h264.c

FFmpeg is an audio/video conversion tool. It includes libavcodec, the leading open source codec libr

    const int left= h->intra4x4_pred_mode_cache[index8 - 1];
    const int top = h->intra4x4_pred_mode_cache[index8 - 8];
    const int min= FFMIN(left, top);

    tprintf("mode:%d %d min:%d\n", left ,top, min);

    if(min<0) return DC_PRED;
    else      return min;
}

static inline void write_back_non_zero_count(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;

    h->non_zero_count[mb_xy][0]= h->non_zero_count_cache[4+8*4];
    h->non_zero_count[mb_xy][1]= h->non_zero_count_cache[5+8*4];
    h->non_zero_count[mb_xy][2]= h->non_zero_count_cache[6+8*4];
    h->non_zero_count[mb_xy][3]= h->non_zero_count_cache[7+8*4];
    h->non_zero_count[mb_xy][4]= h->non_zero_count_cache[7+8*3];
    h->non_zero_count[mb_xy][5]= h->non_zero_count_cache[7+8*2];
    h->non_zero_count[mb_xy][6]= h->non_zero_count_cache[7+8*1];
    
    h->non_zero_count[mb_xy][7]= h->non_zero_count_cache[1+8*2];
    h->non_zero_count[mb_xy][8]= h->non_zero_count_cache[2+8*2];
    h->non_zero_count[mb_xy][9]= h->non_zero_count_cache[2+8*1];

    h->non_zero_count[mb_xy][10]=h->non_zero_count_cache[1+8*5];
    h->non_zero_count[mb_xy][11]=h->non_zero_count_cache[2+8*5];
    h->non_zero_count[mb_xy][12]=h->non_zero_count_cache[2+8*4];
}

/**
 * gets the predicted number of non zero coefficients.
 * @param n block index
 */
static inline int pred_non_zero_count(H264Context *h, int n){
    const int index8= scan8[n];
    const int left= h->non_zero_count_cache[index8 - 1];
    const int top = h->non_zero_count_cache[index8 - 8];
    int i= left + top;
    
    if(i<64) i= (i+1)>>1;

    tprintf("pred_nnz L%X T%X n%d s%d P%X\n", left, top, n, scan8[n], i&31);

    return i&31;
}

static inline int fetch_diagonal_mv(H264Context *h, const int16_t **C, int i, int list, int part_width){
    const int topright_ref= h->ref_cache[list][ i - 8 + part_width ];

    if(topright_ref != PART_NOT_AVAILABLE){
        *C= h->mv_cache[list][ i - 8 + part_width ];
        return topright_ref;
    }else{
        tprintf("topright MV not available\n");

        *C= h->mv_cache[list][ i - 8 - 1 ];
        return h->ref_cache[list][ i - 8 - 1 ];
    }
}

/**
 * gets the predicted MV.
 * @param n the block index
 * @param part_width the width of the partition (4, 8,16) -> (1, 2, 4)
 * @param mx the x component of the predicted motion vector
 * @param my the y component of the predicted motion vector
 */
static inline void pred_motion(H264Context * const h, int n, int part_width, int list, int ref, int * const mx, int * const my){
    const int index8= scan8[n];
    const int top_ref=      h->ref_cache[list][ index8 - 8 ];
    const int left_ref=     h->ref_cache[list][ index8 - 1 ];
    const int16_t * const A= h->mv_cache[list][ index8 - 1 ];
    const int16_t * const B= h->mv_cache[list][ index8 - 8 ];
    const int16_t * C;
    int diagonal_ref, match_count;

    assert(part_width==1 || part_width==2 || part_width==4);

/* mv_cache
  B . . A T T T T 
  U . . L . . , .
  U . . L . . . .
  U . . L . . , .
  . . . L . . . .
*/

    diagonal_ref= fetch_diagonal_mv(h, &C, index8, list, part_width);
    match_count= (diagonal_ref==ref) + (top_ref==ref) + (left_ref==ref);
    if(match_count > 1){ //most common
        *mx= mid_pred(A[0], B[0], C[0]);
        *my= mid_pred(A[1], B[1], C[1]);
    }else if(match_count==1){
        if(left_ref==ref){
            *mx= A[0];
            *my= A[1];        
        }else if(top_ref==ref){
            *mx= B[0];
            *my= B[1];        
        }else{
            *mx= C[0];
            *my= C[1];        
        }
    }else{
        if(top_ref == PART_NOT_AVAILABLE && diagonal_ref == PART_NOT_AVAILABLE && left_ref != PART_NOT_AVAILABLE){
            *mx= A[0];
            *my= A[1];        
        }else{
            *mx= mid_pred(A[0], B[0], C[0]);
            *my= mid_pred(A[1], B[1], C[1]);
        }
    }
        
    tprintf("pred_motion (%2d %2d %2d) (%2d %2d %2d) (%2d %2d %2d) -> (%2d %2d %2d) at %2d %2d %d list %d\n", top_ref, B[0], B[1],                    diagonal_ref, C[0], C[1], left_ref, A[0], A[1], ref, *mx, *my, h->s.mb_x, h->s.mb_y, n, list);
}

/**
 * gets the directionally predicted 16x8 MV.
 * @param n the block index
 * @param mx the x component of the predicted motion vector
 * @param my the y component of the predicted motion vector
 */
static inline void pred_16x8_motion(H264Context * const h, int n, int list, int ref, int * const mx, int * const my){
    if(n==0){
        const int top_ref=      h->ref_cache[list][ scan8[0] - 8 ];
        const int16_t * const B= h->mv_cache[list][ scan8[0] - 8 ];

        tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d", top_ref, B[0], B[1], h->s.mb_x, h->s.mb_y, n, list);
        
        if(top_ref == ref){
            *mx= B[0];
            *my= B[1];
            return;
        }
    }else{
        const int left_ref=     h->ref_cache[list][ scan8[8] - 1 ];
        const int16_t * const A= h->mv_cache[list][ scan8[8] - 1 ];
        
        tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d", left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);

        if(left_ref == ref){
            *mx= A[0];
            *my= A[1];
            return;
        }
    }

    //RARE
    pred_motion(h, n, 4, list, ref, mx, my);
}

/**
 * gets the directionally predicted 8x16 MV.
 * @param n the block index
 * @param mx the x component of the predicted motion vector
 * @param my the y component of the predicted motion vector
 */
static inline void pred_8x16_motion(H264Context * const h, int n, int list, int ref, int * const mx, int * const my){
    if(n==0){
        const int left_ref=      h->ref_cache[list][ scan8[0] - 1 ];
        const int16_t * const A=  h->mv_cache[list][ scan8[0] - 1 ];
        
        tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d", left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);

        if(left_ref == ref){
            *mx= A[0];
            *my= A[1];
            return;
        }
    }else{
        const int16_t * C;
        int diagonal_ref;

        diagonal_ref= fetch_diagonal_mv(h, &C, scan8[4], list, 2);
        
        tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d", diagonal_ref, C[0], C[1], h->s.mb_x, h->s.mb_y, n, list);

        if(diagonal_ref == ref){ 
            *mx= C[0];
            *my= C[1];
            return;
        }
    }

    //RARE
    pred_motion(h, n, 2, list, ref, mx, my);
}

static inline void pred_pskip_motion(H264Context * const h, int * const mx, int * const my){
    const int top_ref = h->ref_cache[0][ scan8[0] - 8 ];
    const int left_ref= h->ref_cache[0][ scan8[0] - 1 ];

    tprintf("pred_pskip: (%d) (%d) at %2d %2d", top_ref, left_ref, h->s.mb_x, h->s.mb_y);

    if(top_ref == PART_NOT_AVAILABLE || left_ref == PART_NOT_AVAILABLE
       || (top_ref == 0  && *(uint32_t*)h->mv_cache[0][ scan8[0] - 8 ] == 0)
       || (left_ref == 0 && *(uint32_t*)h->mv_cache[0][ scan8[0] - 1 ] == 0)){
       
        *mx = *my = 0;
        return;
    }
        
    pred_motion(h, 0, 4, 0, 0, mx, my);

    return;
}

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;
    const int b8_xy= 2*s->mb_x + 2*s->mb_y*h->b8_stride;
    int list;

    for(list=0; list<2; list++){
        int y;
        if((!IS_8X8(mb_type)) && !USES_LIST(mb_type, list)){
            if(1){ //FIXME skip or never read if mb_type doesnt use it
                for(y=0; y<4; y++){
                    *(uint64_t*)s->current_picture.motion_val[list][b_xy + 0 + y*h->b_stride]=
                    *(uint64_t*)s->current_picture.motion_val[list][b_xy + 2 + y*h->b_stride]= 0;
                }
                if( h->pps.cabac ) {
                    /* FIXME needed ? */
                    for(y=0; y<4; y++){
                        *(uint64_t*)h->mvd_table[list][b_xy + 0 + y*h->b_stride]=
                        *(uint64_t*)h->mvd_table[list][b_xy + 2 + y*h->b_stride]= 0;
                    }
                }
                for(y=0; y<2; y++){
                    *(uint16_t*)s->current_picture.motion_val[list][b8_xy + y*h->b8_stride]= (LIST_NOT_USED&0xFF)*0x0101;
                }
            }
            continue; //FIXME direct mode ...
        }
        
        for(y=0; y<4; y++){
            *(uint64_t*)s->current_picture.motion_val[list][b_xy + 0 + y*h->b_stride]= *(uint64_t*)h->mv_cache[list][scan8[0]+0 + 8*y];
            *(uint64_t*)s->current_picture.motion_val[list][b_xy + 2 + y*h->b_stride]= *(uint64_t*)h->mv_cache[list][scan8[0]+2 + 8*y];
        }
        if( h->pps.cabac ) {
            for(y=0; y<4; y++){
                *(uint64_t*)h->mvd_table[list][b_xy + 0 + y*h->b_stride]= *(uint64_t*)h->mvd_cache[list][scan8[0]+0 + 8*y];
                *(uint64_t*)h->mvd_table[list][b_xy + 2 + y*h->b_stride]= *(uint64_t*)h->mvd_cache[list][scan8[0]+2 + 8*y];
            }
        }
        for(y=0; y<2; y++){
            s->current_picture.ref_index[list][b8_xy + 0 + y*h->b8_stride]= h->ref_cache[list][scan8[0]+0 + 16*y];
            s->current_picture.ref_index[list][b8_xy + 1 + y*h->b8_stride]= h->ref_cache[list][scan8[0]+2 + 16*y];
        }
    }
}

/**
 * Decodes a network abstraction layer unit.
 * @param consumed is the number of bytes used as input
 * @param length is the length of the array
 * @param dst_length is the number of decoded bytes FIXME here or a decode rbsp ttailing?
 * @returns decoded bytes, might be src+1 if no escapes 
 */
static uint8_t *decode_nal(H264Context *h, uint8_t *src, int *dst_length, int *consumed, int length){
    int i, si, di;
    uint8_t *dst;

//    src[0]&0x80;		//forbidden bit
    h->nal_ref_idc= src[0]>>5;
    h->nal_unit_type= src[0]&0x1F;

    src++; length--;
#if 0    
    for(i=0; i<length; i++)
        printf("%2X ", src[i]);
#endif
    for(i=0; i+1<length; i+=2){
        if(src[i]) continue;
        if(i>0 && src[i-1]==0) i--;
        if(i+2<length && src[i+1]==0 && src[i+2]<=3){
            if(src[i+2]!=3){
                /* startcode, so we must be past the end */
                length=i;
            }
            break;
        }
    }

    if(i>=length-1){ //no escaped 0
        *dst_length= length;
        *consumed= length+1; //+1 for the header
        return src; 
    }

    h->rbsp_buffer= av_fast_realloc(h->rbsp_buffer, &h->rbsp_buffer_size, length);
    dst= h->rbsp_buffer;

//printf("deoding esc\n");
    si=di=0;
    while(si<length){ 
        //remove escapes (very rare 1:2^22)
        if(si+2<length && src[si]==0 && src[si+1]==0 && src[si+2]<=3){
            if(src[si+2]==3){ //escape
                dst[di++]= 0;
                dst[di++]= 0;
                si+=3;
                continue;
            }else //next start code
                break;
        }

        dst[di++]= src[si++];
    }

    *dst_length= di;
    *consumed= si + 1;//+1 for the header
//FIXME store exact number of bits in the getbitcontext (its needed for decoding)
    return dst;
}

#if 0
/**
 * @param src the data which should be escaped
 * @param dst the target buffer, dst+1 == src is allowed as a special case
 * @param length the length of the src data
 * @param dst_length the length of the dst array
 * @returns length of escaped data in bytes or -1 if an error occured
 */
static int encode_nal(H264Context *h, uint8_t *dst, uint8_t *src, int length, int dst_length){
    int i, escape_count, si, di;
    uint8_t *temp;
    
    assert(length>=0);
    assert(dst_length>0);
    
    dst[0]= (h->nal_ref_idc<<5) + h->nal_unit_type;

    if(length==0) return 1;

    escape_count= 0;
    for(i=0; i<length; i+=2){
        if(src[i]) continue;
        if(i>0 && src[i-1]==0) 
            i--;
        if(i+2<length && src[i+1]==0 && src[i+2]<=3){
            escape_count++;
            i+=2;
        }
    }
    
    if(escape_count==0){ 
        if(dst+1 != src)
            memcpy(dst+1, src, length);
        return length + 1;
    }
    
    if(length + escape_count + 1> dst_length)
        return -1;

    //this should be damn rare (hopefully)

    h->rbsp_buffer= av_fast_realloc(h->rbsp_buffer, &h->rbsp_buffer_size, length + escape_count);
    temp= h->rbsp_buffer;
//printf("encoding esc\n");
    
    si= 0;
    di= 0;
    while(si < length){
        if(si+2<length && src[si]==0 && src[si+1]==0 && src[si+2]<=3){
            temp[di++]= 0; si++;
            temp[di++]= 0; si++;
            temp[di++]= 3; 
            temp[di++]= src[si++];
        }
        else
            temp[di++]= src[si++];
    }
    memcpy(dst+1, temp, length+escape_count);
    
    assert(di == length+escape_count);
    
    return di + 1;
}

/**