h264.c

linux下的MPEG1

    /**
     * num_ref_idx_l0/1_active_minus1 + 1
     */
    unsigned int ref_count[2];   ///< counts frames or fields, depending on current mb mode
    Picture *short_ref[32];
    Picture *long_ref[32];
    Picture default_ref_list[2][32];
    Picture ref_list[2][48];     ///< 0..15: frame refs, 16..47: mbaff field refs
    Picture *delayed_pic[18]; //FIXME size?
    Picture *delayed_output_pic;

    /**
     * memory management control operations buffer.
     */
    MMCO mmco[MAX_MMCO_COUNT];
    int mmco_index;

    int long_ref_count;  ///< number of actual long term references
    int short_ref_count; ///< number of actual short term references

    //data partitioning
    GetBitContext intra_gb;
    GetBitContext inter_gb;
    GetBitContext *intra_gb_ptr;
    GetBitContext *inter_gb_ptr;

    DECLARE_ALIGNED_8(DCTELEM, mb[16*24]);
    DCTELEM mb_padding[256];        ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either check that i is not to large or ensure that there is some unused stuff after mb

    /**
     * Cabac
     */
    CABACContext cabac;
    uint8_t      cabac_state[460];
    int          cabac_init_idc;

    /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
    uint16_t     *cbp_table;
    int cbp;
    int top_cbp;
    int left_cbp;
    /* chroma_pred_mode for i4x4 or i16x16, else 0 */
    uint8_t     *chroma_pred_mode_table;
    int         last_qscale_diff;
    int16_t     (*mvd_table[2])[2];
    DECLARE_ALIGNED_8(int16_t, mvd_cache[2][5*8][2]);
    uint8_t     *direct_table;
    uint8_t     direct_cache[5*8];

    uint8_t zigzag_scan[16];
    uint8_t zigzag_scan8x8[64];
    uint8_t zigzag_scan8x8_cavlc[64];
    uint8_t field_scan[16];
    uint8_t field_scan8x8[64];
    uint8_t field_scan8x8_cavlc[64];
    const uint8_t *zigzag_scan_q0;
    const uint8_t *zigzag_scan8x8_q0;
    const uint8_t *zigzag_scan8x8_cavlc_q0;
    const uint8_t *field_scan_q0;
    const uint8_t *field_scan8x8_q0;
    const uint8_t *field_scan8x8_cavlc_q0;

    int x264_build;
}H264Context;

static VLC coeff_token_vlc[4];
static VLC chroma_dc_coeff_token_vlc;

static VLC total_zeros_vlc[15];
static VLC chroma_dc_total_zeros_vlc[3];

static VLC run_vlc[6];
static VLC run7_vlc;

static void svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp);
static void svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
static void 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);
static void 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);

static av_always_inline uint32_t pack16to32(int a, int b){
#ifdef WORDS_BIGENDIAN
   return (b&0xFFFF) + (a<<16);
#else
   return (a&0xFFFF) + (b<<16);
#endif
}

const uint8_t ff_rem6[52]={
0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3,
};

const uint8_t ff_div6[52]={
0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8,
};


/**
 * fill a rectangle.
 * @param h height of the rectangle, should be a constant
 * @param w width of the rectangle, should be a constant
 * @param size the size of val (1 or 4), should be a constant
 */
static av_always_inline void fill_rectangle(void *vp, int w, int h, int stride, uint32_t val, int size){
    uint8_t *p= (uint8_t*)vp;
    assert(size==1 || size==4);
    assert(w<=4);

    w      *= size;
    stride *= size;

    assert((((long)vp)&(FFMIN(w, STRIDE_ALIGN)-1)) == 0);
    assert((stride&(w-1))==0);
    if(w==2){
        const uint16_t v= size==4 ? val : val*0x0101;
        *(uint16_t*)(p + 0*stride)= v;
        if(h==1) return;
        *(uint16_t*)(p + 1*stride)= v;
        if(h==2) return;
        *(uint16_t*)(p + 2*stride)=
        *(uint16_t*)(p + 3*stride)= v;
    }else if(w==4){
        const uint32_t v= size==4 ? val : val*0x01010101;
        *(uint32_t*)(p + 0*stride)= v;
        if(h==1) return;
        *(uint32_t*)(p + 1*stride)= v;
        if(h==2) return;
        *(uint32_t*)(p + 2*stride)=
        *(uint32_t*)(p + 3*stride)= v;
    }else if(w==8){
    //gcc can't optimize 64bit math on x86_32
#if defined(ARCH_X86_64) || (defined(MP_WORDSIZE) && MP_WORDSIZE >= 64)
        const uint64_t v= val*0x0100000001ULL;
        *(uint64_t*)(p + 0*stride)= v;
        if(h==1) return;
        *(uint64_t*)(p + 1*stride)= v;
        if(h==2) return;
        *(uint64_t*)(p + 2*stride)=
        *(uint64_t*)(p + 3*stride)= v;
    }else if(w==16){
        const uint64_t v= val*0x0100000001ULL;
        *(uint64_t*)(p + 0+0*stride)=
        *(uint64_t*)(p + 8+0*stride)=
        *(uint64_t*)(p + 0+1*stride)=
        *(uint64_t*)(p + 8+1*stride)= v;
        if(h==2) return;
        *(uint64_t*)(p + 0+2*stride)=
        *(uint64_t*)(p + 8+2*stride)=
        *(uint64_t*)(p + 0+3*stride)=
        *(uint64_t*)(p + 8+3*stride)= v;
#else
        *(uint32_t*)(p + 0+0*stride)=
        *(uint32_t*)(p + 4+0*stride)= val;
        if(h==1) return;
        *(uint32_t*)(p + 0+1*stride)=
        *(uint32_t*)(p + 4+1*stride)= val;
        if(h==2) return;
        *(uint32_t*)(p + 0+2*stride)=
        *(uint32_t*)(p + 4+2*stride)=
        *(uint32_t*)(p + 0+3*stride)=
        *(uint32_t*)(p + 4+3*stride)= val;
    }else if(w==16){
        *(uint32_t*)(p + 0+0*stride)=
        *(uint32_t*)(p + 4+0*stride)=
        *(uint32_t*)(p + 8+0*stride)=
        *(uint32_t*)(p +12+0*stride)=
        *(uint32_t*)(p + 0+1*stride)=
        *(uint32_t*)(p + 4+1*stride)=
        *(uint32_t*)(p + 8+1*stride)=
        *(uint32_t*)(p +12+1*stride)= val;
        if(h==2) return;
        *(uint32_t*)(p + 0+2*stride)=
        *(uint32_t*)(p + 4+2*stride)=
        *(uint32_t*)(p + 8+2*stride)=
        *(uint32_t*)(p +12+2*stride)=
        *(uint32_t*)(p + 0+3*stride)=
        *(uint32_t*)(p + 4+3*stride)=
        *(uint32_t*)(p + 8+3*stride)=
        *(uint32_t*)(p +12+3*stride)= val;
#endif
    }else
        assert(0);
    assert(h==4);
}

static void fill_caches(H264Context *h, int mb_type, int for_deblock){
    MpegEncContext * const s = &h->s;
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
    int topleft_xy, top_xy, topright_xy, left_xy[2];
    int topleft_type, top_type, topright_type, left_type[2];
    int left_block[8];
    int i;

    //FIXME deblocking could skip the intra and nnz parts.
    if(for_deblock && (h->slice_num == 1 || h->slice_table[mb_xy] == h->slice_table[mb_xy-s->mb_stride]) && !FRAME_MBAFF)
        return;

    //wow what a mess, why didn't they simplify the interlacing&intra stuff, i can't imagine that these complex rules are worth it

    top_xy     = mb_xy  - s->mb_stride;
    topleft_xy = top_xy - 1;
    topright_xy= top_xy + 1;
    left_xy[1] = left_xy[0] = mb_xy-1;
    left_block[0]= 0;
    left_block[1]= 1;
    left_block[2]= 2;
    left_block[3]= 3;
    left_block[4]= 7;
    left_block[5]= 10;
    left_block[6]= 8;
    left_block[7]= 11;
    if(FRAME_MBAFF){
        const int pair_xy          = s->mb_x     + (s->mb_y & ~1)*s->mb_stride;
        const int top_pair_xy      = pair_xy     - s->mb_stride;
        const int topleft_pair_xy  = top_pair_xy - 1;
        const int topright_pair_xy = top_pair_xy + 1;
        const int topleft_mb_frame_flag  = !IS_INTERLACED(s->current_picture.mb_type[topleft_pair_xy]);
        const int top_mb_frame_flag      = !IS_INTERLACED(s->current_picture.mb_type[top_pair_xy]);
        const int topright_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[topright_pair_xy]);
        const int left_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[pair_xy-1]);
        const int curr_mb_frame_flag = !IS_INTERLACED(mb_type);
        const int bottom = (s->mb_y & 1);
        tprintf("fill_caches: curr_mb_frame_flag:%d, left_mb_frame_flag:%d, topleft_mb_frame_flag:%d, top_mb_frame_flag:%d, topright_mb_frame_flag:%d\n", curr_mb_frame_flag, left_mb_frame_flag, topleft_mb_frame_flag, top_mb_frame_flag, topright_mb_frame_flag);
        if (bottom
                ? !curr_mb_frame_flag // bottom macroblock
                : (!curr_mb_frame_flag && !top_mb_frame_flag) // top macroblock
                ) {
            top_xy -= s->mb_stride;
        }
        if (bottom
                ? !curr_mb_frame_flag // bottom macroblock
                : (!curr_mb_frame_flag && !topleft_mb_frame_flag) // top macroblock
                ) {
            topleft_xy -= s->mb_stride;
        }
        if (bottom
                ? !curr_mb_frame_flag // bottom macroblock
                : (!curr_mb_frame_flag && !topright_mb_frame_flag) // top macroblock
                ) {
            topright_xy -= s->mb_stride;
        }
        if (left_mb_frame_flag != curr_mb_frame_flag) {
            left_xy[1] = left_xy[0] = pair_xy - 1;
            if (curr_mb_frame_flag) {
                if (bottom) {
                    left_block[0]= 2;
                    left_block[1]= 2;
                    left_block[2]= 3;
                    left_block[3]= 3;
                    left_block[4]= 8;
                    left_block[5]= 11;
                    left_block[6]= 8;
                    left_block[7]= 11;
                } else {
                    left_block[0]= 0;
                    left_block[1]= 0;
                    left_block[2]= 1;
                    left_block[3]= 1;
                    left_block[4]= 7;
                    left_block[5]= 10;
                    left_block[6]= 7;
                    left_block[7]= 10;
                }
            } else {
                left_xy[1] += s->mb_stride;
                //left_block[0]= 0;
                left_block[1]= 2;
                left_block[2]= 0;
                left_block[3]= 2;
                //left_block[4]= 7;
                left_block[5]= 10;
                left_block[6]= 7;
                left_block[7]= 10;
            }
        }
    }

    h->top_mb_xy = top_xy;
    h->left_mb_xy[0] = left_xy[0];
    h->left_mb_xy[1] = left_xy[1];
    if(for_deblock){
        topleft_type = 0;
        topright_type = 0;
        top_type     = h->slice_table[top_xy     ] < 255 ? s->current_picture.mb_type[top_xy]     : 0;
        left_type[0] = h->slice_table[left_xy[0] ] < 255 ? s->current_picture.mb_type[left_xy[0]] : 0;
        left_type[1] = h->slice_table[left_xy[1] ] < 255 ? s->current_picture.mb_type[left_xy[1]] : 0;

        if(FRAME_MBAFF && !IS_INTRA(mb_type)){
            int list;
            int v = *(uint16_t*)&h->non_zero_count[mb_xy][14];
            for(i=0; i<16; i++)
                h->non_zero_count_cache[scan8[i]] = (v>>i)&1;
            for(list=0; list<1+(h->slice_type==B_TYPE); list++){
                if(USES_LIST(mb_type,list)){
                    uint32_t *src = (uint32_t*)s->current_picture.motion_val[list][h->mb2b_xy[mb_xy]];
                    uint32_t *dst = (uint32_t*)h->mv_cache[list][scan8[0]];
                    int8_t *ref = &s->current_picture.ref_index[list][h->mb2b8_xy[mb_xy]];
                    for(i=0; i<4; i++, dst+=8, src+=h->b_stride){
                        dst[0] = src[0];
                        dst[1] = src[1];
                        dst[2] = src[2];
                        dst[3] = src[3];
                    }
                    *(uint32_t*)&h->ref_cache[list][scan8[ 0]] =
                    *(uint32_t*)&h->ref_cache[list][scan8[ 2]] = pack16to32(ref[0],ref[1])*0x0101;
                    ref += h->b8_stride;
                    *(uint32_t*)&h->ref_cache[list][scan8[ 8]] =
                    *(uint32_t*)&h->ref_cache[list][scan8[10]] = pack16to32(ref[0],ref[1])*0x0101;
                }else{
                    fill_rectangle(&h-> mv_cache[list][scan8[ 0]], 4, 4, 8, 0, 4);
                    fill_rectangle(&h->ref_cache[list][scan8[ 0]], 4, 4, 8, (uint8_t)LIST_NOT_USED, 1);
                }
            }
        }
    }else{
        topleft_type = h->slice_table[topleft_xy ] == h->slice_num ? s->current_picture.mb_type[topleft_xy] : 0;
        top_type     = h->slice_table[top_xy     ] == h->slice_num ? s->current_picture.mb_type[top_xy]     : 0;
        topright_type= h->slice_table[topright_xy] == h->slice_num ? s->current_picture.mb_type[topright_xy]: 0;
        left_type[0] = h->slice_table[left_xy[0] ] == h->slice_num ? s->current_picture.mb_type[left_xy[0]] : 0;
        left_type[1] = h->slice_table[left_xy[1] ] == h->slice_num ? s->current_picture.mb_type[left_xy[1]] : 0;
    }

    if(IS_INTRA(mb_type)){
        h->topleft_samples_available=
        h->top_samples_available=