comparison h264.c @ 6783:df0893f4fd86 libavcodec

Store mb_xy in H264Context and only calculate it once per MB.

6782:781bfad6e40e

};


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 topleft_partition= -1;
    int i;

    h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[0]);
}

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

    h->intra4x4_pred_mode[mb_xy][0]= h->intra4x4_pred_mode_cache[7+8*1];
    h->intra4x4_pred_mode[mb_xy][1]= h->intra4x4_pred_mode_cache[7+8*2];
    h->intra4x4_pred_mode[mb_xy][2]= h->intra4x4_pred_mode_cache[7+8*3];
    h->intra4x4_pred_mode[mb_xy][3]= h->intra4x4_pred_mode_cache[7+8*4];

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

    }
}

static inline void pred_direct_motion(H264Context * const h, int *mb_type){
    MpegEncContext * const s = &h->s;
    const int mb_xy =   s->mb_x +   s->mb_y*s->mb_stride;
    const int b8_xy = 2*s->mb_x + 2*s->mb_y*h->b8_stride;
    const int b4_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride;
    const int mb_type_col = h->ref_list[1][0].mb_type[mb_xy];
    const int16_t (*l1mv0)[2] = (const int16_t (*)[2]) &h->ref_list[1][0].motion_val[0][b4_xy];
    const int16_t (*l1mv1)[2] = (const int16_t (*)[2]) &h->ref_list[1][0].motion_val[1][b4_xy];

static void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                      qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put),
                      qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg),
                      h264_weight_func *weight_op, h264_biweight_func *weight_avg){
    MpegEncContext * const s = &h->s;
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
    const int mb_type= s->current_picture.mb_type[mb_xy];

    assert(IS_INTER(mb_type));

    prefetch_motion(h, 0);

    int deblock_left;
    int deblock_top;
    int mb_xy;

    if(h->deblocking_filter == 2) {
        mb_xy = s->mb_x + s->mb_y*s->mb_stride;
        deblock_left = h->slice_table[mb_xy] == h->slice_table[mb_xy - 1];
        deblock_top  = h->slice_table[mb_xy] == h->slice_table[h->top_mb_xy];
    } else {
        deblock_left = (s->mb_x > 0);
        deblock_top =  (s->mb_y > 0);

static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple){
    MpegEncContext * const s = &h->s;
    const int mb_x= s->mb_x;
    const int mb_y= s->mb_y;
    const int mb_xy= mb_x + mb_y*s->mb_stride;
    const int mb_type= s->current_picture.mb_type[mb_xy];
    uint8_t  *dest_y, *dest_cb, *dest_cr;
    int linesize, uvlinesize /*dct_offset*/;
    int i;
    int *block_offset = &h->block_offset[0];

                xchg_pair_border(h, pair_dest_y, pair_dest_cb, pair_dest_cr, s->linesize, s->uvlinesize, 0);

            backup_pair_border(h, pair_dest_y, pair_dest_cb, pair_dest_cr, s->linesize, s->uvlinesize);
            // deblock a pair
            // top
            s->mb_y--;
            tprintf(h->s.avctx, "call mbaff filter_mb mb_x:%d mb_y:%d pair_dest_y = %p, dest_y = %p\n", mb_x, mb_y, pair_dest_y, dest_y);
            fill_caches(h, mb_type_top, 1); //FIXME don't fill stuff which isn't used by filter_mb
            h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.qscale_table[mb_xy]);
            h->chroma_qp[1] = get_chroma_qp(h, 1, s->current_picture.qscale_table[mb_xy]);
            filter_mb(h, mb_x, mb_y, pair_dest_y, pair_dest_cb, pair_dest_cr, linesize, uvlinesize);
            // bottom
            s->mb_y++;
            tprintf(h->s.avctx, "call mbaff filter_mb\n");
            fill_caches(h, mb_type_bottom, 1); //FIXME don't fill stuff which isn't used by filter_mb
            h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.qscale_table[mb_xy+s->mb_stride]);
            h->chroma_qp[1] = get_chroma_qp(h, 1, s->current_picture.qscale_table[mb_xy+s->mb_stride]);
            filter_mb(h, mb_x, mb_y+1, dest_y, dest_cb, dest_cr, linesize, uvlinesize);

static void hl_decode_mb(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int mb_x= s->mb_x;
    const int mb_y= s->mb_y;
    const int mb_xy= mb_x + mb_y*s->mb_stride;
    const int mb_type= s->current_picture.mb_type[mb_xy];
    int is_complex = FRAME_MBAFF || MB_FIELD || IS_INTRA_PCM(mb_type) || s->codec_id != CODEC_ID_H264 || (ENABLE_GRAY && (s->flags&CODEC_FLAG_GRAY)) || s->encoding;

    if(!s->decode)
        return;

    return 0;
}

static void predict_field_decoding_flag(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
    int mb_type = (h->slice_table[mb_xy-1] == h->slice_num)
                ? s->current_picture.mb_type[mb_xy-1]
                : (h->slice_table[mb_xy-s->mb_stride] == h->slice_num)
                ? s->current_picture.mb_type[mb_xy-s->mb_stride]
                : 0;

/**
 * decodes a P_SKIP or B_SKIP macroblock
 */
static void decode_mb_skip(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
    int mb_type=0;

    memset(h->non_zero_count[mb_xy], 0, 16);
    memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui

 * decodes a macroblock
 * @returns 0 if ok, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
 */
static int decode_mb_cavlc(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
    int partition_count;
    unsigned int mb_type, cbp;
    int dct8x8_allowed= h->pps.transform_8x8_mode;

    s->dsp.clear_blocks(h->mb); //FIXME avoid if already clear (move after skip handlong?

    tprintf(s->avctx, "pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
    cbp = 0; /* avoid warning. FIXME: find a solution without slowing

                && IS_INTERLACED( s->current_picture.mb_type[mbb_xy] ) )
                mbb_xy -= s->mb_stride;
        }else
            mbb_xy = mb_x + (mb_y-1)*s->mb_stride;
    }else{
        int mb_xy = mb_x + mb_y*s->mb_stride;
        mba_xy = mb_xy - 1;
        mbb_xy = mb_xy - (s->mb_stride << FIELD_PICTURE);
    }

    if( h->slice_table[mba_xy] == h->slice_num && !IS_SKIP( s->current_picture.mb_type[mba_xy] ))

    3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
    5, 5, 5, 5, 6, 6, 6, 6, 7, 7, 7, 7, 8, 8, 8
};

static void decode_cabac_residual( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff) {
    const int mb_xy  = h->s.mb_x + h->s.mb_y*h->s.mb_stride;
    static const int significant_coeff_flag_offset[2][6] = {
      { 105+0, 105+15, 105+29, 105+44, 105+47, 402 },
      { 277+0, 277+15, 277+29, 277+44, 277+47, 436 }
    };
    static const int last_coeff_flag_offset[2][6] = {

}

static inline void compute_mb_neighbors(H264Context *h)
{
    MpegEncContext * const s = &h->s;
    const int mb_xy  = s->mb_x + s->mb_y*s->mb_stride;
    h->top_mb_xy     = mb_xy - s->mb_stride;
    h->left_mb_xy[0] = mb_xy - 1;
    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;

 * decodes a macroblock
 * @returns 0 if ok, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
 */
static int decode_mb_cabac(H264Context *h) {
    MpegEncContext * const s = &h->s;
    const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
    int mb_type, partition_count, cbp = 0;
    int dct8x8_allowed= h->pps.transform_8x8_mode;

    s->dsp.clear_blocks(h->mb); //FIXME avoid if already clear (move after skip handlong?)

    tprintf(s->avctx, "pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
    if( h->slice_type != FF_I_TYPE && h->slice_type != FF_SI_TYPE ) {

    MpegEncContext * const s = &h->s;
    int mb_y_firstrow = s->picture_structure == PICT_BOTTOM_FIELD;
    int mb_xy, mb_type;
    int qp, qp0, qp1, qpc, qpc0, qpc1, qp_thresh;

    mb_xy = mb_x + mb_y*s->mb_stride;

    if(mb_x==0 || mb_y==mb_y_firstrow || !s->dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff ||
       (h->deblocking_filter == 2 && (h->slice_table[mb_xy] != h->slice_table[h->top_mb_xy] ||
                                      h->slice_table[mb_xy] != h->slice_table[mb_xy - 1]))) {
        filter_mb(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize);

6783:df0893f4fd86

};


static void fill_caches(H264Context *h, int mb_type, int for_deblock){
    MpegEncContext * const s = &h->s;
    const int mb_xy= h->mb_xy;
    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 topleft_partition= -1;
    int i;

    h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[0]);
}

static inline void write_back_intra_pred_mode(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int mb_xy= h->mb_xy;

    h->intra4x4_pred_mode[mb_xy][0]= h->intra4x4_pred_mode_cache[7+8*1];
    h->intra4x4_pred_mode[mb_xy][1]= h->intra4x4_pred_mode_cache[7+8*2];
    h->intra4x4_pred_mode[mb_xy][2]= h->intra4x4_pred_mode_cache[7+8*3];
    h->intra4x4_pred_mode[mb_xy][3]= h->intra4x4_pred_mode_cache[7+8*4];

    else      return min;
}

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

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

    }
}

static inline void pred_direct_motion(H264Context * const h, int *mb_type){
    MpegEncContext * const s = &h->s;
    const int mb_xy =   h->mb_xy;
    const int b8_xy = 2*s->mb_x + 2*s->mb_y*h->b8_stride;
    const int b4_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride;
    const int mb_type_col = h->ref_list[1][0].mb_type[mb_xy];
    const int16_t (*l1mv0)[2] = (const int16_t (*)[2]) &h->ref_list[1][0].motion_val[0][b4_xy];
    const int16_t (*l1mv1)[2] = (const int16_t (*)[2]) &h->ref_list[1][0].motion_val[1][b4_xy];

static void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                      qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put),
                      qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg),
                      h264_weight_func *weight_op, h264_biweight_func *weight_avg){
    MpegEncContext * const s = &h->s;
    const int mb_xy= h->mb_xy;
    const int mb_type= s->current_picture.mb_type[mb_xy];

    assert(IS_INTER(mb_type));

    prefetch_motion(h, 0);

    int deblock_left;
    int deblock_top;
    int mb_xy;

    if(h->deblocking_filter == 2) {
        mb_xy = h->mb_xy;
        deblock_left = h->slice_table[mb_xy] == h->slice_table[mb_xy - 1];
        deblock_top  = h->slice_table[mb_xy] == h->slice_table[h->top_mb_xy];
    } else {
        deblock_left = (s->mb_x > 0);
        deblock_top =  (s->mb_y > 0);

static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple){
    MpegEncContext * const s = &h->s;
    const int mb_x= s->mb_x;
    const int mb_y= s->mb_y;
    const int mb_xy= h->mb_xy;
    const int mb_type= s->current_picture.mb_type[mb_xy];
    uint8_t  *dest_y, *dest_cb, *dest_cr;
    int linesize, uvlinesize /*dct_offset*/;
    int i;
    int *block_offset = &h->block_offset[0];

                xchg_pair_border(h, pair_dest_y, pair_dest_cb, pair_dest_cr, s->linesize, s->uvlinesize, 0);

            backup_pair_border(h, pair_dest_y, pair_dest_cb, pair_dest_cr, s->linesize, s->uvlinesize);
            // deblock a pair
            // top
            s->mb_y--; h->mb_xy -= s->mb_stride;
            tprintf(h->s.avctx, "call mbaff filter_mb mb_x:%d mb_y:%d pair_dest_y = %p, dest_y = %p\n", mb_x, mb_y, pair_dest_y, dest_y);
            fill_caches(h, mb_type_top, 1); //FIXME don't fill stuff which isn't used by filter_mb
            h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.qscale_table[mb_xy]);
            h->chroma_qp[1] = get_chroma_qp(h, 1, s->current_picture.qscale_table[mb_xy]);
            filter_mb(h, mb_x, mb_y, pair_dest_y, pair_dest_cb, pair_dest_cr, linesize, uvlinesize);
            // bottom
            s->mb_y++; h->mb_xy += s->mb_stride;
            tprintf(h->s.avctx, "call mbaff filter_mb\n");
            fill_caches(h, mb_type_bottom, 1); //FIXME don't fill stuff which isn't used by filter_mb
            h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.qscale_table[mb_xy+s->mb_stride]);
            h->chroma_qp[1] = get_chroma_qp(h, 1, s->current_picture.qscale_table[mb_xy+s->mb_stride]);
            filter_mb(h, mb_x, mb_y+1, dest_y, dest_cb, dest_cr, linesize, uvlinesize);

static void hl_decode_mb(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int mb_x= s->mb_x;
    const int mb_y= s->mb_y;
    const int mb_xy= h->mb_xy;
    const int mb_type= s->current_picture.mb_type[mb_xy];
    int is_complex = FRAME_MBAFF || MB_FIELD || IS_INTRA_PCM(mb_type) || s->codec_id != CODEC_ID_H264 || (ENABLE_GRAY && (s->flags&CODEC_FLAG_GRAY)) || s->encoding;

    if(!s->decode)
        return;

    return 0;
}

static void predict_field_decoding_flag(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int mb_xy= h->mb_xy;
    int mb_type = (h->slice_table[mb_xy-1] == h->slice_num)
                ? s->current_picture.mb_type[mb_xy-1]
                : (h->slice_table[mb_xy-s->mb_stride] == h->slice_num)
                ? s->current_picture.mb_type[mb_xy-s->mb_stride]
                : 0;

/**
 * decodes a P_SKIP or B_SKIP macroblock
 */
static void decode_mb_skip(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int mb_xy= h->mb_xy;
    int mb_type=0;

    memset(h->non_zero_count[mb_xy], 0, 16);
    memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui

 * decodes a macroblock
 * @returns 0 if ok, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
 */
static int decode_mb_cavlc(H264Context *h){
    MpegEncContext * const s = &h->s;
    int mb_xy;
    int partition_count;
    unsigned int mb_type, cbp;
    int dct8x8_allowed= h->pps.transform_8x8_mode;

    mb_xy = h->mb_xy = s->mb_x + s->mb_y*s->mb_stride;

    s->dsp.clear_blocks(h->mb); //FIXME avoid if already clear (move after skip handlong?

    tprintf(s->avctx, "pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
    cbp = 0; /* avoid warning. FIXME: find a solution without slowing

                && IS_INTERLACED( s->current_picture.mb_type[mbb_xy] ) )
                mbb_xy -= s->mb_stride;
        }else
            mbb_xy = mb_x + (mb_y-1)*s->mb_stride;
    }else{
        int mb_xy = h->mb_xy;
        mba_xy = mb_xy - 1;
        mbb_xy = mb_xy - (s->mb_stride << FIELD_PICTURE);
    }

    if( h->slice_table[mba_xy] == h->slice_num && !IS_SKIP( s->current_picture.mb_type[mba_xy] ))

    3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
    5, 5, 5, 5, 6, 6, 6, 6, 7, 7, 7, 7, 8, 8, 8
};

static void decode_cabac_residual( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff) {
    const int mb_xy  = h->mb_xy;
    static const int significant_coeff_flag_offset[2][6] = {
      { 105+0, 105+15, 105+29, 105+44, 105+47, 402 },
      { 277+0, 277+15, 277+29, 277+44, 277+47, 436 }
    };
    static const int last_coeff_flag_offset[2][6] = {

}

static inline void compute_mb_neighbors(H264Context *h)
{
    MpegEncContext * const s = &h->s;
    const int mb_xy  = h->mb_xy;
    h->top_mb_xy     = mb_xy - s->mb_stride;
    h->left_mb_xy[0] = mb_xy - 1;
    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;

 * decodes a macroblock
 * @returns 0 if ok, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
 */
static int decode_mb_cabac(H264Context *h) {
    MpegEncContext * const s = &h->s;
    int mb_xy;
    int mb_type, partition_count, cbp = 0;
    int dct8x8_allowed= h->pps.transform_8x8_mode;

    mb_xy = h->mb_xy = s->mb_x + s->mb_y*s->mb_stride;

    s->dsp.clear_blocks(h->mb); //FIXME avoid if already clear (move after skip handlong?)

    tprintf(s->avctx, "pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
    if( h->slice_type != FF_I_TYPE && h->slice_type != FF_SI_TYPE ) {

    MpegEncContext * const s = &h->s;
    int mb_y_firstrow = s->picture_structure == PICT_BOTTOM_FIELD;
    int mb_xy, mb_type;
    int qp, qp0, qp1, qpc, qpc0, qpc1, qp_thresh;

    mb_xy = h->mb_xy;

    if(mb_x==0 || mb_y==mb_y_firstrow || !s->dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff ||
       (h->deblocking_filter == 2 && (h->slice_table[mb_xy] != h->slice_table[h->top_mb_xy] ||
                                      h->slice_table[mb_xy] != h->slice_table[mb_xy - 1]))) {
        filter_mb(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize);