;
-            b8_stride = 0;
-        }else if(!(s->picture_structure & h->ref_list[1][0].reference) && !h->ref_list[1][0].mbaff){// FL -> FL & differ parity
-            int fieldoff= 2*(h->ref_list[1][0].reference)-3;
-            mb_xy += s->mb_stride*fieldoff;
-        }
-        goto single_col;
-    }else{                                               // AFL/AFR/FR/FL -> AFR/FR
-        if(IS_INTERLACED(*mb_type)){                     // AFL       /FL -> AFR/FR
-            mb_xy= s->mb_x + (s->mb_y&~1)*s->mb_stride;
-            mb_type_col[0] = h->ref_list[1][0].mb_type[mb_xy];
-            mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy + s->mb_stride];
-            b8_stride *= 3;
-            b4_stride *= 6;
-            //FIXME IS_8X8(mb_type_col[0]) && !h->sps.direct_8x8_inference_flag
-            if(    (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)
-                && (mb_type_col[1] & MB_TYPE_16x16_OR_INTRA)
-                && !is_b8x8){
-                sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
-                *mb_type   |= MB_TYPE_16x8 |MB_TYPE_L0L1|MB_TYPE_DIRECT2; /* B_16x8 */
-            }else{
-                sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
-                *mb_type   |= MB_TYPE_8x8|MB_TYPE_L0L1;
-            }
-        }else{                                           //     AFR/FR    -> AFR/FR
-single_col:
-            mb_type_col[0] =
-            mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy];
-            if(IS_8X8(mb_type_col[0]) && !h->sps.direct_8x8_inference_flag){
-                /* FIXME save sub mb types from previous frames (or derive from MVs)
-                * so we know exactly what block size to use */
-                sub_mb_type = MB_TYPE_8x8|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_4x4 */
-                *mb_type   |= MB_TYPE_8x8|MB_TYPE_L0L1;
-            }else if(!is_b8x8 && (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)){
-                sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
-                *mb_type   |= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_16x16 */
-            }else{
-                sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
-                *mb_type   |= MB_TYPE_8x8|MB_TYPE_L0L1;
-            }
-        }
-    }
-
-    l1mv0  = &h->ref_list[1][0].motion_val[0][h->mb2b_xy [mb_xy]];
-    l1mv1  = &h->ref_list[1][0].motion_val[1][h->mb2b_xy [mb_xy]];
-    l1ref0 = &h->ref_list[1][0].ref_index [0][h->mb2b8_xy[mb_xy]];
-    l1ref1 = &h->ref_list[1][0].ref_index [1][h->mb2b8_xy[mb_xy]];
-    if(!b8_stride){
-        if(s->mb_y&1){
-            l1ref0 += h->b8_stride;
-            l1ref1 += h->b8_stride;
-            l1mv0  +=  2*b4_stride;
-            l1mv1  +=  2*b4_stride;
-        }
-    }
-
-    if(h->direct_spatial_mv_pred){
-        int ref[2];
-        int mv[2][2];
-        int list;
-
-        /* FIXME interlacing + spatial direct uses wrong colocated block positions */
-
-        /* ref = min(neighbors) */
-        for(list=0; list<2; list++){
-            int refa = h->ref_cache[list][scan8[0] - 1];
-            int refb = h->ref_cache[list][scan8[0] - 8];
-            int refc = h->ref_cache[list][scan8[0] - 8 + 4];
-            if(refc == PART_NOT_AVAILABLE)
-                refc = h->ref_cache[list][scan8[0] - 8 - 1];
-            ref[list] = FFMIN3((unsigned)refa, (unsigned)refb, (unsigned)refc);
-            if(ref[list] < 0)
-                ref[list] = -1;
-        }
-
-        if(ref[0] < 0 && ref[1] < 0){
-            ref[0] = ref[1] = 0;
-            mv[0][0] = mv[0][1] =
-            mv[1][0] = mv[1][1] = 0;
-        }else{
-            for(list=0; list<2; list++){
-                if(ref[list] >= 0)
-                    pred_motion(h, 0, 4, list, ref[list], &mv[list][0], &mv[list][1]);
-                else
-                    mv[list][0] = mv[list][1] = 0;
-            }
-        }
-
-        if(ref[1] < 0){
-            if(!is_b8x8)
-                *mb_type &= ~MB_TYPE_L1;
-            sub_mb_type &= ~MB_TYPE_L1;
-        }else if(ref[0] < 0){
-            if(!is_b8x8)
-                *mb_type &= ~MB_TYPE_L0;
-            sub_mb_type &= ~MB_TYPE_L0;
-        }
-
-        if(IS_INTERLACED(*mb_type) != IS_INTERLACED(mb_type_col[0])){
-            for(i8=0; i8<4; i8++){
-                int x8 = i8&1;
-                int y8 = i8>>1;
-                int xy8 = x8+y8*b8_stride;
-                int xy4 = 3*x8+y8*b4_stride;
-                int a=0, b=0;
-
-                if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
-                    continue;
-                h->sub_mb_type[i8] = sub_mb_type;
-
-                fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[0], 1);
-                fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[1], 1);
-                if(!IS_INTRA(mb_type_col[y8])
-                   && (   (l1ref0[xy8] == 0 && FFABS(l1mv0[xy4][0]) <= 1 && FFABS(l1mv0[xy4][1]) <= 1)
-                       || (l1ref0[xy8]  < 0 && l1ref1[xy8] == 0 && FFABS(l1mv1[xy4][0]) <= 1 && FFABS(l1mv1[xy4][1]) <= 1))){
-                    if(ref[0] > 0)
-                        a= pack16to32(mv[0][0],mv[0][1]);
-                    if(ref[1] > 0)
-                        b= pack16to32(mv[1][0],mv[1][1]);
-                }else{
-                    a= pack16to32(mv[0][0],mv[0][1]);
-                    b= pack16to32(mv[1][0],mv[1][1]);
-                }
-                fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, a, 4);
-                fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, b, 4);
-            }
-        }else if(IS_16X16(*mb_type)){
-            int a=0, b=0;
-
-            fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, (uint8_t)ref[0], 1);
-            fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, (uint8_t)ref[1], 1);
-            if(!IS_INTRA(mb_type_col[0])
-               && (   (l1ref0[0] == 0 && FFABS(l1mv0[0][0]) <= 1 && FFABS(l1mv0[0][1]) <= 1)
-                   || (l1ref0[0]  < 0 && l1ref1[0] == 0 && FFABS(l1mv1[0][0]) <= 1 && FFABS(l1mv1[0][1]) <= 1
-                       && (h->x264_build>33 || !h->x264_build)))){
-                if(ref[0] > 0)
-                    a= pack16to32(mv[0][0],mv[0][1]);
-                if(ref[1] > 0)
-                    b= pack16to32(mv[1][0],mv[1][1]);
-            }else{
-                a= pack16to32(mv[0][0],mv[0][1]);
-                b= pack16to32(mv[1][0],mv[1][1]);
-            }
-            fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, a, 4);
-            fill_rectangle(&h->mv_cache[1][scan8[0]], 4, 4, 8, b, 4);
-        }else{
-            for(i8=0; i8<4; i8++){
-                const int x8 = i8&1;
-                const int y8 = i8>>1;
-
-                if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
-                    continue;
-                h->sub_mb_type[i8] = sub_mb_type;
-
-                fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mv[0][0],mv[0][1]), 4);
-                fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mv[1][0],mv[1][1]), 4);
-                fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[0], 1);
-                fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[1], 1);
-
-                /* col_zero_flag */
-                if(!IS_INTRA(mb_type_col[0]) && (   l1ref0[x8 + y8*b8_stride] == 0
-                                              || (l1ref0[x8 + y8*b8_stride] < 0 && l1ref1[x8 + y8*b8_stride] == 0
-                                                  && (h->x264_build>33 || !h->x264_build)))){
-                    const int16_t (*l1mv)[2]= l1ref0[x8 + y8*b8_stride] == 0 ? l1mv0 : l1mv1;
-                    if(IS_SUB_8X8(sub_mb_type)){
-                        const int16_t *mv_col = l1mv[x8*3 + y8*3*b4_stride];
-                        if(FFABS(mv_col[0]) <= 1 && FFABS(mv_col[1]) <= 1){
-                            if(ref[0] == 0)
-                                fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
-                            if(ref[1] == 0)
-                                fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
-                        }
-                    }else
-                    for(i4=0; i4<4; i4++){
-                        const int16_t *mv_col = l1mv[x8*2 + (i4&1) + (y8*2 + (i4>>1))*b4_stride];
-                        if(FFABS(mv_col[0]) <= 1 && FFABS(mv_col[1]) <= 1){
-                            if(ref[0] == 0)
-                                *(uint32_t*)h->mv_cache[0][scan8[i8*4+i4]] = 0;
-                            if(ref[1] == 0)
-                                *(uint32_t*)h->mv_cache[1][scan8[i8*4+i4]] = 0;
-                        }
-                    }
-                }
-            }
-        }
-    }else{ /* direct temporal mv pred */
-        const int *map_col_to_list0[2] = {h->map_col_to_list0[0], h->map_col_to_list0[1]};
-        const int *dist_scale_factor = h->dist_scale_factor;
-        int ref_offset= 0;
-
-        if(FRAME_MBAFF && IS_INTERLACED(*mb_type)){
-            map_col_to_list0[0] = h->map_col_to_list0_field[s->mb_y&1][0];
-            map_col_to_list0[1] = h->map_col_to_list0_field[s->mb_y&1][1];
-            dist_scale_factor   =h->dist_scale_factor_field[s->mb_y&1];
-        }
-        if(h->ref_list[1][0].mbaff && IS_INTERLACED(mb_type_col[0]))
-            ref_offset += 16;
-
-        if(IS_INTERLACED(*mb_type) != IS_INTERLACED(mb_type_col[0])){
-            /* FIXME assumes direct_8x8_inference == 1 */
-            int y_shift  = 2*!IS_INTERLACED(*mb_type);
-
-            for(i8=0; i8<4; i8++){
-                const int x8 = i8&1;
-                const int y8 = i8>>1;
-                int ref0, scale;
-                const int16_t (*l1mv)[2]= l1mv0;
-
-                if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
-                    continue;
-                h->sub_mb_type[i8] = sub_mb_type;
-
-                fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
-                if(IS_INTRA(mb_type_col[y8])){
-                    fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, 0, 1);
-                    fill_rectangle(&h-> mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
-                    fill_rectangle(&h-> mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
-                    continue;
-                }
-
-                ref0 = l1ref0[x8 + y8*b8_stride];
-                if(ref0 >= 0)
-                    ref0 = map_col_to_list0[0][ref0 + ref_offset];
-                else{
-                    ref0 = map_col_to_list0[1][l1ref1[x8 + y8*b8_stride] + ref_offset];
-                    l1mv= l1mv1;
-                }
-                scale = dist_scale_factor[ref0];
-                fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref0, 1);
-
-                {
-                    const int16_t *mv_col = l1mv[x8*3 + y8*b4_stride];
-                    int my_col = (mv_col[1]<<y_shift)/2;
-                    int mx = (scale * mv_col[0] + 128) >> 8;
-                    int my = (scale * my_col + 128) >> 8;
-                    fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mx,my), 4);
-                    fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mx-mv_col[0],my-my_col), 4);
-                }
-            }
-            return;
-        }
-
-        /* one-to-one mv scaling */
-
-        if(IS_16X16(*mb_type)){
-            int ref, mv0, mv1;
-
-            fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, 0, 1);
-            if(IS_INTRA(mb_type_col[0])){
-                ref=mv0=mv1=0;
-            }else{
-                const int ref0 = l1ref0[0] >= 0 ? map_col_to_list0[0][l1ref0[0] + ref_offset]
-                                                : map_col_to_list0[1][l1ref1[0] + ref_offset];
-                const int scale = dist_scale_factor[ref0];
-                const int16_t *mv_col = l1ref0[0] >= 0 ? l1mv0[0] : l1mv1[0];
-                int mv_l0[2];
-                mv_l0[0] = (scale * mv_col[0] + 128) >> 8;
-                mv_l0[1] = (scale * mv_col[1] + 128) >> 8;
-                ref= ref0;
-                mv0= pack16to32(mv_l0[0],mv_l0[1]);
-                mv1= pack16to32(mv_l0[0]-mv_col[0],mv_l0[1]-mv_col[1]);
-            }
-            fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, ref, 1);
-            fill_rectangle(&h-> mv_cache[0][scan8[0]], 4, 4, 8, mv0, 4);
-            fill_rectangle(&h-> mv_cache[1][scan8[0]], 4, 4, 8, mv1, 4);
-        }else{
-            for(i8=0; i8<4; i8++){
-                const int x8 = i8&1;
-                const int y8 = i8>>1;
-                int ref0, scale;
-                const int16_t (*l1mv)[2]= l1mv0;
-
-                if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
-                    continue;
-                h->sub_mb_type[i8] = sub_mb_type;
-                fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
-                if(IS_INTRA(mb_type_col[0])){
-                    fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, 0, 1);
-                    fill_rectangle(&h-> mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
-                    fill_rectangle(&h-> mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
-                    continue;
-                }
-
-                ref0 = l1ref0[x8 + y8*b8_stride] + ref_offset;
-                if(ref0 >= 0)
-                    ref0 = map_col_to_list0[0][ref0];
-                else{
-                    ref0 = map_col_to_list0[1][l1ref1[x8 + y8*b8_stride] + ref_offset];
-                    l1mv= l1mv1;
-                }
-                scale = dist_scale_factor[ref0];
-
-                fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref0, 1);
-                if(IS_SUB_8X8(sub_mb_type)){
-                    const int16_t *mv_col = l1mv[x8*3 + y8*3*b4_stride];
-                    int mx = (scale * mv_col[0] + 128) >> 8;
-                    int my = (scale * mv_col[1] + 128) >> 8;
-                    fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mx,my), 4);
-                    fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mx-mv_col[0],my-mv_col[1]), 4);
-                }else
-                for(i4=0; i4<4; i4++){
-                    const int16_t *mv_col = l1mv[x8*2 + (i4&1) + (y8*2 + (i4>>1))*b4_stride];
-                    int16_t *mv_l0 = h->mv_cache[0][scan8[i8*4+i4]];
-                    mv_l0[0] = (scale * mv_col[0] + 128) >> 8;
-                    mv_l0[1] = (scale * mv_col[1] + 128) >> 8;
-                    *(uint32_t*)h->mv_cache[1][scan8[i8*4+i4]] =
-                        pack16to32(mv_l0[0]-mv_col[0],mv_l0[1]-mv_col[1]);
-                }
-            }
-        }
-    }
-}
-
 static inline void write_back_motion(H264Context *h, int mb_type){
     MpegEncContext * const s = &h->s;
     const int b_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride;
@@ -3871,8 +3444,8 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
         fill_mbaff_ref_list(h);
 
     if(h->slice_type_nos==FF_B_TYPE && !h->direct_spatial_mv_pred)
-        direct_dist_scale_factor(h);
-    direct_ref_list_init(h);
+        ff_h264_direct_dist_scale_factor(h);
+    ff_h264_direct_ref_list_init(h);
 
     if( h->slice_type_nos != FF_I_TYPE && h->pps.cabac ){
         tmp = get_ue_golomb_31(&s->gb);
@@ -4239,7 +3812,7 @@ static void decode_mb_skip(H264Context *h){
         mb_type|= MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2|MB_TYPE_SKIP;
 
         fill_caches(h, mb_type, 0); //FIXME check what is needed and what not ...
-        pred_direct_motion(h, &mb_type);
+        ff_h264_pred_direct_motion(h, &mb_type);
         mb_type|= MB_TYPE_SKIP;
     }
     else
@@ -4417,7 +3990,7 @@ decode_intra_mb:
             }
             if(   IS_DIRECT(h->sub_mb_type[0]) || IS_DIRECT(h->sub_mb_type[1])
                || IS_DIRECT(h->sub_mb_type[2]) || IS_DIRECT(h->sub_mb_type[3])) {
-                pred_direct_motion(h, &mb_type);
+                ff_h264_pred_direct_motion(h, &mb_type);
                 h->ref_cache[0][scan8[4]] =
                 h->ref_cache[1][scan8[4]] =
                 h->ref_cache[0][scan8[12]] =
@@ -4508,7 +4081,7 @@ decode_intra_mb:
             }
         }
     }else if(IS_DIRECT(mb_type)){
-        pred_direct_motion(h, &mb_type);
+        ff_h264_pred_direct_motion(h, &mb_type);
         dct8x8_allowed &= h->sps.direct_8x8_inference_flag;
     }else{
         int list, mx, my, i;
@@ -5514,7 +5087,7 @@ decode_intra_mb:
             }
             if( IS_DIRECT(h->sub_mb_type[0] | h->sub_mb_type[1] |
                           h->sub_mb_type[2] | h->sub_mb_type[3]) ) {
-                pred_direct_motion(h, &mb_type);
+                ff_h264_pred_direct_motion(h, &mb_type);
                 h->ref_cache[0][scan8[4]] =
                 h->ref_cache[1][scan8[4]] =
                 h->ref_cache[0][scan8[12]] =
@@ -5617,7 +5190,7 @@ decode_intra_mb:
             }
         }
     } else if( IS_DIRECT(mb_type) ) {
-        pred_direct_motion(h, &mb_type);
+        ff_h264_pred_direct_motion(h, &mb_type);
         fill_rectangle(h->mvd_cache[0][scan8[0]], 4, 4, 8, 0, 4);
         fill_rectangle(h->mvd_cache[1][scan8[0]], 4, 4, 8, 0, 4);
         dct8x8_allowed &= h->sps.direct_8x8_inference_flag;
index ef2bdb0..93259ec 100644 (file)
@@ -610,6 +610,10 @@ int ff_h264_frame_start(H264Context *h);
 av_cold int ff_h264_decode_init(AVCodecContext *avctx);
 av_cold int ff_h264_decode_end(AVCodecContext *avctx);
 
+void ff_h264_direct_dist_scale_factor(H264Context * const h);
+void ff_h264_direct_ref_list_init(H264Context * const h);
+void ff_h264_pred_direct_motion(H264Context * const h, int *mb_type);
+
 void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize);
 void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize);
 
diff --git a/libavcodec/h264_direct.c b/libavcodec/h264_direct.c
new file mode 100644 (file)
index 0000000..323c4cf
--- /dev/null
@@ -0,0 +1,468 @@
+/*
+ * H.26L/H.264/AVC/JVT/14496-10/... direct mb/block decoding
+ * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file libavcodec/h264_direct.c
+ * H.264 / AVC / MPEG4 part10 direct mb/block decoding.
+ * @author Michael Niedermayer <michaelni@gmx.at>
+ */
+
+#include "internal.h"
+#include "dsputil.h"
+#include "avcodec.h"
+#include "mpegvideo.h"
+#include "h264.h"
+#include "rectangle.h"
+
+#if ARCH_X86
+#include "x86/h264_i386.h"
+#endif
+
+//#undef NDEBUG
+#include <assert.h>
+
+
+static int get_scale_factor(H264Context * const h, int poc, int poc1, int i){
+    int poc0 = h->ref_list[0][i].poc;
+    int td = av_clip(poc1 - poc0, -128, 127);
+    if(td == 0 || h->ref_list[0][i].long_ref){
+        return 256;
+    }else{
+        int tb = av_clip(poc - poc0, -128, 127);
+        int tx = (16384 + (FFABS(td) >> 1)) / td;
+        return av_clip((tb*tx + 32) >> 6, -1024, 1023);
+    }
+}
+
+void ff_h264_direct_dist_scale_factor(H264Context * const h){
+    MpegEncContext * const s = &h->s;
+    const int poc = h->s.current_picture_ptr->field_poc[ s->picture_structure == PICT_BOTTOM_FIELD ];
+    const int poc1 = h->ref_list[1][0].poc;
+    int i, field;
+    for(field=0; field<2; field++){
+        const int poc  = h->s.current_picture_ptr->field_poc[field];
+        const int poc1 = h->ref_list[1][0].field_poc[field];
+        for(i=0; i < 2*h->ref_count[0]; i++)
+            h->dist_scale_factor_field[field][i^field] = get_scale_factor(h, poc, poc1, i+16);
+    }
+
+    for(i=0; i<h->ref_count[0]; i++){
+        h->dist_scale_factor[i] = get_scale_factor(h, poc, poc1, i);
+    }
+}
+
+static void fill_colmap(H264Context *h, int map[2][16+32], int list, int field, int colfield, int mbafi){
+    MpegEncContext * const s = &h->s;
+    Picture * const ref1 = &h->ref_list[1][0];
+    int j, old_ref, rfield;
+    int start= mbafi ? 16                      : 0;
+    int end  = mbafi ? 16+2*h->ref_count[list] : h->ref_count[list];
+    int interl= mbafi || s->picture_structure != PICT_FRAME;
+
+    /* bogus; fills in for missing frames */
+    memset(map[list], 0, sizeof(map[list]));
+
+    for(rfield=0; rfield<2; rfield++){
+        for(old_ref=0; old_ref<ref1->ref_count[colfield][list]; old_ref++){
+            int poc = ref1->ref_poc[colfield][list][old_ref];
+
+            if     (!interl)
+                poc |= 3;
+            else if( interl && (poc&3) == 3) //FIXME store all MBAFF references so this isnt needed
+                poc= (poc&~3) + rfield + 1;
+
+            for(j=start; j<end; j++){
+                if(4*h->ref_list[list][j].frame_num + (h->ref_list[list][j].reference&3) == poc){
+                    int cur_ref= mbafi ? (j-16)^field : j;
+                    map[list][2*old_ref + (rfield^field) + 16] = cur_ref;
+                    if(rfield == field)
+                        map[list][old_ref] = cur_ref;
+                    break;
+                }
+            }
+        }
+    }
+}
+
+void ff_h264_direct_ref_list_init(H264Context * const h){
+    MpegEncContext * const s = &h->s;
+    Picture * const ref1 = &h->ref_list[1][0];
+    Picture * const cur = s->current_picture_ptr;
+    int list, j, field;
+    int sidx= (s->picture_structure&1)^1;
+    int ref1sidx= (ref1->reference&1)^1;
+
+    for(list=0; list<2; list++){
+        cur->ref_count[sidx][list] = h->ref_count[list];
+        for(j=0; j<h->ref_count[list]; j++)
+            cur->ref_poc[sidx][list][j] = 4*h->ref_list[list][j].frame_num + (h->ref_list[list][j].reference&3);
+    }
+
+    if(s->picture_structure == PICT_FRAME){
+        memcpy(cur->ref_count[1], cur->ref_count[0], sizeof(cur->ref_count[0]));
+        memcpy(cur->ref_poc  [1], cur->ref_poc  [0], sizeof(cur->ref_poc  [0]));
+    }
+
+    cur->mbaff= FRAME_MBAFF;
+
+    if(cur->pict_type != FF_B_TYPE || h->direct_spatial_mv_pred)
+        return;
+
+    for(list=0; list<2; list++){
+        fill_colmap(h, h->map_col_to_list0, list, sidx, ref1sidx, 0);
+        for(field=0; field<2; field++)
+            fill_colmap(h, h->map_col_to_list0_field[field], list, field, field, 1);
+    }
+}
+
+void ff_h264_pred_direct_motion(H264Context * const h, int *mb_type){
+    MpegEncContext * const s = &h->s;
+    int b8_stride = h->b8_stride;
+    int b4_stride = h->b_stride;
+    int mb_xy = h->mb_xy;
+    int mb_type_col[2];
+    const int16_t (*l1mv0)[2], (*l1mv1)[2];
+    const int8_t *l1ref0, *l1ref1;
+    const int is_b8x8 = IS_8X8(*mb_type);
+    unsigned int sub_mb_type;
+    int i8, i4;
+
+    assert(h->ref_list[1][0].reference&3);
+
+#define MB_TYPE_16x16_OR_INTRA (MB_TYPE_16x16|MB_TYPE_INTRA4x4|MB_TYPE_INTRA16x16|MB_TYPE_INTRA_PCM)
+
+    if(IS_INTERLACED(h->ref_list[1][0].mb_type[mb_xy])){ // AFL/AFR/FR/FL -> AFL/FL
+        if(!IS_INTERLACED(*mb_type)){                    //     AFR/FR    -> AFL/FL
+            int cur_poc = s->current_picture_ptr->poc;
+            int *col_poc = h->ref_list[1]->field_poc;
+            int col_parity = FFABS(col_poc[0] - cur_poc) >= FFABS(col_poc[1] - cur_poc);
+            mb_xy= s->mb_x + ((s->mb_y&~1) + col_parity)*s->mb_stride;
+            b8_stride = 0;
+        }else if(!(s->picture_structure & h->ref_list[1][0].reference) && !h->ref_list[1][0].mbaff){// FL -> FL & differ parity
+            int fieldoff= 2*(h->ref_list[1][0].reference)-3;
+            mb_xy += s->mb_stride*fieldoff;
+        }
+        goto single_col;
+    }else{                                               // AFL/AFR/FR/FL -> AFR/FR
+        if(IS_INTERLACED(*mb_type)){                     // AFL       /FL -> AFR/FR
+            mb_xy= s->mb_x + (s->mb_y&~1)*s->mb_stride;
+            mb_type_col[0] = h->ref_list[1][0].mb_type[mb_xy];
+            mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy + s->mb_stride];
+            b8_stride *= 3;
+            b4_stride *= 6;
+            //FIXME IS_8X8(mb_type_col[0]) && !h->sps.direct_8x8_inference_flag
+            if(    (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)
+                && (mb_type_col[1] & MB_TYPE_16x16_OR_INTRA)
+                && !is_b8x8){
+                sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
+                *mb_type   |= MB_TYPE_16x8 |MB_TYPE_L0L1|MB_TYPE_DIRECT2; /* B_16x8 */
+            }else{
+                sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
+                *mb_type   |= MB_TYPE_8x8|MB_TYPE_L0L1;
+            }
+        }else{                                           //     AFR/FR    -> AFR/FR
+single_col:
+            mb_type_col[0] =
+            mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy];
+            if(IS_8X8(mb_type_col[0]) && !h->sps.direct_8x8_inference_flag){
+                /* FIXME save sub mb types from previous frames (or derive from MVs)
+                * so we know exactly what block size to use */
+                sub_mb_type = MB_TYPE_8x8|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_4x4 */
+                *mb_type   |= MB_TYPE_8x8|MB_TYPE_L0L1;
+            }else if(!is_b8x8 && (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)){
+                sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
+                *mb_type   |= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_16x16 */
+            }else{
+                sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
+                *mb_type   |= MB_TYPE_8x8|MB_TYPE_L0L1;
+            }
+        }
+    }
+
+    l1mv0  = &h->ref_list[1][0].motion_val[0][h->mb2b_xy [mb_xy]];
+    l1mv1  = &h->ref_list[1][0].motion_val[1][h->mb2b_xy [mb_xy]];
+    l1ref0 = &h->ref_list[1][0].ref_index [0][h->mb2b8_xy[mb_xy]];
+    l1ref1 = &h->ref_list[1][0].ref_index [1][h->mb2b8_xy[mb_xy]];
+    if(!b8_stride){
+        if(s->mb_y&1){
+            l1ref0 += h->b8_stride;
+            l1ref1 += h->b8_stride;
+            l1mv0  +=  2*b4_stride;
+            l1mv1  +=  2*b4_stride;
+        }
+    }
+
+    if(h->direct_spatial_mv_pred){
+        int ref[2];
+        int mv[2][2];
+        int list;
+
+        /* FIXME interlacing + spatial direct uses wrong colocated block positions */
+
+        /* ref = min(neighbors) */
+        for(list=0; list<2; list++){
+            int refa = h->ref_cache[list][scan8[0] - 1];
+            int refb = h->ref_cache[list][scan8[0] - 8];
+            int refc = h->ref_cache[list][scan8[0] - 8 + 4];
+            if(refc == PART_NOT_AVAILABLE)
+                refc = h->ref_cache[list][scan8[0] - 8 - 1];
+            ref[list] = FFMIN3((unsigned)refa, (unsigned)refb, (unsigned)refc);
+            if(ref[list] < 0)
+                ref[list] = -1;
+        }
+
+        if(ref[0] < 0 && ref[1] < 0){
+            ref[0] = ref[1] = 0;
+            mv[0][0] = mv[0][1] =
+            mv[1][0] = mv[1][1] = 0;
+        }else{
+            for(list=0; list<2; list++){
+                if(ref[list] >= 0)
+                    pred_motion(h, 0, 4, list, ref[list], &mv[list][0], &mv[list][1]);
+                else
+                    mv[list][0] = mv[list][1] = 0;
+            }
+        }
+
+        if(ref[1] < 0){
+            if(!is_b8x8)
+                *mb_type &= ~MB_TYPE_L1;
+            sub_mb_type &= ~MB_TYPE_L1;
+        }else if(ref[0] < 0){
+            if(!is_b8x8)
+                *mb_type &= ~MB_TYPE_L0;
+            sub_mb_type &= ~MB_TYPE_L0;
+        }
+
+        if(IS_INTERLACED(*mb_type) != IS_INTERLACED(mb_type_col[0])){
+            for(i8=0; i8<4; i8++){
+                int x8 = i8&1;
+                int y8 = i8>>1;
+                int xy8 = x8+y8*b8_stride;
+                int xy4 = 3*x8+y8*b4_stride;
+                int a=0, b=0;
+
+                if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
+                    continue;
+                h->sub_mb_type[i8] = sub_mb_type;
+
+                fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[0], 1);
+                fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[1], 1);
+                if(!IS_INTRA(mb_type_col[y8])
+                   && (   (l1ref0[xy8] == 0 && FFABS(l1mv0[xy4][0]) <= 1 && FFABS(l1mv0[xy4][1]) <= 1)
+                       || (l1ref0[xy8]  < 0 && l1ref1[xy8] == 0 && FFABS(l1mv1[xy4][0]) <= 1 && FFABS(l1mv1[xy4][1]) <= 1))){
+                    if(ref[0] > 0)
+                        a= pack16to32(mv[0][0],mv[0][1]);
+                    if(ref[1] > 0)
+                        b= pack16to32(mv[1][0],mv[1][1]);
+                }else{
+                    a= pack16to32(mv[0][0],mv[0][1]);
+                    b= pack16to32(mv[1][0],mv[1][1]);
+                }
+                fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, a, 4);
+                fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, b, 4);
+            }
+        }else if(IS_16X16(*mb_type)){
+            int a=0, b=0;
+
+            fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, (uint8_t)ref[0], 1);
+            fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, (uint8_t)ref[1], 1);
+            if(!IS_INTRA(mb_type_col[0])
+               && (   (l1ref0[0] == 0 && FFABS(l1mv0[0][0]) <= 1 && FFABS(l1mv0[0][1]) <= 1)
+                   || (l1ref0[0]  < 0 && l1ref1[0] == 0 && FFABS(l1mv1[0][0]) <= 1 && FFABS(l1mv1[0][1]) <= 1
+                       && (h->x264_build>33 || !h->x264_build)))){
+                if(ref[0] > 0)
+                    a= pack16to32(mv[0][0],mv[0][1]);
+                if(ref[1] > 0)
+                    b= pack16to32(mv[1][0],mv[1][1]);
+            }else{
+                a= pack16to32(mv[0][0],mv[0][1]);
+                b= pack16to32(mv[1][0],mv[1][1]);
+            }
+            fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, a, 4);
+            fill_rectangle(&h->mv_cache[1][scan8[0]], 4, 4, 8, b, 4);
+        }else{
+            for(i8=0; i8<4; i8++){
+                const int x8 = i8&1;
+                const int y8 = i8>>1;
+
+                if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
+                    continue;
+                h->sub_mb_type[i8] = sub_mb_type;
+
+                fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mv[0][0],mv[0][1]), 4);
+                fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mv[1][0],mv[1][1]), 4);
+                fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[0], 1);
+                fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[1], 1);
+
+                /* col_zero_flag */
+                if(!IS_INTRA(mb_type_col[0]) && (   l1ref0[x8 + y8*b8_stride] == 0
+                                              || (l1ref0[x8 + y8*b8_stride] < 0 && l1ref1[x8 + y8*b8_stride] == 0
+                                                  && (h->x264_build>33 || !h->x264_build)))){
+                    const int16_t (*l1mv)[2]= l1ref0[x8 + y8*b8_stride] == 0 ? l1mv0 : l1mv1;
+                    if(IS_SUB_8X8(sub_mb_type)){
+                        const int16_t *mv_col = l1mv[x8*3 + y8*3*b4_stride];
+                        if(FFABS(mv_col[0]) <= 1 && FFABS(mv_col[1]) <= 1){
+                            if(ref[0] == 0)
+                                fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
+                            if(ref[1] == 0)
+                                fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
+                        }
+                    }else
+                    for(i4=0; i4<4; i4++){
+                        const int16_t *mv_col = l1mv[x8*2 + (i4&1) + (y8*2 + (i4>>1))*b4_stride];
+                        if(FFABS(mv_col[0]) <= 1 && FFABS(mv_col[1]) <= 1){
+                            if(ref[0] == 0)
+                                *(uint32_t*)h->mv_cache[0][scan8[i8*4+i4]] = 0;
+                            if(ref[1] == 0)
+                                *(uint32_t*)h->mv_cache[1][scan8[i8*4+i4]] = 0;
+                        }
+                    }
+                }
+            }
+        }
+    }else{ /* direct temporal mv pred */
+        const int *map_col_to_list0[2] = {h->map_col_to_list0[0], h->map_col_to_list0[1]};
+        const int *dist_scale_factor = h->dist_scale_factor;
+        int ref_offset= 0;
+
+        if(FRAME_MBAFF && IS_INTERLACED(*mb_type)){
+            map_col_to_list0[0] = h->map_col_to_list0_field[s->mb_y&1][0];
+            map_col_to_list0[1] = h->map_col_to_list0_field[s->mb_y&1][1];
+            dist_scale_factor   =h->dist_scale_factor_field[s->mb_y&1];
+        }
+        if(h->ref_list[1][0].mbaff && IS_INTERLACED(mb_type_col[0]))
+            ref_offset += 16;
+
+        if(IS_INTERLACED(*mb_type) != IS_INTERLACED(mb_type_col[0])){
+            /* FIXME assumes direct_8x8_inference == 1 */
+            int y_shift  = 2*!IS_INTERLACED(*mb_type);
+
+            for(i8=0; i8<4; i8++){
+                const int x8 = i8&1;
+                const int y8 = i8>>1;
+                int ref0, scale;
+                const int16_t (*l1mv)[2]= l1mv0;
+
+                if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
+                    continue;
+                h->sub_mb_type[i8] = sub_mb_type;
+
+                fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
+                if(IS_INTRA(mb_type_col[y8])){
+                    fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, 0, 1);
+                    fill_rectangle(&h-> mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
+                    fill_rectangle(&h-> mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
+                    continue;
+                }
+
+                ref0 = l1ref0[x8 + y8*b8_stride];
+                if(ref0 >= 0)
+                    ref0 = map_col_to_list0[0][ref0 + ref_offset];
+                else{
+                    ref0 = map_col_to_list0[1][l1ref1[x8 + y8*b8_stride] + ref_offset];
+                    l1mv= l1mv1;
+                }
+                scale = dist_scale_factor[ref0];
+                fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref0, 1);
+
+                {
+                    const int16_t *mv_col = l1mv[x8*3 + y8*b4_stride];
+                    int my_col = (mv_col[1]<<y_shift)/2;
+                    int mx = (scale * mv_col[0] + 128) >> 8;
+                    int my = (scale * my_col + 128) >> 8;
+                    fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mx,my), 4);
+                    fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mx-mv_col[0],my-my_col), 4);
+                }
+            }
+            return;
+        }
+
+        /* one-to-one mv scaling */
+
+        if(IS_16X16(*mb_type)){
+            int ref, mv0, mv1;
+
+            fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, 0, 1);
+            if(IS_INTRA(mb_type_col[0])){
+                ref=mv0=mv1=0;
+            }else{
+                const int ref0 = l1ref0[0] >= 0 ? map_col_to_list0[0][l1ref0[0] + ref_offset]
+                                                : map_col_to_list0[1][l1ref1[0] + ref_offset];
+                const int scale = dist_scale_factor[ref0];
+                const int16_t *mv_col = l1ref0[0] >= 0 ? l1mv0[0] : l1mv1[0];
+                int mv_l0[2];
+                mv_l0[0] = (scale * mv_col[0] + 128) >> 8;
+                mv_l0[1] = (scale * mv_col[1] + 128) >> 8;
+                ref= ref0;
+                mv0= pack16to32(mv_l0[0],mv_l0[1]);
+                mv1= pack16to32(mv_l0[0]-mv_col[0],mv_l0[1]-mv_col[1]);
+            }
+            fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, ref, 1);
+            fill_rectangle(&h-> mv_cache[0][scan8[0]], 4, 4, 8, mv0, 4);
+            fill_rectangle(&h-> mv_cache[1][scan8[0]], 4, 4, 8, mv1, 4);
+        }else{
+            for(i8=0; i8<4; i8++){
+                const int x8 = i8&1;
+                const int y8 = i8>>1;
+                int ref0, scale;
+                const int16_t (*l1mv)[2]= l1mv0;
+
+                if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
+                    continue;
+                h->sub_mb_type[i8] = sub_mb_type;
+                fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
+                if(IS_INTRA(mb_type_col[0])){
+                    fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, 0, 1);
+                    fill_rectangle(&h-> mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
+                    fill_rectangle(&h-> mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
+                    continue;
+                }
+
+                ref0 = l1ref0[x8 + y8*b8_stride] + ref_offset;
+                if(ref0 >= 0)
+                    ref0 = map_col_to_list0[0][ref0];
+                else{
+                    ref0 = map_col_to_list0[1][l1ref1[x8 + y8*b8_stride] + ref_offset];
+                    l1mv= l1mv1;
+                }
+                scale = dist_scale_factor[ref0];
+
+                fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref0, 1);
+                if(IS_SUB_8X8(sub_mb_type)){
+                    const int16_t *mv_col = l1mv[x8*3 + y8*3*b4_stride];
+                    int mx = (scale * mv_col[0] + 128) >> 8;
+                    int my = (scale * mv_col[1] + 128) >> 8;
+                    fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mx,my), 4);
+                    fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mx-mv_col[0],my-mv_col[1]), 4);
+                }else
+                for(i4=0; i4<4; i4++){
+                    const int16_t *mv_col = l1mv[x8*2 + (i4&1) + (y8*2 + (i4>>1))*b4_stride];
+                    int16_t *mv_l0 = h->mv_cache[0][scan8[i8*4+i4]];
+                    mv_l0[0] = (scale * mv_col[0] + 128) >> 8;
+                    mv_l0[1] = (scale * mv_col[1] + 128) >> 8;
+                    *(uint32_t*)h->mv_cache[1][scan8[i8*4+i4]] =
+                        pack16to32(mv_l0[0]-mv_col[0],mv_l0[1]-mv_col[1]);
+                }
+            }
+        }
+    }
+}