comparison h264.c @ 5642:d2598034f2a9 libavcodec

Add slice-based parallel H.264 decoding Patch by Andreas �ªÓman % andreas A olebyn P nu % NB: depends on having a thread library activated at config time, and on having a source encoded with multiple slices Original threads: date: May 18, 2007 11:00 PM subject: [FFmpeg-devel] Parallelized h264 proof-of-concept date: Jun 15, 2007 10:10 PM subject: [FFmpeg-devel] [PATCH] h264 parallelized, (was: Parallelized h264 proof-of-concept) date: Jun 25, 2007 7:02 PM subject: Re: [FFmpeg-devel] [PATCH] h264 parallelized

5641:1e93e637fa21

    }
}

static void free_tables(H264Context *h){
    int i;
    av_freep(&h->intra4x4_pred_mode);
    av_freep(&h->chroma_pred_mode_table);
    av_freep(&h->cbp_table);
    av_freep(&h->mvd_table[0]);
    av_freep(&h->mvd_table[1]);
    av_freep(&h->direct_table);
    av_freep(&h->non_zero_count);
    av_freep(&h->slice_table_base);
    av_freep(&h->top_borders[1]);
    av_freep(&h->top_borders[0]);
    h->slice_table= NULL;

    av_freep(&h->mb2b_xy);
    av_freep(&h->mb2b8_xy);

    av_freep(&h->s.obmc_scratchpad);

    for(i = 0; i < MAX_SPS_COUNT; i++)
        av_freep(h->sps_buffers + i);

    for(i = 0; i < MAX_PPS_COUNT; i++)
        av_freep(h->pps_buffers + i);
}

static void init_dequant8_coeff_table(H264Context *h){
    int i,q,x;
    const int transpose = (h->s.dsp.h264_idct8_add != ff_h264_idct8_add_c); //FIXME ugly

    CHECKED_ALLOCZ(h->intra4x4_pred_mode, big_mb_num * 8  * sizeof(uint8_t))

    CHECKED_ALLOCZ(h->non_zero_count    , big_mb_num * 16 * sizeof(uint8_t))
    CHECKED_ALLOCZ(h->slice_table_base  , (big_mb_num+s->mb_stride) * sizeof(uint8_t))
    CHECKED_ALLOCZ(h->top_borders[0]    , s->mb_width * (16+8+8) * sizeof(uint8_t))
    CHECKED_ALLOCZ(h->top_borders[1]    , s->mb_width * (16+8+8) * sizeof(uint8_t))
    CHECKED_ALLOCZ(h->cbp_table, big_mb_num * sizeof(uint16_t))

    if( h->pps.cabac ) {
        CHECKED_ALLOCZ(h->chroma_pred_mode_table, big_mb_num * sizeof(uint8_t))
        CHECKED_ALLOCZ(h->mvd_table[0], 32*big_mb_num * sizeof(uint16_t));

fail:
    free_tables(h);
    return -1;
}

static void common_init(H264Context *h){
    MpegEncContext * const s = &h->s;

    s->width = s->avctx->width;
    s->height = s->avctx->height;

        h->got_avcC = 0;
    } else {
        h->is_avc = 0;
    }

    return 0;
}

static int frame_start(H264Context *h){
    MpegEncContext * const s = &h->s;

        h->block_offset[24+20+i]= 4*((scan8[i] - scan8[0])&7) + 8*s->uvlinesize*((scan8[i] - scan8[0])>>3);
    }

    /* can't be in alloc_tables because linesize isn't known there.
     * FIXME: redo bipred weight to not require extra buffer? */
    if(!s->obmc_scratchpad)
        s->obmc_scratchpad = av_malloc(16*2*s->linesize + 8*2*s->uvlinesize);

    /* some macroblocks will be accessed before they're available */
    if(FRAME_MBAFF)
        memset(h->slice_table, -1, (s->mb_height*s->mb_stride-1) * sizeof(uint8_t));

//    s->decode= (s->flags&CODEC_FLAG_PSNR) || !s->encoding || s->current_picture.reference /*|| h->contains_intra*/ || 1;
    return 0;
}

        h->field_scan_q0           = h->field_scan;
        h->field_scan8x8_q0        = h->field_scan8x8;
        h->field_scan8x8_cavlc_q0  = h->field_scan8x8_cavlc;
    }
}
/**
 * decodes a slice header.
 * this will allso call MPV_common_init() and frame_start() as needed
 */
static int decode_slice_header(H264Context *h){
    MpegEncContext * const s = &h->s;
    unsigned int first_mb_in_slice;
    unsigned int pps_id;
    int num_ref_idx_active_override_flag;
    static const uint8_t slice_type_map[5]= {P_TYPE, B_TYPE, I_TYPE, SP_TYPE, SI_TYPE};
    unsigned int slice_type, tmp;
    int default_ref_list_done = 0;

    s->current_picture.reference= h->nal_ref_idc != 0;
    s->dropable= h->nal_ref_idc == 0;

    first_mb_in_slice= get_ue_golomb(&s->gb);

    if((s->flags2 & CODEC_FLAG2_CHUNKS) && first_mb_in_slice == 0){
        h->slice_num = 0;
        s->current_picture_ptr= NULL;
    }

    slice_type= get_ue_golomb(&s->gb);
    if(slice_type > 9){

    }else
        h->slice_type_fixed=0;

    slice_type= slice_type_map[ slice_type ];
    if (slice_type == I_TYPE
        || (h->slice_num != 0 && slice_type == h->slice_type) ) {
        default_ref_list_done = 1;
    }
    h->slice_type= slice_type;

    s->pict_type= h->slice_type; // to make a few old func happy, it's wrong though

    pps_id= get_ue_golomb(&s->gb);
    if(pps_id>=MAX_PPS_COUNT){
        av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
        return -1;
    }
    if(!h->pps_buffers[pps_id]) {
        av_log(h->s.avctx, AV_LOG_ERROR, "non existing PPS referenced\n");
        return -1;
    }
    h->pps= *h->pps_buffers[pps_id];

    if(!h->sps_buffers[h->pps.sps_id]) {
        av_log(h->s.avctx, AV_LOG_ERROR, "non existing SPS referenced\n");
        return -1;
    }
    h->sps = *h->sps_buffers[h->pps.sps_id];

    if(h->dequant_coeff_pps != pps_id){
        h->dequant_coeff_pps = pps_id;
        init_dequant_tables(h);
    }

    else
        s->height= 16*s->mb_height - 4*(h->sps.crop_top  + h->sps.crop_bottom); //FIXME recheck

    if (s->context_initialized
        && (   s->width != s->avctx->width || s->height != s->avctx->height)) {
        free_tables(h);
        MPV_common_end(s);
    }
    if (!s->context_initialized) {
        if (MPV_common_init(s) < 0)
            return -1;

        init_scan_tables(h);
        alloc_tables(h);

        s->avctx->width = s->width;
        s->avctx->height = s->height;
        s->avctx->sample_aspect_ratio= h->sps.sar;
        if(!s->avctx->sample_aspect_ratio.den)

            av_reduce(&s->avctx->time_base.num, &s->avctx->time_base.den,
                      s->avctx->time_base.num, s->avctx->time_base.den, 1<<30);
        }
    }

    if(h->slice_num == 0){
        if(frame_start(h) < 0)
            return -1;
    }

    s->current_picture_ptr->frame_num= //FIXME frame_num cleanup
    h->frame_num= get_bits(&s->gb, h->sps.log2_max_frame_num);

    h->mb_mbaff = 0;

        implicit_weight_table(h);
    else
        h->use_weight = 0;

    if(s->current_picture.reference)
        decode_ref_pic_marking(h, &s->gb);

    if(FRAME_MBAFF)
        fill_mbaff_ref_list(h);

    if( h->slice_type != I_TYPE && h->slice_type != SI_TYPE && h->pps.cabac ){

        if( h->deblocking_filter ) {
            h->slice_alpha_c0_offset = get_se_golomb(&s->gb) << 1;
            h->slice_beta_offset = get_se_golomb(&s->gb) << 1;
        }
    }
    if(   s->avctx->skip_loop_filter >= AVDISCARD_ALL
       ||(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY && h->slice_type != I_TYPE)
       ||(s->avctx->skip_loop_filter >= AVDISCARD_BIDIR  && h->slice_type == B_TYPE)
       ||(s->avctx->skip_loop_filter >= AVDISCARD_NONREF && h->nal_ref_idc == 0))
        h->deblocking_filter= 0;

#if 0 //FMO
    if( h->pps.num_slice_groups > 1  && h->pps.mb_slice_group_map_type >= 3 && h->pps.mb_slice_group_map_type <= 5)
        slice_group_change_cycle= get_bits(&s->gb, ?);
#endif

    h->slice_num++;

    h->emu_edge_width= (s->flags&CODEC_FLAG_EMU_EDGE) ? 0 : 16;
    h->emu_edge_height= FRAME_MBAFF ? 0 : h->emu_edge_width;

    if(s->avctx->debug&FF_DEBUG_PICT_INFO){

            }
        }
    }
}

static int decode_slice(H264Context *h){
    MpegEncContext * const s = &h->s;
    const int part_mask= s->partitioned_frame ? (AC_END|AC_ERROR) : 0x7F;

    s->mb_skip_run= -1;

    }

    return 0;
}

static int decode_nal_units(H264Context *h, uint8_t *buf, int buf_size){
    MpegEncContext * const s = &h->s;
    AVCodecContext * const avctx= s->avctx;
    int buf_index=0;
#if 0
    int i;
    for(i=0; i<50; i++){
        av_log(NULL, AV_LOG_ERROR,"%02X ", buf[i]);
    }
#endif
    if(!(s->flags2 & CODEC_FLAG2_CHUNKS)){
        h->slice_num = 0;
        s->current_picture_ptr= NULL;
    }

    for(;;){
        int consumed;
        int dst_length;
        int bit_length;
        uint8_t *ptr;
        int i, nalsize = 0;

        if(h->is_avc) {
            if(buf_index >= buf_size) break;
            nalsize = 0;
            for(i = 0; i < h->nal_length_size; i++)

            if(buf_index+3 >= buf_size) break;

            buf_index+=3;
        }

        ptr= decode_nal(h, buf + buf_index, &dst_length, &consumed, h->is_avc ? nalsize : buf_size - buf_index);
        if (ptr==NULL || dst_length < 0){
            return -1;
        }
        while(ptr[dst_length - 1] == 0 && dst_length > 0)
            dst_length--;
        bit_length= !dst_length ? 0 : (8*dst_length - decode_rbsp_trailing(h, ptr + dst_length - 1));

        if(s->avctx->debug&FF_DEBUG_STARTCODE){
            av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d at %d/%d length %d\n", h->nal_unit_type, buf_index, buf_size, dst_length);
        }

        if (h->is_avc && (nalsize != consumed))
            av_log(h->s.avctx, AV_LOG_ERROR, "AVC: Consumed only %d bytes instead of %d\n", consumed, nalsize);

        if(  (s->hurry_up == 1 && h->nal_ref_idc  == 0) //FIXME do not discard SEI id
           ||(avctx->skip_frame >= AVDISCARD_NONREF && h->nal_ref_idc  == 0))
            continue;

        switch(h->nal_unit_type){
        case NAL_IDR_SLICE:
            idr(h); //FIXME ensure we don't loose some frames if there is reordering
        case NAL_SLICE:
            init_get_bits(&s->gb, ptr, bit_length);
            h->intra_gb_ptr=
            h->inter_gb_ptr= &s->gb;
            s->data_partitioning = 0;

            if(decode_slice_header(h) < 0){
                av_log(h->s.avctx, AV_LOG_ERROR, "decode_slice_header error\n");
                break;
            }
            s->current_picture_ptr->key_frame= (h->nal_unit_type == NAL_IDR_SLICE);
            if(h->redundant_pic_count==0 && s->hurry_up < 5
               && (avctx->skip_frame < AVDISCARD_NONREF || h->nal_ref_idc)
               && (avctx->skip_frame < AVDISCARD_BIDIR  || h->slice_type!=B_TYPE)
               && (avctx->skip_frame < AVDISCARD_NONKEY || h->slice_type==I_TYPE)
               && avctx->skip_frame < AVDISCARD_ALL)
                decode_slice(h);
            break;
        case NAL_DPA:
            init_get_bits(&s->gb, ptr, bit_length);
            h->intra_gb_ptr=
            h->inter_gb_ptr= NULL;
            s->data_partitioning = 1;

            if(decode_slice_header(h) < 0){
                av_log(h->s.avctx, AV_LOG_ERROR, "decode_slice_header error\n");
            }
            break;
        case NAL_DPB:
            init_get_bits(&h->intra_gb, ptr, bit_length);
            h->intra_gb_ptr= &h->intra_gb;
            break;
        case NAL_DPC:
            init_get_bits(&h->inter_gb, ptr, bit_length);
            h->inter_gb_ptr= &h->inter_gb;

            if(h->redundant_pic_count==0 && h->intra_gb_ptr && s->data_partitioning
               && s->context_initialized
               && s->hurry_up < 5
               && (avctx->skip_frame < AVDISCARD_NONREF || h->nal_ref_idc)
               && (avctx->skip_frame < AVDISCARD_BIDIR  || h->slice_type!=B_TYPE)
               && (avctx->skip_frame < AVDISCARD_NONKEY || h->slice_type==I_TYPE)
               && avctx->skip_frame < AVDISCARD_ALL)
                decode_slice(h);
            break;
        case NAL_SEI:
            init_get_bits(&s->gb, ptr, bit_length);
            decode_sei(h);
            break;

        case NAL_AUXILIARY_SLICE:
            break;
        default:
            av_log(avctx, AV_LOG_DEBUG, "Unknown NAL code: %d (%d bits)\n", h->nal_unit_type, bit_length);
        }
    }

    return buf_index;
}

/**
 * returns the number of bytes consumed for building the current frame

5642:d2598034f2a9

    }
}

static void free_tables(H264Context *h){
    int i;
    H264Context *hx;
    av_freep(&h->intra4x4_pred_mode);
    av_freep(&h->chroma_pred_mode_table);
    av_freep(&h->cbp_table);
    av_freep(&h->mvd_table[0]);
    av_freep(&h->mvd_table[1]);
    av_freep(&h->direct_table);
    av_freep(&h->non_zero_count);
    av_freep(&h->slice_table_base);
    h->slice_table= NULL;

    av_freep(&h->mb2b_xy);
    av_freep(&h->mb2b8_xy);

    for(i = 0; i < MAX_SPS_COUNT; i++)
        av_freep(h->sps_buffers + i);

    for(i = 0; i < MAX_PPS_COUNT; i++)
        av_freep(h->pps_buffers + i);

    for(i = 0; i < h->s.avctx->thread_count; i++) {
        hx = h->thread_context[i];
        if(!hx) continue;
        av_freep(&hx->top_borders[1]);
        av_freep(&hx->top_borders[0]);
        av_freep(&hx->s.obmc_scratchpad);
        av_freep(&hx->s.allocated_edge_emu_buffer);
    }
}

static void init_dequant8_coeff_table(H264Context *h){
    int i,q,x;
    const int transpose = (h->s.dsp.h264_idct8_add != ff_h264_idct8_add_c); //FIXME ugly

    CHECKED_ALLOCZ(h->intra4x4_pred_mode, big_mb_num * 8  * sizeof(uint8_t))

    CHECKED_ALLOCZ(h->non_zero_count    , big_mb_num * 16 * sizeof(uint8_t))
    CHECKED_ALLOCZ(h->slice_table_base  , (big_mb_num+s->mb_stride) * sizeof(uint8_t))
    CHECKED_ALLOCZ(h->cbp_table, big_mb_num * sizeof(uint16_t))

    if( h->pps.cabac ) {
        CHECKED_ALLOCZ(h->chroma_pred_mode_table, big_mb_num * sizeof(uint8_t))
        CHECKED_ALLOCZ(h->mvd_table[0], 32*big_mb_num * sizeof(uint16_t));

fail:
    free_tables(h);
    return -1;
}

/**
 * Mimic alloc_tables(), but for every context thread.
 */
static void clone_tables(H264Context *dst, H264Context *src){
    dst->intra4x4_pred_mode       = src->intra4x4_pred_mode;
    dst->non_zero_count           = src->non_zero_count;
    dst->slice_table              = src->slice_table;
    dst->cbp_table                = src->cbp_table;
    dst->mb2b_xy                  = src->mb2b_xy;
    dst->mb2b8_xy                 = src->mb2b8_xy;
    dst->chroma_pred_mode_table   = src->chroma_pred_mode_table;
    dst->mvd_table[0]             = src->mvd_table[0];
    dst->mvd_table[1]             = src->mvd_table[1];
    dst->direct_table             = src->direct_table;

    if(!dst->dequant4_coeff[0])
        init_dequant_tables(dst);
    dst->s.obmc_scratchpad = NULL;
    ff_h264_pred_init(&dst->hpc, src->s.codec_id);
    dst->dequant_coeff_pps= -1;
}

/**
 * Init context
 * Allocate buffers which are not shared amongst multiple threads.
 */
static int context_init(H264Context *h){
    MpegEncContext * const s = &h->s;

    CHECKED_ALLOCZ(h->top_borders[0], h->s.mb_width * (16+8+8) * sizeof(uint8_t))
    CHECKED_ALLOCZ(h->top_borders[1], h->s.mb_width * (16+8+8) * sizeof(uint8_t))

    // edge emu needs blocksize + filter length - 1 (=17x17 for halfpel / 21x21 for h264)
    CHECKED_ALLOCZ(s->allocated_edge_emu_buffer,
                   (s->width+64)*2*21*2); //(width + edge + align)*interlaced*MBsize*tolerance
    s->edge_emu_buffer= s->allocated_edge_emu_buffer + (s->width+64)*2*21;
    return 0;
fail:
    return -1; // free_tables will clean up for us
}

static void common_init(H264Context *h){
    MpegEncContext * const s = &h->s;

    s->width = s->avctx->width;
    s->height = s->avctx->height;

        h->got_avcC = 0;
    } else {
        h->is_avc = 0;
    }

    h->thread_context[0] = h;
    return 0;
}

static int frame_start(H264Context *h){
    MpegEncContext * const s = &h->s;

        h->block_offset[24+20+i]= 4*((scan8[i] - scan8[0])&7) + 8*s->uvlinesize*((scan8[i] - scan8[0])>>3);
    }

    /* can't be in alloc_tables because linesize isn't known there.
     * FIXME: redo bipred weight to not require extra buffer? */
    for(i = 0; i < s->avctx->thread_count; i++)
        if(!h->thread_context[i]->s.obmc_scratchpad)
            h->thread_context[i]->s.obmc_scratchpad = av_malloc(16*2*s->linesize + 8*2*s->uvlinesize);

    /* some macroblocks will be accessed before they're available */
    if(FRAME_MBAFF || s->avctx->thread_count > 1)
        memset(h->slice_table, -1, (s->mb_height*s->mb_stride-1) * sizeof(uint8_t));

//    s->decode= (s->flags&CODEC_FLAG_PSNR) || !s->encoding || s->current_picture.reference /*|| h->contains_intra*/ || 1;
    return 0;
}

        h->field_scan_q0           = h->field_scan;
        h->field_scan8x8_q0        = h->field_scan8x8;
        h->field_scan8x8_cavlc_q0  = h->field_scan8x8_cavlc;
    }
}

/**
 * Replicates H264 "master" context to thread contexts.
 */
static void clone_slice(H264Context *dst, H264Context *src)
{
    memcpy(dst->block_offset,     src->block_offset, sizeof(dst->block_offset));
    dst->s.current_picture_ptr  = src->s.current_picture_ptr;
    dst->s.current_picture      = src->s.current_picture;
    dst->s.linesize             = src->s.linesize;
    dst->s.uvlinesize           = src->s.uvlinesize;

    dst->prev_poc_msb           = src->prev_poc_msb;
    dst->prev_poc_lsb           = src->prev_poc_lsb;
    dst->prev_frame_num_offset  = src->prev_frame_num_offset;
    dst->prev_frame_num         = src->prev_frame_num;
    dst->short_ref_count        = src->short_ref_count;

    memcpy(dst->short_ref,        src->short_ref,        sizeof(dst->short_ref));
    memcpy(dst->long_ref,         src->long_ref,         sizeof(dst->long_ref));
    memcpy(dst->default_ref_list, src->default_ref_list, sizeof(dst->default_ref_list));
    memcpy(dst->ref_list,         src->ref_list,         sizeof(dst->ref_list));
}

/**
 * decodes a slice header.
 * this will allso call MPV_common_init() and frame_start() as needed
 *
 * @param h h264context
 * @param h0 h264 master context (differs from 'h' when doing sliced based parallel decoding)
 *
 * @return 0 if okay, <0 if an error occured, 1 if decoding must not be multithreaded
 */
static int decode_slice_header(H264Context *h, H264Context *h0){
    MpegEncContext * const s = &h->s;
    unsigned int first_mb_in_slice;
    unsigned int pps_id;
    int num_ref_idx_active_override_flag;
    static const uint8_t slice_type_map[5]= {P_TYPE, B_TYPE, I_TYPE, SP_TYPE, SI_TYPE};
    unsigned int slice_type, tmp, i;
    int default_ref_list_done = 0;

    s->current_picture.reference= h->nal_ref_idc != 0;
    s->dropable= h->nal_ref_idc == 0;

    first_mb_in_slice= get_ue_golomb(&s->gb);

    if((s->flags2 & CODEC_FLAG2_CHUNKS) && first_mb_in_slice == 0){
        h0->current_slice = 0;
        s->current_picture_ptr= NULL;
    }

    slice_type= get_ue_golomb(&s->gb);
    if(slice_type > 9){

    }else
        h->slice_type_fixed=0;

    slice_type= slice_type_map[ slice_type ];
    if (slice_type == I_TYPE
        || (h0->current_slice != 0 && slice_type == h0->last_slice_type) ) {
        default_ref_list_done = 1;
    }
    h->slice_type= slice_type;

    s->pict_type= h->slice_type; // to make a few old func happy, it's wrong though

    pps_id= get_ue_golomb(&s->gb);
    if(pps_id>=MAX_PPS_COUNT){
        av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
        return -1;
    }
    if(!h0->pps_buffers[pps_id]) {
        av_log(h->s.avctx, AV_LOG_ERROR, "non existing PPS referenced\n");
        return -1;
    }
    h->pps= *h0->pps_buffers[pps_id];

    if(!h0->sps_buffers[h->pps.sps_id]) {
        av_log(h->s.avctx, AV_LOG_ERROR, "non existing SPS referenced\n");
        return -1;
    }
    h->sps = *h0->sps_buffers[h->pps.sps_id];

    if(h->dequant_coeff_pps != pps_id){
        h->dequant_coeff_pps = pps_id;
        init_dequant_tables(h);
    }

    else
        s->height= 16*s->mb_height - 4*(h->sps.crop_top  + h->sps.crop_bottom); //FIXME recheck

    if (s->context_initialized
        && (   s->width != s->avctx->width || s->height != s->avctx->height)) {
        if(h != h0)
            return -1;   // width / height changed during parallelized decoding
        free_tables(h);
        MPV_common_end(s);
    }
    if (!s->context_initialized) {
        if(h != h0)
            return -1;  // we cant (re-)initialize context during parallel decoding
        if (MPV_common_init(s) < 0)
            return -1;

        init_scan_tables(h);
        alloc_tables(h);

        for(i = 1; i < s->avctx->thread_count; i++) {
            H264Context *c;
            c = h->thread_context[i] = av_malloc(sizeof(H264Context));
            memcpy(c, h, sizeof(MpegEncContext));
            memset(&c->s + 1, 0, sizeof(H264Context) - sizeof(MpegEncContext));
            c->sps = h->sps;
            c->pps = h->pps;
            init_scan_tables(c);
            clone_tables(c, h);
        }

        for(i = 0; i < s->avctx->thread_count; i++)
            if(context_init(h->thread_context[i]) < 0)
                return -1;

        s->avctx->width = s->width;
        s->avctx->height = s->height;
        s->avctx->sample_aspect_ratio= h->sps.sar;
        if(!s->avctx->sample_aspect_ratio.den)

            av_reduce(&s->avctx->time_base.num, &s->avctx->time_base.den,
                      s->avctx->time_base.num, s->avctx->time_base.den, 1<<30);
        }
    }

    if(h0->current_slice == 0){
        if(frame_start(h) < 0)
            return -1;
    }
    if(h != h0)
        clone_slice(h, h0);

    s->current_picture_ptr->frame_num= //FIXME frame_num cleanup
    h->frame_num= get_bits(&s->gb, h->sps.log2_max_frame_num);

    h->mb_mbaff = 0;

        implicit_weight_table(h);
    else
        h->use_weight = 0;

    if(s->current_picture.reference)
        decode_ref_pic_marking(h0, &s->gb);

    if(FRAME_MBAFF)
        fill_mbaff_ref_list(h);

    if( h->slice_type != I_TYPE && h->slice_type != SI_TYPE && h->pps.cabac ){

        if( h->deblocking_filter ) {
            h->slice_alpha_c0_offset = get_se_golomb(&s->gb) << 1;
            h->slice_beta_offset = get_se_golomb(&s->gb) << 1;
        }
    }

    if(h->deblocking_filter == 1 && h0->max_contexts > 1) {
        h0->max_contexts = 1;
        if(!h0->single_decode_warning) {
            av_log(s->avctx, AV_LOG_INFO, "Cannot parallelize deblocking type 1, decoding such frames in sequential order\n");
            h0->single_decode_warning = 1;
        }
        if(h != h0)
            return 1; // deblocking switched inside frame
    }

    if(   s->avctx->skip_loop_filter >= AVDISCARD_ALL
       ||(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY && h->slice_type != I_TYPE)
       ||(s->avctx->skip_loop_filter >= AVDISCARD_BIDIR  && h->slice_type == B_TYPE)
       ||(s->avctx->skip_loop_filter >= AVDISCARD_NONREF && h->nal_ref_idc == 0))
        h->deblocking_filter= 0;

#if 0 //FMO
    if( h->pps.num_slice_groups > 1  && h->pps.mb_slice_group_map_type >= 3 && h->pps.mb_slice_group_map_type <= 5)
        slice_group_change_cycle= get_bits(&s->gb, ?);
#endif

    h0->last_slice_type = slice_type;
    h->slice_num = ++h0->current_slice;

    h->emu_edge_width= (s->flags&CODEC_FLAG_EMU_EDGE) ? 0 : 16;
    h->emu_edge_height= FRAME_MBAFF ? 0 : h->emu_edge_width;

    if(s->avctx->debug&FF_DEBUG_PICT_INFO){

            }
        }
    }
}

static int decode_slice(struct AVCodecContext *avctx, H264Context *h){
    MpegEncContext * const s = &h->s;
    const int part_mask= s->partitioned_frame ? (AC_END|AC_ERROR) : 0x7F;

    s->mb_skip_run= -1;

    }

    return 0;
}

/**
 * Call decode_slice() for each context.
 *
 * @param h h264 master context
 * @param context_count number of contexts to execute
 */
static void execute_decode_slices(H264Context *h, int context_count){
    MpegEncContext * const s = &h->s;
    AVCodecContext * const avctx= s->avctx;
    H264Context *hx;
    int i;

    if(context_count == 1) {
        decode_slice(avctx, h);
    } else {
        for(i = 1; i < context_count; i++) {
            hx = h->thread_context[i];
            hx->s.error_resilience = avctx->error_resilience;
            hx->s.error_count = 0;
        }

        avctx->execute(avctx, (void *)decode_slice,
                       (void **)h->thread_context, NULL, context_count);

        /* pull back stuff from slices to master context */
        hx = h->thread_context[context_count - 1];
        s->mb_x = hx->s.mb_x;
        s->mb_y = hx->s.mb_y;
        for(i = 1; i < context_count; i++)
            h->s.error_count += h->thread_context[i]->s.error_count;
    }
}


static int decode_nal_units(H264Context *h, uint8_t *buf, int buf_size){
    MpegEncContext * const s = &h->s;
    AVCodecContext * const avctx= s->avctx;
    int buf_index=0;
    H264Context *hx; ///< thread context
    int context_count = 0;

    h->max_contexts = avctx->thread_count;
#if 0
    int i;
    for(i=0; i<50; i++){
        av_log(NULL, AV_LOG_ERROR,"%02X ", buf[i]);
    }
#endif
    if(!(s->flags2 & CODEC_FLAG2_CHUNKS)){
        h->current_slice = 0;
        s->current_picture_ptr= NULL;
    }

    for(;;){
        int consumed;
        int dst_length;
        int bit_length;
        uint8_t *ptr;
        int i, nalsize = 0;
        int err;

        if(h->is_avc) {
            if(buf_index >= buf_size) break;
            nalsize = 0;
            for(i = 0; i < h->nal_length_size; i++)

            if(buf_index+3 >= buf_size) break;

            buf_index+=3;
        }

        hx = h->thread_context[context_count];

        ptr= decode_nal(hx, buf + buf_index, &dst_length, &consumed, h->is_avc ? nalsize : buf_size - buf_index);
        if (ptr==NULL || dst_length < 0){
            return -1;
        }
        while(ptr[dst_length - 1] == 0 && dst_length > 0)
            dst_length--;
        bit_length= !dst_length ? 0 : (8*dst_length - decode_rbsp_trailing(h, ptr + dst_length - 1));

        if(s->avctx->debug&FF_DEBUG_STARTCODE){
            av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d at %d/%d length %d\n", hx->nal_unit_type, buf_index, buf_size, dst_length);
        }

        if (h->is_avc && (nalsize != consumed))
            av_log(h->s.avctx, AV_LOG_ERROR, "AVC: Consumed only %d bytes instead of %d\n", consumed, nalsize);

        if(  (s->hurry_up == 1 && h->nal_ref_idc  == 0) //FIXME do not discard SEI id
           ||(avctx->skip_frame >= AVDISCARD_NONREF && h->nal_ref_idc  == 0))
            continue;

      again:
        err = 0;
        switch(hx->nal_unit_type){
        case NAL_IDR_SLICE:
            if (h->nal_unit_type != NAL_IDR_SLICE) {
                av_log(h->s.avctx, AV_LOG_ERROR, "Invalid mix of idr and non-idr slices");
                return -1;
            }
            idr(h); //FIXME ensure we don't loose some frames if there is reordering
        case NAL_SLICE:
            init_get_bits(&hx->s.gb, ptr, bit_length);
            hx->intra_gb_ptr=
            hx->inter_gb_ptr= &hx->s.gb;
            hx->s.data_partitioning = 0;

            if((err = decode_slice_header(hx, h)))
               break;

            s->current_picture_ptr->key_frame= (hx->nal_unit_type == NAL_IDR_SLICE);
            if(hx->redundant_pic_count==0 && hx->s.hurry_up < 5
               && (avctx->skip_frame < AVDISCARD_NONREF || hx->nal_ref_idc)
               && (avctx->skip_frame < AVDISCARD_BIDIR  || hx->slice_type!=B_TYPE)
               && (avctx->skip_frame < AVDISCARD_NONKEY || hx->slice_type==I_TYPE)
               && avctx->skip_frame < AVDISCARD_ALL)
                context_count++;
            break;
        case NAL_DPA:
            init_get_bits(&hx->s.gb, ptr, bit_length);
            hx->intra_gb_ptr=
            hx->inter_gb_ptr= NULL;
            hx->s.data_partitioning = 1;

            err = decode_slice_header(hx, h);
            break;
        case NAL_DPB:
            init_get_bits(&hx->intra_gb, ptr, bit_length);
            hx->intra_gb_ptr= &hx->intra_gb;
            break;
        case NAL_DPC:
            init_get_bits(&hx->inter_gb, ptr, bit_length);
            hx->inter_gb_ptr= &hx->inter_gb;

            if(hx->redundant_pic_count==0 && hx->intra_gb_ptr && hx->s.data_partitioning
               && s->context_initialized
               && s->hurry_up < 5
               && (avctx->skip_frame < AVDISCARD_NONREF || hx->nal_ref_idc)
               && (avctx->skip_frame < AVDISCARD_BIDIR  || hx->slice_type!=B_TYPE)
               && (avctx->skip_frame < AVDISCARD_NONKEY || hx->slice_type==I_TYPE)
               && avctx->skip_frame < AVDISCARD_ALL)
                context_count++;
            break;
        case NAL_SEI:
            init_get_bits(&s->gb, ptr, bit_length);
            decode_sei(h);
            break;

        case NAL_AUXILIARY_SLICE:
            break;
        default:
            av_log(avctx, AV_LOG_DEBUG, "Unknown NAL code: %d (%d bits)\n", h->nal_unit_type, bit_length);
        }

        if(context_count == h->max_contexts) {
            execute_decode_slices(h, context_count);
            context_count = 0;
        }

        if (err < 0)
            av_log(h->s.avctx, AV_LOG_ERROR, "decode_slice_header error\n");
        else if(err == 1) {
            /* Slice could not be decoded in parallel mode, copy down
             * NAL unit stuff to context 0 and restart. Note that
             * rbsp_buffer is not transfered, but since we no longer
             * run in parallel mode this should not be an issue. */
            h->nal_unit_type = hx->nal_unit_type;
            h->nal_ref_idc   = hx->nal_ref_idc;
            hx = h;
            goto again;
        }
    }
    if(context_count)
        execute_decode_slices(h, context_count);
    return buf_index;
}

/**
 * returns the number of bytes consumed for building the current frame