Merge remote-tracking branch 'qatar/master'

* qatar/master: h264: deMpegEncContextize Conflicts: libavcodec/dxva2_h264.c libavcodec/h264.c libavcodec/h264.h libavcodec/h264_cabac.c libavcodec/h264_cavlc.c libavcodec/h264_loopfilter.c libavcodec/h264_mb_template.c libavcodec/h264_parser.c libavcodec/h264_ps.c libavcodec/h264_refs.c libavcodec/h264_sei.c libavcodec/svq3.c Merged-by: Michael Niedermayer <michaelni@gmx.at>
12 years ago · b7fe35c9e5
--- a/libavcodec/crystalhd.c
+++ b/libavcodec/crystalhd.c
@@ -932,8 +932,8 @@ static int decode(AVCodecContext *avctx, void *data, int *got_frame, AVPacket *a
                } else {
                    av_log(avctx, AV_LOG_VERBOSE,
                           "CrystalHD: parser picture type %d\n",
                           h->s.picture_structure);
                    pic_type = h->s.picture_structure;
                           h->picture_structure);
                    pic_type = h->picture_structure;
                }
            } else {
                av_log(avctx, AV_LOG_WARNING,
--- a/libavcodec/dxva2.c
+++ b/libavcodec/dxva2.c
@@ -76,7 +76,7 @@ int ff_dxva2_commit_buffer(AVCodecContext *avctx,
    return result;
 }
 int ff_dxva2_common_end_frame(AVCodecContext *avctx, MpegEncContext *s,
 int ff_dxva2_common_end_frame(AVCodecContext *avctx, Picture *pic,
                              const void *pp, unsigned pp_size,
                              const void *qm, unsigned qm_size,
                              int (*commit_bs_si)(AVCodecContext *,
@@ -90,7 +90,7 @@ int ff_dxva2_common_end_frame(AVCodecContext *avctx, MpegEncContext *s,
    int      result;
    if (FAILED(IDirectXVideoDecoder_BeginFrame(ctx->decoder,
                                               ff_dxva2_get_surface(s->current_picture_ptr),
                                               ff_dxva2_get_surface(pic),
                                               NULL))) {
        av_log(avctx, AV_LOG_ERROR, "Failed to begin frame\n");
        return -1;
@@ -146,7 +146,5 @@ end:
        result = -1;
    }
    if (!result)
        ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
    return result;
 }
--- a/libavcodec/dxva2_h264.c
+++ b/libavcodec/dxva2_h264.c
@@ -44,15 +44,14 @@ static void fill_picture_entry(DXVA_PicEntry_H264 *pic,
 static void fill_picture_parameters(struct dxva_context *ctx, const H264Context *h,
                                    DXVA_PicParams_H264 *pp)
 {
    const MpegEncContext *s = &h->s;
    const Picture *current_picture = s->current_picture_ptr;
    const Picture *current_picture = h->cur_pic_ptr;
    int i, j;
    memset(pp, 0, sizeof(*pp));
    /* Configure current picture */
    fill_picture_entry(&pp->CurrPic,
                       ff_dxva2_get_surface_index(ctx, current_picture),
                       s->picture_structure == PICT_BOTTOM_FIELD);
                       h->picture_structure == PICT_BOTTOM_FIELD);
    /* Configure the set of references */
    pp->UsedForReferenceFlags  = 0;
    pp->NonExistingFrameFlags  = 0;
@@ -88,13 +87,13 @@ static void fill_picture_parameters(struct dxva_context *ctx, const H264Context
        }
    }
    pp->wFrameWidthInMbsMinus1        = s->mb_width  - 1;
    pp->wFrameHeightInMbsMinus1       = s->mb_height - 1;
    pp->wFrameWidthInMbsMinus1        = h->mb_width  - 1;
    pp->wFrameHeightInMbsMinus1       = h->mb_height - 1;
    pp->num_ref_frames                = h->sps.ref_frame_count;
    pp->wBitFields                    = ((s->picture_structure != PICT_FRAME) <<  0) |
    pp->wBitFields                    = ((h->picture_structure != PICT_FRAME) <<  0) |
                                        ((h->sps.mb_aff &&
                                        (s->picture_structure == PICT_FRAME)) <<  1) |
                                        (h->picture_structure == PICT_FRAME)) <<  1) |
                                        (h->sps.residual_color_transform_flag <<  2) |
                                        /* sp_for_switch_flag (not implemented by FFmpeg) */
                                        (0                                    <<  3) |
@@ -120,11 +119,11 @@ static void fill_picture_parameters(struct dxva_context *ctx, const H264Context
        pp->Reserved16Bits            = 3; /* FIXME is there a way to detect the right mode ? */
    pp->StatusReportFeedbackNumber    = 1 + ctx->report_id++;
    pp->CurrFieldOrderCnt[0] = 0;
    if ((s->picture_structure & PICT_TOP_FIELD) &&
    if ((h->picture_structure & PICT_TOP_FIELD) &&
        current_picture->field_poc[0] != INT_MAX)
        pp->CurrFieldOrderCnt[0] = current_picture->field_poc[0];
    pp->CurrFieldOrderCnt[1] = 0;
    if ((s->picture_structure & PICT_BOTTOM_FIELD) &&
    if ((h->picture_structure & PICT_BOTTOM_FIELD) &&
        current_picture->field_poc[1] != INT_MAX)
        pp->CurrFieldOrderCnt[1] = current_picture->field_poc[1];
    pp->pic_init_qs_minus26           = h->pps.init_qs - 26;
@@ -200,7 +199,6 @@ static void fill_slice_long(AVCodecContext *avctx, DXVA_Slice_H264_Long *slice,
 {
    const H264Context *h = avctx->priv_data;
    struct dxva_context *ctx = avctx->hwaccel_context;
    const MpegEncContext *s = &h->s;
    unsigned list;
    memset(slice, 0, sizeof(*slice));
@@ -208,9 +206,9 @@ static void fill_slice_long(AVCodecContext *avctx, DXVA_Slice_H264_Long *slice,
    slice->SliceBytesInBuffer    = size;
    slice->wBadSliceChopping     = 0;
    slice->first_mb_in_slice     = (s->mb_y >> FIELD_OR_MBAFF_PICTURE) * s->mb_width + s->mb_x;
    slice->first_mb_in_slice     = (h->mb_y >> FIELD_OR_MBAFF_PICTURE) * h->mb_width + h->mb_x;
    slice->NumMbsForSlice        = 0; /* XXX it is set once we have all slices */
    slice->BitOffsetToSliceData  = get_bits_count(&s->gb);
    slice->BitOffsetToSliceData  = get_bits_count(&h->gb);
    slice->slice_type            = ff_h264_get_slice_type(h);
    if (h->slice_type_fixed)
        slice->slice_type += 5;
@@ -260,7 +258,7 @@ static void fill_slice_long(AVCodecContext *avctx, DXVA_Slice_H264_Long *slice,
        }
    }
    slice->slice_qs_delta    = 0; /* XXX not implemented by FFmpeg */
    slice->slice_qp_delta    = s->qscale - h->pps.init_qp;
    slice->slice_qp_delta    = h->qscale - h->pps.init_qp;
    slice->redundant_pic_cnt = h->redundant_pic_count;
    if (h->slice_type == AV_PICTURE_TYPE_B)
        slice->direct_spatial_mv_pred_flag = h->direct_spatial_mv_pred;
@@ -277,10 +275,9 @@ static int commit_bitstream_and_slice_buffer(AVCodecContext *avctx,
                                             DXVA2_DecodeBufferDesc *sc)
 {
    const H264Context *h = avctx->priv_data;
    const MpegEncContext *s = &h->s;
    const unsigned mb_count = s->mb_width * s->mb_height;
    const unsigned mb_count = h->mb_width * h->mb_height;
    struct dxva_context *ctx = avctx->hwaccel_context;
    const Picture *current_picture = h->s.current_picture_ptr;
    const Picture *current_picture = h->cur_pic_ptr;
    struct dxva2_picture_context *ctx_pic = current_picture->f.hwaccel_picture_private;
    DXVA_Slice_H264_Short *slice = NULL;
    uint8_t  *dxva_data, *current, *end;
@@ -376,7 +373,7 @@ static int start_frame(AVCodecContext *avctx,
 {
    const H264Context *h = avctx->priv_data;
    struct dxva_context *ctx = avctx->hwaccel_context;
    struct dxva2_picture_context *ctx_pic = h->s.current_picture_ptr->f.hwaccel_picture_private;
    struct dxva2_picture_context *ctx_pic = h->cur_pic_ptr->f.hwaccel_picture_private;
    if (!ctx->decoder || !ctx->cfg || ctx->surface_count <= 0)
        return -1;
@@ -399,7 +396,7 @@ static int decode_slice(AVCodecContext *avctx,
 {
    const H264Context *h = avctx->priv_data;
    struct dxva_context *ctx = avctx->hwaccel_context;
    const Picture *current_picture = h->s.current_picture_ptr;
    const Picture *current_picture = h->cur_pic_ptr;
    struct dxva2_picture_context *ctx_pic = current_picture->f.hwaccel_picture_private;
    unsigned position;
@@ -427,16 +424,19 @@ static int decode_slice(AVCodecContext *avctx,
 static int end_frame(AVCodecContext *avctx)
 {
    H264Context *h = avctx->priv_data;
    MpegEncContext *s = &h->s;
    struct dxva2_picture_context *ctx_pic =
        h->s.current_picture_ptr->f.hwaccel_picture_private;
        h->cur_pic_ptr->f.hwaccel_picture_private;
    int ret;
    if (ctx_pic->slice_count <= 0 || ctx_pic->bitstream_size <= 0)
        return -1;
    return ff_dxva2_common_end_frame(avctx, s,
                                     &ctx_pic->pp, sizeof(ctx_pic->pp),
                                     &ctx_pic->qm, sizeof(ctx_pic->qm),
                                     commit_bitstream_and_slice_buffer);
    ret = ff_dxva2_common_end_frame(avctx, h->cur_pic_ptr,
                                    &ctx_pic->pp, sizeof(ctx_pic->pp),
                                    &ctx_pic->qm, sizeof(ctx_pic->qm),
                                    commit_bitstream_and_slice_buffer);
    if (!ret)
        ff_h264_draw_horiz_band(h, 0, h->avctx->height);
    return ret;
 }
 AVHWAccel ff_h264_dxva2_hwaccel = {
--- a/libavcodec/dxva2_internal.h
+++ b/libavcodec/dxva2_internal.h
@@ -54,7 +54,7 @@ int ff_dxva2_commit_buffer(AVCodecContext *, struct dxva_context *,
                           unsigned mb_count);
 int ff_dxva2_common_end_frame(AVCodecContext *, MpegEncContext *,
 int ff_dxva2_common_end_frame(AVCodecContext *, Picture *,
                              const void *pp, unsigned pp_size,
                              const void *qm, unsigned qm_size,
                              int (*commit_bs_si)(AVCodecContext *,
--- a/libavcodec/dxva2_mpeg2.c
+++ b/libavcodec/dxva2_mpeg2.c
@@ -251,13 +251,17 @@ static int end_frame(AVCodecContext *avctx)
    struct MpegEncContext *s = avctx->priv_data;
    struct dxva2_picture_context *ctx_pic =
        s->current_picture_ptr->f.hwaccel_picture_private;
    int ret;
    if (ctx_pic->slice_count <= 0 || ctx_pic->bitstream_size <= 0)
        return -1;
    return ff_dxva2_common_end_frame(avctx, s,
                                     &ctx_pic->pp, sizeof(ctx_pic->pp),
                                     &ctx_pic->qm, sizeof(ctx_pic->qm),
                                     commit_bitstream_and_slice_buffer);
    ret = ff_dxva2_common_end_frame(avctx, s->current_picture_ptr,
                                    &ctx_pic->pp, sizeof(ctx_pic->pp),
                                    &ctx_pic->qm, sizeof(ctx_pic->qm),
                                    commit_bitstream_and_slice_buffer);
    if (!ret)
        ff_mpeg_draw_horiz_band(s, 0, avctx->height);
    return ret;
 }
 AVHWAccel ff_mpeg2_dxva2_hwaccel = {
--- a/libavcodec/dxva2_vc1.c
+++ b/libavcodec/dxva2_vc1.c
@@ -257,14 +257,18 @@ static int end_frame(AVCodecContext *avctx)
 {
    VC1Context *v = avctx->priv_data;
    struct dxva2_picture_context *ctx_pic = v->s.current_picture_ptr->f.hwaccel_picture_private;
    int ret;
    if (ctx_pic->bitstream_size <= 0)
        return -1;
    return ff_dxva2_common_end_frame(avctx, &v->s,
                                     &ctx_pic->pp, sizeof(ctx_pic->pp),
                                     NULL, 0,
                                     commit_bitstream_and_slice_buffer);
    ret = ff_dxva2_common_end_frame(avctx, v->s.current_picture_ptr,
                                    &ctx_pic->pp, sizeof(ctx_pic->pp),
                                    NULL, 0,
                                    commit_bitstream_and_slice_buffer);
    if (!ret)
        ff_mpeg_draw_horiz_band(&v->s, 0, avctx->height);
    return ret;
 }
 #if CONFIG_WMV3_DXVA2_HWACCEL
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
--- a/libavcodec/h264.h
+++ b/libavcodec/h264.h
@@ -30,6 +30,7 @@
 #include "libavutil/intreadwrite.h"
 #include "cabac.h"
 #include "get_bits.h"
 #include "mpegvideo.h"
 #include "h264chroma.h"
 #include "h264dsp.h"
@@ -62,7 +63,7 @@
 #define MB_MBAFF    h->mb_mbaff
 #define MB_FIELD    h->mb_field_decoding_flag
 #define FRAME_MBAFF h->mb_aff_frame
 #define FIELD_PICTURE (s->picture_structure != PICT_FRAME)
 #define FIELD_PICTURE (h->picture_structure != PICT_FRAME)
 #define LEFT_MBS 2
 #define LTOP     0
 #define LBOT     1
@@ -254,15 +255,42 @@ typedef struct MMCO {
 * H264Context
 */
 typedef struct H264Context {
    MpegEncContext s;
    AVCodecContext *avctx;
    DSPContext       dsp;
    VideoDSPContext vdsp;
    H264DSPContext h264dsp;
    H264ChromaContext h264chroma;
    H264QpelContext h264qpel;
    MotionEstContext me;
    ParseContext parse_context;
    GetBitContext gb;
    ERContext er;
    Picture *DPB;
    Picture *cur_pic_ptr;
    Picture cur_pic;
    int picture_count;
    int picture_range_start, picture_range_end;
    int pixel_shift;    ///< 0 for 8-bit H264, 1 for high-bit-depth H264
    int chroma_qp[2];   // QPc
    int qp_thresh;      ///< QP threshold to skip loopfilter
    int width, height;
    int linesize, uvlinesize;
    int chroma_x_shift, chroma_y_shift;
    int qscale;
    int droppable;
    int data_partitioning;
    int coded_picture_number;
    int low_delay;
    int context_initialized;
    int flags;
    int workaround_bugs;
    int prev_mb_skipped;
    int next_mb_skipped;
@@ -352,6 +380,8 @@ typedef struct H264Context {
    int mb_aff_frame;
    int mb_field_decoding_flag;
    int mb_mbaff;               ///< mb_aff_frame && mb_field_decoding_flag
    int picture_structure;
    int first_field;
    DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
@@ -428,6 +458,13 @@ typedef struct H264Context {
    int x264_build;
    int mb_x, mb_y;
    int resync_mb_x;
    int resync_mb_y;
    int mb_skip_run;
    int mb_height, mb_width;
    int mb_stride;
    int mb_num;
    int mb_xy;
    int is_complex;
@@ -452,7 +489,8 @@ typedef struct H264Context {
    int nal_length_size;  ///< Number of bytes used for nal length (1, 2 or 4)
    int got_first;        ///< this flag is != 0 if we've parsed a frame
    int context_reinitialized;
    int bit_depth_luma;         ///< luma bit depth from sps to detect changes
    int chroma_format_idc;      ///< chroma format from sps to detect changes
    SPS *sps_buffers[MAX_SPS_COUNT];
    PPS *pps_buffers[MAX_PPS_COUNT];
@@ -525,12 +563,16 @@ typedef struct H264Context {
     */
    int max_contexts;
    int slice_context_count;
    /**
     *  1 if the single thread fallback warning has already been
     *  displayed, 0 otherwise.
     */
    int single_decode_warning;
    enum AVPictureType pict_type;
    int last_slice_type;
    /** @} */
@@ -602,6 +644,8 @@ typedef struct H264Context {
    uint8_t parse_history[4];
    int parse_history_count;
    int parse_last_mb;
    uint8_t *edge_emu_buffer;
    int16_t *dc_val_base;
 } H264Context;
 extern const uint8_t ff_h264_chroma_qp[7][QP_MAX_NUM + 1]; ///< One chroma qp table for each possible bit depth (8-14).
@@ -810,7 +854,7 @@ static av_always_inline int pred_intra_mode(H264Context *h, int n)
    const int top    = h->intra4x4_pred_mode_cache[index8 - 8];
    const int min    = FFMIN(left, top);
    tprintf(h->s.avctx, "mode:%d %d min:%d\n", left, top, min);
    tprintf(h->avctx, "mode:%d %d min:%d\n", left, top, min);
    if (min < 0)
        return DC_PRED;
@@ -844,7 +888,7 @@ static av_always_inline void write_back_non_zero_count(H264Context *h)
    AV_COPY32(&nnz[32], &nnz_cache[4 + 8 * 11]);
    AV_COPY32(&nnz[36], &nnz_cache[4 + 8 * 12]);
    if (!h->s.chroma_y_shift) {
    if (!h->chroma_y_shift) {
        AV_COPY32(&nnz[24], &nnz_cache[4 + 8 * 8]);
        AV_COPY32(&nnz[28], &nnz_cache[4 + 8 * 9]);
        AV_COPY32(&nnz[40], &nnz_cache[4 + 8 * 13]);
@@ -853,12 +897,11 @@ static av_always_inline void write_back_non_zero_count(H264Context *h)
 }
 static av_always_inline void write_back_motion_list(H264Context *h,
                                                    MpegEncContext *const s,
                                                    int b_stride,
                                                    int b_xy, int b8_xy,
                                                    int mb_type, int list)
 {
    int16_t(*mv_dst)[2] = &s->current_picture.f.motion_val[list][b_xy];
    int16_t(*mv_dst)[2] = &h->cur_pic.f.motion_val[list][b_xy];
    int16_t(*mv_src)[2] = &h->mv_cache[list][scan8[0]];
    AV_COPY128(mv_dst + 0 * b_stride, mv_src + 8 * 0);
    AV_COPY128(mv_dst + 1 * b_stride, mv_src + 8 * 1);
@@ -879,7 +922,7 @@ static av_always_inline void write_back_motion_list(H264Context *h,
    }
    {
        int8_t *ref_index = &s->current_picture.f.ref_index[list][b8_xy];
        int8_t *ref_index = &h->cur_pic.f.ref_index[list][b8_xy];
        int8_t *ref_cache = h->ref_cache[list];
        ref_index[0 + 0 * 2] = ref_cache[scan8[0]];
        ref_index[1 + 0 * 2] = ref_cache[scan8[4]];
@@ -890,19 +933,18 @@ static av_always_inline void write_back_motion_list(H264Context *h,
 static av_always_inline void write_back_motion(H264Context *h, int mb_type)
 {
    MpegEncContext *const s = &h->s;
    const int b_stride      = h->b_stride;
    const int b_xy  = 4 * s->mb_x + 4 * s->mb_y * h->b_stride; // try mb2b(8)_xy
    const int b_xy  = 4 * h->mb_x + 4 * h->mb_y * h->b_stride; // try mb2b(8)_xy
    const int b8_xy = 4 * h->mb_xy;
    if (USES_LIST(mb_type, 0)) {
        write_back_motion_list(h, s, b_stride, b_xy, b8_xy, mb_type, 0);
        write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 0);
    } else {
        fill_rectangle(&s->current_picture.f.ref_index[0][b8_xy],
        fill_rectangle(&h->cur_pic.f.ref_index[0][b8_xy],
                       2, 2, 2, (uint8_t)LIST_NOT_USED, 1);
    }
    if (USES_LIST(mb_type, 1))
        write_back_motion_list(h, s, b_stride, b_xy, b8_xy, mb_type, 1);
        write_back_motion_list(h, b_stride, b_xy, b8_xy, mb_type, 1);
    if (h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC) {
        if (IS_8X8(mb_type)) {
@@ -926,4 +968,6 @@ static av_always_inline int get_dct8x8_allowed(H264Context *h)
                  0x0001000100010001ULL));
 }
 void ff_h264_draw_horiz_band(H264Context *h, int y, int height);
 #endif /* AVCODEC_H264_H */
--- a/libavcodec/h264_cabac.c
+++ b/libavcodec/h264_cabac.c
@@ -1259,10 +1259,9 @@ static const int8_t cabac_context_init_PB[3][1024][2] =
 };
 void ff_h264_init_cabac_states(H264Context *h) {
    MpegEncContext * const s = &h->s;
    int i;
    const int8_t (*tab)[2];
    const int slice_qp = av_clip(s->qscale - 6*(h->sps.bit_depth_luma-8), 0, 51);
    const int slice_qp = av_clip(h->qscale - 6*(h->sps.bit_depth_luma-8), 0, 51);
    if( h->slice_type_nos == AV_PICTURE_TYPE_I ) tab = cabac_context_init_I;
    else                                 tab = cabac_context_init_PB[h->cabac_init_idc];
@@ -1280,13 +1279,12 @@ void ff_h264_init_cabac_states(H264Context *h) {
 }
 static int decode_cabac_field_decoding_flag(H264Context *h) {
    MpegEncContext * const s = &h->s;
    const long mbb_xy = h->mb_xy - 2L*s->mb_stride;
    const long mbb_xy = h->mb_xy - 2L*h->mb_stride;
    unsigned long ctx = 0;
    ctx += h->mb_field_decoding_flag & !!s->mb_x; //for FMO:(s->current_picture.f.mb_type[mba_xy] >> 7) & (h->slice_table[mba_xy] == h->slice_num);
    ctx += (s->current_picture.f.mb_type[mbb_xy] >> 7) & (h->slice_table[mbb_xy] == h->slice_num);
    ctx += h->mb_field_decoding_flag & !!h->mb_x; //for FMO:(s->current_picture.f.mb_type[mba_xy] >> 7) & (h->slice_table[mba_xy] == h->slice_num);
    ctx += (h->cur_pic.f.mb_type[mbb_xy] >> 7) & (h->slice_table[mbb_xy] == h->slice_num);
    return get_cabac_noinline( &h->cabac, &(h->cabac_state+70)[ctx] );
 }
@@ -1322,34 +1320,33 @@ static int decode_cabac_intra_mb_type(H264Context *h, int ctx_base, int intra_sl
 }
 static int decode_cabac_mb_skip( H264Context *h, int mb_x, int mb_y ) {
    MpegEncContext * const s = &h->s;
    int mba_xy, mbb_xy;
    int ctx = 0;
    if(FRAME_MBAFF){ //FIXME merge with the stuff in fill_caches?
        int mb_xy = mb_x + (mb_y&~1)*s->mb_stride;
        int mb_xy = mb_x + (mb_y&~1)*h->mb_stride;
        mba_xy = mb_xy - 1;
        if( (mb_y&1)
            && h->slice_table[mba_xy] == h->slice_num
            && MB_FIELD == !!IS_INTERLACED( s->current_picture.f.mb_type[mba_xy] ) )
            mba_xy += s->mb_stride;
            && MB_FIELD == !!IS_INTERLACED( h->cur_pic.f.mb_type[mba_xy] ) )
            mba_xy += h->mb_stride;
        if( MB_FIELD ){
            mbb_xy = mb_xy - s->mb_stride;
            mbb_xy = mb_xy - h->mb_stride;
            if( !(mb_y&1)
                && h->slice_table[mbb_xy] == h->slice_num
                && IS_INTERLACED( s->current_picture.f.mb_type[mbb_xy] ) )
                mbb_xy -= s->mb_stride;
                && IS_INTERLACED( h->cur_pic.f.mb_type[mbb_xy] ) )
                mbb_xy -= h->mb_stride;
        }else
            mbb_xy = mb_x + (mb_y-1)*s->mb_stride;
            mbb_xy = mb_x + (mb_y-1)*h->mb_stride;
    }else{
        int mb_xy = h->mb_xy;
        mba_xy = mb_xy - 1;
        mbb_xy = mb_xy - (s->mb_stride << FIELD_PICTURE);
        mbb_xy = mb_xy - (h->mb_stride << FIELD_PICTURE);
    }
    if( h->slice_table[mba_xy] == h->slice_num && !IS_SKIP( s->current_picture.f.mb_type[mba_xy] ))
    if( h->slice_table[mba_xy] == h->slice_num && !IS_SKIP(h->cur_pic.f.mb_type[mba_xy] ))
        ctx++;
    if( h->slice_table[mbb_xy] == h->slice_num && !IS_SKIP( s->current_picture.f.mb_type[mbb_xy] ))
    if( h->slice_table[mbb_xy] == h->slice_num && !IS_SKIP(h->cur_pic.f.mb_type[mbb_xy] ))
        ctx++;
    if( h->slice_type_nos == AV_PICTURE_TYPE_B )
@@ -1506,7 +1503,7 @@ static int decode_cabac_mb_mvd( H264Context *h, int ctxbase, int amvd, int *mvda
            mvd += 1 << k;
            k++;
            if(k>24){
                av_log(h->s.avctx, AV_LOG_ERROR, "overflow in decode_cabac_mb_mvd\n");
                av_log(h->avctx, AV_LOG_ERROR, "overflow in decode_cabac_mb_mvd\n");
                return INT_MIN;
            }
        }
@@ -1831,8 +1828,7 @@ static av_always_inline void decode_cabac_luma_residual( H264Context *h, const u
    static const uint8_t ctx_cat[4][3] = {{0,6,10},{1,7,11},{2,8,12},{5,9,13}};
    const uint32_t *qmul;
    int i8x8, i4x4;
    MpegEncContext * const s = &h->s;
    int qscale = p == 0 ? s->qscale : h->chroma_qp[p-1];
    int qscale = p == 0 ? h->qscale : h->chroma_qp[p-1];
    if( IS_INTRA16x16( mb_type ) ) {
        AV_ZERO128(h->mb_luma_dc[p]+0);
        AV_ZERO128(h->mb_luma_dc[p]+8);
@@ -1878,7 +1874,6 @@ static av_always_inline void decode_cabac_luma_residual( H264Context *h, const u
 * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR if an error is noticed
 */
 int ff_h264_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;
@@ -1886,21 +1881,21 @@ int ff_h264_decode_mb_cabac(H264Context *h) {
    const int pixel_shift = h->pixel_shift;
    unsigned local_ref_count[2];
    mb_xy = h->mb_xy = s->mb_x + s->mb_y*s->mb_stride;
    mb_xy = h->mb_xy = h->mb_x + h->mb_y*h->mb_stride;
    tprintf(s->avctx, "pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
    tprintf(h->avctx, "pic:%d mb:%d/%d\n", h->frame_num, h->mb_x, h->mb_y);
    if( h->slice_type_nos != AV_PICTURE_TYPE_I ) {
        int skip;
        /* a skipped mb needs the aff flag from the following mb */
        if( FRAME_MBAFF && (s->mb_y&1)==1 && h->prev_mb_skipped )
        if( FRAME_MBAFF && (h->mb_y&1)==1 && h->prev_mb_skipped )
            skip = h->next_mb_skipped;
        else
            skip = decode_cabac_mb_skip( h, s->mb_x, s->mb_y );
            skip = decode_cabac_mb_skip( h, h->mb_x, h->mb_y );
        /* read skip flags */
        if( skip ) {
            if( FRAME_MBAFF && (s->mb_y&1)==0 ){
                s->current_picture.f.mb_type[mb_xy] = MB_TYPE_SKIP;
                h->next_mb_skipped = decode_cabac_mb_skip( h, s->mb_x, s->mb_y+1 );
            if( FRAME_MBAFF && (h->mb_y&1)==0 ){
                h->cur_pic.f.mb_type[mb_xy] = MB_TYPE_SKIP;
                h->next_mb_skipped = decode_cabac_mb_skip( h, h->mb_x, h->mb_y+1 );
                if(!h->next_mb_skipped)
                    h->mb_mbaff = h->mb_field_decoding_flag = decode_cabac_field_decoding_flag(h);
            }
@@ -1916,7 +1911,7 @@ int ff_h264_decode_mb_cabac(H264Context *h) {
        }
    }
    if(FRAME_MBAFF){
        if( (s->mb_y&1) == 0 )
        if( (h->mb_y&1) == 0 )
            h->mb_mbaff =
            h->mb_field_decoding_flag = decode_cabac_field_decoding_flag(h);
    }
@@ -2017,10 +2012,10 @@ decode_intra_mb:
        h->cbp_table[mb_xy] = 0xf7ef;
        h->chroma_pred_mode_table[mb_xy] = 0;
        // In deblocking, the quantizer is 0
        s->current_picture.f.qscale_table[mb_xy] = 0;
        h->cur_pic.f.qscale_table[mb_xy] = 0;
        // All coeffs are present
        memset(h->non_zero_count[mb_xy], 16, 48);
        s->current_picture.f.mb_type[mb_xy] = mb_type;
        h->cur_pic.f.mb_type[mb_xy] = mb_type;
        h->last_qscale_diff = 0;
        return 0;
    }
@@ -2045,7 +2040,7 @@ decode_intra_mb:
                    int pred = pred_intra_mode( h, i );
                    h->intra4x4_pred_mode_cache[ scan8[i] ] = decode_cabac_mb_intra4x4_pred_mode( h, pred );
                    av_dlog(s->avctx, "i4x4 pred=%d mode=%d\n", pred,
                    av_dlog(h->avctx, "i4x4 pred=%d mode=%d\n", pred,
                            h->intra4x4_pred_mode_cache[scan8[i]]);
                }
            }
@@ -2099,7 +2094,7 @@ decode_intra_mb:
                        if (local_ref_count[list] > 1) {
                            ref[list][i] = decode_cabac_mb_ref( h, list, 4*i );
                            if (ref[list][i] >= (unsigned)local_ref_count[list]) {
                                av_log(s->avctx, AV_LOG_ERROR, "Reference %d >= %d\n", ref[list][i], local_ref_count[list]);
                                av_log(h->avctx, AV_LOG_ERROR, "Reference %d >= %d\n", ref[list][i], local_ref_count[list]);
                                return -1;
                            }
                        }else
@@ -2134,7 +2129,7 @@ decode_intra_mb:
                        uint8_t (* mvd_cache)[2]= &h->mvd_cache[list][ scan8[index] ];
                        pred_motion(h, index, block_width, list, h->ref_cache[list][ scan8[index] ], &mx, &my);
                        DECODE_CABAC_MB_MVD( h, list, index)
                        tprintf(s->avctx, "final mv:%d %d\n", mx, my);
                        tprintf(h->avctx, "final mv:%d %d\n", mx, my);
                        if(IS_SUB_8X8(sub_mb_type)){
                            mv_cache[ 1 ][0]=
@@ -2185,7 +2180,7 @@ decode_intra_mb:
                    if (local_ref_count[list] > 1) {
                        ref= decode_cabac_mb_ref(h, list, 0);
                        if (ref >= (unsigned)local_ref_count[list]) {
                            av_log(s->avctx, AV_LOG_ERROR, "Reference %d >= %d\n", ref, local_ref_count[list]);
                            av_log(h->avctx, AV_LOG_ERROR, "Reference %d >= %d\n", ref, local_ref_count[list]);
                            return -1;
                        }
                    }else
@@ -2198,7 +2193,7 @@ decode_intra_mb:
                    int mx,my,mpx,mpy;
                    pred_motion(h, 0, 4, list, h->ref_cache[list][ scan8[0] ], &mx, &my);
                    DECODE_CABAC_MB_MVD( h, list, 0)
                    tprintf(s->avctx, "final mv:%d %d\n", mx, my);
                    tprintf(h->avctx, "final mv:%d %d\n", mx, my);
                    fill_rectangle(h->mvd_cache[list][ scan8[0] ], 4, 4, 8, pack8to16(mpx,mpy), 2);
                    fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
@@ -2213,7 +2208,7 @@ decode_intra_mb:
                            if (local_ref_count[list] > 1) {
                                ref= decode_cabac_mb_ref( h, list, 8*i );
                                if (ref >= (unsigned)local_ref_count[list]) {
                                    av_log(s->avctx, AV_LOG_ERROR, "Reference %d >= %d\n", ref, local_ref_count[list]);
                                    av_log(h->avctx, AV_LOG_ERROR, "Reference %d >= %d\n", ref, local_ref_count[list]);
                                    return -1;
                                }
                            }else
@@ -2229,7 +2224,7 @@ decode_intra_mb:
                        int mx,my,mpx,mpy;
                        pred_16x8_motion(h, 8*i, list, h->ref_cache[list][scan8[0] + 16*i], &mx, &my);
                        DECODE_CABAC_MB_MVD( h, list, 8*i)
                        tprintf(s->avctx, "final mv:%d %d\n", mx, my);
                        tprintf(h->avctx, "final mv:%d %d\n", mx, my);
                        fill_rectangle(h->mvd_cache[list][ scan8[0] + 16*i ], 4, 2, 8, pack8to16(mpx,mpy), 2);
                        fill_rectangle(h->mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, pack16to32(mx,my), 4);
@@ -2248,7 +2243,7 @@ decode_intra_mb:
                            if (local_ref_count[list] > 1) {
                                ref= decode_cabac_mb_ref( h, list, 4*i );
                                if (ref >= (unsigned)local_ref_count[list]) {
                                    av_log(s->avctx, AV_LOG_ERROR, "Reference %d >= %d\n", ref, local_ref_count[list]);
                                    av_log(h->avctx, AV_LOG_ERROR, "Reference %d >= %d\n", ref, local_ref_count[list]);
                                    return -1;
                                }
                            }else
@@ -2265,7 +2260,7 @@ decode_intra_mb:
                        pred_8x16_motion(h, i*4, list, h->ref_cache[list][ scan8[0] + 2*i ], &mx, &my);
                        DECODE_CABAC_MB_MVD( h, list, 4*i)
                        tprintf(s->avctx, "final mv:%d %d\n", mx, my);
                        tprintf(h->avctx, "final mv:%d %d\n", mx, my);
                        fill_rectangle(h->mvd_cache[list][ scan8[0] + 2*i ], 2, 4, 8, pack8to16(mpx,mpy), 2);
                        fill_rectangle(h->mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, pack16to32(mx,my), 4);
                    }else{
@@ -2288,7 +2283,7 @@ decode_intra_mb:
            cbp |= decode_cabac_mb_cbp_chroma( h ) << 4;
    } else {
        if (!decode_chroma && cbp>15) {
            av_log(s->avctx, AV_LOG_ERROR, "gray chroma\n");
            av_log(h->avctx, AV_LOG_ERROR, "gray chroma\n");
            return AVERROR_INVALIDDATA;
        }
    }
@@ -2321,18 +2316,18 @@ decode_intra_mb:
            AV_WN32A(&nnz_cache[4+8*10], top_empty);
        }
    }
    s->current_picture.f.mb_type[mb_xy] = mb_type;
    h->cur_pic.f.mb_type[mb_xy] = mb_type;
    if( cbp || IS_INTRA16x16( mb_type ) ) {
        const uint8_t *scan, *scan8x8;
        const uint32_t *qmul;
        if(IS_INTERLACED(mb_type)){
            scan8x8= s->qscale ? h->field_scan8x8 : h->field_scan8x8_q0;
            scan= s->qscale ? h->field_scan : h->field_scan_q0;
            scan8x8= h->qscale ? h->field_scan8x8 : h->field_scan8x8_q0;
            scan= h->qscale ? h->field_scan : h->field_scan_q0;
        }else{
            scan8x8= s->qscale ? h->zigzag_scan8x8 : h->zigzag_scan8x8_q0;
            scan= s->qscale ? h->zigzag_scan : h->zigzag_scan_q0;
            scan8x8= h->qscale ? h->zigzag_scan8x8 : h->zigzag_scan8x8_q0;
            scan= h->qscale ? h->zigzag_scan : h->zigzag_scan_q0;
        }
        // decode_cabac_mb_dqp
@@ -2345,7 +2340,7 @@ decode_intra_mb:
                ctx= 3;
                val++;
                if(val > 2*max_qp){ //prevent infinite loop
                    av_log(h->s.avctx, AV_LOG_ERROR, "cabac decode of qscale diff failed at %d %d\n", s->mb_x, s->mb_y);
                    av_log(h->avctx, AV_LOG_ERROR, "cabac decode of qscale diff failed at %d %d\n", h->mb_x, h->mb_y);
                    return -1;
                }
            }
@@ -2355,13 +2350,13 @@ decode_intra_mb:
            else
                val= -((val + 1)>>1);
            h->last_qscale_diff = val;
            s->qscale += val;
            if(((unsigned)s->qscale) > max_qp){
                if(s->qscale<0) s->qscale+= max_qp+1;
                else            s->qscale-= max_qp+1;
            h->qscale += val;
            if(((unsigned)h->qscale) > max_qp){
                if(h->qscale<0) h->qscale+= max_qp+1;
                else            h->qscale-= max_qp+1;
            }
            h->chroma_qp[0] = get_chroma_qp(h, 0, s->qscale);
            h->chroma_qp[1] = get_chroma_qp(h, 1, s->qscale);
            h->chroma_qp[0] = get_chroma_qp(h, 0, h->qscale);
            h->chroma_qp[1] = get_chroma_qp(h, 1, h->qscale);
        }else
            h->last_qscale_diff=0;
@@ -2423,7 +2418,7 @@ decode_intra_mb:
        h->last_qscale_diff = 0;
    }
    s->current_picture.f.qscale_table[mb_xy] = s->qscale;
    h->cur_pic.f.qscale_table[mb_xy] = h->qscale;
    write_back_non_zero_count(h);
    return 0;
--- a/libavcodec/h264_cavlc.c
+++ b/libavcodec/h264_cavlc.c
@@ -292,7 +292,7 @@ static inline int pred_non_zero_count(H264Context *h, int n){
    if(i<64) i= (i+1)>>1;
    tprintf(h->s.avctx, "pred_nnz L%X T%X n%d s%d P%X\n", left, top, n, scan8[n], i&31);
    tprintf(h->avctx, "pred_nnz L%X T%X n%d s%d P%X\n", left, top, n, scan8[n], i&31);
    return i&31;
 }
@@ -443,7 +443,6 @@ static inline int get_level_prefix(GetBitContext *gb){
 * @return <0 if an error occurred
 */
 static int decode_residual(H264Context *h, GetBitContext *gb, int16_t *block, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff){
    MpegEncContext * const s = &h->s;
    static const int coeff_token_table_index[17]= {0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3};
    int level[16];
    int zeros_left, coeff_token, total_coeff, i, trailing_ones, run_before;
@@ -474,12 +473,12 @@ static int decode_residual(H264Context *h, GetBitContext *gb, int16_t *block, in
    if(total_coeff==0)
        return 0;
    if(total_coeff > (unsigned)max_coeff) {
        av_log(h->s.avctx, AV_LOG_ERROR, "corrupted macroblock %d %d (total_coeff=%d)\n", s->mb_x, s->mb_y, total_coeff);
        av_log(h->avctx, AV_LOG_ERROR, "corrupted macroblock %d %d (total_coeff=%d)\n", h->mb_x, h->mb_y, total_coeff);
        return -1;
    }
    trailing_ones= coeff_token&3;
    tprintf(h->s.avctx, "trailing:%d, total:%d\n", trailing_ones, total_coeff);
    tprintf(h->avctx, "trailing:%d, total:%d\n", trailing_ones, total_coeff);
    av_assert2(total_coeff<=16);
    i = show_bits(gb, 3);
@@ -515,7 +514,7 @@ static int decode_residual(H264Context *h, GetBitContext *gb, int16_t *block, in
                level_code= 30;
                if(prefix>=16){
                    if(prefix > 25+3){
                        av_log(h->s.avctx, AV_LOG_ERROR, "Invalid level prefix\n");
                        av_log(h->avctx, AV_LOG_ERROR, "Invalid level prefix\n");
                        return -1;
                    }
                    level_code += (1<<(prefix-3))-4096;
@@ -618,7 +617,7 @@ static int decode_residual(H264Context *h, GetBitContext *gb, int16_t *block, in
    }
    if(zeros_left<0){
        av_log(h->s.avctx, AV_LOG_ERROR, "negative number of zero coeffs at %d %d\n", s->mb_x, s->mb_y);
        av_log(h->avctx, AV_LOG_ERROR, "negative number of zero coeffs at %d %d\n", h->mb_x, h->mb_y);
        return -1;
    }
@@ -627,8 +626,7 @@ static int decode_residual(H264Context *h, GetBitContext *gb, int16_t *block, in
 static av_always_inline int decode_luma_residual(H264Context *h, GetBitContext *gb, const uint8_t *scan, const uint8_t *scan8x8, int pixel_shift, int mb_type, int cbp, int p){
    int i4x4, i8x8;
    MpegEncContext * const s = &h->s;
    int qscale = p == 0 ? s->qscale : h->chroma_qp[p-1];
    int qscale = p == 0 ? h->qscale : h->chroma_qp[p-1];
    if(IS_INTRA16x16(mb_type)){
        AV_ZERO128(h->mb_luma_dc[p]+0);
        AV_ZERO128(h->mb_luma_dc[p]+8);
@@ -693,7 +691,6 @@ static av_always_inline int decode_luma_residual(H264Context *h, GetBitContext *
 }
 int ff_h264_decode_mb_cavlc(H264Context *h){
    MpegEncContext * const s = &h->s;
    int mb_xy;
    int partition_count;
    unsigned int mb_type, cbp;
@@ -702,32 +699,32 @@ int ff_h264_decode_mb_cavlc(H264Context *h){
    const int pixel_shift = h->pixel_shift;
    unsigned local_ref_count[2];
    mb_xy = h->mb_xy = s->mb_x + s->mb_y*s->mb_stride;
    mb_xy = h->mb_xy = h->mb_x + h->mb_y*h->mb_stride;
    tprintf(s->avctx, "pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
    tprintf(h->avctx, "pic:%d mb:%d/%d\n", h->frame_num, h->mb_x, h->mb_y);
    cbp = 0; /* avoid warning. FIXME: find a solution without slowing
                down the code */
    if(h->slice_type_nos != AV_PICTURE_TYPE_I){
        if(s->mb_skip_run==-1)
            s->mb_skip_run= get_ue_golomb(&s->gb);
        if(h->mb_skip_run==-1)
            h->mb_skip_run= get_ue_golomb(&h->gb);
        if (s->mb_skip_run--) {
            if(FRAME_MBAFF && (s->mb_y&1) == 0){
                if(s->mb_skip_run==0)
                    h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&s->gb);
        if (h->mb_skip_run--) {
            if(FRAME_MBAFF && (h->mb_y&1) == 0){
                if(h->mb_skip_run==0)
                    h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&h->gb);
            }
            decode_mb_skip(h);
            return 0;
        }
    }
    if(FRAME_MBAFF){
        if( (s->mb_y&1) == 0 )
            h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&s->gb);
        if( (h->mb_y&1) == 0 )
            h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&h->gb);
    }
    h->prev_mb_skipped= 0;
    mb_type= get_ue_golomb(&s->gb);
    mb_type= get_ue_golomb(&h->gb);
    if(h->slice_type_nos == AV_PICTURE_TYPE_B){
        if(mb_type < 23){
            partition_count= b_mb_type_info[mb_type].partition_count;
@@ -750,7 +747,7 @@ int ff_h264_decode_mb_cavlc(H264Context *h){
            mb_type--;
 decode_intra_mb:
        if(mb_type > 25){
            av_log(h->s.avctx, AV_LOG_ERROR, "mb_type %d in %c slice too large at %d %d\n", mb_type, av_get_picture_type_char(h->slice_type), s->mb_x, s->mb_y);
            av_log(h->avctx, AV_LOG_ERROR, "mb_type %d in %c slice too large at %d %d\n", mb_type, av_get_picture_type_char(h->slice_type), h->mb_x, h->mb_y);
            return -1;
        }
        partition_count=0;
@@ -770,19 +767,19 @@ decode_intra_mb:
                            h->sps.bit_depth_luma >> 3;
        // We assume these blocks are very rare so we do not optimize it.
        align_get_bits(&s->gb);
        align_get_bits(&h->gb);
        // The pixels are stored in the same order as levels in h->mb array.
        for(x=0; x < mb_size; x++){
            ((uint8_t*)h->mb)[x]= get_bits(&s->gb, 8);
            ((uint8_t*)h->mb)[x]= get_bits(&h->gb, 8);
        }
        // In deblocking, the quantizer is 0
        s->current_picture.f.qscale_table[mb_xy] = 0;
        h->cur_pic.f.qscale_table[mb_xy] = 0;
        // All coeffs are present
        memset(h->non_zero_count[mb_xy], 16, 48);
        s->current_picture.f.mb_type[mb_xy] = mb_type;
        h->cur_pic.f.mb_type[mb_xy] = mb_type;
        return 0;
    }
@@ -799,7 +796,7 @@ decode_intra_mb:
        if(IS_INTRA4x4(mb_type)){
            int i;
            int di = 1;
            if(dct8x8_allowed && get_bits1(&s->gb)){
            if(dct8x8_allowed && get_bits1(&h->gb)){
                mb_type |= MB_TYPE_8x8DCT;
                di = 4;
            }
@@ -808,8 +805,8 @@ decode_intra_mb:
            for(i=0; i<16; i+=di){
                int mode= pred_intra_mode(h, i);
                if(!get_bits1(&s->gb)){
                    const int rem_mode= get_bits(&s->gb, 3);
                if(!get_bits1(&h->gb)){
                    const int rem_mode= get_bits(&h->gb, 3);
                    mode = rem_mode + (rem_mode >= mode);
                }
@@ -827,7 +824,7 @@ decode_intra_mb:
                return -1;
        }
        if(decode_chroma){
            pred_mode= ff_h264_check_intra_pred_mode(h, get_ue_golomb_31(&s->gb), 1);
            pred_mode= ff_h264_check_intra_pred_mode(h, get_ue_golomb_31(&h->gb), 1);
            if(pred_mode < 0)
                return -1;
            h->chroma_pred_mode= pred_mode;
@@ -839,9 +836,9 @@ decode_intra_mb:
        if(h->slice_type_nos == AV_PICTURE_TYPE_B){
            for(i=0; i<4; i++){
                h->sub_mb_type[i]= get_ue_golomb_31(&s->gb);
                h->sub_mb_type[i]= get_ue_golomb_31(&h->gb);
                if(h->sub_mb_type[i] >=13){
                    av_log(h->s.avctx, AV_LOG_ERROR, "B sub_mb_type %u out of range at %d %d\n", h->sub_mb_type[i], s->mb_x, s->mb_y);
                    av_log(h->avctx, AV_LOG_ERROR, "B sub_mb_type %u out of range at %d %d\n", h->sub_mb_type[i], h->mb_x, h->mb_y);
                    return -1;
                }
                sub_partition_count[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
@@ -857,9 +854,9 @@ decode_intra_mb:
        }else{
            av_assert2(h->slice_type_nos == AV_PICTURE_TYPE_P); //FIXME SP correct ?
            for(i=0; i<4; i++){
                h->sub_mb_type[i]= get_ue_golomb_31(&s->gb);
                h->sub_mb_type[i]= get_ue_golomb_31(&h->gb);
                if(h->sub_mb_type[i] >=4){
                    av_log(h->s.avctx, AV_LOG_ERROR, "P sub_mb_type %u out of range at %d %d\n", h->sub_mb_type[i], s->mb_x, s->mb_y);
                    av_log(h->avctx, AV_LOG_ERROR, "P sub_mb_type %u out of range at %d %d\n", h->sub_mb_type[i], h->mb_x, h->mb_y);
                    return -1;
                }
                sub_partition_count[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
@@ -876,11 +873,11 @@ decode_intra_mb:
                    if(ref_count == 1){
                        tmp= 0;
                    }else if(ref_count == 2){
                        tmp= get_bits1(&s->gb)^1;
                        tmp= get_bits1(&h->gb)^1;
                    }else{
                        tmp= get_ue_golomb_31(&s->gb);
                        tmp= get_ue_golomb_31(&h->gb);
                        if(tmp>=ref_count){
                            av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", tmp);
                            av_log(h->avctx, AV_LOG_ERROR, "ref %u overflow\n", tmp);
                            return -1;
                        }
                    }
@@ -912,9 +909,9 @@ decode_intra_mb:
                        const int index= 4*i + block_width*j;
                        int16_t (* mv_cache)[2]= &h->mv_cache[list][ scan8[index] ];
                        pred_motion(h, index, block_width, list, h->ref_cache[list][ scan8[index] ], &mx, &my);
                        mx += get_se_golomb(&s->gb);
                        my += get_se_golomb(&s->gb);
                        tprintf(s->avctx, "final mv:%d %d\n", mx, my);
                        mx += get_se_golomb(&h->gb);
                        my += get_se_golomb(&h->gb);
                        tprintf(h->avctx, "final mv:%d %d\n", mx, my);
                        if(IS_SUB_8X8(sub_mb_type)){
                            mv_cache[ 1 ][0]=
@@ -951,11 +948,11 @@ decode_intra_mb:
                        if(local_ref_count[list]==1){
                            val= 0;
                        }else if(local_ref_count[list]==2){
                            val= get_bits1(&s->gb)^1;
                            val= get_bits1(&h->gb)^1;
                        }else{
                            val= get_ue_golomb_31(&s->gb);
                            val= get_ue_golomb_31(&h->gb);
                            if(val >= local_ref_count[list]){
                                av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
                                av_log(h->avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
                                return -1;
                            }
                        }
@@ -965,9 +962,9 @@ decode_intra_mb:
            for(list=0; list<h->list_count; list++){
                if(IS_DIR(mb_type, 0, list)){
                    pred_motion(h, 0, 4, list, h->ref_cache[list][ scan8[0] ], &mx, &my);
                    mx += get_se_golomb(&s->gb);
                    my += get_se_golomb(&s->gb);
                    tprintf(s->avctx, "final mv:%d %d\n", mx, my);
                    mx += get_se_golomb(&h->gb);
                    my += get_se_golomb(&h->gb);
                    tprintf(h->avctx, "final mv:%d %d\n", mx, my);
                    fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
                }
@@ -981,11 +978,11 @@ decode_intra_mb:
                            if(local_ref_count[list] == 1){
                                val= 0;
                            }else if(local_ref_count[list] == 2){
                                val= get_bits1(&s->gb)^1;
                                val= get_bits1(&h->gb)^1;
                            }else{
                                val= get_ue_golomb_31(&s->gb);
                                val= get_ue_golomb_31(&h->gb);
                                if(val >= local_ref_count[list]){
                                    av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
                                    av_log(h->avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
                                    return -1;
                                }
                            }
@@ -999,9 +996,9 @@ decode_intra_mb:
                    unsigned int val;
                    if(IS_DIR(mb_type, i, list)){
                        pred_16x8_motion(h, 8*i, list, h->ref_cache[list][scan8[0] + 16*i], &mx, &my);
                        mx += get_se_golomb(&s->gb);
                        my += get_se_golomb(&s->gb);
                        tprintf(s->avctx, "final mv:%d %d\n", mx, my);
                        mx += get_se_golomb(&h->gb);
                        my += get_se_golomb(&h->gb);
                        tprintf(h->avctx, "final mv:%d %d\n", mx, my);
                        val= pack16to32(mx,my);
                    }else
@@ -1018,11 +1015,11 @@ decode_intra_mb:
                            if(local_ref_count[list]==1){
                                val= 0;
                            }else if(local_ref_count[list]==2){
                                val= get_bits1(&s->gb)^1;
                                val= get_bits1(&h->gb)^1;
                            }else{
                                val= get_ue_golomb_31(&s->gb);
                                val= get_ue_golomb_31(&h->gb);
                                if(val >= local_ref_count[list]){
                                    av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
                                    av_log(h->avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
                                    return -1;
                                }
                            }
@@ -1036,9 +1033,9 @@ decode_intra_mb:
                    unsigned int val;
                    if(IS_DIR(mb_type, i, list)){
                        pred_8x16_motion(h, i*4, list, h->ref_cache[list][ scan8[0] + 2*i ], &mx, &my);
                        mx += get_se_golomb(&s->gb);
                        my += get_se_golomb(&s->gb);
                        tprintf(s->avctx, "final mv:%d %d\n", mx, my);
                        mx += get_se_golomb(&h->gb);
                        my += get_se_golomb(&h->gb);
                        tprintf(h->avctx, "final mv:%d %d\n", mx, my);
                        val= pack16to32(mx,my);
                    }else
@@ -1053,18 +1050,18 @@ decode_intra_mb:
        write_back_motion(h, mb_type);
    if(!IS_INTRA16x16(mb_type)){
        cbp= get_ue_golomb(&s->gb);
        cbp= get_ue_golomb(&h->gb);
        if(decode_chroma){
            if(cbp > 47){
                av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y);
                av_log(h->avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, h->mb_x, h->mb_y);
                return -1;
            }
            if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp[cbp];
            else                     cbp= golomb_to_inter_cbp   [cbp];
        }else{
            if(cbp > 15){
                av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y);
                av_log(h->avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, h->mb_x, h->mb_y);
                return -1;
            }
            if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp_gray[cbp];
@@ -1072,17 +1069,17 @@ decode_intra_mb:
        }
    } else {
        if (!decode_chroma && cbp>15) {
            av_log(s->avctx, AV_LOG_ERROR, "gray chroma\n");
            av_log(h->avctx, AV_LOG_ERROR, "gray chroma\n");
            return AVERROR_INVALIDDATA;
        }
    }
    if(dct8x8_allowed && (cbp&15) && !IS_INTRA(mb_type)){
        mb_type |= MB_TYPE_8x8DCT*get_bits1(&s->gb);
        mb_type |= MB_TYPE_8x8DCT*get_bits1(&h->gb);
    }
    h->cbp=
    h->cbp_table[mb_xy]= cbp;
    s->current_picture.f.mb_type[mb_xy] = mb_type;
    h->cur_pic.f.mb_type[mb_xy] = mb_type;
    if(cbp || IS_INTRA16x16(mb_type)){
        int i4x4, i8x8, chroma_idx;
@@ -1093,28 +1090,28 @@ decode_intra_mb:
        const int max_qp = 51 + 6*(h->sps.bit_depth_luma-8);
        if(IS_INTERLACED(mb_type)){
            scan8x8= s->qscale ? h->field_scan8x8_cavlc : h->field_scan8x8_cavlc_q0;
            scan= s->qscale ? h->field_scan : h->field_scan_q0;
            scan8x8= h->qscale ? h->field_scan8x8_cavlc : h->field_scan8x8_cavlc_q0;
            scan= h->qscale ? h->field_scan : h->field_scan_q0;
        }else{
            scan8x8= s->qscale ? h->zigzag_scan8x8_cavlc : h->zigzag_scan8x8_cavlc_q0;
            scan= s->qscale ? h->zigzag_scan : h->zigzag_scan_q0;
            scan8x8= h->qscale ? h->zigzag_scan8x8_cavlc : h->zigzag_scan8x8_cavlc_q0;
            scan= h->qscale ? h->zigzag_scan : h->zigzag_scan_q0;
        }
        dquant= get_se_golomb(&s->gb);
        dquant= get_se_golomb(&h->gb);
        s->qscale += dquant;
        h->qscale += dquant;
        if(((unsigned)s->qscale) > max_qp){
            if(s->qscale<0) s->qscale+= max_qp+1;
            else            s->qscale-= max_qp+1;
            if(((unsigned)s->qscale) > max_qp){
                av_log(h->s.avctx, AV_LOG_ERROR, "dquant out of range (%d) at %d %d\n", dquant, s->mb_x, s->mb_y);
        if(((unsigned)h->qscale) > max_qp){
            if(h->qscale<0) h->qscale+= max_qp+1;
            else            h->qscale-= max_qp+1;
            if(((unsigned)h->qscale) > max_qp){
                av_log(h->avctx, AV_LOG_ERROR, "dquant out of range (%d) at %d %d\n", dquant, h->mb_x, h->mb_y);
                return -1;
            }
        }
        h->chroma_qp[0]= get_chroma_qp(h, 0, s->qscale);
        h->chroma_qp[1]= get_chroma_qp(h, 1, s->qscale);
        h->chroma_qp[0]= get_chroma_qp(h, 0, h->qscale);
        h->chroma_qp[1]= get_chroma_qp(h, 1, h->qscale);
        if( (ret = decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 0)) < 0 ){
            return -1;
@@ -1163,7 +1160,7 @@ decode_intra_mb:
        fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1);
        fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1);
    }
    s->current_picture.f.qscale_table[mb_xy] = s->qscale;
    h->cur_pic.f.qscale_table[mb_xy] = h->qscale;
    write_back_non_zero_count(h);
    return 0;
--- a/libavcodec/h264_direct.c
+++ b/libavcodec/h264_direct.c
@@ -50,14 +50,13 @@ static int get_scale_factor(H264Context * const h, int poc, int poc1, int i){
 }
 void ff_h264_direct_dist_scale_factor(H264Context * const h){
    MpegEncContext * const s = &h->s;
    const int poc = h->s.current_picture_ptr->field_poc[ s->picture_structure == PICT_BOTTOM_FIELD ];
    const int poc = h->cur_pic_ptr->field_poc[h->picture_structure == PICT_BOTTOM_FIELD];
    const int poc1 = h->ref_list[1][0].poc;
    int i, field;
    if (FRAME_MBAFF)
        for (field = 0; field < 2; field++){
            const int poc  = h->s.current_picture_ptr->field_poc[field];
            const int poc  = h->cur_pic_ptr->field_poc[field];
            const int poc1 = h->ref_list[1][0].field_poc[field];
            for (i = 0; i < 2 * h->ref_count[0]; i++)
                h->dist_scale_factor_field[field][i^field] =
@@ -70,12 +69,11 @@ void ff_h264_direct_dist_scale_factor(H264Context * const h){
 }
 static void fill_colmap(H264Context *h, int map[2][16+32], int list, int field, int colfield, int mbafi){
    MpegEncContext * const s = &h->s;
    Picture * const ref1 = &h->ref_list[1][0];
    int j, old_ref, rfield;
    int start= mbafi ? 16                      : 0;
    int end  = mbafi ? 16+2*h->ref_count[0]    : h->ref_count[0];
    int interl= mbafi || s->picture_structure != PICT_FRAME;
    int interl= mbafi || h->picture_structure != PICT_FRAME;
    /* bogus; fills in for missing frames */
    memset(map[list], 0, sizeof(map[list]));
@@ -104,11 +102,10 @@ static void fill_colmap(H264Context *h, int map[2][16+32], int list, int field,
 }
 void ff_h264_direct_ref_list_init(H264Context * const h){
    MpegEncContext * const s = &h->s;
    Picture * const ref1 = &h->ref_list[1][0];
    Picture * const cur = s->current_picture_ptr;
    Picture * const cur = h->cur_pic_ptr;
    int list, j, field;
    int sidx= (s->picture_structure&1)^1;
    int sidx= (h->picture_structure&1)^1;
    int ref1sidx = (ref1->f.reference&1)^1;
    for(list=0; list<2; list++){
@@ -117,7 +114,7 @@ void ff_h264_direct_ref_list_init(H264Context * const h){
            cur->ref_poc[sidx][list][j] = 4 * h->ref_list[list][j].frame_num + (h->ref_list[list][j].f.reference & 3);
    }
    if(s->picture_structure == PICT_FRAME){
    if(h->picture_structure == PICT_FRAME){
        memcpy(cur->ref_count[1], cur->ref_count[0], sizeof(cur->ref_count[0]));
        memcpy(cur->ref_poc  [1], cur->ref_poc  [0], sizeof(cur->ref_poc  [0]));
    }
@@ -125,12 +122,12 @@ void ff_h264_direct_ref_list_init(H264Context * const h){
    cur->mbaff= FRAME_MBAFF;
    h->col_fieldoff= 0;
    if(s->picture_structure == PICT_FRAME){
        int cur_poc = s->current_picture_ptr->poc;
    if(h->picture_structure == PICT_FRAME){
        int cur_poc = h->cur_pic_ptr->poc;
        int *col_poc = h->ref_list[1]->field_poc;
        h->col_parity= (FFABS(col_poc[0] - cur_poc) >= FFABS(col_poc[1] - cur_poc));
        ref1sidx=sidx= h->col_parity;
    } else if (!(s->picture_structure & h->ref_list[1][0].f.reference) && !h->ref_list[1][0].mbaff) { // FL -> FL & differ parity
    } else if (!(h->picture_structure & h->ref_list[1][0].f.reference) && !h->ref_list[1][0].mbaff) { // FL -> FL & differ parity
        h->col_fieldoff = 2 * h->ref_list[1][0].f.reference - 3;
    }
@@ -149,9 +146,9 @@ static void await_reference_mb_row(H264Context * const h, Picture *ref, int mb_y
 {
    int ref_field = ref->f.reference - 1;
    int ref_field_picture = ref->field_picture;
    int ref_height = 16*h->s.mb_height >> ref_field_picture;
    int ref_height = 16*h->mb_height >> ref_field_picture;
    if(!HAVE_THREADS || !(h->s.avctx->active_thread_type&FF_THREAD_FRAME))
    if(!HAVE_THREADS || !(h->avctx->active_thread_type&FF_THREAD_FRAME))
        return;
    //FIXME it can be safe to access mb stuff
@@ -163,10 +160,9 @@ static void await_reference_mb_row(H264Context * const h, Picture *ref, int mb_y
 }
 static void pred_spatial_direct_motion(H264Context * const h, int *mb_type){
    MpegEncContext * const s = &h->s;
    int b8_stride = 2;
    int b4_stride = h->b_stride;
    int mb_xy = h->mb_xy, mb_y = s->mb_y;
    int mb_xy = h->mb_xy, mb_y = h->mb_y;
    int mb_type_col[2];
    const int16_t (*l1mv0)[2], (*l1mv1)[2];
    const int8_t *l1ref0, *l1ref1;
@@ -179,7 +175,7 @@ static void pred_spatial_direct_motion(H264Context * const h, int *mb_type){
    assert(h->ref_list[1][0].f.reference & 3);
    await_reference_mb_row(h, &h->ref_list[1][0], s->mb_y + !!IS_INTERLACED(*mb_type));
    await_reference_mb_row(h, &h->ref_list[1][0], h->mb_y + !!IS_INTERLACED(*mb_type));
 #define MB_TYPE_16x16_OR_INTRA (MB_TYPE_16x16|MB_TYPE_INTRA4x4|MB_TYPE_INTRA16x16|MB_TYPE_INTRA_PCM)
@@ -241,21 +237,21 @@ static void pred_spatial_direct_motion(H264Context * const h, int *mb_type){
    if (IS_INTERLACED(h->ref_list[1][0].f.mb_type[mb_xy])) { // AFL/AFR/FR/FL -> AFL/FL
        if (!IS_INTERLACED(*mb_type)) {                          //     AFR/FR    -> AFL/FL
            mb_y = (s->mb_y&~1) + h->col_parity;
            mb_xy= s->mb_x + ((s->mb_y&~1) + h->col_parity)*s->mb_stride;
            mb_y = (h->mb_y&~1) + h->col_parity;
            mb_xy= h->mb_x + ((h->mb_y&~1) + h->col_parity)*h->mb_stride;
            b8_stride = 0;
        }else{
            mb_y  += h->col_fieldoff;
            mb_xy += s->mb_stride*h->col_fieldoff; // non zero for FL -> FL & differ parity
            mb_xy += h->mb_stride*h->col_fieldoff; // non zero for FL -> FL & differ parity
        }
        goto single_col;
    }else{                                               // AFL/AFR/FR/FL -> AFR/FR
        if(IS_INTERLACED(*mb_type)){                     // AFL       /FL -> AFR/FR
            mb_y = s->mb_y&~1;
            mb_xy= s->mb_x + (s->mb_y&~1)*s->mb_stride;
            mb_y = h->mb_y&~1;
            mb_xy= h->mb_x + (h->mb_y&~1)*h->mb_stride;
            mb_type_col[0] = h->ref_list[1][0].f.mb_type[mb_xy];
            mb_type_col[1] = h->ref_list[1][0].f.mb_type[mb_xy + s->mb_stride];
            b8_stride = 2+4*s->mb_stride;
            mb_type_col[1] = h->ref_list[1][0].f.mb_type[mb_xy + h->mb_stride];
            b8_stride = 2+4*h->mb_stride;
            b4_stride *= 6;
            if (IS_INTERLACED(mb_type_col[0]) != IS_INTERLACED(mb_type_col[1])) {
                mb_type_col[0] &= ~MB_TYPE_INTERLACED;
@@ -298,7 +294,7 @@ single_col:
    l1ref0 = &h->ref_list[1][0].f.ref_index [0][4 * mb_xy];
    l1ref1 = &h->ref_list[1][0].f.ref_index [1][4 * mb_xy];
    if(!b8_stride){
        if(s->mb_y&1){
        if(h->mb_y&1){
            l1ref0 += 2;
            l1ref1 += 2;
            l1mv0  +=  2*b4_stride;
@@ -414,10 +410,9 @@ single_col:
 }
 static void pred_temp_direct_motion(H264Context * const h, int *mb_type){
    MpegEncContext * const s = &h->s;
    int b8_stride = 2;
    int b4_stride = h->b_stride;
    int mb_xy = h->mb_xy, mb_y = s->mb_y;
    int mb_xy = h->mb_xy, mb_y = h->mb_y;
    int mb_type_col[2];
    const int16_t (*l1mv0)[2], (*l1mv1)[2];
    const int8_t *l1ref0, *l1ref1;
@@ -427,25 +422,25 @@ static void pred_temp_direct_motion(H264Context * const h, int *mb_type){
    assert(h->ref_list[1][0].f.reference & 3);
    await_reference_mb_row(h, &h->ref_list[1][0], s->mb_y + !!IS_INTERLACED(*mb_type));
    await_reference_mb_row(h, &h->ref_list[1][0], h->mb_y + !!IS_INTERLACED(*mb_type));
    if (IS_INTERLACED(h->ref_list[1][0].f.mb_type[mb_xy])) { // AFL/AFR/FR/FL -> AFL/FL
        if (!IS_INTERLACED(*mb_type)) {                          //     AFR/FR    -> AFL/FL
            mb_y = (s->mb_y&~1) + h->col_parity;
            mb_xy= s->mb_x + ((s->mb_y&~1) + h->col_parity)*s->mb_stride;
            mb_y = (h->mb_y&~1) + h->col_parity;
            mb_xy= h->mb_x + ((h->mb_y&~1) + h->col_parity)*h->mb_stride;
            b8_stride = 0;
        }else{
            mb_y  += h->col_fieldoff;
            mb_xy += s->mb_stride*h->col_fieldoff; // non zero for FL -> FL & differ parity
            mb_xy += h->mb_stride*h->col_fieldoff; // non zero for FL -> FL & differ parity
        }
        goto single_col;
    }else{                                               // AFL/AFR/FR/FL -> AFR/FR
        if(IS_INTERLACED(*mb_type)){                     // AFL       /FL -> AFR/FR
            mb_y = s->mb_y&~1;
            mb_xy= s->mb_x + (s->mb_y&~1)*s->mb_stride;
            mb_y = h->mb_y&~1;
            mb_xy= h->mb_x + (h->mb_y&~1)*h->mb_stride;
            mb_type_col[0] = h->ref_list[1][0].f.mb_type[mb_xy];
            mb_type_col[1] = h->ref_list[1][0].f.mb_type[mb_xy + s->mb_stride];
            b8_stride = 2+4*s->mb_stride;
            mb_type_col[1] = h->ref_list[1][0].f.mb_type[mb_xy + h->mb_stride];
            b8_stride = 2+4*h->mb_stride;
            b4_stride *= 6;
            if (IS_INTERLACED(mb_type_col[0]) != IS_INTERLACED(mb_type_col[1])) {
                mb_type_col[0] &= ~MB_TYPE_INTERLACED;
@@ -489,7 +484,7 @@ single_col:
    l1ref0 = &h->ref_list[1][0].f.ref_index [0][4 * mb_xy];
    l1ref1 = &h->ref_list[1][0].f.ref_index [1][4 * mb_xy];
    if(!b8_stride){
        if(s->mb_y&1){
        if(h->mb_y&1){
            l1ref0 += 2;
            l1ref1 += 2;
            l1mv0  +=  2*b4_stride;
@@ -503,9 +498,9 @@ single_col:
        int ref_offset;
        if(FRAME_MBAFF && IS_INTERLACED(*mb_type)){
            map_col_to_list0[0] = h->map_col_to_list0_field[s->mb_y&1][0];
            map_col_to_list0[1] = h->map_col_to_list0_field[s->mb_y&1][1];
            dist_scale_factor   =h->dist_scale_factor_field[s->mb_y&1];
            map_col_to_list0[0] = h->map_col_to_list0_field[h->mb_y&1][0];
            map_col_to_list0[1] = h->map_col_to_list0_field[h->mb_y&1][1];
            dist_scale_factor   =h->dist_scale_factor_field[h->mb_y&1];
        }
        ref_offset = (h->ref_list[1][0].mbaff<<4) & (mb_type_col[0]>>3); //if(h->ref_list[1][0].mbaff && IS_INTERLACED(mb_type_col[0])) ref_offset=16 else 0
--- a/libavcodec/h264_loopfilter.c
+++ b/libavcodec/h264_loopfilter.c
@@ -241,8 +241,7 @@ static av_always_inline void h264_filter_mb_fast_internal(H264Context *h,
                                                          unsigned int uvlinesize,
                                                          int pixel_shift)
 {
    MpegEncContext * const s = &h->s;
    int chroma = CHROMA && !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));
    int chroma = CHROMA && !(CONFIG_GRAY && (h->flags&CODEC_FLAG_GRAY));
    int chroma444 = CHROMA444;
    int chroma422 = CHROMA422;
@@ -254,10 +253,10 @@ static av_always_inline void h264_filter_mb_fast_internal(H264Context *h,
    int a = h->slice_alpha_c0_offset - qp_bd_offset;
    int b = h->slice_beta_offset - qp_bd_offset;
    int mb_type = s->current_picture.f.mb_type[mb_xy];
    int qp      = s->current_picture.f.qscale_table[mb_xy];
    int qp0     = s->current_picture.f.qscale_table[mb_xy - 1];
    int qp1     = s->current_picture.f.qscale_table[h->top_mb_xy];
    int mb_type = h->cur_pic.f.mb_type[mb_xy];
    int qp      = h->cur_pic.f.qscale_table[mb_xy];
    int qp0     = h->cur_pic.f.qscale_table[mb_xy - 1];
    int qp1     = h->cur_pic.f.qscale_table[h->top_mb_xy];
    int qpc = get_chroma_qp( h, 0, qp );
    int qpc0 = get_chroma_qp( h, 0, qp0 );
    int qpc1 = get_chroma_qp( h, 0, qp1 );
@@ -462,7 +461,6 @@ static int check_mv(H264Context *h, long b_idx, long bn_idx, int mvy_limit){
 }
 static av_always_inline void filter_mb_dir(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, int mb_xy, int mb_type, int mvy_limit, int first_vertical_edge_done, int a, int b, int chroma, int dir) {
    MpegEncContext * const s = &h->s;
    int edge;
    int chroma_qp_avg[2];
    int chroma444 = CHROMA444;
@@ -490,16 +488,16 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
            //
            unsigned int tmp_linesize   = 2 *   linesize;
            unsigned int tmp_uvlinesize = 2 * uvlinesize;
            int mbn_xy = mb_xy - 2 * s->mb_stride;
            int mbn_xy = mb_xy - 2 * h->mb_stride;
            int j;
            for(j=0; j<2; j++, mbn_xy += s->mb_stride){
            for(j=0; j<2; j++, mbn_xy += h->mb_stride){
                DECLARE_ALIGNED(8, int16_t, bS)[4];
                int qp;
                if (IS_INTRA(mb_type | s->current_picture.f.mb_type[mbn_xy])) {
                if (IS_INTRA(mb_type | h->cur_pic.f.mb_type[mbn_xy])) {
                    AV_WN64A(bS, 0x0003000300030003ULL);
                } else {
                    if (!CABAC && IS_8x8DCT(s->current_picture.f.mb_type[mbn_xy])) {
                    if (!CABAC && IS_8x8DCT(h->cur_pic.f.mb_type[mbn_xy])) {
                        bS[0]= 1+((h->cbp_table[mbn_xy] & 0x4000)||h->non_zero_count_cache[scan8[0]+0]);
                        bS[1]= 1+((h->cbp_table[mbn_xy] & 0x4000)||h->non_zero_count_cache[scan8[0]+1]);
                        bS[2]= 1+((h->cbp_table[mbn_xy] & 0x8000)||h->non_zero_count_cache[scan8[0]+2]);
@@ -514,12 +512,12 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
                }
                // Do not use s->qscale as luma quantizer because it has not the same
                // value in IPCM macroblocks.
                qp = (s->current_picture.f.qscale_table[mb_xy] + s->current_picture.f.qscale_table[mbn_xy] + 1) >> 1;
                tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize);
                { int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
                qp = (h->cur_pic.f.qscale_table[mb_xy] + h->cur_pic.f.qscale_table[mbn_xy] + 1) >> 1;
                tprintf(h->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize);
                { int i; for (i = 0; i < 4; i++) tprintf(h->avctx, " bS[%d]:%d", i, bS[i]); tprintf(h->avctx, "\n"); }
                filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, a, b, h, 0 );
                chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1;
                chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1;
                chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, h->cur_pic.f.qscale_table[mbn_xy]) + 1) >> 1;
                chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, h->cur_pic.f.qscale_table[mbn_xy]) + 1) >> 1;
                if (chroma) {
                    if (chroma444) {
                        filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);
@@ -537,7 +535,7 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
            if( IS_INTRA(mb_type|mbm_type)) {
                AV_WN64A(bS, 0x0003000300030003ULL);
                if (   (!IS_INTERLACED(mb_type|mbm_type))
                    || ((FRAME_MBAFF || (s->picture_structure != PICT_FRAME)) && (dir == 0))
                    || ((FRAME_MBAFF || (h->picture_structure != PICT_FRAME)) && (dir == 0))
                )
                    AV_WN64A(bS, 0x0004000400040004ULL);
            } else {
@@ -579,12 +577,12 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
            // Do not use s->qscale as luma quantizer because it has not the same
            // value in IPCM macroblocks.
            if(bS[0]+bS[1]+bS[2]+bS[3]){
                qp = (s->current_picture.f.qscale_table[mb_xy] + s->current_picture.f.qscale_table[mbm_xy] + 1) >> 1;
                //tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp[0], s->current_picture.qscale_table[mbn_xy]);
                tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
                //{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
                chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1;
                chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1;
                qp = (h->cur_pic.f.qscale_table[mb_xy] + h->cur_pic.f.qscale_table[mbm_xy] + 1) >> 1;
                //tprintf(h->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp[0], h->cur_pic.qscale_table[mbn_xy]);
                tprintf(h->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
                //{ int i; for (i = 0; i < 4; i++) tprintf(h->avctx, " bS[%d]:%d", i, bS[i]); tprintf(h->avctx, "\n"); }
                chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, h->cur_pic.f.qscale_table[mbm_xy]) + 1) >> 1;
                chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, h->cur_pic.f.qscale_table[mbm_xy]) + 1) >> 1;
                if( dir == 0 ) {
                    filter_mb_edgev( &img_y[0], linesize, bS, qp, a, b, h, 1 );
                    if (chroma) {
@@ -664,10 +662,10 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
        /* Filter edge */
        // Do not use s->qscale as luma quantizer because it has not the same
        // value in IPCM macroblocks.
        qp = s->current_picture.f.qscale_table[mb_xy];
        //tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp[0], s->current_picture.qscale_table[mbn_xy]);
        tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
        //{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
        qp = h->cur_pic.f.qscale_table[mb_xy];
        //tprintf(h->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp[0], h->cur_pic.qscale_table[mbn_xy]);
        tprintf(h->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
        //{ int i; for (i = 0; i < 4; i++) tprintf(h->avctx, " bS[%d]:%d", i, bS[i]); tprintf(h->avctx, "\n"); }
        if( dir == 0 ) {
            filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, a, b, h, 0 );
            if (chroma) {
@@ -704,13 +702,12 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
 }
 void ff_h264_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) {
    MpegEncContext * const s = &h->s;
    const int mb_xy= mb_x + mb_y*s->mb_stride;
    const int mb_type = s->current_picture.f.mb_type[mb_xy];
    const int mb_xy= mb_x + mb_y*h->mb_stride;
    const int mb_type = h->cur_pic.f.mb_type[mb_xy];
    const int mvy_limit = IS_INTERLACED(mb_type) ? 2 : 4;
    int first_vertical_edge_done = 0;
    av_unused int dir;
    int chroma = CHROMA && !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));
    int chroma = CHROMA && !(CONFIG_GRAY && (h->flags&CODEC_FLAG_GRAY));
    int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
    int a = h->slice_alpha_c0_offset - qp_bd_offset;
    int b = h->slice_beta_offset - qp_bd_offset;
@@ -762,9 +759,9 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
            }
        }
        mb_qp   = s->current_picture.f.qscale_table[mb_xy];
        mbn0_qp = s->current_picture.f.qscale_table[h->left_mb_xy[0]];
        mbn1_qp = s->current_picture.f.qscale_table[h->left_mb_xy[1]];
        mb_qp   = h->cur_pic.f.qscale_table[mb_xy];
        mbn0_qp = h->cur_pic.f.qscale_table[h->left_mb_xy[0]];
        mbn1_qp = h->cur_pic.f.qscale_table[h->left_mb_xy[1]];
        qp[0] = ( mb_qp + mbn0_qp + 1 ) >> 1;
        bqp[0] = ( get_chroma_qp( h, 0, mb_qp ) +
                   get_chroma_qp( h, 0, mbn0_qp ) + 1 ) >> 1;
@@ -777,8 +774,8 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
                   get_chroma_qp( h, 1, mbn1_qp ) + 1 ) >> 1;
        /* Filter edge */
        tprintf(s->avctx, "filter mb:%d/%d MBAFF, QPy:%d/%d, QPb:%d/%d QPr:%d/%d ls:%d uvls:%d", mb_x, mb_y, qp[0], qp[1], bqp[0], bqp[1], rqp[0], rqp[1], linesize, uvlinesize);
        { int i; for (i = 0; i < 8; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
        tprintf(h->avctx, "filter mb:%d/%d MBAFF, QPy:%d/%d, QPb:%d/%d QPr:%d/%d ls:%d uvls:%d", mb_x, mb_y, qp[0], qp[1], bqp[0], bqp[1], rqp[0], rqp[1], linesize, uvlinesize);
        { int i; for (i = 0; i < 8; i++) tprintf(h->avctx, " bS[%d]:%d", i, bS[i]); tprintf(h->avctx, "\n"); }
        if(MB_FIELD){
            filter_mb_mbaff_edgev ( h, img_y                ,   linesize, bS  , 1, qp [0], a, b, 1 );
            filter_mb_mbaff_edgev ( h, img_y  + 8*  linesize,   linesize, bS+4, 1, qp [1], a, b, 1 );
--- a/libavcodec/h264_mb_template.c
+++ b/libavcodec/h264_mb_template.c
@@ -40,39 +40,38 @@
 static av_noinline void FUNC(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_x    = h->mb_x;
    const int mb_y    = h->mb_y;
    const int mb_xy   = h->mb_xy;
    const int mb_type = s->current_picture.f.mb_type[mb_xy];
    const int mb_type = h->cur_pic.f.mb_type[mb_xy];
    uint8_t *dest_y, *dest_cb, *dest_cr;
    int linesize, uvlinesize /*dct_offset*/;
    int i, j;
    int *block_offset = &h->block_offset[0];
    const int transform_bypass = !SIMPLE && (s->qscale == 0 && h->sps.transform_bypass);
    const int transform_bypass = !SIMPLE && (h->qscale == 0 && h->sps.transform_bypass);
    /* is_h264 should always be true if SVQ3 is disabled. */
    const int is_h264 = !CONFIG_SVQ3_DECODER || SIMPLE || s->codec_id == AV_CODEC_ID_H264;
    const int is_h264 = !CONFIG_SVQ3_DECODER || SIMPLE || h->avctx->codec_id == AV_CODEC_ID_H264;
    void (*idct_add)(uint8_t *dst, int16_t *block, int stride);
    const int block_h   = 16 >> s->chroma_y_shift;
    const int block_h   = 16 >> h->chroma_y_shift;
    const int chroma422 = CHROMA422;
    dest_y  = s->current_picture.f.data[0] + ((mb_x << PIXEL_SHIFT)     + mb_y * s->linesize)  * 16;
    dest_cb = s->current_picture.f.data[1] +  (mb_x << PIXEL_SHIFT) * 8 + mb_y * s->uvlinesize * block_h;
    dest_cr = s->current_picture.f.data[2] +  (mb_x << PIXEL_SHIFT) * 8 + mb_y * s->uvlinesize * block_h;
    dest_y  = h->cur_pic.f.data[0] + ((mb_x << PIXEL_SHIFT)     + mb_y * h->linesize)  * 16;
    dest_cb = h->cur_pic.f.data[1] +  (mb_x << PIXEL_SHIFT) * 8 + mb_y * h->uvlinesize * block_h;
    dest_cr = h->cur_pic.f.data[2] +  (mb_x << PIXEL_SHIFT) * 8 + mb_y * h->uvlinesize * block_h;
    s->vdsp.prefetch(dest_y  + (s->mb_x & 3) * 4 * s->linesize   + (64 << PIXEL_SHIFT), s->linesize,       4);
    s->vdsp.prefetch(dest_cb + (s->mb_x & 7)     * s->uvlinesize + (64 << PIXEL_SHIFT), dest_cr - dest_cb, 2);
    h->vdsp.prefetch(dest_y  + (h->mb_x & 3) * 4 * h->linesize   + (64 << PIXEL_SHIFT), h->linesize,       4);
    h->vdsp.prefetch(dest_cb + (h->mb_x & 7)     * h->uvlinesize + (64 << PIXEL_SHIFT), dest_cr - dest_cb, 2);
    h->list_counts[mb_xy] = h->list_count;
    if (!SIMPLE && MB_FIELD) {
        linesize     = h->mb_linesize = s->linesize * 2;
        uvlinesize   = h->mb_uvlinesize = s->uvlinesize * 2;
        linesize     = h->mb_linesize = h->linesize * 2;
        uvlinesize   = h->mb_uvlinesize = h->uvlinesize * 2;
        block_offset = &h->block_offset[48];
        if (mb_y & 1) { // FIXME move out of this function?
            dest_y  -= s->linesize * 15;
            dest_cb -= s->uvlinesize * (block_h - 1);
            dest_cr -= s->uvlinesize * (block_h - 1);
            dest_y  -= h->linesize * 15;
            dest_cb -= h->uvlinesize * (block_h - 1);
            dest_cr -= h->uvlinesize * (block_h - 1);
        }
        if (FRAME_MBAFF) {
            int list;
@@ -81,20 +80,20 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
                    continue;
                if (IS_16X16(mb_type)) {
                    int8_t *ref = &h->ref_cache[list][scan8[0]];
                    fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (s->mb_y & 1), 1);
                    fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (h->mb_y & 1), 1);
                } else {
                    for (i = 0; i < 16; i += 4) {
                        int ref = h->ref_cache[list][scan8[i]];
                        if (ref >= 0)
                            fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2,
                                           8, (16 + ref) ^ (s->mb_y & 1), 1);
                                           8, (16 + ref) ^ (h->mb_y & 1), 1);
                    }
                }
            }
        }
    } else {
        linesize   = h->mb_linesize   = s->linesize;
        uvlinesize = h->mb_uvlinesize = s->uvlinesize;
        linesize   = h->mb_linesize   = h->linesize;
        uvlinesize = h->mb_uvlinesize = h->uvlinesize;
        // dct_offset = s->linesize * 16;
    }
@@ -111,7 +110,7 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
                for (j = 0; j < 16; j++)
                    tmp_y[j] = get_bits(&gb, bit_depth);
            }
            if (SIMPLE || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) {
            if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) {
                if (!h->sps.chroma_format_idc) {
                    for (i = 0; i < block_h; i++) {
                        uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize);
@@ -136,7 +135,7 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
        } else {
            for (i = 0; i < 16; i++)
                memcpy(dest_y + i * linesize, (uint8_t *)h->mb + i * 16, 16);
            if (SIMPLE || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) {
            if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) {
                if (!h->sps.chroma_format_idc) {
                    for (i = 0; i < 8; i++) {
                        memset(dest_cb + i*uvlinesize, 1 << (bit_depth - 1), 8);
@@ -158,7 +157,7 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
                xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize,
                               uvlinesize, 1, 0, SIMPLE, PIXEL_SHIFT);
            if (SIMPLE || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) {
            if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) {
                h->hpc.pred8x8[h->chroma_pred_mode](dest_cb, uvlinesize);
                h->hpc.pred8x8[h->chroma_pred_mode](dest_cr, uvlinesize);
            }
@@ -173,14 +172,14 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
        } else if (is_h264) {
            if (chroma422) {
                FUNC(hl_motion_422)(h, dest_y, dest_cb, dest_cr,
                              s->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
                              s->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
                              h->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
                              h->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
                              h->h264dsp.weight_h264_pixels_tab,
                              h->h264dsp.biweight_h264_pixels_tab);
            } else {
                FUNC(hl_motion_420)(h, dest_y, dest_cb, dest_cr,
                              s->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
                              s->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
                              h->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
                              h->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
                              h->h264dsp.weight_h264_pixels_tab,
                              h->h264dsp.biweight_h264_pixels_tab);
            }
@@ -189,7 +188,7 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
        hl_decode_mb_idct_luma(h, mb_type, is_h264, SIMPLE, transform_bypass,
                               PIXEL_SHIFT, block_offset, linesize, dest_y, 0);
        if ((SIMPLE || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) &&
        if ((SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) &&
            (h->cbp & 0x30)) {
            uint8_t *dest[2] = { dest_cb, dest_cr };
            if (transform_bypass) {
@@ -253,7 +252,7 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
                                uint8_t *const ptr = dest[j - 1] + block_offset[i];
                                ff_svq3_add_idct_c(ptr, h->mb + i * 16,
                                                   uvlinesize,
                                                   ff_h264_chroma_qp[0][s->qscale + 12] - 12, 2);
                                                   ff_h264_chroma_qp[0][h->qscale + 12] - 12, 2);
                            }
                    }
                }
@@ -261,8 +260,8 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
        }
    }
    if (h->cbp || IS_INTRA(mb_type)) {
        s->dsp.clear_blocks(h->mb);
        s->dsp.clear_blocks(h->mb + (24 * 16 << PIXEL_SHIFT));
        h->dsp.clear_blocks(h->mb);
        h->dsp.clear_blocks(h->mb + (24 * 16 << PIXEL_SHIFT));
    }
 }
@@ -274,33 +273,32 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
 static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h)
 {
    MpegEncContext *const s = &h->s;
    const int mb_x    = s->mb_x;
    const int mb_y    = s->mb_y;
    const int mb_x    = h->mb_x;
    const int mb_y    = h->mb_y;
    const int mb_xy   = h->mb_xy;
    const int mb_type = s->current_picture.f.mb_type[mb_xy];
    const int mb_type = h->cur_pic.f.mb_type[mb_xy];
    uint8_t *dest[3];
    int linesize;
    int i, j, p;
    int *block_offset = &h->block_offset[0];
    const int transform_bypass = !SIMPLE && (s->qscale == 0 && h->sps.transform_bypass);
    const int plane_count      = (SIMPLE || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) ? 3 : 1;
    const int transform_bypass = !SIMPLE && (h->qscale == 0 && h->sps.transform_bypass);
    const int plane_count      = (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) ? 3 : 1;
    for (p = 0; p < plane_count; p++) {
        dest[p] = s->current_picture.f.data[p] +
                  ((mb_x << PIXEL_SHIFT) + mb_y * s->linesize) * 16;
        s->vdsp.prefetch(dest[p] + (s->mb_x & 3) * 4 * s->linesize + (64 << PIXEL_SHIFT),
                         s->linesize, 4);
        dest[p] = h->cur_pic.f.data[p] +
                  ((mb_x << PIXEL_SHIFT) + mb_y * h->linesize) * 16;
        h->vdsp.prefetch(dest[p] + (h->mb_x & 3) * 4 * h->linesize + (64 << PIXEL_SHIFT),
                         h->linesize, 4);
    }
    h->list_counts[mb_xy] = h->list_count;
    if (!SIMPLE && MB_FIELD) {
        linesize     = h->mb_linesize = h->mb_uvlinesize = s->linesize * 2;
        linesize     = h->mb_linesize = h->mb_uvlinesize = h->linesize * 2;
        block_offset = &h->block_offset[48];
        if (mb_y & 1) // FIXME move out of this function?
            for (p = 0; p < 3; p++)
                dest[p] -= s->linesize * 15;
                dest[p] -= h->linesize * 15;
        if (FRAME_MBAFF) {
            int list;
            for (list = 0; list < h->list_count; list++) {
@@ -308,19 +306,19 @@ static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h)
                    continue;
                if (IS_16X16(mb_type)) {
                    int8_t *ref = &h->ref_cache[list][scan8[0]];
                    fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (s->mb_y & 1), 1);
                    fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (h->mb_y & 1), 1);
                } else {
                    for (i = 0; i < 16; i += 4) {
                        int ref = h->ref_cache[list][scan8[i]];
                        if (ref >= 0)
                            fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2,
                                           8, (16 + ref) ^ (s->mb_y & 1), 1);
                                           8, (16 + ref) ^ (h->mb_y & 1), 1);
                    }
                }
            }
        }
    } else {
        linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize;
        linesize = h->mb_linesize = h->mb_uvlinesize = h->linesize;
    }
    if (!SIMPLE && IS_INTRA_PCM(mb_type)) {
@@ -357,8 +355,8 @@ static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h)
                               linesize, 0, 1, SIMPLE, PIXEL_SHIFT);
        } else {
            FUNC(hl_motion_444)(h, dest[0], dest[1], dest[2],
                      s->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
                      s->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
                      h->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
                      h->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
                      h->h264dsp.weight_h264_pixels_tab,
                      h->h264dsp.biweight_h264_pixels_tab);
        }
@@ -369,8 +367,8 @@ static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h)
                                   dest[p], p);
    }
    if (h->cbp || IS_INTRA(mb_type)) {
        s->dsp.clear_blocks(h->mb);
        s->dsp.clear_blocks(h->mb + (24 * 16 << PIXEL_SHIFT));
        h->dsp.clear_blocks(h->mb);
        h->dsp.clear_blocks(h->mb + (24 * 16 << PIXEL_SHIFT));
    }
 }
--- a/libavcodec/h264_mc_template.c
+++ b/libavcodec/h264_mc_template.c
@@ -46,7 +46,7 @@ static void mc_part(H264Context *h, int n, int square,
                    int list0, int list1)
 {
    if ((h->use_weight == 2 && list0 && list1 &&
         (h->implicit_weight[h->ref_cache[0][scan8[n]]][h->ref_cache[1][scan8[n]]][h->s.mb_y & 1] != 32)) ||
         (h->implicit_weight[h->ref_cache[0][scan8[n]]][h->ref_cache[1][scan8[n]]][h->mb_y & 1] != 32)) ||
        h->use_weight == 1)
        mc_part_weighted(h, n, square, height, delta, dest_y, dest_cb, dest_cr,
                         x_offset, y_offset, qpix_put, chroma_put,
@@ -67,13 +67,12 @@ static void MCFUNC(hl_motion)(H264Context *h, uint8_t *dest_y,
                              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.f.mb_type[mb_xy];
    const int mb_type = h->cur_pic.f.mb_type[mb_xy];
    av_assert2(IS_INTER(mb_type));
    if (HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME))
    if (HAVE_THREADS && (h->avctx->active_thread_type & FF_THREAD_FRAME))
        await_references(h);
    prefetch_motion(h, 0, PIXEL_SHIFT, CHROMA_IDC);
--- a/libavcodec/h264_mvpred.h
+++ b/libavcodec/h264_mvpred.h
@@ -38,32 +38,31 @@ static av_always_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];
    MpegEncContext *s      = &h->s;
    /* there is no consistent mapping of mvs to neighboring locations that will
     * make mbaff happy, so we can't move all this logic to fill_caches */
    if (FRAME_MBAFF) {
 #define SET_DIAG_MV(MV_OP, REF_OP, XY, Y4)                              \
        const int xy = XY, y4 = Y4;                                     \
        const int mb_type = mb_types[xy + (y4 >> 2) * s->mb_stride];    \
        const int mb_type = mb_types[xy + (y4 >> 2) * h->mb_stride];    \
        if (!USES_LIST(mb_type, list))                                  \
            return LIST_NOT_USED;                                       \
        mv = s->current_picture_ptr->f.motion_val[list][h->mb2b_xy[xy] + 3 + y4 * h->b_stride]; \
        mv = h->cur_pic_ptr->f.motion_val[list][h->mb2b_xy[xy] + 3 + y4 * h->b_stride]; \
        h->mv_cache[list][scan8[0] - 2][0] = mv[0];                     \
        h->mv_cache[list][scan8[0] - 2][1] = mv[1] MV_OP;               \
        return s->current_picture_ptr->f.ref_index[list][4 * xy + 1 + (y4 & ~1)] REF_OP;
        return h->cur_pic_ptr->f.ref_index[list][4 * xy + 1 + (y4 & ~1)] REF_OP;
        if (topright_ref == PART_NOT_AVAILABLE
            && i >= scan8[0] + 8 && (i & 7) == 4
            && h->ref_cache[list][scan8[0] - 1] != PART_NOT_AVAILABLE) {
            const uint32_t *mb_types = s->current_picture_ptr->f.mb_type;
            const uint32_t *mb_types = h->cur_pic_ptr->f.mb_type;
            const int16_t *mv;
            AV_ZERO32(h->mv_cache[list][scan8[0] - 2]);
            *C = h->mv_cache[list][scan8[0] - 2];
            if (!MB_FIELD && IS_INTERLACED(h->left_type[0])) {
                SET_DIAG_MV(* 2, >> 1, h->left_mb_xy[0] + s->mb_stride,
                            (s->mb_y & 1) * 2 + (i >> 5));
                SET_DIAG_MV(* 2, >> 1, h->left_mb_xy[0] + h->mb_stride,
                            (h->mb_y & 1) * 2 + (i >> 5));
            }
            if (MB_FIELD && !IS_INTERLACED(h->left_type[0])) {
                // left shift will turn LIST_NOT_USED into PART_NOT_AVAILABLE, but that's OK.
@@ -77,7 +76,7 @@ static av_always_inline int fetch_diagonal_mv(H264Context *h, const int16_t **C,
        *C = h->mv_cache[list][i - 8 + part_width];
        return topright_ref;
    } else {
        tprintf(s->avctx, "topright MV not available\n");
        tprintf(h->avctx, "topright MV not available\n");
        *C = h->mv_cache[list][i - 8 - 1];
        return h->ref_cache[list][i - 8 - 1];
@@ -115,7 +114,7 @@ static av_always_inline void pred_motion(H264Context *const h, int n,
    diagonal_ref = fetch_diagonal_mv(h, &C, index8, list, part_width);
    match_count  = (diagonal_ref == ref) + (top_ref == ref) + (left_ref == ref);
    tprintf(h->s.avctx, "pred_motion match_count=%d\n", match_count);
    tprintf(h->avctx, "pred_motion match_count=%d\n", match_count);
    if (match_count > 1) { //most common
        *mx = mid_pred(A[0], B[0], C[0]);
        *my = mid_pred(A[1], B[1], C[1]);
@@ -142,10 +141,10 @@ static av_always_inline void pred_motion(H264Context *const h, int n,
        }
    }
    tprintf(h->s.avctx,
    tprintf(h->avctx,
            "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);
            A[0], A[1], ref, *mx, *my, h->mb_x, h->mb_y, n, list);
 }
 /**
@@ -162,8 +161,8 @@ static av_always_inline void pred_16x8_motion(H264Context *const h,
        const int top_ref      = h->ref_cache[list][scan8[0] - 8];
        const int16_t *const B = h->mv_cache[list][scan8[0] - 8];
        tprintf(h->s.avctx, "pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d\n",
                top_ref, B[0], B[1], h->s.mb_x, h->s.mb_y, n, list);
        tprintf(h->avctx, "pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d\n",
                top_ref, B[0], B[1], h->mb_x, h->mb_y, n, list);
        if (top_ref == ref) {
            *mx = B[0];
@@ -174,8 +173,8 @@ static av_always_inline void pred_16x8_motion(H264Context *const h,
        const int left_ref     = h->ref_cache[list][scan8[8] - 1];
        const int16_t *const A = h->mv_cache[list][scan8[8] - 1];
        tprintf(h->s.avctx, "pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d\n",
                left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);
        tprintf(h->avctx, "pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d\n",
                left_ref, A[0], A[1], h->mb_x, h->mb_y, n, list);
        if (left_ref == ref) {
            *mx = A[0];
@@ -202,8 +201,8 @@ static av_always_inline void pred_8x16_motion(H264Context *const h,
        const int left_ref     = h->ref_cache[list][scan8[0] - 1];
        const int16_t *const A = h->mv_cache[list][scan8[0] - 1];
        tprintf(h->s.avctx, "pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d\n",
                left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);
        tprintf(h->avctx, "pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d\n",
                left_ref, A[0], A[1], h->mb_x, h->mb_y, n, list);
        if (left_ref == ref) {
            *mx = A[0];
@@ -216,8 +215,8 @@ static av_always_inline void pred_8x16_motion(H264Context *const h,
        diagonal_ref = fetch_diagonal_mv(h, &C, scan8[4], list, 2);
        tprintf(h->s.avctx, "pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d\n",
                diagonal_ref, C[0], C[1], h->s.mb_x, h->s.mb_y, n, list);
        tprintf(h->avctx, "pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d\n",
                diagonal_ref, C[0], C[1], h->mb_x, h->mb_y, n, list);
        if (diagonal_ref == ref) {
            *mx = C[0];
@@ -253,9 +252,8 @@ static av_always_inline void pred_pskip_motion(H264Context *const h)
 {
    DECLARE_ALIGNED(4, static const int16_t, zeromv)[2] = { 0 };
    DECLARE_ALIGNED(4, int16_t, mvbuf)[3][2];
    MpegEncContext *const s = &h->s;
    int8_t *ref     = s->current_picture.f.ref_index[0];
    int16_t(*mv)[2] = s->current_picture.f.motion_val[0];
    int8_t *ref     = h->cur_pic.f.ref_index[0];
    int16_t(*mv)[2] = h->cur_pic.f.motion_val[0];
    int top_ref, left_ref, diagonal_ref, match_count, mx, my;
    const int16_t *A, *B, *C;
    int b_stride = h->b_stride;
@@ -293,8 +291,8 @@ static av_always_inline void pred_pskip_motion(H264Context *const h)
        goto zeromv;
    }
    tprintf(h->s.avctx, "pred_pskip: (%d) (%d) at %2d %2d\n",
            top_ref, left_ref, h->s.mb_x, h->s.mb_y);
    tprintf(h->avctx, "pred_pskip: (%d) (%d) at %2d %2d\n",
            top_ref, left_ref, h->mb_x, h->mb_y);
    if (USES_LIST(h->topright_type, 0)) {
        diagonal_ref = ref[4 * h->topright_mb_xy + 2];
@@ -320,7 +318,7 @@ static av_always_inline void pred_pskip_motion(H264Context *const h)
    }
    match_count = !diagonal_ref + !top_ref + !left_ref;
    tprintf(h->s.avctx, "pred_pskip_motion match_count=%d\n", match_count);
    tprintf(h->avctx, "pred_pskip_motion match_count=%d\n", match_count);
    if (match_count > 1) {
        mx = mid_pred(A[0], B[0], C[0]);
        my = mid_pred(A[1], B[1], C[1]);
@@ -350,7 +348,6 @@ zeromv:
 static void fill_decode_neighbors(H264Context *h, int mb_type)
 {
    MpegEncContext *const s = &h->s;
    const int mb_xy = h->mb_xy;
    int topleft_xy, top_xy, topright_xy, left_xy[LEFT_MBS];
    static const uint8_t left_block_options[4][32] = {
@@ -362,7 +359,7 @@ static void fill_decode_neighbors(H264Context *h, int mb_type)
    h->topleft_partition = -1;
    top_xy = mb_xy - (s->mb_stride << MB_FIELD);
    top_xy = mb_xy - (h->mb_stride << MB_FIELD);
    /* Wow, what a mess, why didn't they simplify the interlacing & intra
     * stuff, I can't imagine that these complex rules are worth it. */
@@ -372,16 +369,16 @@ static void fill_decode_neighbors(H264Context *h, int mb_type)
    left_xy[LBOT] = left_xy[LTOP] = mb_xy - 1;
    h->left_block = left_block_options[0];
    if (FRAME_MBAFF) {
        const int left_mb_field_flag = IS_INTERLACED(s->current_picture.f.mb_type[mb_xy - 1]);
        const int left_mb_field_flag = IS_INTERLACED(h->cur_pic.f.mb_type[mb_xy - 1]);
        const int curr_mb_field_flag = IS_INTERLACED(mb_type);
        if (s->mb_y & 1) {
        if (h->mb_y & 1) {
            if (left_mb_field_flag != curr_mb_field_flag) {
                left_xy[LBOT] = left_xy[LTOP] = mb_xy - s->mb_stride - 1;
                left_xy[LBOT] = left_xy[LTOP] = mb_xy - h->mb_stride - 1;
                if (curr_mb_field_flag) {
                    left_xy[LBOT] += s->mb_stride;
                    left_xy[LBOT] += h->mb_stride;
                    h->left_block  = left_block_options[3];
                } else {
                    topleft_xy += s->mb_stride;
                    topleft_xy += h->mb_stride;
                    /* take top left mv from the middle of the mb, as opposed
                     * to all other modes which use the bottom right partition */
                    h->topleft_partition = 0;
@@ -390,13 +387,13 @@ static void fill_decode_neighbors(H264Context *h, int mb_type)
            }
        } else {
            if (curr_mb_field_flag) {
                topleft_xy  += s->mb_stride & (((s->current_picture.f.mb_type[top_xy - 1] >> 7) & 1) - 1);
                topright_xy += s->mb_stride & (((s->current_picture.f.mb_type[top_xy + 1] >> 7) & 1) - 1);
                top_xy      += s->mb_stride & (((s->current_picture.f.mb_type[top_xy]     >> 7) & 1) - 1);
                topleft_xy  += h->mb_stride & (((h->cur_pic.f.mb_type[top_xy - 1] >> 7) & 1) - 1);
                topright_xy += h->mb_stride & (((h->cur_pic.f.mb_type[top_xy + 1] >> 7) & 1) - 1);
                top_xy      += h->mb_stride & (((h->cur_pic.f.mb_type[top_xy]     >> 7) & 1) - 1);
            }
            if (left_mb_field_flag != curr_mb_field_flag) {
                if (curr_mb_field_flag) {
                    left_xy[LBOT] += s->mb_stride;
                    left_xy[LBOT] += h->mb_stride;
                    h->left_block  = left_block_options[3];
                } else {
                    h->left_block = left_block_options[2];
@@ -412,11 +409,11 @@ static void fill_decode_neighbors(H264Context *h, int mb_type)
    h->left_mb_xy[LBOT] = left_xy[LBOT];
    //FIXME do we need all in the context?
    h->topleft_type    = s->current_picture.f.mb_type[topleft_xy];
    h->top_type        = s->current_picture.f.mb_type[top_xy];
    h->topright_type   = s->current_picture.f.mb_type[topright_xy];
    h->left_type[LTOP] = s->current_picture.f.mb_type[left_xy[LTOP]];
    h->left_type[LBOT] = s->current_picture.f.mb_type[left_xy[LBOT]];
    h->topleft_type    = h->cur_pic.f.mb_type[topleft_xy];
    h->top_type        = h->cur_pic.f.mb_type[top_xy];
    h->topright_type   = h->cur_pic.f.mb_type[topright_xy];
    h->left_type[LTOP] = h->cur_pic.f.mb_type[left_xy[LTOP]];
    h->left_type[LBOT] = h->cur_pic.f.mb_type[left_xy[LBOT]];
    if (FMO) {
        if (h->slice_table[topleft_xy] != h->slice_num)
@@ -440,7 +437,6 @@ static void fill_decode_neighbors(H264Context *h, int mb_type)
 static void fill_decode_caches(H264Context *h, int mb_type)
 {
    MpegEncContext *const s = &h->s;
    int topleft_xy, top_xy, topright_xy, left_xy[LEFT_MBS];
    int topleft_type, top_type, topright_type, left_type[LEFT_MBS];
    const uint8_t *left_block = h->left_block;
@@ -483,7 +479,7 @@ static void fill_decode_caches(H264Context *h, int mb_type)
                        h->left_samples_available    &= 0xFF5F;
                    }
                } else {
                    int left_typei = s->current_picture.f.mb_type[left_xy[LTOP] + s->mb_stride];
                    int left_typei = h->cur_pic.f.mb_type[left_xy[LTOP] + h->mb_stride];
                    av_assert2(left_xy[LTOP] == left_xy[LBOT]);
                    if (!((left_typei & type_mask) && (left_type[LTOP] & type_mask))) {
@@ -540,7 +536,7 @@ static void fill_decode_caches(H264Context *h, int mb_type)
        if (top_type) {
            nnz = h->non_zero_count[top_xy];
            AV_COPY32(&nnz_cache[4 + 8 * 0], &nnz[4 * 3]);
            if (!s->chroma_y_shift) {
            if (!h->chroma_y_shift) {
                AV_COPY32(&nnz_cache[4 + 8 *  5], &nnz[4 *  7]);
                AV_COPY32(&nnz_cache[4 + 8 * 10], &nnz[4 * 11]);
            } else {
@@ -605,9 +601,9 @@ static void fill_decode_caches(H264Context *h, int mb_type)
        int b_stride = h->b_stride;
        for (list = 0; list < h->list_count; list++) {
            int8_t *ref_cache = &h->ref_cache[list][scan8[0]];
            int8_t *ref       = s->current_picture.f.ref_index[list];
            int8_t *ref       = h->cur_pic.f.ref_index[list];
            int16_t(*mv_cache)[2] = &h->mv_cache[list][scan8[0]];
            int16_t(*mv)[2]       = s->current_picture.f.motion_val[list];
            int16_t(*mv)[2]       = h->cur_pic.f.motion_val[list];
            if (!USES_LIST(mb_type, list))
                continue;
            av_assert2(!(IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred));
@@ -799,7 +795,6 @@ static void fill_decode_caches(H264Context *h, int mb_type)
 */
 static void av_unused decode_mb_skip(H264Context *h)
 {
    MpegEncContext *const s = &h->s;
    const int mb_xy = h->mb_xy;
    int mb_type     = 0;
@@ -825,10 +820,10 @@ static void av_unused decode_mb_skip(H264Context *h)
    }
    write_back_motion(h, mb_type);
    s->current_picture.f.mb_type[mb_xy]      = mb_type;
    s->current_picture.f.qscale_table[mb_xy] = s->qscale;
    h->slice_table[mb_xy]                    = h->slice_num;
    h->prev_mb_skipped                       = 1;
    h->cur_pic.f.mb_type[mb_xy]      = mb_type;
    h->cur_pic.f.qscale_table[mb_xy] = h->qscale;
    h->slice_table[mb_xy]            = h->slice_num;
    h->prev_mb_skipped               = 1;
 }
 #endif /* AVCODEC_H264_MVPRED_H */
--- a/libavcodec/h264_parser.c
+++ b/libavcodec/h264_parser.c
@@ -36,7 +36,7 @@ static int ff_h264_find_frame_end(H264Context *h, const uint8_t *buf, int buf_si
 {
    int i, j;
    uint32_t state;
    ParseContext *pc = &(h->s.parse_context);
    ParseContext *pc = &h->parse_context;
    int next_avc= h->is_avc ? 0 : buf_size;
 //    mb_addr= pc->mb_addr - 1;
@@ -45,7 +45,7 @@ static int ff_h264_find_frame_end(H264Context *h, const uint8_t *buf, int buf_si
        state= 7;
    if(h->is_avc && !h->nal_length_size)
        av_log(h->s.avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
        av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
    for(i=0; i<buf_size; i++){
        if(i >= next_avc) {
@@ -54,7 +54,7 @@ static int ff_h264_find_frame_end(H264Context *h, const uint8_t *buf, int buf_si
            for(j = 0; j < h->nal_length_size; j++)
                nalsize = (nalsize << 8) | buf[i++];
            if(nalsize <= 0 || nalsize > buf_size - i){
                av_log(h->s.avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
                av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
                return buf_size;
            }
            next_avc= i + nalsize;
@@ -153,7 +153,7 @@ static inline int parse_nal_units(AVCodecParserContext *s,
    s->pict_type = AV_PICTURE_TYPE_I;
    s->key_frame = 0;
    h->s.avctx= avctx;
    h->avctx= avctx;
    h->sei_recovery_frame_cnt = -1;
    h->sei_dpb_output_delay         =  0;
    h->sei_cpb_removal_delay        = -1;
@@ -171,7 +171,7 @@ static inline int parse_nal_units(AVCodecParserContext *s,
            for (i = 0; i < h->nal_length_size; i++)
                nalsize = (nalsize << 8) | *buf++;
            if (nalsize <= 0 || nalsize > buf_end - buf) {
                av_log(h->s.avctx, AV_LOG_ERROR, "AVC: nal size %d\n", nalsize);
                av_log(h->avctx, AV_LOG_ERROR, "AVC: nal size %d\n", nalsize);
                break;
            }
            src_length = nalsize;
@@ -194,13 +194,13 @@ static inline int parse_nal_units(AVCodecParserContext *s,
        if (ptr==NULL || dst_length < 0)
            break;
        init_get_bits(&h->s.gb, ptr, 8*dst_length);
        init_get_bits(&h->gb, ptr, 8*dst_length);
        switch(h->nal_unit_type) {
        case NAL_SPS:
            ff_h264_decode_seq_parameter_set(h);
            break;
        case NAL_PPS:
            ff_h264_decode_picture_parameter_set(h, h->s.gb.size_in_bits);
            ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
            break;
        case NAL_SEI:
            ff_h264_decode_sei(h);
@@ -209,40 +209,40 @@ static inline int parse_nal_units(AVCodecParserContext *s,
            s->key_frame = 1;
            /* fall through */
        case NAL_SLICE:
            get_ue_golomb_long(&h->s.gb);  // skip first_mb_in_slice
            slice_type = get_ue_golomb_31(&h->s.gb);
            get_ue_golomb_long(&h->gb);  // skip first_mb_in_slice
            slice_type = get_ue_golomb_31(&h->gb);
            s->pict_type = golomb_to_pict_type[slice_type % 5];
            if (h->sei_recovery_frame_cnt >= 0) {
                /* key frame, since recovery_frame_cnt is set */
                s->key_frame = 1;
            }
            pps_id= get_ue_golomb(&h->s.gb);
            pps_id= get_ue_golomb(&h->gb);
            if(pps_id>=MAX_PPS_COUNT) {
                av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
                av_log(h->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");
                av_log(h->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");
                av_log(h->avctx, AV_LOG_ERROR, "non-existing SPS referenced\n");
                return -1;
            }
            h->sps = *h->sps_buffers[h->pps.sps_id];
            h->frame_num = get_bits(&h->s.gb, h->sps.log2_max_frame_num);
            h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num);
            avctx->profile = ff_h264_get_profile(&h->sps);
            avctx->level   = h->sps.level_idc;
            if(h->sps.frame_mbs_only_flag){
                h->s.picture_structure= PICT_FRAME;
                h->picture_structure= PICT_FRAME;
            }else{
                if(get_bits1(&h->s.gb)) { //field_pic_flag
                    h->s.picture_structure= PICT_TOP_FIELD + get_bits1(&h->s.gb); //bottom_field_flag
                if(get_bits1(&h->gb)) { //field_pic_flag
                    h->picture_structure= PICT_TOP_FIELD + get_bits1(&h->gb); //bottom_field_flag
                } else {
                    h->s.picture_structure= PICT_FRAME;
                    h->picture_structure= PICT_FRAME;
                }
            }
@@ -268,11 +268,11 @@ static inline int parse_nal_units(AVCodecParserContext *s,
                        s->repeat_pict = 5;
                        break;
                    default:
                        s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
                        s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
                        break;
                }
            } else {
                s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
                s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
            }
            return 0; /* no need to evaluate the rest */
@@ -282,7 +282,7 @@ static inline int parse_nal_units(AVCodecParserContext *s,
    if (q264)
        return 0;
    /* didn't find a picture! */
    av_log(h->s.avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
    av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
    return -1;
 }
@@ -292,19 +292,19 @@ static int h264_parse(AVCodecParserContext *s,
                      const uint8_t *buf, int buf_size)
 {
    H264Context *h = s->priv_data;
    ParseContext *pc = &h->s.parse_context;
    ParseContext *pc = &h->parse_context;
    int next;
    if (!h->got_first) {
        h->got_first = 1;
        if (avctx->extradata_size) {
            h->s.avctx = avctx;
            h->avctx = avctx;
            // must be done like in decoder, otherwise opening the parser,
            // letting it create extradata and then closing and opening again
            // will cause has_b_frames to be always set.
            // Note that estimate_timings_from_pts does exactly this.
            if (!avctx->has_b_frames)
                h->s.low_delay = 1;
                h->low_delay = 1;
            ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
        }
    }
@@ -374,7 +374,7 @@ static int h264_split(AVCodecContext *avctx,
 static void close(AVCodecParserContext *s)
 {
    H264Context *h = s->priv_data;
    ParseContext *pc = &h->s.parse_context;
    ParseContext *pc = &h->parse_context;
    av_free(pc->buffer);
    ff_h264_free_context(h);
@@ -384,7 +384,7 @@ static int init(AVCodecParserContext *s)
 {
    H264Context *h = s->priv_data;
    h->thread_context[0] = h;
    h->s.slice_context_count = 1;
    h->slice_context_count = 1;
    return 0;
 }
--- a/libavcodec/h264_ps.c
+++ b/libavcodec/h264_ps.c
@@ -151,46 +151,44 @@ static const uint8_t default_scaling8[2][64]={
 }};
 static inline int decode_hrd_parameters(H264Context *h, SPS *sps){
    MpegEncContext * const s = &h->s;
    int cpb_count, i;
    cpb_count = get_ue_golomb_31(&s->gb) + 1;
    cpb_count = get_ue_golomb_31(&h->gb) + 1;
    if(cpb_count > 32U){
        av_log(h->s.avctx, AV_LOG_ERROR, "cpb_count %d invalid\n", cpb_count);
        av_log(h->avctx, AV_LOG_ERROR, "cpb_count %d invalid\n", cpb_count);
        return -1;
    }
    get_bits(&s->gb, 4); /* bit_rate_scale */
    get_bits(&s->gb, 4); /* cpb_size_scale */
    get_bits(&h->gb, 4); /* bit_rate_scale */
    get_bits(&h->gb, 4); /* cpb_size_scale */
    for(i=0; i<cpb_count; i++){
        get_ue_golomb_long(&s->gb); /* bit_rate_value_minus1 */
        get_ue_golomb_long(&s->gb); /* cpb_size_value_minus1 */
        get_bits1(&s->gb);     /* cbr_flag */
        get_ue_golomb_long(&h->gb); /* bit_rate_value_minus1 */
        get_ue_golomb_long(&h->gb); /* cpb_size_value_minus1 */
        get_bits1(&h->gb);     /* cbr_flag */
    }
    sps->initial_cpb_removal_delay_length = get_bits(&s->gb, 5) + 1;
    sps->cpb_removal_delay_length = get_bits(&s->gb, 5) + 1;
    sps->dpb_output_delay_length = get_bits(&s->gb, 5) + 1;
    sps->time_offset_length = get_bits(&s->gb, 5);
    sps->initial_cpb_removal_delay_length = get_bits(&h->gb, 5) + 1;
    sps->cpb_removal_delay_length = get_bits(&h->gb, 5) + 1;
    sps->dpb_output_delay_length = get_bits(&h->gb, 5) + 1;
    sps->time_offset_length = get_bits(&h->gb, 5);
    sps->cpb_cnt = cpb_count;
    return 0;
 }
 static inline int decode_vui_parameters(H264Context *h, SPS *sps){
    MpegEncContext * const s = &h->s;
    int aspect_ratio_info_present_flag;
    unsigned int aspect_ratio_idc;
    aspect_ratio_info_present_flag= get_bits1(&s->gb);
    aspect_ratio_info_present_flag= get_bits1(&h->gb);
    if( aspect_ratio_info_present_flag ) {
        aspect_ratio_idc= get_bits(&s->gb, 8);
        aspect_ratio_idc= get_bits(&h->gb, 8);
        if( aspect_ratio_idc == EXTENDED_SAR ) {
            sps->sar.num= get_bits(&s->gb, 16);
            sps->sar.den= get_bits(&s->gb, 16);
            sps->sar.num= get_bits(&h->gb, 16);
            sps->sar.den= get_bits(&h->gb, 16);
        }else if(aspect_ratio_idc < FF_ARRAY_ELEMS(pixel_aspect)){
            sps->sar=  pixel_aspect[aspect_ratio_idc];
        }else{
            av_log(h->s.avctx, AV_LOG_ERROR, "illegal aspect ratio\n");
            av_log(h->avctx, AV_LOG_ERROR, "illegal aspect ratio\n");
            return -1;
        }
    }else{
@@ -199,20 +197,20 @@ static inline int decode_vui_parameters(H264Context *h, SPS *sps){
    }
 //            s->avctx->aspect_ratio= sar_width*s->width / (float)(s->height*sar_height);
    if(get_bits1(&s->gb)){      /* overscan_info_present_flag */
        get_bits1(&s->gb);      /* overscan_appropriate_flag */
    if(get_bits1(&h->gb)){      /* overscan_info_present_flag */
        get_bits1(&h->gb);      /* overscan_appropriate_flag */
    }
    sps->video_signal_type_present_flag = get_bits1(&s->gb);
    sps->video_signal_type_present_flag = get_bits1(&h->gb);
    if(sps->video_signal_type_present_flag){
        get_bits(&s->gb, 3);    /* video_format */
        sps->full_range = get_bits1(&s->gb); /* video_full_range_flag */
        get_bits(&h->gb, 3);    /* video_format */
        sps->full_range = get_bits1(&h->gb); /* video_full_range_flag */
        sps->colour_description_present_flag = get_bits1(&s->gb);
        sps->colour_description_present_flag = get_bits1(&h->gb);
        if(sps->colour_description_present_flag){
            sps->color_primaries = get_bits(&s->gb, 8); /* colour_primaries */
            sps->color_trc       = get_bits(&s->gb, 8); /* transfer_characteristics */
            sps->colorspace      = get_bits(&s->gb, 8); /* matrix_coefficients */
            sps->color_primaries = get_bits(&h->gb, 8); /* colour_primaries */
            sps->color_trc       = get_bits(&h->gb, 8); /* transfer_characteristics */
            sps->colorspace      = get_bits(&h->gb, 8); /* matrix_coefficients */
            if (sps->color_primaries >= AVCOL_PRI_NB)
                sps->color_primaries  = AVCOL_PRI_UNSPECIFIED;
            if (sps->color_trc >= AVCOL_TRC_NB)
@@ -222,58 +220,58 @@ static inline int decode_vui_parameters(H264Context *h, SPS *sps){
        }
    }
    if(get_bits1(&s->gb)){      /* chroma_location_info_present_flag */
        s->avctx->chroma_sample_location = get_ue_golomb(&s->gb)+1;  /* chroma_sample_location_type_top_field */
        get_ue_golomb(&s->gb);  /* chroma_sample_location_type_bottom_field */
    if(get_bits1(&h->gb)){      /* chroma_location_info_present_flag */
        h->avctx->chroma_sample_location = get_ue_golomb(&h->gb)+1;  /* chroma_sample_location_type_top_field */
        get_ue_golomb(&h->gb);  /* chroma_sample_location_type_bottom_field */
    }
    sps->timing_info_present_flag = get_bits1(&s->gb);
    sps->timing_info_present_flag = get_bits1(&h->gb);
    if(sps->timing_info_present_flag){
        sps->num_units_in_tick = get_bits_long(&s->gb, 32);
        sps->time_scale = get_bits_long(&s->gb, 32);
        sps->num_units_in_tick = get_bits_long(&h->gb, 32);
        sps->time_scale = get_bits_long(&h->gb, 32);
        if(!sps->num_units_in_tick || !sps->time_scale){
            av_log(h->s.avctx, AV_LOG_ERROR, "time_scale/num_units_in_tick invalid or unsupported (%d/%d)\n", sps->time_scale, sps->num_units_in_tick);
            av_log(h->avctx, AV_LOG_ERROR, "time_scale/num_units_in_tick invalid or unsupported (%d/%d)\n", sps->time_scale, sps->num_units_in_tick);
            return -1;
        }
        sps->fixed_frame_rate_flag = get_bits1(&s->gb);
        sps->fixed_frame_rate_flag = get_bits1(&h->gb);
    }
    sps->nal_hrd_parameters_present_flag = get_bits1(&s->gb);
    sps->nal_hrd_parameters_present_flag = get_bits1(&h->gb);
    if(sps->nal_hrd_parameters_present_flag)
        if(decode_hrd_parameters(h, sps) < 0)
            return -1;
    sps->vcl_hrd_parameters_present_flag = get_bits1(&s->gb);
    sps->vcl_hrd_parameters_present_flag = get_bits1(&h->gb);
    if(sps->vcl_hrd_parameters_present_flag)
        if(decode_hrd_parameters(h, sps) < 0)
            return -1;
    if(sps->nal_hrd_parameters_present_flag || sps->vcl_hrd_parameters_present_flag)
        get_bits1(&s->gb);     /* low_delay_hrd_flag */
    sps->pic_struct_present_flag = get_bits1(&s->gb);
    if(!get_bits_left(&s->gb))
        get_bits1(&h->gb);     /* low_delay_hrd_flag */
    sps->pic_struct_present_flag = get_bits1(&h->gb);
    if(!get_bits_left(&h->gb))
        return 0;
    sps->bitstream_restriction_flag = get_bits1(&s->gb);
    sps->bitstream_restriction_flag = get_bits1(&h->gb);
    if(sps->bitstream_restriction_flag){
        get_bits1(&s->gb);     /* motion_vectors_over_pic_boundaries_flag */
        get_ue_golomb(&s->gb); /* max_bytes_per_pic_denom */
        get_ue_golomb(&s->gb); /* max_bits_per_mb_denom */
        get_ue_golomb(&s->gb); /* log2_max_mv_length_horizontal */
        get_ue_golomb(&s->gb); /* log2_max_mv_length_vertical */
        sps->num_reorder_frames= get_ue_golomb(&s->gb);
        get_ue_golomb(&s->gb); /*max_dec_frame_buffering*/
        if (get_bits_left(&s->gb) < 0) {
        get_bits1(&h->gb);     /* motion_vectors_over_pic_boundaries_flag */
        get_ue_golomb(&h->gb); /* max_bytes_per_pic_denom */
        get_ue_golomb(&h->gb); /* max_bits_per_mb_denom */
        get_ue_golomb(&h->gb); /* log2_max_mv_length_horizontal */
        get_ue_golomb(&h->gb); /* log2_max_mv_length_vertical */
        sps->num_reorder_frames= get_ue_golomb(&h->gb);
        get_ue_golomb(&h->gb); /*max_dec_frame_buffering*/
        if (get_bits_left(&h->gb) < 0) {
            sps->num_reorder_frames=0;
            sps->bitstream_restriction_flag= 0;
        }
        if(sps->num_reorder_frames > 16U /*max_dec_frame_buffering || max_dec_frame_buffering > 16*/){
            av_log(h->s.avctx, AV_LOG_ERROR, "illegal num_reorder_frames %d\n", sps->num_reorder_frames);
            av_log(h->avctx, AV_LOG_ERROR, "illegal num_reorder_frames %d\n", sps->num_reorder_frames);
            return -1;
        }
    }
    if (get_bits_left(&s->gb) < 0) {
        av_log(h->s.avctx, AV_LOG_ERROR, "Overread VUI by %d bits\n", -get_bits_left(&s->gb));
    if (get_bits_left(&h->gb) < 0) {
        av_log(h->avctx, AV_LOG_ERROR, "Overread VUI by %d bits\n", -get_bits_left(&h->gb));
        return AVERROR_INVALIDDATA;
    }
@@ -282,15 +280,14 @@ static inline int decode_vui_parameters(H264Context *h, SPS *sps){
 static void decode_scaling_list(H264Context *h, uint8_t *factors, int size,
                                const uint8_t *jvt_list, const uint8_t *fallback_list){
    MpegEncContext * const s = &h->s;
    int i, last = 8, next = 8;
    const uint8_t *scan = size == 16 ? zigzag_scan : ff_zigzag_direct;
    if(!get_bits1(&s->gb)) /* matrix not written, we use the predicted one */
    if(!get_bits1(&h->gb)) /* matrix not written, we use the predicted one */
        memcpy(factors, fallback_list, size*sizeof(uint8_t));
    else
    for(i=0;i<size;i++){
        if(next)
            next = (last + get_se_golomb(&s->gb)) & 0xff;
            next = (last + get_se_golomb(&h->gb)) & 0xff;
        if(!i && !next){ /* matrix not written, we use the preset one */
            memcpy(factors, jvt_list, size*sizeof(uint8_t));
            break;
@@ -301,7 +298,6 @@ static void decode_scaling_list(H264Context *h, uint8_t *factors, int size,
 static void decode_scaling_matrices(H264Context *h, SPS *sps, PPS *pps, int is_sps,
                                   uint8_t (*scaling_matrix4)[16], uint8_t (*scaling_matrix8)[64]){
    MpegEncContext * const s = &h->s;
    int fallback_sps = !is_sps && sps->scaling_matrix_present;
    const uint8_t *fallback[4] = {
        fallback_sps ? sps->scaling_matrix4[0] : default_scaling4[0],
@@ -309,7 +305,7 @@ static void decode_scaling_matrices(H264Context *h, SPS *sps, PPS *pps, int is_s
        fallback_sps ? sps->scaling_matrix8[0] : default_scaling8[0],
        fallback_sps ? sps->scaling_matrix8[3] : default_scaling8[1]
    };
    if(get_bits1(&s->gb)){
    if(get_bits1(&h->gb)){
        sps->scaling_matrix_present |= is_sps;
        decode_scaling_list(h,scaling_matrix4[0],16,default_scaling4[0],fallback[0]); // Intra, Y
        decode_scaling_list(h,scaling_matrix4[1],16,default_scaling4[0],scaling_matrix4[0]); // Intra, Cr
@@ -331,25 +327,24 @@ static void decode_scaling_matrices(H264Context *h, SPS *sps, PPS *pps, int is_s
 }
 int ff_h264_decode_seq_parameter_set(H264Context *h){
    MpegEncContext * const s = &h->s;
    int profile_idc, level_idc, constraint_set_flags = 0;
    unsigned int sps_id;
    int i, log2_max_frame_num_minus4;
    SPS *sps;
    profile_idc= get_bits(&s->gb, 8);
    constraint_set_flags |= get_bits1(&s->gb) << 0;   //constraint_set0_flag
    constraint_set_flags |= get_bits1(&s->gb) << 1;   //constraint_set1_flag
    constraint_set_flags |= get_bits1(&s->gb) << 2;   //constraint_set2_flag
    constraint_set_flags |= get_bits1(&s->gb) << 3;   //constraint_set3_flag
    constraint_set_flags |= get_bits1(&s->gb) << 4;   //constraint_set4_flag
    constraint_set_flags |= get_bits1(&s->gb) << 5;   //constraint_set5_flag
    get_bits(&s->gb, 2); // reserved
    level_idc= get_bits(&s->gb, 8);
    sps_id= get_ue_golomb_31(&s->gb);
    profile_idc= get_bits(&h->gb, 8);
    constraint_set_flags |= get_bits1(&h->gb) << 0;   //constraint_set0_flag
    constraint_set_flags |= get_bits1(&h->gb) << 1;   //constraint_set1_flag
    constraint_set_flags |= get_bits1(&h->gb) << 2;   //constraint_set2_flag
    constraint_set_flags |= get_bits1(&h->gb) << 3;   //constraint_set3_flag
    constraint_set_flags |= get_bits1(&h->gb) << 4;   //constraint_set4_flag
    constraint_set_flags |= get_bits1(&h->gb) << 5;   //constraint_set5_flag
    get_bits(&h->gb, 2); // reserved
    level_idc= get_bits(&h->gb, 8);
    sps_id= get_ue_golomb_31(&h->gb);
    if(sps_id >= MAX_SPS_COUNT) {
        av_log(h->s.avctx, AV_LOG_ERROR, "sps_id (%d) out of range\n", sps_id);
        av_log(h->avctx, AV_LOG_ERROR, "sps_id (%d) out of range\n", sps_id);
        return -1;
    }
    sps= av_mallocz(sizeof(SPS));
@@ -372,25 +367,25 @@ int ff_h264_decode_seq_parameter_set(H264Context *h){
        sps->profile_idc ==  44 || sps->profile_idc ==  83 ||
        sps->profile_idc ==  86 || sps->profile_idc == 118 ||
        sps->profile_idc == 128 || sps->profile_idc == 144) {
        sps->chroma_format_idc= get_ue_golomb_31(&s->gb);
        sps->chroma_format_idc= get_ue_golomb_31(&h->gb);
        if (sps->chroma_format_idc > 3U) {
            av_log(h->s.avctx, AV_LOG_ERROR, "chroma_format_idc %d is illegal\n", sps->chroma_format_idc);
            av_log(h->avctx, AV_LOG_ERROR, "chroma_format_idc %d is illegal\n", sps->chroma_format_idc);
            goto fail;
        } else if(sps->chroma_format_idc == 3) {
            sps->residual_color_transform_flag = get_bits1(&s->gb);
            sps->residual_color_transform_flag = get_bits1(&h->gb);
            if(sps->residual_color_transform_flag) {
                av_log(h->s.avctx, AV_LOG_ERROR, "separate color planes are not supported\n");
                av_log(h->avctx, AV_LOG_ERROR, "separate color planes are not supported\n");
                goto fail;
            }
        }
        sps->bit_depth_luma   = get_ue_golomb(&s->gb) + 8;
        sps->bit_depth_chroma = get_ue_golomb(&s->gb) + 8;
        sps->bit_depth_luma   = get_ue_golomb(&h->gb) + 8;
        sps->bit_depth_chroma = get_ue_golomb(&h->gb) + 8;
        if (sps->bit_depth_luma > 14U || sps->bit_depth_chroma > 14U) {
            av_log(h->s.avctx, AV_LOG_ERROR, "illegal bit depth value (%d, %d)\n",
            av_log(h->avctx, AV_LOG_ERROR, "illegal bit depth value (%d, %d)\n",
                   sps->bit_depth_luma, sps->bit_depth_chroma);
            goto fail;
        }
        sps->transform_bypass = get_bits1(&s->gb);
        sps->transform_bypass = get_bits1(&h->gb);
        decode_scaling_matrices(h, sps, NULL, 1, sps->scaling_matrix4, sps->scaling_matrix8);
    }else{
        sps->chroma_format_idc= 1;
@@ -398,81 +393,81 @@ int ff_h264_decode_seq_parameter_set(H264Context *h){
        sps->bit_depth_chroma = 8;
    }
    log2_max_frame_num_minus4 = get_ue_golomb(&s->gb);
    log2_max_frame_num_minus4 = get_ue_golomb(&h->gb);
    if (log2_max_frame_num_minus4 < MIN_LOG2_MAX_FRAME_NUM - 4 ||
        log2_max_frame_num_minus4 > MAX_LOG2_MAX_FRAME_NUM - 4) {
        av_log(h->s.avctx, AV_LOG_ERROR,
        av_log(h->avctx, AV_LOG_ERROR,
               "log2_max_frame_num_minus4 out of range (0-12): %d\n",
               log2_max_frame_num_minus4);
        goto fail;
    }
    sps->log2_max_frame_num = log2_max_frame_num_minus4 + 4;
    sps->poc_type= get_ue_golomb_31(&s->gb);
    sps->poc_type= get_ue_golomb_31(&h->gb);
    if(sps->poc_type == 0){ //FIXME #define
        unsigned t = get_ue_golomb(&s->gb);
        unsigned t = get_ue_golomb(&h->gb);
        if(t>12){
            av_log(h->s.avctx, AV_LOG_ERROR, "log2_max_poc_lsb (%d) is out of range\n", t);
            av_log(h->avctx, AV_LOG_ERROR, "log2_max_poc_lsb (%d) is out of range\n", t);
            goto fail;
        }
        sps->log2_max_poc_lsb= t + 4;
    } else if(sps->poc_type == 1){//FIXME #define
        sps->delta_pic_order_always_zero_flag= get_bits1(&s->gb);
        sps->offset_for_non_ref_pic= get_se_golomb(&s->gb);
        sps->offset_for_top_to_bottom_field= get_se_golomb(&s->gb);
        sps->poc_cycle_length                = get_ue_golomb(&s->gb);
        sps->delta_pic_order_always_zero_flag= get_bits1(&h->gb);
        sps->offset_for_non_ref_pic= get_se_golomb(&h->gb);
        sps->offset_for_top_to_bottom_field= get_se_golomb(&h->gb);
        sps->poc_cycle_length                = get_ue_golomb(&h->gb);
        if((unsigned)sps->poc_cycle_length >= FF_ARRAY_ELEMS(sps->offset_for_ref_frame)){
            av_log(h->s.avctx, AV_LOG_ERROR, "poc_cycle_length overflow %u\n", sps->poc_cycle_length);
            av_log(h->avctx, AV_LOG_ERROR, "poc_cycle_length overflow %u\n", sps->poc_cycle_length);
            goto fail;
        }
        for(i=0; i<sps->poc_cycle_length; i++)
            sps->offset_for_ref_frame[i]= get_se_golomb(&s->gb);
            sps->offset_for_ref_frame[i]= get_se_golomb(&h->gb);
    }else if(sps->poc_type != 2){
        av_log(h->s.avctx, AV_LOG_ERROR, "illegal POC type %d\n", sps->poc_type);
        av_log(h->avctx, AV_LOG_ERROR, "illegal POC type %d\n", sps->poc_type);
        goto fail;
    }
    sps->ref_frame_count= get_ue_golomb_31(&s->gb);
    if (h->s.avctx->codec_tag == MKTAG('S', 'M', 'V', '2'))
    sps->ref_frame_count= get_ue_golomb_31(&h->gb);
    if (h->avctx->codec_tag == MKTAG('S', 'M', 'V', '2'))
        sps->ref_frame_count= FFMAX(2, sps->ref_frame_count);
    if(sps->ref_frame_count > MAX_PICTURE_COUNT-2 || sps->ref_frame_count > 16U){
        av_log(h->s.avctx, AV_LOG_ERROR, "too many reference frames\n");
        av_log(h->avctx, AV_LOG_ERROR, "too many reference frames\n");
        goto fail;
    }
    sps->gaps_in_frame_num_allowed_flag= get_bits1(&s->gb);
    sps->mb_width = get_ue_golomb(&s->gb) + 1;
    sps->mb_height= get_ue_golomb(&s->gb) + 1;
    sps->gaps_in_frame_num_allowed_flag= get_bits1(&h->gb);
    sps->mb_width = get_ue_golomb(&h->gb) + 1;
    sps->mb_height= get_ue_golomb(&h->gb) + 1;
    if((unsigned)sps->mb_width >= INT_MAX/16 || (unsigned)sps->mb_height >= INT_MAX/16 ||
       av_image_check_size(16*sps->mb_width, 16*sps->mb_height, 0, h->s.avctx)){
        av_log(h->s.avctx, AV_LOG_ERROR, "mb_width/height overflow\n");
       av_image_check_size(16*sps->mb_width, 16*sps->mb_height, 0, h->avctx)){
        av_log(h->avctx, AV_LOG_ERROR, "mb_width/height overflow\n");
        goto fail;
    }
    sps->frame_mbs_only_flag= get_bits1(&s->gb);
    sps->frame_mbs_only_flag= get_bits1(&h->gb);
    if(!sps->frame_mbs_only_flag)
        sps->mb_aff= get_bits1(&s->gb);
        sps->mb_aff= get_bits1(&h->gb);
    else
        sps->mb_aff= 0;
    sps->direct_8x8_inference_flag= get_bits1(&s->gb);
    sps->direct_8x8_inference_flag= get_bits1(&h->gb);
 #ifndef ALLOW_INTERLACE
    if(sps->mb_aff)
        av_log(h->s.avctx, AV_LOG_ERROR, "MBAFF support not included; enable it at compile-time.\n");
        av_log(h->avctx, AV_LOG_ERROR, "MBAFF support not included; enable it at compile-time.\n");
 #endif
    sps->crop= get_bits1(&s->gb);
    sps->crop= get_bits1(&h->gb);
    if(sps->crop){
        int crop_vertical_limit   = sps->chroma_format_idc  & 2 ? 16 : 8;
        int crop_horizontal_limit = sps->chroma_format_idc == 3 ? 16 : 8;
        sps->crop_left  = get_ue_golomb(&s->gb);
        sps->crop_right = get_ue_golomb(&s->gb);
        sps->crop_top   = get_ue_golomb(&s->gb);
        sps->crop_bottom= get_ue_golomb(&s->gb);
        if (h->s.avctx->flags2 & CODEC_FLAG2_IGNORE_CROP) {
            av_log(h->s.avctx, AV_LOG_DEBUG,
        sps->crop_left  = get_ue_golomb(&h->gb);
        sps->crop_right = get_ue_golomb(&h->gb);
        sps->crop_top   = get_ue_golomb(&h->gb);
        sps->crop_bottom= get_ue_golomb(&h->gb);
        if (h->avctx->flags2 & CODEC_FLAG2_IGNORE_CROP) {
            av_log(h->avctx, AV_LOG_DEBUG,
                   "discarding sps cropping, "
                   "original values are l:%u r:%u t:%u b:%u\n",
                   sps->crop_left,
@@ -486,10 +481,10 @@ int ff_h264_decode_seq_parameter_set(H264Context *h){
            sps->crop_bottom = 0;
        }
        if(sps->crop_left || sps->crop_top){
            av_log(h->s.avctx, AV_LOG_ERROR, "insane cropping not completely supported, this could look slightly wrong ... (left: %d, top: %d)\n", sps->crop_left, sps->crop_top);
            av_log(h->avctx, AV_LOG_ERROR, "insane cropping not completely supported, this could look slightly wrong ... (left: %d, top: %d)\n", sps->crop_left, sps->crop_top);
        }
        if(sps->crop_right >= crop_horizontal_limit || sps->crop_bottom >= crop_vertical_limit){
            av_log(h->s.avctx, AV_LOG_ERROR, "brainfart cropping not supported, cropping disabled (right: %d, bottom: %d)\n", sps->crop_right, sps->crop_bottom);
            av_log(h->avctx, AV_LOG_ERROR, "brainfart cropping not supported, cropping disabled (right: %d, bottom: %d)\n", sps->crop_right, sps->crop_bottom);
        /* It is very unlikely that partial cropping will make anybody happy.
         * Not cropping at all fixes for example playback of Sisvel 3D streams
         * in applications supporting Sisvel 3D. */
@@ -505,7 +500,7 @@ int ff_h264_decode_seq_parameter_set(H264Context *h){
        sps->crop_bottom= 0;
    }
    sps->vui_parameters_present_flag= get_bits1(&s->gb);
    sps->vui_parameters_present_flag= get_bits1(&h->gb);
    if( sps->vui_parameters_present_flag )
        if (decode_vui_parameters(h, sps) < 0)
            goto fail;
@@ -513,9 +508,9 @@ int ff_h264_decode_seq_parameter_set(H264Context *h){
    if(!sps->sar.den)
        sps->sar.den= 1;
    if(s->avctx->debug&FF_DEBUG_PICT_INFO){
    if(h->avctx->debug&FF_DEBUG_PICT_INFO){
        static const char csp[4][5] = { "Gray", "420", "422", "444" };
        av_log(h->s.avctx, AV_LOG_DEBUG, "sps:%u profile:%d/%d poc:%d ref:%d %dx%d %s %s crop:%d/%d/%d/%d %s %s %d/%d b%d reo:%d\n",
        av_log(h->avctx, AV_LOG_DEBUG, "sps:%u profile:%d/%d poc:%d ref:%d %dx%d %s %s crop:%d/%d/%d/%d %s %s %d/%d b%d reo:%d\n",
               sps_id, sps->profile_idc, sps->level_idc,
               sps->poc_type,
               sps->ref_frame_count,
@@ -561,7 +556,7 @@ static int more_rbsp_data_in_pps(H264Context *h, PPS *pps)
    if ((profile_idc == 66 || profile_idc == 77 ||
         profile_idc == 88) && (sps->constraint_set_flags & 7)) {
        av_log(h->s.avctx, AV_LOG_VERBOSE,
        av_log(h->avctx, AV_LOG_VERBOSE,
               "Current profile doesn't provide more RBSP data in PPS, skipping\n");
        return 0;
    }
@@ -570,38 +565,37 @@ static int more_rbsp_data_in_pps(H264Context *h, PPS *pps)
 }
 int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length){
    MpegEncContext * const s = &h->s;
    unsigned int pps_id= get_ue_golomb(&s->gb);
    unsigned int pps_id= get_ue_golomb(&h->gb);
    PPS *pps;
    const int qp_bd_offset = 6*(h->sps.bit_depth_luma-8);
    int bits_left;
    if(pps_id >= MAX_PPS_COUNT) {
        av_log(h->s.avctx, AV_LOG_ERROR, "pps_id (%d) out of range\n", pps_id);
        av_log(h->avctx, AV_LOG_ERROR, "pps_id (%d) out of range\n", pps_id);
        return -1;
    } else if (h->sps.bit_depth_luma > 14) {
        av_log(h->s.avctx, AV_LOG_ERROR, "Invalid luma bit depth=%d\n", h->sps.bit_depth_luma);
        av_log(h->avctx, AV_LOG_ERROR, "Invalid luma bit depth=%d\n", h->sps.bit_depth_luma);
        return AVERROR_INVALIDDATA;
    } else if (h->sps.bit_depth_luma == 11 || h->sps.bit_depth_luma == 13) {
        av_log(h->s.avctx, AV_LOG_ERROR, "Unimplemented luma bit depth=%d\n", h->sps.bit_depth_luma);
        av_log(h->avctx, AV_LOG_ERROR, "Unimplemented luma bit depth=%d\n", h->sps.bit_depth_luma);
        return AVERROR_PATCHWELCOME;
    }
    pps= av_mallocz(sizeof(PPS));
    if(pps == NULL)
        return -1;
    pps->sps_id= get_ue_golomb_31(&s->gb);
    pps->sps_id= get_ue_golomb_31(&h->gb);
    if((unsigned)pps->sps_id>=MAX_SPS_COUNT || h->sps_buffers[pps->sps_id] == NULL){
        av_log(h->s.avctx, AV_LOG_ERROR, "sps_id out of range\n");
        av_log(h->avctx, AV_LOG_ERROR, "sps_id out of range\n");
        goto fail;
    }
    pps->cabac= get_bits1(&s->gb);
    pps->pic_order_present= get_bits1(&s->gb);
    pps->slice_group_count= get_ue_golomb(&s->gb) + 1;
    pps->cabac= get_bits1(&h->gb);
    pps->pic_order_present= get_bits1(&h->gb);
    pps->slice_group_count= get_ue_golomb(&h->gb) + 1;
    if(pps->slice_group_count > 1 ){
        pps->mb_slice_group_map_type= get_ue_golomb(&s->gb);
        av_log(h->s.avctx, AV_LOG_ERROR, "FMO not supported\n");
        pps->mb_slice_group_map_type= get_ue_golomb(&h->gb);
        av_log(h->avctx, AV_LOG_ERROR, "FMO not supported\n");
        switch(pps->mb_slice_group_map_type){
        case 0:
 #if 0
@@ -636,32 +630,32 @@ int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length){
            break;
        }
    }
    pps->ref_count[0]= get_ue_golomb(&s->gb) + 1;
    pps->ref_count[1]= get_ue_golomb(&s->gb) + 1;
    pps->ref_count[0]= get_ue_golomb(&h->gb) + 1;
    pps->ref_count[1]= get_ue_golomb(&h->gb) + 1;
    if(pps->ref_count[0]-1 > 32-1 || pps->ref_count[1]-1 > 32-1){
        av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow (pps)\n");
        av_log(h->avctx, AV_LOG_ERROR, "reference overflow (pps)\n");
        goto fail;
    }
    pps->weighted_pred= get_bits1(&s->gb);
    pps->weighted_bipred_idc= get_bits(&s->gb, 2);
    pps->init_qp= get_se_golomb(&s->gb) + 26 + qp_bd_offset;
    pps->init_qs= get_se_golomb(&s->gb) + 26 + qp_bd_offset;
    pps->chroma_qp_index_offset[0]= get_se_golomb(&s->gb);
    pps->deblocking_filter_parameters_present= get_bits1(&s->gb);
    pps->constrained_intra_pred= get_bits1(&s->gb);
    pps->redundant_pic_cnt_present = get_bits1(&s->gb);
    pps->weighted_pred= get_bits1(&h->gb);
    pps->weighted_bipred_idc= get_bits(&h->gb, 2);
    pps->init_qp= get_se_golomb(&h->gb) + 26 + qp_bd_offset;
    pps->init_qs= get_se_golomb(&h->gb) + 26 + qp_bd_offset;
    pps->chroma_qp_index_offset[0]= get_se_golomb(&h->gb);
    pps->deblocking_filter_parameters_present= get_bits1(&h->gb);
    pps->constrained_intra_pred= get_bits1(&h->gb);
    pps->redundant_pic_cnt_present = get_bits1(&h->gb);
    pps->transform_8x8_mode= 0;
    h->dequant_coeff_pps= -1; //contents of sps/pps can change even if id doesn't, so reinit
    memcpy(pps->scaling_matrix4, h->sps_buffers[pps->sps_id]->scaling_matrix4, sizeof(pps->scaling_matrix4));
    memcpy(pps->scaling_matrix8, h->sps_buffers[pps->sps_id]->scaling_matrix8, sizeof(pps->scaling_matrix8));
    bits_left = bit_length - get_bits_count(&s->gb);
    bits_left = bit_length - get_bits_count(&h->gb);
    if(bits_left > 0 && more_rbsp_data_in_pps(h, pps)){
        pps->transform_8x8_mode= get_bits1(&s->gb);
        pps->transform_8x8_mode= get_bits1(&h->gb);
        decode_scaling_matrices(h, h->sps_buffers[pps->sps_id], pps, 0, pps->scaling_matrix4, pps->scaling_matrix8);
        pps->chroma_qp_index_offset[1]= get_se_golomb(&s->gb); //second_chroma_qp_index_offset
        pps->chroma_qp_index_offset[1]= get_se_golomb(&h->gb); //second_chroma_qp_index_offset
    } else {
        pps->chroma_qp_index_offset[1]= pps->chroma_qp_index_offset[0];
    }
@@ -671,8 +665,8 @@ int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length){
    if(pps->chroma_qp_index_offset[0] != pps->chroma_qp_index_offset[1])
        pps->chroma_qp_diff= 1;
    if(s->avctx->debug&FF_DEBUG_PICT_INFO){
        av_log(h->s.avctx, AV_LOG_DEBUG, "pps:%u sps:%u %s slice_groups:%d ref:%d/%d %s qp:%d/%d/%d/%d %s %s %s %s\n",
    if(h->avctx->debug&FF_DEBUG_PICT_INFO){
        av_log(h->avctx, AV_LOG_DEBUG, "pps:%u sps:%u %s slice_groups:%d ref:%d/%d %s qp:%d/%d/%d/%d %s %s %s %s\n",
               pps_id, pps->sps_id,
               pps->cabac ? "CABAC" : "CAVLC",
               pps->slice_group_count,
--- a/libavcodec/h264_refs.c
+++ b/libavcodec/h264_refs.c
@@ -107,7 +107,6 @@ static int add_sorted(Picture **sorted, Picture **src, int len, int limit, int d
 }
 int ff_h264_fill_default_ref_list(H264Context *h){
    MpegEncContext * const s = &h->s;
    int i, len;
    if(h->slice_type_nos==AV_PICTURE_TYPE_B){
@@ -116,16 +115,16 @@ int ff_h264_fill_default_ref_list(H264Context *h){
        int lens[2];
        if(FIELD_PICTURE)
            cur_poc= s->current_picture_ptr->field_poc[ s->picture_structure == PICT_BOTTOM_FIELD ];
            cur_poc= h->cur_pic_ptr->field_poc[h->picture_structure == PICT_BOTTOM_FIELD];
        else
            cur_poc= s->current_picture_ptr->poc;
            cur_poc= h->cur_pic_ptr->poc;
        for(list= 0; list<2; list++){
            len= add_sorted(sorted    , h->short_ref, h->short_ref_count, cur_poc, 1^list);
            len+=add_sorted(sorted+len, h->short_ref, h->short_ref_count, cur_poc, 0^list);
            av_assert0(len<=32);
            len= build_def_list(h->default_ref_list[list]    , sorted     , len, 0, s->picture_structure);
            len+=build_def_list(h->default_ref_list[list]+len, h->long_ref, 16 , 1, s->picture_structure);
            len= build_def_list(h->default_ref_list[list]    , sorted     , len, 0, h->picture_structure);
            len+=build_def_list(h->default_ref_list[list]+len, h->long_ref, 16 , 1, h->picture_structure);
            av_assert0(len<=32);
            if(len < h->ref_count[list])
@@ -139,19 +138,19 @@ int ff_h264_fill_default_ref_list(H264Context *h){
                FFSWAP(Picture, h->default_ref_list[1][0], h->default_ref_list[1][1]);
        }
    }else{
        len = build_def_list(h->default_ref_list[0]    , h->short_ref, h->short_ref_count, 0, s->picture_structure);
        len+= build_def_list(h->default_ref_list[0]+len, h-> long_ref, 16                , 1, s->picture_structure);
        len = build_def_list(h->default_ref_list[0]    , h->short_ref, h->short_ref_count, 0, h->picture_structure);
        len+= build_def_list(h->default_ref_list[0]+len, h-> long_ref, 16                , 1, h->picture_structure);
        av_assert0(len<=32);
        if(len < h->ref_count[0])
            memset(&h->default_ref_list[0][len], 0, sizeof(Picture)*(h->ref_count[0] - len));
    }
 #ifdef TRACE
    for (i=0; i<h->ref_count[0]; i++) {
        tprintf(h->s.avctx, "List0: %s fn:%d 0x%p\n", (h->default_ref_list[0][i].long_ref ? "LT" : "ST"), h->default_ref_list[0][i].pic_id, h->default_ref_list[0][i].f.data[0]);
        tprintf(h->avctx, "List0: %s fn:%d 0x%p\n", (h->default_ref_list[0][i].long_ref ? "LT" : "ST"), h->default_ref_list[0][i].pic_id, h->default_ref_list[0][i].f.data[0]);
    }
    if(h->slice_type_nos==AV_PICTURE_TYPE_B){
        for (i=0; i<h->ref_count[1]; i++) {
            tprintf(h->s.avctx, "List1: %s fn:%d 0x%p\n", (h->default_ref_list[1][i].long_ref ? "LT" : "ST"), h->default_ref_list[1][i].pic_id, h->default_ref_list[1][i].f.data[0]);
            tprintf(h->avctx, "List1: %s fn:%d 0x%p\n", (h->default_ref_list[1][i].long_ref ? "LT" : "ST"), h->default_ref_list[1][i].pic_id, h->default_ref_list[1][i].f.data[0]);
        }
    }
 #endif
@@ -172,9 +171,7 @@ static void print_long_term(H264Context *h);
 *         described by pic_num
 */
 static int pic_num_extract(H264Context *h, int pic_num, int *structure){
    MpegEncContext * const s = &h->s;
    *structure = s->picture_structure;
    *structure = h->picture_structure;
    if(FIELD_PICTURE){
        if (!(pic_num & 1))
            /* opposite field */
@@ -186,7 +183,6 @@ static int pic_num_extract(H264Context *h, int pic_num, int *structure){
 }
 int ff_h264_decode_ref_pic_list_reordering(H264Context *h){
    MpegEncContext * const s = &h->s;
    int list, index, pic_structure;
    print_short_term(h);
@@ -195,11 +191,11 @@ int ff_h264_decode_ref_pic_list_reordering(H264Context *h){
    for(list=0; list<h->list_count; list++){
        memcpy(h->ref_list[list], h->default_ref_list[list], sizeof(Picture)*h->ref_count[list]);
        if(get_bits1(&s->gb)){
        if(get_bits1(&h->gb)){
            int pred= h->curr_pic_num;
            for(index=0; ; index++){
                unsigned int reordering_of_pic_nums_idc= get_ue_golomb_31(&s->gb);
                unsigned int reordering_of_pic_nums_idc= get_ue_golomb_31(&h->gb);
                unsigned int pic_id;
                int i;
                Picture *ref = NULL;
@@ -208,17 +204,17 @@ int ff_h264_decode_ref_pic_list_reordering(H264Context *h){
                    break;
                if(index >= h->ref_count[list]){
                    av_log(h->s.avctx, AV_LOG_ERROR, "reference count overflow\n");
                    av_log(h->avctx, AV_LOG_ERROR, "reference count overflow\n");
                    return -1;
                }
                if(reordering_of_pic_nums_idc<3){
                    if(reordering_of_pic_nums_idc<2){
                        const unsigned int abs_diff_pic_num= get_ue_golomb(&s->gb) + 1;
                        const unsigned int abs_diff_pic_num= get_ue_golomb(&h->gb) + 1;
                        int frame_num;
                        if(abs_diff_pic_num > h->max_pic_num){
                            av_log(h->s.avctx, AV_LOG_ERROR, "abs_diff_pic_num overflow\n");
                            av_log(h->avctx, AV_LOG_ERROR, "abs_diff_pic_num overflow\n");
                            return -1;
                        }
@@ -242,12 +238,12 @@ int ff_h264_decode_ref_pic_list_reordering(H264Context *h){
                            ref->pic_id= pred;
                    }else{
                        int long_idx;
                        pic_id= get_ue_golomb(&s->gb); //long_term_pic_idx
                        pic_id= get_ue_golomb(&h->gb); //long_term_pic_idx
                        long_idx= pic_num_extract(h, pic_id, &pic_structure);
                        if(long_idx>31){
                            av_log(h->s.avctx, AV_LOG_ERROR, "long_term_pic_idx overflow\n");
                            av_log(h->avctx, AV_LOG_ERROR, "long_term_pic_idx overflow\n");
                            return -1;
                        }
                        ref = h->long_ref[long_idx];
@@ -262,7 +258,7 @@ int ff_h264_decode_ref_pic_list_reordering(H264Context *h){
                    }
                    if (i < 0) {
                        av_log(h->s.avctx, AV_LOG_ERROR, "reference picture missing during reorder\n");
                        av_log(h->avctx, AV_LOG_ERROR, "reference picture missing during reorder\n");
                        memset(&h->ref_list[list][index], 0, sizeof(Picture)); //FIXME
                    } else {
                        for(i=index; i+1<h->ref_count[list]; i++){
@@ -278,7 +274,7 @@ int ff_h264_decode_ref_pic_list_reordering(H264Context *h){
                        }
                    }
                }else{
                    av_log(h->s.avctx, AV_LOG_ERROR, "illegal reordering_of_pic_nums_idc\n");
                    av_log(h->avctx, AV_LOG_ERROR, "illegal reordering_of_pic_nums_idc\n");
                    return -1;
                }
            }
@@ -288,7 +284,7 @@ int ff_h264_decode_ref_pic_list_reordering(H264Context *h){
        for(index= 0; index < h->ref_count[list]; index++){
            if (!h->ref_list[list][index].f.data[0]) {
                int i;
                av_log(h->s.avctx, AV_LOG_ERROR, "Missing reference picture, default is %d\n", h->default_ref_list[list][0].poc);
                av_log(h->avctx, AV_LOG_ERROR, "Missing reference picture, default is %d\n", h->default_ref_list[list][0].poc);
                for (i=0; i<FF_ARRAY_ELEMS(h->last_pocs); i++)
                    h->last_pocs[i] = INT_MIN;
                if (h->default_ref_list[list][0].f.data[0])
@@ -363,13 +359,12 @@ static inline int unreference_pic(H264Context *h, Picture *pic, int refmask){
 *                 frame number is found
 */
 static Picture * find_short(H264Context *h, int frame_num, int *idx){
    MpegEncContext * const s = &h->s;
    int i;
    for(i=0; i<h->short_ref_count; i++){
        Picture *pic= h->short_ref[i];
        if(s->avctx->debug&FF_DEBUG_MMCO)
            av_log(h->s.avctx, AV_LOG_DEBUG, "%d %d %p\n", i, pic->frame_num, pic);
        if(h->avctx->debug&FF_DEBUG_MMCO)
            av_log(h->avctx, AV_LOG_DEBUG, "%d %d %p\n", i, pic->frame_num, pic);
        if(pic->frame_num == frame_num) {
            *idx = i;
            return pic;
@@ -396,12 +391,11 @@ static void remove_short_at_index(H264Context *h, int i){
 * @return the removed picture or NULL if an error occurs
 */
 static Picture * remove_short(H264Context *h, int frame_num, int ref_mask){
    MpegEncContext * const s = &h->s;
    Picture *pic;
    int i;
    if(s->avctx->debug&FF_DEBUG_MMCO)
        av_log(h->s.avctx, AV_LOG_DEBUG, "remove short %d count %d\n", frame_num, h->short_ref_count);
    if(h->avctx->debug&FF_DEBUG_MMCO)
        av_log(h->avctx, AV_LOG_DEBUG, "remove short %d count %d\n", frame_num, h->short_ref_count);
    pic = find_short(h, frame_num, &i);
    if (pic){
@@ -456,11 +450,11 @@ void ff_h264_remove_all_refs(H264Context *h){
 */
 static void print_short_term(H264Context *h) {
    uint32_t i;
    if(h->s.avctx->debug&FF_DEBUG_MMCO) {
        av_log(h->s.avctx, AV_LOG_DEBUG, "short term list:\n");
    if(h->avctx->debug&FF_DEBUG_MMCO) {
        av_log(h->avctx, AV_LOG_DEBUG, "short term list:\n");
        for(i=0; i<h->short_ref_count; i++){
            Picture *pic= h->short_ref[i];
            av_log(h->s.avctx, AV_LOG_DEBUG, "%d fn:%d poc:%d %p\n",
            av_log(h->avctx, AV_LOG_DEBUG, "%d fn:%d poc:%d %p\n",
                   i, pic->frame_num, pic->poc, pic->f.data[0]);
        }
    }
@@ -471,12 +465,12 @@ static void print_short_term(H264Context *h) {
 */
 static void print_long_term(H264Context *h) {
    uint32_t i;
    if(h->s.avctx->debug&FF_DEBUG_MMCO) {
        av_log(h->s.avctx, AV_LOG_DEBUG, "long term list:\n");
    if(h->avctx->debug&FF_DEBUG_MMCO) {
        av_log(h->avctx, AV_LOG_DEBUG, "long term list:\n");
        for(i = 0; i < 16; i++){
            Picture *pic= h->long_ref[i];
            if (pic) {
                av_log(h->s.avctx, AV_LOG_DEBUG, "%d fn:%d poc:%d %p\n",
                av_log(h->avctx, AV_LOG_DEBUG, "%d fn:%d poc:%d %p\n",
                       i, pic->frame_num, pic->poc, pic->f.data[0]);
            }
        }
@@ -500,14 +494,12 @@ static int check_opcodes(MMCO *mmco1, MMCO *mmco2, int n_mmcos)
 int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice)
 {
    MpegEncContext * const s = &h->s;
    MMCO mmco_temp[MAX_MMCO_COUNT], *mmco = first_slice ? h->mmco : mmco_temp;
    int mmco_index = 0, i;
    if (h->short_ref_count &&
        h->long_ref_count + h->short_ref_count >= h->sps.ref_frame_count &&
        !(FIELD_PICTURE && !s->first_field &&
          s->current_picture_ptr->f.reference)) {
        !(FIELD_PICTURE && !h->first_field && h->cur_pic_ptr->f.reference)) {
        mmco[0].opcode = MMCO_SHORT2UNUSED;
        mmco[0].short_pic_num = h->short_ref[h->short_ref_count - 1]->frame_num;
        mmco_index = 1;
@@ -524,7 +516,7 @@ int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice)
    } else if (!first_slice && mmco_index >= 0 &&
               (mmco_index != h->mmco_index ||
                (i = check_opcodes(h->mmco, mmco_temp, mmco_index)))) {
        av_log(h->s.avctx, AV_LOG_ERROR,
        av_log(h->avctx, AV_LOG_ERROR,
               "Inconsistent MMCO state between slices [%d, %d]\n",
               mmco_index, h->mmco_index);
        return AVERROR_INVALIDDATA;
@@ -533,18 +525,17 @@ int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice)
 }
 int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
    MpegEncContext * const s = &h->s;
    int i, av_uninit(j);
    int current_ref_assigned=0, err=0;
    Picture *av_uninit(pic);
    if((s->avctx->debug&FF_DEBUG_MMCO) && mmco_count==0)
        av_log(h->s.avctx, AV_LOG_DEBUG, "no mmco here\n");
    if((h->avctx->debug&FF_DEBUG_MMCO) && mmco_count==0)
        av_log(h->avctx, AV_LOG_DEBUG, "no mmco here\n");
    for(i=0; i<mmco_count; i++){
        int av_uninit(structure), av_uninit(frame_num);
        if(s->avctx->debug&FF_DEBUG_MMCO)
            av_log(h->s.avctx, AV_LOG_DEBUG, "mmco:%d %d %d\n", h->mmco[i].opcode, h->mmco[i].short_pic_num, h->mmco[i].long_arg);
        if(h->avctx->debug&FF_DEBUG_MMCO)
            av_log(h->avctx, AV_LOG_DEBUG, "mmco:%d %d %d\n", h->mmco[i].opcode, h->mmco[i].short_pic_num, h->mmco[i].long_arg);
        if(   mmco[i].opcode == MMCO_SHORT2UNUSED
           || mmco[i].opcode == MMCO_SHORT2LONG){
@@ -553,7 +544,7 @@ int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
            if(!pic){
                if(mmco[i].opcode != MMCO_SHORT2LONG || !h->long_ref[mmco[i].long_arg]
                   || h->long_ref[mmco[i].long_arg]->frame_num != frame_num) {
                    av_log(h->s.avctx, AV_LOG_ERROR, "mmco: unref short failure\n");
                    av_log(h->avctx, AV_LOG_ERROR, "mmco: unref short failure\n");
                    err = AVERROR_INVALIDDATA;
                }
                continue;
@@ -562,8 +553,8 @@ int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
        switch(mmco[i].opcode){
        case MMCO_SHORT2UNUSED:
            if(s->avctx->debug&FF_DEBUG_MMCO)
                av_log(h->s.avctx, AV_LOG_DEBUG, "mmco: unref short %d count %d\n", h->mmco[i].short_pic_num, h->short_ref_count);
            if(h->avctx->debug&FF_DEBUG_MMCO)
                av_log(h->avctx, AV_LOG_DEBUG, "mmco: unref short %d count %d\n", h->mmco[i].short_pic_num, h->short_ref_count);
            remove_short(h, frame_num, structure ^ PICT_FRAME);
            break;
        case MMCO_SHORT2LONG:
@@ -582,8 +573,8 @@ int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
            pic = h->long_ref[j];
            if (pic) {
                remove_long(h, j, structure ^ PICT_FRAME);
            } else if(s->avctx->debug&FF_DEBUG_MMCO)
                av_log(h->s.avctx, AV_LOG_DEBUG, "mmco: unref long failure\n");
            } else if(h->avctx->debug&FF_DEBUG_MMCO)
                av_log(h->avctx, AV_LOG_DEBUG, "mmco: unref long failure\n");
            break;
        case MMCO_LONG:
                    // Comment below left from previous code as it is an interresting note.
@@ -594,15 +585,15 @@ int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
                     * and mark this field valid.
                     */
            if (h->long_ref[mmco[i].long_arg] != s->current_picture_ptr) {
            if (h->long_ref[mmco[i].long_arg] != h->cur_pic_ptr) {
                remove_long(h, mmco[i].long_arg, 0);
                h->long_ref[ mmco[i].long_arg ]= s->current_picture_ptr;
                h->long_ref[ mmco[i].long_arg ]= h->cur_pic_ptr;
                h->long_ref[ mmco[i].long_arg ]->long_ref=1;
                h->long_ref_count++;
            }
            s->current_picture_ptr->f.reference |= s->picture_structure;
            h->cur_pic_ptr->f.reference |= h->picture_structure;
            current_ref_assigned=1;
            break;
        case MMCO_SET_MAX_LONG:
@@ -620,9 +611,9 @@ int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
                remove_long(h, j, 0);
            }
            h->frame_num=
            s->current_picture_ptr->frame_num= 0;
            h->cur_pic_ptr->frame_num= 0;
            h->mmco_reset = 1;
            s->current_picture_ptr->mmco_reset=1;
            h->cur_pic_ptr->mmco_reset=1;
            for (j = 0; j < MAX_DELAYED_PIC_COUNT; j++)
                h->last_pocs[j] = INT_MIN;
            break;
@@ -637,28 +628,28 @@ int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
         * in long_ref; trying to put it on the short list here is an
         * error in the encoded bit stream (ref: 7.4.3.3, NOTE 2 and 3).
         */
        if (h->short_ref_count && h->short_ref[0] == s->current_picture_ptr) {
        if (h->short_ref_count && h->short_ref[0] == h->cur_pic_ptr) {
            /* Just mark the second field valid */
            s->current_picture_ptr->f.reference = PICT_FRAME;
        } else if (s->current_picture_ptr->long_ref) {
            av_log(h->s.avctx, AV_LOG_ERROR, "illegal short term reference "
                                             "assignment for second field "
                                             "in complementary field pair "
                                             "(first field is long term)\n");
            h->cur_pic_ptr->f.reference = PICT_FRAME;
        } else if (h->cur_pic_ptr->long_ref) {
            av_log(h->avctx, AV_LOG_ERROR, "illegal short term reference "
                                           "assignment for second field "
                                           "in complementary field pair "
                                           "(first field is long term)\n");
            err = AVERROR_INVALIDDATA;
        } else {
            pic= remove_short(h, s->current_picture_ptr->frame_num, 0);
            pic= remove_short(h, h->cur_pic_ptr->frame_num, 0);
            if(pic){
                av_log(h->s.avctx, AV_LOG_ERROR, "illegal short term buffer state detected\n");
                av_log(h->avctx, AV_LOG_ERROR, "illegal short term buffer state detected\n");
                err = AVERROR_INVALIDDATA;
            }
            if(h->short_ref_count)
                memmove(&h->short_ref[1], &h->short_ref[0], h->short_ref_count*sizeof(Picture*));
            h->short_ref[0]= s->current_picture_ptr;
            h->short_ref[0]= h->cur_pic_ptr;
            h->short_ref_count++;
            s->current_picture_ptr->f.reference |= s->picture_structure;
            h->cur_pic_ptr->f.reference |= h->picture_structure;
        }
    }
@@ -668,7 +659,7 @@ int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
         * stream. Need to discard one frame. Prevents overrun of the
         * short_ref and long_ref buffers.
         */
        av_log(h->s.avctx, AV_LOG_ERROR,
        av_log(h->avctx, AV_LOG_ERROR,
               "number of reference frames (%d+%d) exceeds max (%d; probably "
               "corrupt input), discarding one\n",
               h->long_ref_count, h->short_ref_count, h->sps.ref_frame_count);
@@ -690,25 +681,24 @@ int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count){
    print_short_term(h);
    print_long_term(h);
    if(err >= 0 && h->long_ref_count==0 && h->short_ref_count<=2 && h->pps.ref_count[0]<=1 + (s->picture_structure != PICT_FRAME) && s->current_picture_ptr->f.pict_type == AV_PICTURE_TYPE_I){
        s->current_picture_ptr->sync |= 1;
        if(!h->s.avctx->has_b_frames)
    if(err >= 0 && h->long_ref_count==0 && h->short_ref_count<=2 && h->pps.ref_count[0]<=1 + (h->picture_structure != PICT_FRAME) && h->cur_pic_ptr->f.pict_type == AV_PICTURE_TYPE_I){
        h->cur_pic_ptr->sync |= 1;
        if(!h->avctx->has_b_frames)
            h->sync = 2;
    }
    return (h->s.avctx->err_recognition & AV_EF_EXPLODE) ? err : 0;
    return (h->avctx->err_recognition & AV_EF_EXPLODE) ? err : 0;
 }
 int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
                                   int first_slice)
 {
    MpegEncContext * const s = &h->s;
    int i, ret;
    MMCO mmco_temp[MAX_MMCO_COUNT], *mmco = first_slice ? h->mmco : mmco_temp;
    int mmco_index = 0;
    if (h->nal_unit_type == NAL_IDR_SLICE){ // FIXME fields
        s->broken_link = get_bits1(gb) - 1;
        skip_bits1(gb); // broken_link
        if (get_bits1(gb)){
            mmco[0].opcode = MMCO_LONG;
            mmco[0].long_arg = 0;
@@ -741,7 +731,7 @@ int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
                        (long_arg >= 16 && !(opcode == MMCO_SET_MAX_LONG &&
                                             long_arg == 16) &&
                         !(opcode == MMCO_LONG2UNUSED && FIELD_PICTURE))){
                        av_log(h->s.avctx, AV_LOG_ERROR,
                        av_log(h->avctx, AV_LOG_ERROR,
                               "illegal long ref in memory management control "
                               "operation %d\n", opcode);
                        return -1;
@@ -750,7 +740,7 @@ int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
                }
                if (opcode > (unsigned) MMCO_LONG){
                    av_log(h->s.avctx, AV_LOG_ERROR,
                    av_log(h->avctx, AV_LOG_ERROR,
                           "illegal memory management control operation %d\n",
                           opcode);
                    return -1;
@@ -762,7 +752,7 @@ int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
        } else {
            if (first_slice) {
                ret = ff_generate_sliding_window_mmcos(h, first_slice);
                if (ret < 0 && s->avctx->err_recognition & AV_EF_EXPLODE)
                if (ret < 0 && h->avctx->err_recognition & AV_EF_EXPLODE)
                    return ret;
            }
            mmco_index = -1;
@@ -774,7 +764,7 @@ int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb,
    } else if (!first_slice && mmco_index >= 0 &&
               (mmco_index != h->mmco_index ||
                (i = check_opcodes(h->mmco, mmco_temp, mmco_index)))) {
        av_log(h->s.avctx, AV_LOG_ERROR,
        av_log(h->avctx, AV_LOG_ERROR,
               "Inconsistent MMCO state between slices [%d, %d]\n",
               mmco_index, h->mmco_index);
        return AVERROR_INVALIDDATA;
--- a/libavcodec/h264_sei.c
+++ b/libavcodec/h264_sei.c
@@ -45,14 +45,13 @@ void ff_h264_reset_sei(H264Context *h) {
 }
 static int decode_picture_timing(H264Context *h){
    MpegEncContext * const s = &h->s;
    if(h->sps.nal_hrd_parameters_present_flag || h->sps.vcl_hrd_parameters_present_flag){
        h->sei_cpb_removal_delay = get_bits_long(&s->gb, h->sps.cpb_removal_delay_length);
        h->sei_dpb_output_delay = get_bits_long(&s->gb, h->sps.dpb_output_delay_length);
        h->sei_cpb_removal_delay = get_bits_long(&h->gb, h->sps.cpb_removal_delay_length);
        h->sei_dpb_output_delay = get_bits_long(&h->gb, h->sps.dpb_output_delay_length);
    }
    if(h->sps.pic_struct_present_flag){
        unsigned int i, num_clock_ts;
        h->sei_pic_struct = get_bits(&s->gb, 4);
        h->sei_pic_struct = get_bits(&h->gb, 4);
        h->sei_ct_type    = 0;
        if (h->sei_pic_struct > SEI_PIC_STRUCT_FRAME_TRIPLING)
@@ -61,42 +60,41 @@ static int decode_picture_timing(H264Context *h){
        num_clock_ts = sei_num_clock_ts_table[h->sei_pic_struct];
        for (i = 0 ; i < num_clock_ts ; i++){
            if(get_bits(&s->gb, 1)){                  /* clock_timestamp_flag */
            if(get_bits(&h->gb, 1)){                  /* clock_timestamp_flag */
                unsigned int full_timestamp_flag;
                h->sei_ct_type |= 1<<get_bits(&s->gb, 2);
                skip_bits(&s->gb, 1);                 /* nuit_field_based_flag */
                skip_bits(&s->gb, 5);                 /* counting_type */
                full_timestamp_flag = get_bits(&s->gb, 1);
                skip_bits(&s->gb, 1);                 /* discontinuity_flag */
                skip_bits(&s->gb, 1);                 /* cnt_dropped_flag */
                skip_bits(&s->gb, 8);                 /* n_frames */
                h->sei_ct_type |= 1<<get_bits(&h->gb, 2);
                skip_bits(&h->gb, 1);                 /* nuit_field_based_flag */
                skip_bits(&h->gb, 5);                 /* counting_type */
                full_timestamp_flag = get_bits(&h->gb, 1);
                skip_bits(&h->gb, 1);                 /* discontinuity_flag */
                skip_bits(&h->gb, 1);                 /* cnt_dropped_flag */
                skip_bits(&h->gb, 8);                 /* n_frames */
                if(full_timestamp_flag){
                    skip_bits(&s->gb, 6);             /* seconds_value 0..59 */
                    skip_bits(&s->gb, 6);             /* minutes_value 0..59 */
                    skip_bits(&s->gb, 5);             /* hours_value 0..23 */
                    skip_bits(&h->gb, 6);             /* seconds_value 0..59 */
                    skip_bits(&h->gb, 6);             /* minutes_value 0..59 */
                    skip_bits(&h->gb, 5);             /* hours_value 0..23 */
                }else{
                    if(get_bits(&s->gb, 1)){          /* seconds_flag */
                        skip_bits(&s->gb, 6);         /* seconds_value range 0..59 */
                        if(get_bits(&s->gb, 1)){      /* minutes_flag */
                            skip_bits(&s->gb, 6);     /* minutes_value 0..59 */
                            if(get_bits(&s->gb, 1))   /* hours_flag */
                                skip_bits(&s->gb, 5); /* hours_value 0..23 */
                    if(get_bits(&h->gb, 1)){          /* seconds_flag */
                        skip_bits(&h->gb, 6);         /* seconds_value range 0..59 */
                        if(get_bits(&h->gb, 1)){      /* minutes_flag */
                            skip_bits(&h->gb, 6);     /* minutes_value 0..59 */
                            if(get_bits(&h->gb, 1))   /* hours_flag */
                                skip_bits(&h->gb, 5); /* hours_value 0..23 */
                        }
                    }
                }
                if(h->sps.time_offset_length > 0)
                    skip_bits(&s->gb, h->sps.time_offset_length); /* time_offset */
                    skip_bits(&h->gb, h->sps.time_offset_length); /* time_offset */
            }
        }
        if(s->avctx->debug & FF_DEBUG_PICT_INFO)
            av_log(s->avctx, AV_LOG_DEBUG, "ct_type:%X pic_struct:%d\n", h->sei_ct_type, h->sei_pic_struct);
        if(h->avctx->debug & FF_DEBUG_PICT_INFO)
            av_log(h->avctx, AV_LOG_DEBUG, "ct_type:%X pic_struct:%d\n", h->sei_ct_type, h->sei_pic_struct);
    }
    return 0;
 }
 static int decode_user_data_itu_t_t35(H264Context *h, int size) {
    MpegEncContext * const s = &h->s;
    uint32_t user_identifier;
    int dtg_active_format;
@@ -104,28 +102,28 @@ static int decode_user_data_itu_t_t35(H264Context *h, int size) {
        return -1;
    size -= 7;
    skip_bits(&s->gb, 8);   // country_code
    skip_bits(&s->gb, 16);  // provider_code
    user_identifier = get_bits_long(&s->gb, 32);
    skip_bits(&h->gb, 8);   // country_code
    skip_bits(&h->gb, 16);  // provider_code
    user_identifier = get_bits_long(&h->gb, 32);
    switch (user_identifier) {
        case 0x44544731:    // "DTG1" - AFD_data
            if (size < 1)
                return -1;
            skip_bits(&s->gb, 1);
            if (get_bits(&s->gb, 1)) {
                skip_bits(&s->gb, 6);
            skip_bits(&h->gb, 1);
            if (get_bits(&h->gb, 1)) {
                skip_bits(&h->gb, 6);
                if (size < 2)
                    return -1;
                skip_bits(&s->gb, 4);
                dtg_active_format = get_bits(&s->gb, 4);
                s->avctx->dtg_active_format = dtg_active_format;
                skip_bits(&h->gb, 4);
                dtg_active_format = get_bits(&h->gb, 4);
                h->avctx->dtg_active_format = dtg_active_format;
            } else {
                skip_bits(&s->gb, 6);
                skip_bits(&h->gb, 6);
            }
            break;
        default:
            skip_bits(&s->gb, size * 8);
            skip_bits(&h->gb, size * 8);
            break;
    }
@@ -133,7 +131,6 @@ static int decode_user_data_itu_t_t35(H264Context *h, int size) {
 }
 static int decode_unregistered_user_data(H264Context *h, int size){
    MpegEncContext * const s = &h->s;
    uint8_t user_data[16+256];
    int e, build, i;
@@ -141,7 +138,7 @@ static int decode_unregistered_user_data(H264Context *h, int size){
        return -1;
    for(i=0; i<sizeof(user_data)-1 && i<size; i++){
        user_data[i]= get_bits(&s->gb, 8);
        user_data[i]= get_bits(&h->gb, 8);
    }
    user_data[i]= 0;
@@ -151,33 +148,30 @@ static int decode_unregistered_user_data(H264Context *h, int size){
    if(e==1 && build==1 && !strncmp(user_data+16, "x264 - core 0000", 16))
        h->x264_build = 67;
    if(s->avctx->debug & FF_DEBUG_BUGS)
        av_log(s->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data+16);
    if(h->avctx->debug & FF_DEBUG_BUGS)
        av_log(h->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data+16);
    for(; i<size; i++)
        skip_bits(&s->gb, 8);
        skip_bits(&h->gb, 8);
    return 0;
 }
 static int decode_recovery_point(H264Context *h){
    MpegEncContext * const s = &h->s;
    h->sei_recovery_frame_cnt = get_ue_golomb(&s->gb);
    skip_bits(&s->gb, 4);       /* 1b exact_match_flag, 1b broken_link_flag, 2b changing_slice_group_idc */
    h->sei_recovery_frame_cnt = get_ue_golomb(&h->gb);
    skip_bits(&h->gb, 4);       /* 1b exact_match_flag, 1b broken_link_flag, 2b changing_slice_group_idc */
    return 0;
 }
 static int decode_buffering_period(H264Context *h){
    MpegEncContext * const s = &h->s;
    unsigned int sps_id;
    int sched_sel_idx;
    SPS *sps;
    sps_id = get_ue_golomb_31(&s->gb);
    sps_id = get_ue_golomb_31(&h->gb);
    if(sps_id > 31 || !h->sps_buffers[sps_id]) {
        av_log(h->s.avctx, AV_LOG_ERROR, "non-existing SPS %d referenced in buffering period\n", sps_id);
        av_log(h->avctx, AV_LOG_ERROR, "non-existing SPS %d referenced in buffering period\n", sps_id);
        return -1;
    }
    sps = h->sps_buffers[sps_id];
@@ -185,14 +179,14 @@ static int decode_buffering_period(H264Context *h){
    // NOTE: This is really so duplicated in the standard... See H.264, D.1.1
    if (sps->nal_hrd_parameters_present_flag) {
        for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
            h->initial_cpb_removal_delay[sched_sel_idx] = get_bits_long(&s->gb, sps->initial_cpb_removal_delay_length);
            skip_bits(&s->gb, sps->initial_cpb_removal_delay_length); // initial_cpb_removal_delay_offset
            h->initial_cpb_removal_delay[sched_sel_idx] = get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
            skip_bits(&h->gb, sps->initial_cpb_removal_delay_length); // initial_cpb_removal_delay_offset
        }
    }
    if (sps->vcl_hrd_parameters_present_flag) {
        for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
            h->initial_cpb_removal_delay[sched_sel_idx] = get_bits_long(&s->gb, sps->initial_cpb_removal_delay_length);
            skip_bits(&s->gb, sps->initial_cpb_removal_delay_length); // initial_cpb_removal_delay_offset
            h->initial_cpb_removal_delay[sched_sel_idx] = get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
            skip_bits(&h->gb, sps->initial_cpb_removal_delay_length); // initial_cpb_removal_delay_offset
        }
    }
@@ -201,27 +195,25 @@ static int decode_buffering_period(H264Context *h){
 }
 int ff_h264_decode_sei(H264Context *h){
    MpegEncContext * const s = &h->s;
    while (get_bits_left(&s->gb) > 16) {
    while (get_bits_left(&h->gb) > 16) {
        int size, type;
        type=0;
        do{
            if (get_bits_left(&s->gb) < 8)
            if (get_bits_left(&h->gb) < 8)
                return -1;
            type+= show_bits(&s->gb, 8);
        }while(get_bits(&s->gb, 8) == 255);
            type+= show_bits(&h->gb, 8);
        }while(get_bits(&h->gb, 8) == 255);
        size=0;
        do{
            if (get_bits_left(&s->gb) < 8)
            if (get_bits_left(&h->gb) < 8)
                return -1;
            size+= show_bits(&s->gb, 8);
        }while(get_bits(&s->gb, 8) == 255);
            size+= show_bits(&h->gb, 8);
        }while(get_bits(&h->gb, 8) == 255);
        if(s->avctx->debug&FF_DEBUG_STARTCODE)
            av_log(h->s.avctx, AV_LOG_DEBUG, "SEI %d len:%d\n", type, size);
        if(h->avctx->debug&FF_DEBUG_STARTCODE)
            av_log(h->avctx, AV_LOG_DEBUG, "SEI %d len:%d\n", type, size);
        switch(type){
        case SEI_TYPE_PIC_TIMING: // Picture timing SEI
@@ -245,11 +237,11 @@ int ff_h264_decode_sei(H264Context *h){
                return -1;
            break;
        default:
            skip_bits(&s->gb, 8*size);
            skip_bits(&h->gb, 8*size);
        }
        //FIXME check bits here
        align_get_bits(&s->gb);
        align_get_bits(&h->gb);
    }
    return 0;
--- a/libavcodec/h264pred.c
+++ b/libavcodec/h264pred.c
@@ -410,8 +410,6 @@ static void pred8x8_tm_vp8_c(uint8_t *src, ptrdiff_t stride)
 void ff_h264_pred_init(H264PredContext *h, int codec_id, const int bit_depth,
                       int chroma_format_idc)
 {
 //    MpegEncContext * const s = &h->s;
 #undef FUNC
 #undef FUNCC
 #define FUNC(a, depth) a ## _ ## depth
--- a/libavcodec/mpegvideo.h
+++ b/libavcodec/mpegvideo.h
@@ -143,7 +143,7 @@ typedef struct Picture{
    uint16_t *mc_mb_var;        ///< Table for motion compensated MB variances
    uint8_t *mb_mean;           ///< Table for MB luminance
    int b_frame_score;          /* */
    struct MpegEncContext *owner2; ///< pointer to the MpegEncContext that allocated this picture
    void *owner2;               ///< pointer to the context that allocated this picture
    int needs_realloc;          ///< Picture needs to be reallocated (eg due to a frame size change)
    int period_since_free;      ///< "cycles" since this Picture has been freed
 } Picture;
--- a/libavcodec/svq3.c
+++ b/libavcodec/svq3.c
--- a/libavcodec/vaapi.c
+++ b/libavcodec/vaapi.c
@@ -21,6 +21,7 @@
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */
 #include "h264.h"
 #include "vaapi_internal.h"
 /**
@@ -40,7 +41,7 @@ static void destroy_buffers(VADisplay display, VABufferID *buffers, unsigned int
    }
 }
 static int render_picture(struct vaapi_context *vactx, VASurfaceID surface)
 int ff_vaapi_render_picture(struct vaapi_context *vactx, VASurfaceID surface)
 {
    VABufferID va_buffers[3];
    unsigned int n_va_buffers = 0;
@@ -77,7 +78,7 @@ static int render_picture(struct vaapi_context *vactx, VASurfaceID surface)
    return 0;
 }
 static int commit_slices(struct vaapi_context *vactx)
 int ff_vaapi_commit_slices(struct vaapi_context *vactx)
 {
    VABufferID *slice_buf_ids;
    VABufferID slice_param_buf_id, slice_data_buf_id;
@@ -152,7 +153,7 @@ VASliceParameterBufferBase *ff_vaapi_alloc_slice(struct vaapi_context *vactx, co
    if (!vactx->slice_data)
        vactx->slice_data = buffer;
    if (vactx->slice_data + vactx->slice_data_size != buffer) {
        if (commit_slices(vactx) < 0)
        if (ff_vaapi_commit_slices(vactx) < 0)
            return NULL;
        vactx->slice_data = buffer;
    }
@@ -175,23 +176,12 @@ VASliceParameterBufferBase *ff_vaapi_alloc_slice(struct vaapi_context *vactx, co
    return slice_param;
 }
 int ff_vaapi_common_end_frame(MpegEncContext *s)
 void ff_vaapi_common_end_frame(AVCodecContext *avctx)
 {
    struct vaapi_context * const vactx = s->avctx->hwaccel_context;
    int ret = -1;
    struct vaapi_context * const vactx = avctx->hwaccel_context;
    av_dlog(s->avctx, "ff_vaapi_common_end_frame()\n");
    av_dlog(avctx, "ff_vaapi_common_end_frame()\n");
    if (commit_slices(vactx) < 0)
        goto done;
    if (vactx->n_slice_buf_ids > 0) {
        if (render_picture(vactx, ff_vaapi_get_surface_id(s->current_picture_ptr)) < 0)
            goto done;
        ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
    }
    ret = 0;
 done:
    destroy_buffers(vactx->display, &vactx->pic_param_buf_id, 1);
    destroy_buffers(vactx->display, &vactx->iq_matrix_buf_id, 1);
    destroy_buffers(vactx->display, &vactx->bitplane_buf_id, 1);
@@ -202,6 +192,27 @@ done:
    vactx->slice_buf_ids_alloc = 0;
    vactx->slice_count         = 0;
    vactx->slice_params_alloc  = 0;
 }
 int ff_vaapi_mpeg_end_frame(AVCodecContext *avctx)
 {
    struct vaapi_context * const vactx = avctx->hwaccel_context;
    MpegEncContext *s = avctx->priv_data;
    int ret;
    ret = ff_vaapi_commit_slices(vactx);
    if (ret < 0)
        goto finish;
    ret = ff_vaapi_render_picture(vactx,
                                  ff_vaapi_get_surface_id(s->current_picture_ptr));
    if (ret < 0)
        goto finish;
    ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
 finish:
    ff_vaapi_common_end_frame(avctx->priv_data);
    return ret;
 }
--- a/libavcodec/vaapi_h264.c
+++ b/libavcodec/vaapi_h264.c
@@ -224,7 +224,6 @@ static int start_frame(AVCodecContext          *avctx,
                       av_unused uint32_t       size)
 {
    H264Context * const h = avctx->priv_data;
    MpegEncContext * const s = &h->s;
    struct vaapi_context * const vactx = avctx->hwaccel_context;
    VAPictureParameterBufferH264 *pic_param;
    VAIQMatrixBufferH264 *iq_matrix;
@@ -237,11 +236,11 @@ static int start_frame(AVCodecContext          *avctx,
    pic_param = ff_vaapi_alloc_pic_param(vactx, sizeof(VAPictureParameterBufferH264));
    if (!pic_param)
        return -1;
    fill_vaapi_pic(&pic_param->CurrPic, s->current_picture_ptr, s->picture_structure);
    fill_vaapi_pic(&pic_param->CurrPic, h->cur_pic_ptr, h->picture_structure);
    if (fill_vaapi_ReferenceFrames(pic_param, h) < 0)
        return -1;
    pic_param->picture_width_in_mbs_minus1                      = s->mb_width - 1;
    pic_param->picture_height_in_mbs_minus1                     = s->mb_height - 1;
    pic_param->picture_width_in_mbs_minus1                      = h->mb_width - 1;
    pic_param->picture_height_in_mbs_minus1                     = h->mb_height - 1;
    pic_param->bit_depth_luma_minus8                            = h->sps.bit_depth_luma - 8;
    pic_param->bit_depth_chroma_minus8                          = h->sps.bit_depth_chroma - 8;
    pic_param->num_ref_frames                                   = h->sps.ref_frame_count;
@@ -269,7 +268,7 @@ static int start_frame(AVCodecContext          *avctx,
    pic_param->pic_fields.bits.weighted_pred_flag               = h->pps.weighted_pred;
    pic_param->pic_fields.bits.weighted_bipred_idc              = h->pps.weighted_bipred_idc;
    pic_param->pic_fields.bits.transform_8x8_mode_flag          = h->pps.transform_8x8_mode;
    pic_param->pic_fields.bits.field_pic_flag                   = s->picture_structure != PICT_FRAME;
    pic_param->pic_fields.bits.field_pic_flag                   = h->picture_structure != PICT_FRAME;
    pic_param->pic_fields.bits.constrained_intra_pred_flag      = h->pps.constrained_intra_pred;
    pic_param->pic_fields.bits.pic_order_present_flag           = h->pps.pic_order_present;
    pic_param->pic_fields.bits.deblocking_filter_control_present_flag = h->pps.deblocking_filter_parameters_present;
@@ -290,10 +289,24 @@ static int start_frame(AVCodecContext          *avctx,
 /** End a hardware decoding based frame. */
 static int end_frame(AVCodecContext *avctx)
 {
    struct vaapi_context * const vactx = avctx->hwaccel_context;
    H264Context * const h = avctx->priv_data;
    int ret;
    av_dlog(avctx, "end_frame()\n");
    return ff_vaapi_common_end_frame(&h->s);
    ret = ff_vaapi_commit_slices(vactx);
    if (ret < 0)
        goto finish;
    ret = ff_vaapi_render_picture(vactx, ff_vaapi_get_surface_id(h->cur_pic_ptr));
    if (ret < 0)
        goto finish;
    ff_h264_draw_horiz_band(h, 0, h->avctx->height);
 finish:
    ff_vaapi_common_end_frame(avctx);
    return ret;
 }
 /** Decode the given H.264 slice with VA API. */
@@ -302,7 +315,6 @@ static int decode_slice(AVCodecContext *avctx,
                        uint32_t        size)
 {
    H264Context * const h = avctx->priv_data;
    MpegEncContext * const s = &h->s;
    VASliceParameterBufferH264 *slice_param;
    av_dlog(avctx, "decode_slice(): buffer %p, size %d\n", buffer, size);
@@ -311,14 +323,14 @@ static int decode_slice(AVCodecContext *avctx,
    slice_param = (VASliceParameterBufferH264 *)ff_vaapi_alloc_slice(avctx->hwaccel_context, buffer, size);
    if (!slice_param)
        return -1;
    slice_param->slice_data_bit_offset          = get_bits_count(&h->s.gb) + 8; /* bit buffer started beyond nal_unit_type */
    slice_param->first_mb_in_slice              = (s->mb_y >> FIELD_OR_MBAFF_PICTURE) * s->mb_width + s->mb_x;
    slice_param->slice_data_bit_offset          = get_bits_count(&h->gb) + 8; /* bit buffer started beyond nal_unit_type */
    slice_param->first_mb_in_slice              = (h->mb_y >> FIELD_OR_MBAFF_PICTURE) * h->mb_width + h->mb_x;
    slice_param->slice_type                     = ff_h264_get_slice_type(h);
    slice_param->direct_spatial_mv_pred_flag    = h->slice_type == AV_PICTURE_TYPE_B ? h->direct_spatial_mv_pred : 0;
    slice_param->num_ref_idx_l0_active_minus1   = h->list_count > 0 ? h->ref_count[0] - 1 : 0;
    slice_param->num_ref_idx_l1_active_minus1   = h->list_count > 1 ? h->ref_count[1] - 1 : 0;
    slice_param->cabac_init_idc                 = h->cabac_init_idc;
    slice_param->slice_qp_delta                 = s->qscale - h->pps.init_qp;
    slice_param->slice_qp_delta                 = h->qscale - h->pps.init_qp;
    slice_param->disable_deblocking_filter_idc  = h->deblocking_filter < 2 ? !h->deblocking_filter : h->deblocking_filter;
    slice_param->slice_alpha_c0_offset_div2     = h->slice_alpha_c0_offset / 2 - 26;
    slice_param->slice_beta_offset_div2         = h->slice_beta_offset     / 2 - 26;
--- a/libavcodec/vaapi_internal.h
+++ b/libavcodec/vaapi_internal.h
@@ -42,7 +42,7 @@ static inline VASurfaceID ff_vaapi_get_surface_id(Picture *pic)
 }
 /** Common AVHWAccel.end_frame() implementation */
 int ff_vaapi_common_end_frame(MpegEncContext *s);
 void ff_vaapi_common_end_frame(AVCodecContext *avctx);
 /** Allocate a new picture parameter buffer */
 void *ff_vaapi_alloc_pic_param(struct vaapi_context *vactx, unsigned int size);
@@ -63,6 +63,10 @@ uint8_t *ff_vaapi_alloc_bitplane(struct vaapi_context *vactx, uint32_t size);
 */
 VASliceParameterBufferBase *ff_vaapi_alloc_slice(struct vaapi_context *vactx, const uint8_t *buffer, uint32_t size);
 int ff_vaapi_mpeg_end_frame(AVCodecContext *avctx);
 int ff_vaapi_commit_slices(struct vaapi_context *vactx);
 int ff_vaapi_render_picture(struct vaapi_context *vactx, VASurfaceID surface);
 /* @} */
 #endif /* AVCODEC_VAAPI_INTERNAL_H */
--- a/libavcodec/vaapi_mpeg2.c
+++ b/libavcodec/vaapi_mpeg2.c
@@ -99,11 +99,6 @@ static int vaapi_mpeg2_start_frame(AVCodecContext *avctx, av_unused const uint8_
    return 0;
 }
 static int vaapi_mpeg2_end_frame(AVCodecContext *avctx)
 {
    return ff_vaapi_common_end_frame(avctx->priv_data);
 }
 static int vaapi_mpeg2_decode_slice(AVCodecContext *avctx, const uint8_t *buffer, uint32_t size)
 {
    MpegEncContext * const s = avctx->priv_data;
@@ -144,6 +139,6 @@ AVHWAccel ff_mpeg2_vaapi_hwaccel = {
    .id             = AV_CODEC_ID_MPEG2VIDEO,
    .pix_fmt        = AV_PIX_FMT_VAAPI_VLD,
    .start_frame    = vaapi_mpeg2_start_frame,
    .end_frame      = vaapi_mpeg2_end_frame,
    .end_frame      = ff_vaapi_mpeg_end_frame,
    .decode_slice   = vaapi_mpeg2_decode_slice,
 };
--- a/libavcodec/vaapi_mpeg4.c
+++ b/libavcodec/vaapi_mpeg4.c
@@ -115,11 +115,6 @@ static int vaapi_mpeg4_start_frame(AVCodecContext *avctx, av_unused const uint8_
    return 0;
 }
 static int vaapi_mpeg4_end_frame(AVCodecContext *avctx)
 {
    return ff_vaapi_common_end_frame(avctx->priv_data);
 }
 static int vaapi_mpeg4_decode_slice(AVCodecContext *avctx, const uint8_t *buffer, uint32_t size)
 {
    MpegEncContext * const s = avctx->priv_data;
@@ -156,7 +151,7 @@ AVHWAccel ff_mpeg4_vaapi_hwaccel = {
    .id             = AV_CODEC_ID_MPEG4,
    .pix_fmt        = AV_PIX_FMT_VAAPI_VLD,
    .start_frame    = vaapi_mpeg4_start_frame,
    .end_frame      = vaapi_mpeg4_end_frame,
    .end_frame      = ff_vaapi_mpeg_end_frame,
    .decode_slice   = vaapi_mpeg4_decode_slice,
 };
 #endif
@@ -168,7 +163,7 @@ AVHWAccel ff_h263_vaapi_hwaccel = {
    .id             = AV_CODEC_ID_H263,
    .pix_fmt        = AV_PIX_FMT_VAAPI_VLD,
    .start_frame    = vaapi_mpeg4_start_frame,
    .end_frame      = vaapi_mpeg4_end_frame,
    .end_frame      = ff_vaapi_mpeg_end_frame,
    .decode_slice   = vaapi_mpeg4_decode_slice,
 };
 #endif
--- a/libavcodec/vaapi_vc1.c
+++ b/libavcodec/vaapi_vc1.c
@@ -310,13 +310,6 @@ static int vaapi_vc1_start_frame(AVCodecContext *avctx, av_unused const uint8_t
    return 0;
 }
 static int vaapi_vc1_end_frame(AVCodecContext *avctx)
 {
    VC1Context * const v = avctx->priv_data;
    return ff_vaapi_common_end_frame(&v->s);
 }
 static int vaapi_vc1_decode_slice(AVCodecContext *avctx, const uint8_t *buffer, uint32_t size)
 {
    VC1Context * const v = avctx->priv_data;
@@ -347,7 +340,7 @@ AVHWAccel ff_wmv3_vaapi_hwaccel = {
    .id             = AV_CODEC_ID_WMV3,
    .pix_fmt        = AV_PIX_FMT_VAAPI_VLD,
    .start_frame    = vaapi_vc1_start_frame,
    .end_frame      = vaapi_vc1_end_frame,
    .end_frame      = ff_vaapi_mpeg_end_frame,
    .decode_slice   = vaapi_vc1_decode_slice,
 };
 #endif
@@ -358,6 +351,6 @@ AVHWAccel ff_vc1_vaapi_hwaccel = {
    .id             = AV_CODEC_ID_VC1,
    .pix_fmt        = AV_PIX_FMT_VAAPI_VLD,
    .start_frame    = vaapi_vc1_start_frame,
    .end_frame      = vaapi_vc1_end_frame,
    .end_frame      = ff_vaapi_mpeg_end_frame,
    .decode_slice   = vaapi_vc1_decode_slice,
 };
--- a/libavcodec/vda_h264.c
+++ b/libavcodec/vda_h264.c
@@ -243,7 +243,7 @@ static int end_frame(AVCodecContext *avctx)
 {
    H264Context *h                      = avctx->priv_data;
    struct vda_context *vda_ctx         = avctx->hwaccel_context;
    AVFrame *frame                      = &h->s.current_picture_ptr->f;
    AVFrame *frame                      = &h->cur_pic_ptr->f;
    int status;
    if (!vda_ctx->decoder || !vda_ctx->priv_bitstream)
--- a/libavcodec/vdpau.c
+++ b/libavcodec/vdpau.c
@@ -48,20 +48,18 @@ int ff_vdpau_common_start_frame(AVCodecContext *avctx,
    return 0;
 }
 int ff_vdpau_common_end_frame(AVCodecContext *avctx)
 int ff_vdpau_mpeg_end_frame(AVCodecContext *avctx)
 {
    MpegEncContext * const s = avctx->priv_data;
    AVVDPAUContext *hwctx = avctx->hwaccel_context;
    MpegEncContext *s = avctx->priv_data;
    VdpVideoSurface surf = ff_vdpau_get_surface_id(s->current_picture_ptr);
    if (hwctx->bitstream_buffers_used) {
        VdpVideoSurface surf = ff_vdpau_get_surface_id(s->current_picture_ptr);
    hwctx->render(hwctx->decoder, surf, (void *)&hwctx->info,
                  hwctx->bitstream_buffers_used, hwctx->bitstream_buffers);
        hwctx->render(hwctx->decoder, surf, (void *)&hwctx->info,
                      hwctx->bitstream_buffers_used, hwctx->bitstream_buffers);
    ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
    hwctx->bitstream_buffers_used = 0;
        ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
        hwctx->bitstream_buffers_used = 0;
    }
    return 0;
 }
@@ -87,15 +85,14 @@ int ff_vdpau_add_buffer(AVCodecContext *avctx,
 /* Obsolete non-hwaccel VDPAU support below... */
 void ff_vdpau_h264_set_reference_frames(MpegEncContext *s)
 void ff_vdpau_h264_set_reference_frames(H264Context *h)
 {
    H264Context *h = s->avctx->priv_data;
    struct vdpau_render_state *render, *render_ref;
    VdpReferenceFrameH264 *rf, *rf2;
    Picture *pic;
    int i, list, pic_frame_idx;
    render = (struct vdpau_render_state *)s->current_picture_ptr->f.data[0];
    render = (struct vdpau_render_state *)h->cur_pic_ptr->f.data[0];
    assert(render);
    rf = &render->info.h264.referenceFrames[0];
@@ -156,12 +153,9 @@ void ff_vdpau_h264_set_reference_frames(MpegEncContext *s)
    }
 }
 void ff_vdpau_add_data_chunk(MpegEncContext *s,
                             const uint8_t *buf, int buf_size)
 void ff_vdpau_add_data_chunk(uint8_t *data, const uint8_t *buf, int buf_size)
 {
    struct vdpau_render_state *render;
    render = (struct vdpau_render_state *)s->current_picture_ptr->f.data[0];
    struct vdpau_render_state *render = (struct vdpau_render_state*)data;
    assert(render);
    render->bitstream_buffers= av_fast_realloc(
@@ -176,17 +170,16 @@ void ff_vdpau_add_data_chunk(MpegEncContext *s,
    render->bitstream_buffers_used++;
 }
 void ff_vdpau_h264_picture_start(MpegEncContext *s)
 void ff_vdpau_h264_picture_start(H264Context *h)
 {
    H264Context *h = s->avctx->priv_data;
    struct vdpau_render_state *render;
    int i;
    render = (struct vdpau_render_state *)s->current_picture_ptr->f.data[0];
    render = (struct vdpau_render_state *)h->cur_pic_ptr->f.data[0];
    assert(render);
    for (i = 0; i < 2; ++i) {
        int foc = s->current_picture_ptr->field_poc[i];
        int foc = h->cur_pic_ptr->field_poc[i];
        if (foc == INT_MAX)
            foc = 0;
        render->info.h264.field_order_cnt[i] = foc;
@@ -195,21 +188,20 @@ void ff_vdpau_h264_picture_start(MpegEncContext *s)
    render->info.h264.frame_num = h->frame_num;
 }
 void ff_vdpau_h264_picture_complete(MpegEncContext *s)
 void ff_vdpau_h264_picture_complete(H264Context *h)
 {
    H264Context *h = s->avctx->priv_data;
    struct vdpau_render_state *render;
    render = (struct vdpau_render_state *)s->current_picture_ptr->f.data[0];
    render = (struct vdpau_render_state *)h->cur_pic_ptr->f.data[0];
    assert(render);
    render->info.h264.slice_count = h->slice_num;
    if (render->info.h264.slice_count < 1)
        return;
    render->info.h264.is_reference                           = (s->current_picture_ptr->f.reference & 3) ? VDP_TRUE : VDP_FALSE;
    render->info.h264.field_pic_flag                         = s->picture_structure != PICT_FRAME;
    render->info.h264.bottom_field_flag                      = s->picture_structure == PICT_BOTTOM_FIELD;
    render->info.h264.is_reference                           = (h->cur_pic_ptr->f.reference & 3) ? VDP_TRUE : VDP_FALSE;
    render->info.h264.field_pic_flag                         = h->picture_structure != PICT_FRAME;
    render->info.h264.bottom_field_flag                      = h->picture_structure == PICT_BOTTOM_FIELD;
    render->info.h264.num_ref_frames                         = h->sps.ref_frame_count;
    render->info.h264.mb_adaptive_frame_field_flag           = h->sps.mb_aff && !render->info.h264.field_pic_flag;
    render->info.h264.constrained_intra_pred_flag            = h->pps.constrained_intra_pred;
@@ -235,7 +227,7 @@ void ff_vdpau_h264_picture_complete(MpegEncContext *s)
    memcpy(render->info.h264.scaling_lists_8x8[0], h->pps.scaling_matrix8[0], sizeof(render->info.h264.scaling_lists_8x8[0]));
    memcpy(render->info.h264.scaling_lists_8x8[1], h->pps.scaling_matrix8[3], sizeof(render->info.h264.scaling_lists_8x8[0]));
    ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
    ff_h264_draw_horiz_band(h, 0, h->avctx->height);
    render->bitstream_buffers_used = 0;
 }
@@ -287,7 +279,7 @@ void ff_vdpau_mpeg_picture_complete(MpegEncContext *s, const uint8_t *buf,
        render->info.mpeg.forward_reference      = last->surface;
    }
    ff_vdpau_add_data_chunk(s, buf, buf_size);
    ff_vdpau_add_data_chunk(s->current_picture_ptr->f.data[0], buf, buf_size);
    render->info.mpeg.slice_count                = slice_count;
@@ -357,7 +349,7 @@ void ff_vdpau_vc1_decode_picture(MpegEncContext *s, const uint8_t *buf,
        render->info.vc1.forward_reference = last->surface;
    }
    ff_vdpau_add_data_chunk(s, buf, buf_size);
    ff_vdpau_add_data_chunk(s->current_picture_ptr->f.data[0], buf, buf_size);
    render->info.vc1.slice_count          = 1;
@@ -413,7 +405,7 @@ void ff_vdpau_mpeg4_decode_picture(MpegEncContext *s, const uint8_t *buf,
        render->info.mpeg4.forward_reference      = last->surface;
    }
    ff_vdpau_add_data_chunk(s, buf, buf_size);
    ff_vdpau_add_data_chunk(s->current_picture_ptr->f.data[0], buf, buf_size);
    ff_mpeg_draw_horiz_band(s, 0, s->avctx->height);
    render->bitstream_buffers_used = 0;
--- a/libavcodec/vdpau_h264.c
+++ b/libavcodec/vdpau_h264.c
@@ -119,9 +119,8 @@ static int vdpau_h264_start_frame(AVCodecContext *avctx,
 {
    H264Context * const h = avctx->priv_data;
    AVVDPAUContext *hwctx = avctx->hwaccel_context;
    MpegEncContext * const s = &h->s;
    VdpPictureInfoH264 *info = &hwctx->info.h264;
    Picture *pic = s->current_picture_ptr;
    Picture *pic = h->cur_pic_ptr;
    /* init VdpPictureInfoH264 */
    info->slice_count                            = 0;
@@ -129,8 +128,8 @@ static int vdpau_h264_start_frame(AVCodecContext *avctx,
    info->field_order_cnt[1]                     = h264_foc(pic->field_poc[1]);
    info->is_reference                           = h->nal_ref_idc != 0;
    info->frame_num                              = h->frame_num;
    info->field_pic_flag                         = s->picture_structure != PICT_FRAME;
    info->bottom_field_flag                      = s->picture_structure == PICT_BOTTOM_FIELD;
    info->field_pic_flag                         = h->picture_structure != PICT_FRAME;
    info->bottom_field_flag                      = h->picture_structure == PICT_BOTTOM_FIELD;
    info->num_ref_frames                         = h->sps.ref_frame_count;
    info->mb_adaptive_frame_field_flag           = h->sps.mb_aff && !info->field_pic_flag;
    info->constrained_intra_pred_flag            = h->pps.constrained_intra_pred;
@@ -185,12 +184,27 @@ static int vdpau_h264_decode_slice(AVCodecContext *avctx,
    return 0;
 }
 static int vdpau_h264_end_frame(AVCodecContext *avctx)
 {
    AVVDPAUContext *hwctx = avctx->hwaccel_context;
    H264Context *h = avctx->priv_data;
    VdpVideoSurface surf = ff_vdpau_get_surface_id(h->cur_pic_ptr);
    hwctx->render(hwctx->decoder, surf, (void *)&hwctx->info,
                  hwctx->bitstream_buffers_used, hwctx->bitstream_buffers);
    ff_h264_draw_horiz_band(h, 0, h->avctx->height);
    hwctx->bitstream_buffers_used = 0;
    return 0;
 }
 AVHWAccel ff_h264_vdpau_hwaccel = {
    .name           = "h264_vdpau",
    .type           = AVMEDIA_TYPE_VIDEO,
    .id             = AV_CODEC_ID_H264,
    .pix_fmt        = AV_PIX_FMT_VDPAU,
    .start_frame    = vdpau_h264_start_frame,
    .end_frame      = ff_vdpau_common_end_frame,
    .end_frame      = vdpau_h264_end_frame,
    .decode_slice   = vdpau_h264_decode_slice,
 };
--- a/libavcodec/vdpau_internal.h
+++ b/libavcodec/vdpau_internal.h
@@ -25,6 +25,7 @@
 #define AVCODEC_VDPAU_INTERNAL_H
 #include <stdint.h>
 #include "h264.h"
 #include "mpegvideo.h"
 /** Extract VdpVideoSurface from a Picture */
@@ -35,20 +36,20 @@ static inline uintptr_t ff_vdpau_get_surface_id(Picture *pic)
 int ff_vdpau_common_start_frame(AVCodecContext *avctx,
                                const uint8_t *buffer, uint32_t size);
 int ff_vdpau_common_end_frame(AVCodecContext *avctx);
 int ff_vdpau_mpeg_end_frame(AVCodecContext *avctx);
 int ff_vdpau_add_buffer(AVCodecContext *avctx,
                        const uint8_t *buf, uint32_t buf_size);
 void ff_vdpau_add_data_chunk(MpegEncContext *s, const uint8_t *buf,
 void ff_vdpau_add_data_chunk(uint8_t *data, const uint8_t *buf,
                             int buf_size);
 void ff_vdpau_mpeg_picture_complete(MpegEncContext *s, const uint8_t *buf,
                                    int buf_size, int slice_count);
 void ff_vdpau_h264_picture_start(MpegEncContext *s);
 void ff_vdpau_h264_set_reference_frames(MpegEncContext *s);
 void ff_vdpau_h264_picture_complete(MpegEncContext *s);
 void ff_vdpau_h264_picture_start(H264Context *h);
 void ff_vdpau_h264_set_reference_frames(H264Context *h);
 void ff_vdpau_h264_picture_complete(H264Context *h);
 void ff_vdpau_vc1_decode_picture(MpegEncContext *s, const uint8_t *buf,
                                 int buf_size);
--- a/libavcodec/vdpau_mpeg12.c
+++ b/libavcodec/vdpau_mpeg12.c
@@ -98,7 +98,7 @@ AVHWAccel ff_mpeg1_vdpau_hwaccel = {
    .id             = AV_CODEC_ID_MPEG1VIDEO,
    .pix_fmt        = AV_PIX_FMT_VDPAU,
    .start_frame    = vdpau_mpeg_start_frame,
    .end_frame      = ff_vdpau_common_end_frame,
    .end_frame      = ff_vdpau_mpeg_end_frame,
    .decode_slice   = vdpau_mpeg_decode_slice,
 };
 #endif
@@ -110,7 +110,7 @@ AVHWAccel ff_mpeg2_vdpau_hwaccel = {
    .id             = AV_CODEC_ID_MPEG2VIDEO,
    .pix_fmt        = AV_PIX_FMT_VDPAU,
    .start_frame    = vdpau_mpeg_start_frame,
    .end_frame      = ff_vdpau_common_end_frame,
    .end_frame      = ff_vdpau_mpeg_end_frame,
    .decode_slice   = vdpau_mpeg_decode_slice,
 };
 #endif
--- a/libavcodec/vdpau_mpeg4.c
+++ b/libavcodec/vdpau_mpeg4.c
@@ -92,7 +92,7 @@ AVHWAccel ff_h263_vdpau_hwaccel = {
    .id             = AV_CODEC_ID_H263,
    .pix_fmt        = AV_PIX_FMT_VDPAU,
    .start_frame    = vdpau_mpeg4_start_frame,
    .end_frame      = ff_vdpau_common_end_frame,
    .end_frame      = ff_vdpau_mpeg_end_frame,
    .decode_slice   = vdpau_mpeg4_decode_slice,
 };
 #endif
@@ -104,7 +104,7 @@ AVHWAccel ff_mpeg4_vdpau_hwaccel = {
    .id             = AV_CODEC_ID_MPEG4,
    .pix_fmt        = AV_PIX_FMT_VDPAU,
    .start_frame    = vdpau_mpeg4_start_frame,
    .end_frame      = ff_vdpau_common_end_frame,
    .end_frame      = ff_vdpau_mpeg_end_frame,
    .decode_slice   = vdpau_mpeg4_decode_slice,
 };
 #endif
--- a/libavcodec/vdpau_vc1.c
+++ b/libavcodec/vdpau_vc1.c
@@ -112,7 +112,7 @@ AVHWAccel ff_wmv3_vdpau_hwaccel = {
    .id             = AV_CODEC_ID_WMV3,
    .pix_fmt        = AV_PIX_FMT_VDPAU,
    .start_frame    = vdpau_vc1_start_frame,
    .end_frame      = ff_vdpau_common_end_frame,
    .end_frame      = ff_vdpau_mpeg_end_frame,
    .decode_slice   = vdpau_vc1_decode_slice,
 };
 #endif
@@ -123,6 +123,6 @@ AVHWAccel ff_vc1_vdpau_hwaccel = {
    .id             = AV_CODEC_ID_VC1,
    .pix_fmt        = AV_PIX_FMT_VDPAU,
    .start_frame    = vdpau_vc1_start_frame,
    .end_frame      = ff_vdpau_common_end_frame,
    .end_frame      = ff_vdpau_mpeg_end_frame,
    .decode_slice   = vdpau_vc1_decode_slice,
 };