h264.c

arm平台下的H264编码和解码源代码

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;
            }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;
}

/**
 * write 1,10,100,1000,... for alignment, yes its exactly inverse to mpeg4
 */
static void encode_rbsp_trailing(PutBitContext *pb){
    int length;
    put_bits(pb, 1, 1);
    length= (-put_bits_count(pb))&7;
    if(length) put_bits(pb, length, 0);
}
#endif

/**
 * identifies the exact end of the bitstream
 * @return the length of the trailing, or 0 if damaged
 */
static int decode_rbsp_trailing(uint8_t *src){
    int v= *src;
    int r;

    tprintf("rbsp trailing %X\n", v);

    for(r=1; r<9; r++){
        if(v&1) return r;
        v>>=1;
    }
    return 0;
}

/**
 * idct tranforms the 16 dc values and dequantize them.
 * @param qp quantization parameter
 */
static void h264_luma_dc_dequant_idct_c(DCTELEM *block, int qp){
    const int qmul= dequant_coeff[qp][0];
#define stride 16
    int i;
    int temp[16]; //FIXME check if this is a good idea
    static const int x_offset[4]={0, 1*stride, 4* stride,  5*stride};
    static const int y_offset[4]={0, 2*stride, 8* stride, 10*stride};

//memset(block, 64, 2*256);
//return;
    for(i=0; i<4; i++){
        const int offset= y_offset[i];
        const int z0= block[offset+stride*0] + block[offset+stride*4];
        const int z1= block[offset+stride*0] - block[offset+stride*4];
        const int z2= block[offset+stride*1] - block[offset+stride*5];
        const int z3= block[offset+stride*1] + block[offset+stride*5];

        temp[4*i+0]= z0+z3;
        temp[4*i+1]= z1+z2;
        temp[4*i+2]= z1-z2;
        temp[4*i+3]= z0-z3;
    }

    for(i=0; i<4; i++){
        const int offset= x_offset[i];
        const int z0= temp[4*0+i] + temp[4*2+i];
        const int z1= temp[4*0+i] - temp[4*2+i];
        const int z2= temp[4*1+i] - temp[4*3+i];
        const int z3= temp[4*1+i] + temp[4*3+i];

        block[stride*0 +offset]= ((z0 + z3)*qmul + 2)>>2; //FIXME think about merging this into decode_resdual
        block[stride*2 +offset]= ((z1 + z2)*qmul + 2)>>2;
        block[stride*8 +offset]= ((z1 - z2)*qmul + 2)>>2;
        block[stride*10+offset]= ((z0 - z3)*qmul + 2)>>2;
    }
}

#if 0
/**
 * dct tranforms the 16 dc values.
 * @param qp quantization parameter ??? FIXME
 */
static void h264_luma_dc_dct_c(DCTELEM *block/*, int qp*/){
//    const int qmul= dequant_coeff[qp][0];
    int i;
    int temp[16]; //FIXME check if this is a good idea
    static const int x_offset[4]={0, 1*stride, 4* stride,  5*stride};
    static const int y_offset[4]={0, 2*stride, 8* stride, 10*stride};

    for(i=0; i<4; i++){
        const int offset= y_offset[i];
        const int z0= block[offset+stride*0] + block[offset+stride*4];
        const int z1= block[offset+stride*0] - block[offset+stride*4];
        const int z2= block[offset+stride*1] - block[offset+stride*5];
        const int z3= block[offset+stride*1] + block[offset+stride*5];

        temp[4*i+0]= z0+z3;
        temp[4*i+1]= z1+z2;
        temp[4*i+2]= z1-z2;
        temp[4*i+3]= z0-z3;
    }

    for(i=0; i<4; i++){
        const int offset= x_offset[i];
        const int z0= temp[4*0+i] + temp[4*2+i];
        const int z1= temp[4*0+i] - temp[4*2+i];
        const int z2= temp[4*1+i] - temp[4*3+i];
        const int z3= temp[4*1+i] + temp[4*3+i];

        block[stride*0 +offset]= (z0 + z3)>>1;
        block[stride*2 +offset]= (z1 + z2)>>1;
        block[stride*8 +offset]= (z1 - z2)>>1;
        block[stride*10+offset]= (z0 - z3)>>1;
    }
}
#endif

#undef xStride
#undef stride

static void chroma_dc_dequant_idct_c(DCTELEM *block, int qp){
    const int qmul= dequant_coeff[qp][0];
    const int stride= 16*2;
    const int xStride= 16;
    int a,b,c,d,e;

    a= block[stride*0 + xStride*0];
    b= block[stride*0 + xStride*1];
    c= block[stride*1 + xStride*0];
    d= block[stride*1 + xStride*1];

    e= a-b;
    a= a+b;
    b= c-d;
    c= c+d;

    block[stride*0 + xStride*0]= ((a+c)*qmul + 0)>>1;
    block[stride*0 + xStride*1]= ((e+b)*qmul + 0)>>1;
    block[stride*1 + xStride*0]= ((a-c)*qmul + 0)>>1;
    block[stride*1 + xStride*1]= ((e-b)*qmul + 0)>>1;
}

#if 0
static void chroma_dc_dct_c(DCTELEM *block){
    const int stride= 16*2;
    const int xStride= 16;
    int a,b,c,d,e;

    a= block[stride*0 + xStride*0];
    b= block[stride*0 + xStride*1];
    c= block[stride*1 + xStride*0];
    d= block[stride*1 + xStride*1];

    e= a-b;
    a= a+b;
    b= c-d;
    c= c+d;

    block[stride*0 + xStride*0]= (a+c);
    block[stride*0 + xStride*1]= (e+b);
    block[stride*1 + xStride*0]= (a-c);
    block[stride*1 + xStride*1]= (e-b);
}
#endif

/**
 * gets the chroma qp.
 */
static inline int get_chroma_qp(H264Context *h, int qscale){
    
    return chroma_qp[clip(qscale + h->pps.chroma_qp_index_offset, 0, 51)];
}


/**
 *
 */
static void h264_add_idct_c(uint8_t *dst, DCTELEM *block, int stride){
    int i;
    uint8_t *cm = cropTbl + MAX_NEG_CROP;

    block[0] += 32;

    for(i=0; i<4; i++){
        const int z0=  block[0 + 4*i]     +  block[2 + 4*i];
        const int z1=  block[0 + 4*i]     -  block[2 + 4*i];
        const int z2= (block[1 + 4*i]>>1) -  block[3 + 4*i];
        const int z3=  block[1 + 4*i]     + (block[3 + 4*i]>>1);

        block[0 + 4*i]= z0 + z3;
        block[1 + 4*i]= z1 + z2;
        block[2 + 4*i]= z1 - z2;
        block[3 + 4*i]= z0 - z3;
    }

    for(i=0; i<4; i++){
        const int z0=  block[i + 4*0]     +  block[i + 4*2];
        const int z1=  block[i + 4*0]     -  block[i + 4*2];
        const int z2= (block[i + 4*1]>>1) -  block[i + 4*3];
        const int z3=  block[i + 4*1]     + (block[i + 4*3]>>1);

        dst[i + 0*stride]= cm[ dst[i + 0*stride] + ((z0 + z3) >> 6) ];
        dst[i + 1*stride]= cm[ dst[i + 1*stride] + ((z1 + z2) >> 6) ];
        dst[i + 2*stride]= cm[ dst[i + 2*stride] + ((z1 - z2) >> 6) ];