18 files changed, 986 insertions, 296 deletions

diff --git a/libavcodec/avcodec.h b/libavcodec/avcodec.h
index 3ea1e76..0f9201d 100644
--- a/libavcodec/avcodec.h
+++ b/libavcodec/avcodec.h
@@ -725,10 +725,12 @@ typedef struct RcOverride{
  * Codec should fill in channel configuration and samplerate instead of container
  */
 #define CODEC_CAP_CHANNEL_CONF     0x0400
+
 /**
  * Codec is able to deal with negative linesizes
  */
 #define CODEC_CAP_NEG_LINESIZES    0x0800
+
 /**
  * Codec supports frame-level multithreading.
  */
@@ -2983,7 +2985,9 @@ typedef struct AVCodec {
     const enum AVSampleFormat *sample_fmts; ///< array of supported sample formats, or NULL if unknown, array is terminated by -1
     const int64_t *channel_layouts;         ///< array of support channel layouts, or NULL if unknown. array is terminated by 0
     uint8_t max_lowres;                     ///< maximum value for lowres supported by the decoder
+
     AVClass *priv_class;                    ///< AVClass for the private context
+
     const AVProfile *profiles;              ///< array of recognized profiles, or NULL if unknown, array is terminated by {FF_PROFILE_UNKNOWN}
 
     /**
diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index b293642..d4e538c 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -298,7 +298,7 @@ static int sse16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
 
 /* draw the edges of width 'w' of an image of size width, height */
 //FIXME check that this is ok for mpeg4 interlaced
-static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w)
+static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w, int sides)
 {
     uint8_t *ptr, *last_line;
     int i;
@@ -306,8 +306,8 @@ static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w)
     last_line = buf + (height - 1) * wrap;
     for(i=0;i<w;i++) {
         /* top and bottom */
-        memcpy(buf - (i + 1) * wrap, buf, width);
-        memcpy(last_line + (i + 1) * wrap, last_line, width);
+        if (sides&EDGE_TOP)    memcpy(buf - (i + 1) * wrap, buf, width);
+        if (sides&EDGE_BOTTOM) memcpy(last_line + (i + 1) * wrap, last_line, width);
     }
     /* left and right */
     ptr = buf;
@@ -318,10 +318,15 @@ static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w)
     }
     /* corners */
     for(i=0;i<w;i++) {
-        memset(buf - (i + 1) * wrap - w, buf[0], w); /* top left */
-        memset(buf - (i + 1) * wrap + width, buf[width-1], w); /* top right */
-        memset(last_line + (i + 1) * wrap - w, last_line[0], w); /* top left */
-        memset(last_line + (i + 1) * wrap + width, last_line[width-1], w); /* top right */
+        if (sides&EDGE_TOP) {
+            memset(buf - (i + 1) * wrap - w, buf[0], w); /* top left */
+            memset(buf - (i + 1) * wrap + width, buf[width-1], w); /* top right */
+        }
+
+        if (sides&EDGE_BOTTOM) {
+            memset(last_line + (i + 1) * wrap - w, last_line[0], w); /* top left */
+            memset(last_line + (i + 1) * wrap + width, last_line[width-1], w); /* top right */
+        }
     }
 }
 
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index 017219b..d42abe8 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -492,8 +492,10 @@ typedef struct DSPContext {
 #define BASIS_SHIFT 16
 #define RECON_SHIFT 6
 
-    void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
+    void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int sides);
 #define EDGE_WIDTH 16
+#define EDGE_TOP    1
+#define EDGE_BOTTOM 2
 
     void (*prefetch)(void *mem, int stride, int h);
 
diff --git a/libavcodec/h263dec.c b/libavcodec/h263dec.c
index 8085b4a..4830202 100644
--- a/libavcodec/h263dec.c
+++ b/libavcodec/h263dec.c
@@ -35,6 +35,7 @@
 #include "mpeg4video_parser.h"
 #include "msmpeg4.h"
 #include "vdpau_internal.h"
+#include "thread.h"
 #include "flv.h"
 #include "mpeg4video.h"
 
@@ -235,6 +236,7 @@ static int decode_slice(MpegEncContext *s){
                     if(++s->mb_x >= s->mb_width){
                         s->mb_x=0;
                         ff_draw_horiz_band(s, s->mb_y*mb_size, mb_size);
+                        MPV_report_decode_progress(s);
                         s->mb_y++;
                     }
                     return 0;
@@ -255,6 +257,7 @@ static int decode_slice(MpegEncContext *s){
         }
 
         ff_draw_horiz_band(s, s->mb_y*mb_size, mb_size);
+        MPV_report_decode_progress(s);
 
         s->mb_x= 0;
     }
@@ -639,6 +642,8 @@ retry:
     if(MPV_frame_start(s, avctx) < 0)
         return -1;
 
+    if (!s->divx_packed) ff_thread_finish_setup(avctx);
+
     if (CONFIG_MPEG4_VDPAU_DECODER && (s->avctx->codec->capabilities & CODEC_CAP_HWACCEL_VDPAU)) {
         ff_vdpau_mpeg4_decode_picture(s, s->gb.buffer, s->gb.buffer_end - s->gb.buffer);
         goto frame_end;
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index 5ebf929..d25b310 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -36,6 +36,7 @@
 #include "golomb.h"
 #include "mathops.h"
 #include "rectangle.h"
+#include "thread.h"
 #include "vdpau_internal.h"
 #include "libavutil/avassert.h"
 
@@ -249,6 +250,141 @@ static int ff_h264_decode_rbsp_trailing(H264Context *h, const uint8_t *src){
     return 0;
 }
 
+static inline int get_lowest_part_list_y(H264Context *h, Picture *pic, int n, int height,
+                                 int y_offset, int list){
+    int raw_my= h->mv_cache[list][ scan8[n] ][1];
+    int filter_height= (raw_my&3) ? 2 : 0;
+    int full_my= (raw_my>>2) + y_offset;
+    int top = full_my - filter_height, bottom = full_my + height + filter_height;
+
+    return FFMAX(abs(top), bottom);
+}
+
+static inline void get_lowest_part_y(H264Context *h, int refs[2][48], int n, int height,
+                               int y_offset, int list0, int list1, int *nrefs){
+    MpegEncContext * const s = &h->s;
+    int my;
+
+    y_offset += 16*(s->mb_y >> MB_FIELD);
+
+    if(list0){
+        int ref_n = h->ref_cache[0][ scan8[n] ];
+        Picture *ref= &h->ref_list[0][ref_n];
+
+        // Error resilience puts the current picture in the ref list.
+        // Don't try to wait on these as it will cause a deadlock.
+        // Fields can wait on each other, though.
+        if(ref->thread_opaque != s->current_picture.thread_opaque ||
+           (ref->reference&3) != s->picture_structure) {
+            my = get_lowest_part_list_y(h, ref, n, height, y_offset, 0);
+            if (refs[0][ref_n] < 0) nrefs[0] += 1;
+            refs[0][ref_n] = FFMAX(refs[0][ref_n], my);
+        }
+    }
+
+    if(list1){
+        int ref_n = h->ref_cache[1][ scan8[n] ];
+        Picture *ref= &h->ref_list[1][ref_n];
+
+        if(ref->thread_opaque != s->current_picture.thread_opaque ||
+           (ref->reference&3) != s->picture_structure) {
+            my = get_lowest_part_list_y(h, ref, n, height, y_offset, 1);
+            if (refs[1][ref_n] < 0) nrefs[1] += 1;
+            refs[1][ref_n] = FFMAX(refs[1][ref_n], my);
+        }
+    }
+}
+
+/**
+ * Wait until all reference frames are available for MC operations.
+ *
+ * @param h the H264 context
+ */
+static void await_references(H264Context *h){
+    MpegEncContext * const s = &h->s;
+    const int mb_xy= h->mb_xy;
+    const int mb_type= s->current_picture.mb_type[mb_xy];
+    int refs[2][48];
+    int nrefs[2] = {0};
+    int ref, list;
+
+    memset(refs, -1, sizeof(refs));
+
+    if(IS_16X16(mb_type)){
+        get_lowest_part_y(h, refs, 0, 16, 0,
+                  IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), nrefs);
+    }else if(IS_16X8(mb_type)){
+        get_lowest_part_y(h, refs, 0, 8, 0,
+                  IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), nrefs);
+        get_lowest_part_y(h, refs, 8, 8, 8,
+                  IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), nrefs);
+    }else if(IS_8X16(mb_type)){
+        get_lowest_part_y(h, refs, 0, 16, 0,
+                  IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), nrefs);
+        get_lowest_part_y(h, refs, 4, 16, 0,
+                  IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), nrefs);
+    }else{
+        int i;
+
+        assert(IS_8X8(mb_type));
+
+        for(i=0; i<4; i++){
+            const int sub_mb_type= h->sub_mb_type[i];
+            const int n= 4*i;
+            int y_offset= (i&2)<<2;
+
+            if(IS_SUB_8X8(sub_mb_type)){
+                get_lowest_part_y(h, refs, n  , 8, y_offset,
+                          IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs);
+            }else if(IS_SUB_8X4(sub_mb_type)){
+                get_lowest_part_y(h, refs, n  , 4, y_offset,
+                          IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs);
+                get_lowest_part_y(h, refs, n+2, 4, y_offset+4,
+                          IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs);
+            }else if(IS_SUB_4X8(sub_mb_type)){
+                get_lowest_part_y(h, refs, n  , 8, y_offset,
+                          IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs);
+                get_lowest_part_y(h, refs, n+1, 8, y_offset,
+                          IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs);
+            }else{
+                int j;
+                assert(IS_SUB_4X4(sub_mb_type));
+                for(j=0; j<4; j++){
+                    int sub_y_offset= y_offset + 2*(j&2);
+                    get_lowest_part_y(h, refs, n+j, 4, sub_y_offset,
+                              IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs);
+                }
+            }
+        }
+    }
+
+    for(list=h->list_count-1; list>=0; list--){
+        for(ref=0; ref<48 && nrefs[list]; ref++){
+            int row = refs[list][ref];
+            if(row >= 0){
+                Picture *ref_pic = &h->ref_list[list][ref];
+                int ref_field = ref_pic->reference - 1;
+                int ref_field_picture = ref_pic->field_picture;
+                int pic_height = 16*s->mb_height >> ref_field_picture;
+
+                row <<= MB_MBAFF;
+                nrefs[list]--;
+
+                if(!FIELD_PICTURE && ref_field_picture){ // frame referencing two fields
+                    ff_thread_await_progress((AVFrame*)ref_pic, FFMIN((row >> 1) - !(row&1), pic_height-1), 1);
+                    ff_thread_await_progress((AVFrame*)ref_pic, FFMIN((row >> 1)           , pic_height-1), 0);
+                }else if(FIELD_PICTURE && !ref_field_picture){ // field referencing one field of a frame
+                    ff_thread_await_progress((AVFrame*)ref_pic, FFMIN(row*2 + ref_field    , pic_height-1), 0);
+                }else if(FIELD_PICTURE){
+                    ff_thread_await_progress((AVFrame*)ref_pic, FFMIN(row, pic_height-1), ref_field);
+                }else{
+                    ff_thread_await_progress((AVFrame*)ref_pic, FFMIN(row, pic_height-1), 0);
+                }
+            }
+        }
+    }
+}
+
 #if 0
 /**
  * DCT transforms the 16 dc values.
@@ -539,6 +675,8 @@ static void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t
 
     assert(IS_INTER(mb_type));
 
+    if(HAVE_PTHREADS && s->avctx->active_thread_type&FF_THREAD_FRAME)
+        await_references(h);
     prefetch_motion(h, 0);
 
     if(IS_16X16(mb_type)){
@@ -887,7 +1025,7 @@ av_cold int ff_h264_decode_init(AVCodecContext *avctx){
     ff_h264_decode_init_vlc();
 
     h->thread_context[0] = h;
-    h->outputed_poc = INT_MIN;
+    h->outputed_poc = h->next_outputed_poc = INT_MIN;
     h->prev_poc_msb= 1<<16;
     h->x264_build = -1;
     ff_h264_reset_sei(h);
@@ -910,6 +1048,125 @@ av_cold int ff_h264_decode_init(AVCodecContext *avctx){
     return 0;
 }
 
+static void copy_picture_range(Picture **to, Picture **from, int count, MpegEncContext *new_base, MpegEncContext *old_base)
+{
+    int i;
+
+    for (i=0; i<count; i++){
+        to[i] = REBASE_PICTURE(from[i], new_base, old_base);
+    }
+}
+
+static void copy_parameter_set(void **to, void **from, int count, int size)
+{
+    int i;
+
+    for (i=0; i<count; i++){
+        if (to[i] && !from[i]) av_freep(&to[i]);
+        else if (from[i] && !to[i]) to[i] = av_malloc(size);
+
+        if (from[i]) memcpy(to[i], from[i], size);
+    }
+}
+
+static int decode_init_thread_copy(AVCodecContext *avctx){
+    H264Context *h= avctx->priv_data;
+
+    if (!avctx->is_copy) return 0;
+    memset(h->sps_buffers, 0, sizeof(h->sps_buffers));
+    memset(h->pps_buffers, 0, sizeof(h->pps_buffers));
+
+    return 0;
+}
+
+#define copy_fields(to, from, start_field, end_field) memcpy(&to->start_field, &from->start_field, (char*)&to->end_field - (char*)&to->start_field)
+static int decode_update_thread_context(AVCodecContext *dst, const AVCodecContext *src){
+    H264Context *h= dst->priv_data, *h1= src->priv_data;
+    MpegEncContext * const s = &h->s, * const s1 = &h1->s;
+    int inited = s->context_initialized, err;
+    int i;
+
+    if(dst == src || !s1->context_initialized) return 0;
+
+    err = ff_mpeg_update_thread_context(dst, src);
+    if(err) return err;
+
+    //FIXME handle width/height changing
+    if(!inited){
+        for(i = 0; i < MAX_SPS_COUNT; i++)
+            av_freep(h->sps_buffers + i);
+
+        for(i = 0; i < MAX_PPS_COUNT; i++)
+            av_freep(h->pps_buffers + i);
+
+        memcpy(&h->s + 1, &h1->s + 1, sizeof(H264Context) - sizeof(MpegEncContext)); //copy all fields after MpegEnc
+        memset(h->sps_buffers, 0, sizeof(h->sps_buffers));
+        memset(h->pps_buffers, 0, sizeof(h->pps_buffers));
+        ff_h264_alloc_tables(h);
+        context_init(h);
+
+        for(i=0; i<2; i++){
+            h->rbsp_buffer[i] = NULL;
+            h->rbsp_buffer_size[i] = 0;
+        }
+
+        h->thread_context[0] = h;
+
+        // frame_start may not be called for the next thread (if it's decoding a bottom field)
+        // so this has to be allocated here
+        h->s.obmc_scratchpad = av_malloc(16*2*s->linesize + 8*2*s->uvlinesize);
+
+        s->dsp.clear_blocks(h->mb);
+    }
+
+    //extradata/NAL handling
+    h->is_avc          = h1->is_avc;
+
+    //SPS/PPS
+    copy_parameter_set((void**)h->sps_buffers, (void**)h1->sps_buffers, MAX_SPS_COUNT, sizeof(SPS));
+    h->sps             = h1->sps;
+    copy_parameter_set((void**)h->pps_buffers, (void**)h1->pps_buffers, MAX_PPS_COUNT, sizeof(PPS));
+    h->pps             = h1->pps;
+
+    //Dequantization matrices
+    //FIXME these are big - can they be only copied when PPS changes?
+    copy_fields(h, h1, dequant4_buffer, dequant4_coeff);
+
+    for(i=0; i<6; i++)
+        h->dequant4_coeff[i] = h->dequant4_buffer[0] + (h1->dequant4_coeff[i] - h1->dequant4_buffer[0]);
+
+    for(i=0; i<2; i++)
+        h->dequant8_coeff[i] = h->dequant8_buffer[0] + (h1->dequant8_coeff[i] - h1->dequant8_buffer[0]);
+
+    h->dequant_coeff_pps = h1->dequant_coeff_pps;
+
+    //POC timing
+    copy_fields(h, h1, poc_lsb, redundant_pic_count);
+
+    //reference lists
+    copy_fields(h, h1, ref_count, intra_gb);
+    copy_fields(h, h1, short_ref, cabac_init_idc);
+
+    copy_picture_range(h->short_ref,   h1->short_ref,   32, s, s1);
+    copy_picture_range(h->long_ref,    h1->long_ref,    32,  s, s1);
+    copy_picture_range(h->delayed_pic, h1->delayed_pic, MAX_DELAYED_PIC_COUNT+2, s, s1);
+
+    h->last_slice_type = h1->last_slice_type;
+
+    if(!s->current_picture_ptr) return 0;
+
+    if(!s->dropable) {
+        ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
+        h->prev_poc_msb     = h->poc_msb;
+        h->prev_poc_lsb     = h->poc_lsb;
+    }
+    h->prev_frame_num_offset= h->frame_num_offset;
+    h->prev_frame_num       = h->frame_num;
+    h->outputed_poc         = h->next_outputed_poc;
+
+    return 0;
+}
+
 int ff_h264_frame_start(H264Context *h){
     MpegEncContext * const s = &h->s;
     int i;
@@ -961,11 +1218,167 @@ int ff_h264_frame_start(H264Context *h){
 
     s->current_picture_ptr->field_poc[0]=
     s->current_picture_ptr->field_poc[1]= INT_MAX;
+
+    h->next_output_pic = NULL;
+
     assert(s->current_picture_ptr->long_ref==0);
 
     return 0;
 }
 
+/**
+  * Run setup operations that must be run after slice header decoding.
+  * This includes finding the next displayed frame.
+  *
+  * @param h h264 master context
+  */
+static void decode_postinit(H264Context *h){
+    MpegEncContext * const s = &h->s;
+    Picture *out = s->current_picture_ptr;
+    Picture *cur = s->current_picture_ptr;
+    int i, pics, out_of_order, out_idx;
+
+    s->current_picture_ptr->qscale_type= FF_QSCALE_TYPE_H264;
+    s->current_picture_ptr->pict_type= s->pict_type;
+
+    if (h->next_output_pic) return;
+
+    if (cur->field_poc[0]==INT_MAX || cur->field_poc[1]==INT_MAX) {
+        //FIXME this allows the next thread to start once we encounter the first field of a PAFF packet
+        //This works if the next packet contains the second field. It does not work if both fields are
+        //in the same packet.
+        //ff_thread_finish_setup(s->avctx);
+        return;
+    }
+
+    cur->interlaced_frame = 0;
+    cur->repeat_pict = 0;
+
+    /* Signal interlacing information externally. */
+    /* Prioritize picture timing SEI information over used decoding process if it exists. */
+
+    if(h->sps.pic_struct_present_flag){
+        switch (h->sei_pic_struct)
+        {
+        case SEI_PIC_STRUCT_FRAME:
+            break;
+        case SEI_PIC_STRUCT_TOP_FIELD:
+        case SEI_PIC_STRUCT_BOTTOM_FIELD:
+            cur->interlaced_frame = 1;
+            break;
+        case SEI_PIC_STRUCT_TOP_BOTTOM:
+        case SEI_PIC_STRUCT_BOTTOM_TOP:
+            if (FIELD_OR_MBAFF_PICTURE)
+                cur->interlaced_frame = 1;
+            else
+                // try to flag soft telecine progressive
+                cur->interlaced_frame = h->prev_interlaced_frame;
+            break;
+        case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
+        case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
+            // Signal the possibility of telecined film externally (pic_struct 5,6)
+            // From these hints, let the applications decide if they apply deinterlacing.
+            cur->repeat_pict = 1;
+            break;
+        case SEI_PIC_STRUCT_FRAME_DOUBLING:
+            // Force progressive here, as doubling interlaced frame is a bad idea.
+            cur->repeat_pict = 2;
+            break;
+        case SEI_PIC_STRUCT_FRAME_TRIPLING:
+            cur->repeat_pict = 4;
+            break;
+        }
+
+        if ((h->sei_ct_type & 3) && h->sei_pic_struct <= SEI_PIC_STRUCT_BOTTOM_TOP)
+            cur->interlaced_frame = (h->sei_ct_type & (1<<1)) != 0;
+    }else{
+        /* Derive interlacing flag from used decoding process. */
+        cur->interlaced_frame = FIELD_OR_MBAFF_PICTURE;
+    }
+    h->prev_interlaced_frame = cur->interlaced_frame;
+
+    if (cur->field_poc[0] != cur->field_poc[1]){
+        /* Derive top_field_first from field pocs. */
+        cur->top_field_first = cur->field_poc[0] < cur->field_poc[1];
+    }else{
+        if(cur->interlaced_frame || h->sps.pic_struct_present_flag){
+            /* Use picture timing SEI information. Even if it is a information of a past frame, better than nothing. */
+            if(h->sei_pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM
+              || h->sei_pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM_TOP)
+                cur->top_field_first = 1;
+            else
+                cur->top_field_first = 0;
+        }else{
+            /* Most likely progressive */
+            cur->top_field_first = 0;
+        }
+    }
+
+    //FIXME do something with unavailable reference frames
+
+    /* Sort B-frames into display order */
+
+    if(h->sps.bitstream_restriction_flag
+       && s->avctx->has_b_frames < h->sps.num_reorder_frames){
+        s->avctx->has_b_frames = h->sps.num_reorder_frames;
+        s->low_delay = 0;
+    }
+
+    if(   s->avctx->strict_std_compliance >= FF_COMPLIANCE_STRICT
+       && !h->sps.bitstream_restriction_flag){
+        s->avctx->has_b_frames= MAX_DELAYED_PIC_COUNT;
+        s->low_delay= 0;
+    }
+
+    pics = 0;
+    while(h->delayed_pic[pics]) pics++;
+
+    assert(pics <= MAX_DELAYED_PIC_COUNT);
+
+    h->delayed_pic[pics++] = cur;
+    if(cur->reference == 0)
+        cur->reference = DELAYED_PIC_REF;
+
+    out = h->delayed_pic[0];
+    out_idx = 0;
+    for(i=1; h->delayed_pic[i] && !h->delayed_pic[i]->key_frame && !h->delayed_pic[i]->mmco_reset; i++)
+        if(h->delayed_pic[i]->poc < out->poc){
+            out = h->delayed_pic[i];
+            out_idx = i;
+        }
+    if(s->avctx->has_b_frames == 0 && (h->delayed_pic[0]->key_frame || h->delayed_pic[0]->mmco_reset))
+        h->next_outputed_poc= INT_MIN;
+    out_of_order = out->poc < h->next_outputed_poc;
+
+    if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames >= h->sps.num_reorder_frames)
+        { }
+    else if((out_of_order && pics-1 == s->avctx->has_b_frames && s->avctx->has_b_frames < MAX_DELAYED_PIC_COUNT)
+       || (s->low_delay &&
+        ((h->next_outputed_poc != INT_MIN && out->poc > h->next_outputed_poc + 2)
+         || cur->pict_type == FF_B_TYPE)))
+    {
+        s->low_delay = 0;
+        s->avctx->has_b_frames++;
+    }
+
+    if(out_of_order || pics > s->avctx->has_b_frames){
+        out->reference &= ~DELAYED_PIC_REF;
+        for(i=out_idx; h->delayed_pic[i]; i++)
+            h->delayed_pic[i] = h->delayed_pic[i+1];
+    }
+    if(!out_of_order && pics > s->avctx->has_b_frames){
+        h->next_output_pic = out;
+        if(out_idx==0 && h->delayed_pic[0] && (h->delayed_pic[0]->key_frame || h->delayed_pic[0]->mmco_reset)) {
+            h->next_outputed_poc = INT_MIN;
+        } else
+            h->next_outputed_poc = out->poc;
+    }else{
+        av_log(s->avctx, AV_LOG_DEBUG, "no picture\n");
+    }
+
+    ff_thread_finish_setup(s->avctx);
+}
+
 static inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int simple){
     MpegEncContext * const s = &h->s;
     uint8_t *top_border;
@@ -1479,7 +1892,7 @@ static void flush_dpb(AVCodecContext *avctx){
             h->delayed_pic[i]->reference= 0;
         h->delayed_pic[i]= NULL;
     }
-    h->outputed_poc= INT_MIN;
+    h->outputed_poc=h->next_outputed_poc= INT_MIN;
     h->prev_interlaced_frame = 1;
     idr(h);
     if(h->s.current_picture_ptr)
@@ -1603,24 +2016,28 @@ static void init_scan_tables(H264Context *h){
     }
 }
 
-static void field_end(H264Context *h){
+static void field_end(H264Context *h, int in_setup){
     MpegEncContext * const s = &h->s;
     AVCodecContext * const avctx= s->avctx;
     s->mb_y= 0;
 
-    s->current_picture_ptr->qscale_type= FF_QSCALE_TYPE_H264;
-    s->current_picture_ptr->pict_type= s->pict_type;
+    if (!in_setup && !s->dropable)
+        ff_thread_report_progress((AVFrame*)s->current_picture_ptr, (16*s->mb_height >> FIELD_PICTURE) - 1,
+                                 s->picture_structure==PICT_BOTTOM_FIELD);
 
     if (CONFIG_H264_VDPAU_DECODER && s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
         ff_vdpau_h264_set_reference_frames(s);
 
-    if(!s->dropable) {
-        ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
-        h->prev_poc_msb= h->poc_msb;
-        h->prev_poc_lsb= h->poc_lsb;
+    if(in_setup || !(avctx->active_thread_type&FF_THREAD_FRAME)){
+        if(!s->dropable) {
+            ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
+            h->prev_poc_msb= h->poc_msb;
+            h->prev_poc_lsb= h->poc_lsb;
+        }
+        h->prev_frame_num_offset= h->frame_num_offset;
+        h->prev_frame_num= h->frame_num;
+        h->outputed_poc = h->next_outputed_poc;
     }
-    h->prev_frame_num_offset= h->frame_num_offset;
-    h->prev_frame_num= h->frame_num;
 
     if (avctx->hwaccel) {
         if (avctx->hwaccel->end_frame(avctx) < 0)
@@ -1737,7 +2154,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
 
     if(first_mb_in_slice == 0){ //FIXME better field boundary detection
         if(h0->current_slice && FIELD_PICTURE){
-            field_end(h);
+            field_end(h, 1);
         }
 
         h0->current_slice = 0;
@@ -1806,8 +2223,10 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
     if (s->context_initialized
         && (   s->width != s->avctx->width || s->height != s->avctx->height
             || av_cmp_q(h->sps.sar, s->avctx->sample_aspect_ratio))) {
-        if(h != h0)
+        if(h != h0) {
+            av_log_missing_feature(s->avctx, "Width/height changing with threads is", 0);
             return -1;   // width / height changed during parallelized decoding
+        }
         free_tables(h, 0);
         flush_dpb(s->avctx);
         MPV_common_end(s);
@@ -1852,21 +2271,26 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
         init_scan_tables(h);
         ff_h264_alloc_tables(h);
 
-        for(i = 1; i < s->avctx->thread_count; i++) {
-            H264Context *c;
-            c = h->thread_context[i] = av_malloc(sizeof(H264Context));
-            memcpy(c, h->s.thread_context[i], sizeof(MpegEncContext));
-            memset(&c->s + 1, 0, sizeof(H264Context) - sizeof(MpegEncContext));
-            c->h264dsp = h->h264dsp;
-            c->sps = h->sps;
-            c->pps = h->pps;
-            init_scan_tables(c);
-            clone_tables(c, h, i);
-        }
-
-        for(i = 0; i < s->avctx->thread_count; i++)
-            if(context_init(h->thread_context[i]) < 0)
+        if (!HAVE_THREADS || !(s->avctx->active_thread_type&FF_THREAD_SLICE)) {
+            if (context_init(h) < 0)
                 return -1;
+        } else {
+            for(i = 1; i < s->avctx->thread_count; i++) {
+                H264Context *c;
+                c = h->thread_context[i] = av_malloc(sizeof(H264Context));
+                memcpy(c, h->s.thread_context[i], sizeof(MpegEncContext));
+                memset(&c->s + 1, 0, sizeof(H264Context) - sizeof(MpegEncContext));
+                c->h264dsp = h->h264dsp;
+                c->sps = h->sps;
+                c->pps = h->pps;
+                init_scan_tables(c);
+                clone_tables(c, h, i);
+            }
+
+            for(i = 0; i < s->avctx->thread_count; i++)
+                if(context_init(h->thread_context[i]) < 0)
+                    return -1;
+        }
     }
 
     h->frame_num= get_bits(&s->gb, h->sps.log2_max_frame_num);
@@ -1887,6 +2311,10 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
     h->mb_field_decoding_flag= s->picture_structure != PICT_FRAME;
 
     if(h0->current_slice == 0){
+        if(h->frame_num != h->prev_frame_num &&
+          (h->prev_frame_num+1)%(1<<h->sps.log2_max_frame_num) < (h->frame_num - h->sps.ref_frame_count))
+            h->prev_frame_num = h->frame_num - h->sps.ref_frame_count - 1;
+
         while(h->frame_num !=  h->prev_frame_num &&
               h->frame_num != (h->prev_frame_num+1)%(1<<h->sps.log2_max_frame_num)){
             Picture *prev = h->short_ref_count ? h->short_ref[0] : NULL;
@@ -1896,6 +2324,8 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
             h->prev_frame_num++;
             h->prev_frame_num %= 1<<h->sps.log2_max_frame_num;
             s->current_picture_ptr->frame_num= h->prev_frame_num;
+            ff_thread_report_progress((AVFrame*)s->current_picture_ptr, INT_MAX, 0);
+            ff_thread_report_progress((AVFrame*)s->current_picture_ptr, INT_MAX, 1);
             ff_generate_sliding_window_mmcos(h);
             ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
             /* Error concealment: if a ref is missing, copy the previous ref in its place.
@@ -2045,6 +2475,9 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
     if(h->slice_type_nos!=FF_I_TYPE && ff_h264_decode_ref_pic_list_reordering(h) < 0)
         return -1;
 
+    //FIXME mt gives valgrind warnings and crashes if this is uncommented
+    /*
+
     if(h->slice_type_nos!=FF_I_TYPE){
         s->last_picture_ptr= &h->ref_list[0][0];
         ff_copy_picture(&s->last_picture, s->last_picture_ptr);
@@ -2054,6 +2487,8 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
         ff_copy_picture(&s->next_picture, s->next_picture_ptr);
     }
 
+    */
+
     if(   (h->pps.weighted_pred          && h->slice_type_nos == FF_P_TYPE )
        ||  (h->pps.weighted_bipred_idc==1 && h->slice_type_nos== FF_B_TYPE ) )
         pred_weight_table(h);
@@ -2200,7 +2635,8 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
                           +(h->ref_list[j][i].reference&3);
     }
 
-    h->emu_edge_width= (s->flags&CODEC_FLAG_EMU_EDGE) ? 0 : 16;
+    //FIXME: fix draw_edges+PAFF+frame threads
+    h->emu_edge_width= (s->flags&CODEC_FLAG_EMU_EDGE || (!h->sps.frame_mbs_only_flag && s->avctx->active_thread_type&FF_THREAD_FRAME)) ? 0 : 16;
     h->emu_edge_height= (FRAME_MBAFF || FIELD_PICTURE) ? 0 : h->emu_edge_width;
 
     if(s->avctx->debug&FF_DEBUG_PICT_INFO){
@@ -2521,6 +2957,40 @@ static void predict_field_decoding_flag(H264Context *h){
     h->mb_mbaff = h->mb_field_decoding_flag = IS_INTERLACED(mb_type) ? 1 : 0;
 }
 
+/**
+ * Draw edges and report progress for the last MB row.
+ */
+static void decode_finish_row(H264Context *h){
+    MpegEncContext * const s = &h->s;
+    int top = 16*(s->mb_y >> FIELD_PICTURE);
+    int height = 16 << FRAME_MBAFF;
+    int deblock_border = (16 + 4) << FRAME_MBAFF;
+    int pic_height = 16*s->mb_height >> FIELD_PICTURE;
+
+    if (h->deblocking_filter) {
+        if((top + height) >= pic_height)
+            height += deblock_border;
+
+        top -= deblock_border;
+    }
+
+    if (top >= pic_height || (top + height) < h->emu_edge_height)
+        return;
+
+    height = FFMIN(height, pic_height - top);
+    if (top < h->emu_edge_height) {
+        height = top+height;
+        top = 0;
+    }
+
+    ff_draw_horiz_band(s, top, height);
+
+    if (s->dropable) return;
+
+    ff_thread_report_progress((AVFrame*)s->current_picture_ptr, top + height - 1,
+                             s->picture_structure==PICT_BOTTOM_FIELD);
+}
+
 static int decode_slice(struct AVCodecContext *avctx, void *arg){
     H264Context *h = *(void**)arg;
     MpegEncContext * const s = &h->s;
@@ -2574,7 +3044,7 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg){
             if( ++s->mb_x >= s->mb_width ) {
                 s->mb_x = 0;
                 loop_filter(h);
-                ff_draw_horiz_band(s, 16*s->mb_y, 16);
+                decode_finish_row(h);
                 ++s->mb_y;
                 if(FIELD_OR_MBAFF_PICTURE) {
                     ++s->mb_y;
@@ -2614,7 +3084,7 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg){
             if(++s->mb_x >= s->mb_width){
                 s->mb_x=0;
                 loop_filter(h);
-                ff_draw_horiz_band(s, 16*s->mb_y, 16);
+                decode_finish_row(h);
                 ++s->mb_y;
                 if(FIELD_OR_MBAFF_PICTURE) {
                     ++s->mb_y;
@@ -2747,7 +3217,7 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){
     int context_count = 0;
     int next_avc= h->is_avc ? 0 : buf_size;
 
-    h->max_contexts = avctx->thread_count;
+    h->max_contexts = (HAVE_THREADS && (s->avctx->active_thread_type&FF_THREAD_SLICE)) ? avctx->thread_count : 1;
 #if 0
     int i;
     for(i=0; i<50; i++){
@@ -2842,16 +3312,21 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){
             if((err = decode_slice_header(hx, h)))
                break;
 
+            s->current_picture_ptr->key_frame |=
+                    (hx->nal_unit_type == NAL_IDR_SLICE) ||
+                    (h->sei_recovery_frame_cnt >= 0);
+
             if (h->current_slice == 1) {
+                if(!(s->flags2 & CODEC_FLAG2_CHUNKS)) {
+                    decode_postinit(h);
+                }
+
                 if (s->avctx->hwaccel && s->avctx->hwaccel->start_frame(s->avctx, NULL, 0) < 0)
                     return -1;
                 if(CONFIG_H264_VDPAU_DECODER && s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
                     ff_vdpau_h264_picture_start(s);
             }
 
-            s->current_picture_ptr->key_frame |=
-                    (hx->nal_unit_type == NAL_IDR_SLICE) ||
-                    (h->sei_recovery_frame_cnt >= 0);
             if(hx->redundant_pic_count==0 && hx->s.hurry_up < 5
                && (avctx->skip_frame < AVDISCARD_NONREF || hx->nal_ref_idc)
                && (avctx->skip_frame < AVDISCARD_BIDIR  || hx->slice_type_nos!=FF_B_TYPE)
@@ -2981,6 +3456,8 @@ static int decode_frame(AVCodecContext *avctx,
         Picture *out;
         int i, out_idx;
 
+        s->current_picture_ptr = NULL;
+
 //FIXME factorize this with the output code below
         out = h->delayed_pic[0];
         out_idx = 0;
@@ -3017,143 +3494,18 @@ static int decode_frame(AVCodecContext *avctx,
     }
 
     if(!(s->flags2 & CODEC_FLAG2_CHUNKS) || (s->mb_y >= s->mb_height && s->mb_height)){
-        Picture *out = s->current_picture_ptr;
-        Picture *cur = s->current_picture_ptr;
-        int i, pics, out_of_order, out_idx;
 
-        field_end(h);
+        if(s->flags2 & CODEC_FLAG2_CHUNKS) decode_postinit(h);
 
-        if (cur->field_poc[0]==INT_MAX || cur->field_poc[1]==INT_MAX) {
+        field_end(h, 0);
+
+        if (!h->next_output_pic) {
             /* Wait for second field. */
             *data_size = 0;
 
         } else {
-            cur->interlaced_frame = 0;
-            cur->repeat_pict = 0;
-
-            /* Signal interlacing information externally. */
-            /* Prioritize picture timing SEI information over used decoding process if it exists. */
-
-            if(h->sps.pic_struct_present_flag){
-                switch (h->sei_pic_struct)
-                {
-                case SEI_PIC_STRUCT_FRAME:
-                    break;
-                case SEI_PIC_STRUCT_TOP_FIELD:
-                case SEI_PIC_STRUCT_BOTTOM_FIELD:
-                    cur->interlaced_frame = 1;
-                    break;
-                case SEI_PIC_STRUCT_TOP_BOTTOM:
-                case SEI_PIC_STRUCT_BOTTOM_TOP:
-                    if (FIELD_OR_MBAFF_PICTURE)
-                        cur->interlaced_frame = 1;
-                    else
-                        // try to flag soft telecine progressive
-                        cur->interlaced_frame = h->prev_interlaced_frame;
-                    break;
-                case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
-                case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
-                    // Signal the possibility of telecined film externally (pic_struct 5,6)
-                    // From these hints, let the applications decide if they apply deinterlacing.
-                    cur->repeat_pict = 1;
-                    break;
-                case SEI_PIC_STRUCT_FRAME_DOUBLING:
-                    // Force progressive here, as doubling interlaced frame is a bad idea.
-                    cur->repeat_pict = 2;
-                    break;
-                case SEI_PIC_STRUCT_FRAME_TRIPLING:
-                    cur->repeat_pict = 4;
-                    break;
-                }
-
-                if ((h->sei_ct_type & 3) && h->sei_pic_struct <= SEI_PIC_STRUCT_BOTTOM_TOP)
-                    cur->interlaced_frame = (h->sei_ct_type & (1<<1)) != 0;
-            }else{
-                /* Derive interlacing flag from used decoding process. */
-                cur->interlaced_frame = FIELD_OR_MBAFF_PICTURE;
-            }
-            h->prev_interlaced_frame = cur->interlaced_frame;
-
-            if (cur->field_poc[0] != cur->field_poc[1]){
-                /* Derive top_field_first from field pocs. */
-                cur->top_field_first = cur->field_poc[0] < cur->field_poc[1];
-            }else{
-                if(cur->interlaced_frame || h->sps.pic_struct_present_flag){
-                    /* Use picture timing SEI information. Even if it is a information of a past frame, better than nothing. */
-                    if(h->sei_pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM
-                      || h->sei_pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM_TOP)
-                        cur->top_field_first = 1;
-                    else
-                        cur->top_field_first = 0;
-                }else{
-                    /* Most likely progressive */
-                    cur->top_field_first = 0;
-                }
-            }
-
-        //FIXME do something with unavailable reference frames
-
-            /* Sort B-frames into display order */
-
-            if(h->sps.bitstream_restriction_flag
-               && s->avctx->has_b_frames < h->sps.num_reorder_frames){
-                s->avctx->has_b_frames = h->sps.num_reorder_frames;
-                s->low_delay = 0;
-            }
-
-            if(   s->avctx->strict_std_compliance >= FF_COMPLIANCE_STRICT
-               && !h->sps.bitstream_restriction_flag){
-                s->avctx->has_b_frames= MAX_DELAYED_PIC_COUNT;
-                s->low_delay= 0;
-            }
-
-            pics = 0;
-            while(h->delayed_pic[pics]) pics++;
-
-            assert(pics <= MAX_DELAYED_PIC_COUNT);
-
-            h->delayed_pic[pics++] = cur;
-            if(cur->reference == 0)
-                cur->reference = DELAYED_PIC_REF;
-
-            out = h->delayed_pic[0];
-            out_idx = 0;
-            for(i=1; h->delayed_pic[i] && !h->delayed_pic[i]->key_frame && !h->delayed_pic[i]->mmco_reset; i++)
-                if(h->delayed_pic[i]->poc < out->poc){
-                    out = h->delayed_pic[i];
-                    out_idx = i;
-                }
-            if(s->avctx->has_b_frames == 0 && (h->delayed_pic[0]->key_frame || h->delayed_pic[0]->mmco_reset))
-                h->outputed_poc= INT_MIN;
-            out_of_order = out->poc < h->outputed_poc;
-
-            if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames >= h->sps.num_reorder_frames)
-                { }
-            else if((out_of_order && pics-1 == s->avctx->has_b_frames && s->avctx->has_b_frames < MAX_DELAYED_PIC_COUNT)
-               || (s->low_delay &&
-                ((h->outputed_poc != INT_MIN && out->poc > h->outputed_poc + 2)
-                 || cur->pict_type == FF_B_TYPE)))
-            {
-                s->low_delay = 0;
-                s->avctx->has_b_frames++;
-            }
-
-            if(out_of_order || pics > s->avctx->has_b_frames){
-                out->reference &= ~DELAYED_PIC_REF;
-                for(i=out_idx; h->delayed_pic[i]; i++)
-                    h->delayed_pic[i] = h->delayed_pic[i+1];
-            }
-            if(!out_of_order && pics > s->avctx->has_b_frames){
-                *data_size = sizeof(AVFrame);
-
-                if(out_idx==0 && h->delayed_pic[0] && (h->delayed_pic[0]->key_frame || h->delayed_pic[0]->mmco_reset)) {
-                    h->outputed_poc = INT_MIN;
-                } else
-                    h->outputed_poc = out->poc;
-                *pict= *(AVFrame*)out;
-            }else{
-                av_log(avctx, AV_LOG_DEBUG, "no picture\n");
-            }
+            *data_size = sizeof(AVFrame);
+            *pict = *(AVFrame*)h->next_output_pic;
         }
     }
 
@@ -3412,9 +3764,11 @@ AVCodec ff_h264_decoder = {
     NULL,
     ff_h264_decode_end,
     decode_frame,
-    /*CODEC_CAP_DRAW_HORIZ_BAND |*/ CODEC_CAP_DR1 | CODEC_CAP_DELAY,
+    /*CODEC_CAP_DRAW_HORIZ_BAND |*/ CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_FRAME_THREADS,
     .flush= flush_dpb,
     .long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
+    .init_thread_copy      = ONLY_IF_THREADS_ENABLED(decode_init_thread_copy),
+    .update_thread_context = ONLY_IF_THREADS_ENABLED(decode_update_thread_context),
     .profiles = NULL_IF_CONFIG_SMALL(profiles),
 };
 
diff --git a/libavcodec/h264.h b/libavcodec/h264.h
index b403968..770a7ec 100644
--- a/libavcodec/h264.h
+++ b/libavcodec/h264.h
@@ -392,9 +392,9 @@ typedef struct H264Context{
     /**
      * num_ref_idx_l0/1_active_minus1 + 1
      */
+    uint8_t *list_counts;            ///< Array of list_count per MB specifying the slice type
     unsigned int ref_count[2];   ///< counts frames or fields, depending on current mb mode
     unsigned int list_count;
-    uint8_t *list_counts;            ///< Array of list_count per MB specifying the slice type
     Picture ref_list[2][48];         /**< 0..15: frame refs, 16..47: mbaff field refs.
                                           Reordered version of default_ref_list
                                           according to picture reordering in slice header */
@@ -504,7 +504,9 @@ typedef struct H264Context{
     Picture *long_ref[32];
     Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
     Picture *delayed_pic[MAX_DELAYED_PIC_COUNT+2]; //FIXME size?
+    Picture *next_output_pic;
     int outputed_poc;
+    int next_outputed_poc;
 
     /**
      * memory management control operations buffer.
diff --git a/libavcodec/h264_direct.c b/libavcodec/h264_direct.c
index d22780d..a26a51a 100644
--- a/libavcodec/h264_direct.c
+++ b/libavcodec/h264_direct.c
@@ -31,6 +31,7 @@
 #include "mpegvideo.h"
 #include "h264.h"
 #include "rectangle.h"
+#include "thread.h"
 
 //#undef NDEBUG
 #include <assert.h>
@@ -126,7 +127,7 @@ void ff_h264_direct_ref_list_init(H264Context * const h){
         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].reference) && !h->ref_list[1][0].mbaff){ // FL -> FL & differ parity
-        h->col_fieldoff= s->mb_stride*(2*(h->ref_list[1][0].reference) - 3);
+        h->col_fieldoff= 2*(h->ref_list[1][0].reference) - 3;
     }
 
     if(cur->pict_type != FF_B_TYPE || h->direct_spatial_mv_pred)
@@ -140,11 +141,27 @@ void ff_h264_direct_ref_list_init(H264Context * const h){
     }
 }
 
+static void await_reference_mb_row(H264Context * const h, Picture *ref, int mb_y)
+{
+    int ref_field = ref->reference - 1;
+    int ref_field_picture = ref->field_picture;
+    int ref_height = 16*h->s.mb_height >> ref_field_picture;
+
+    if(!HAVE_PTHREADS || !(h->s.avctx->active_thread_type&FF_THREAD_FRAME))
+        return;
+
+    //FIXME it can be safe to access mb stuff
+    //even if pixels aren't deblocked yet
+
+    ff_thread_await_progress((AVFrame*)ref, FFMIN(16*mb_y >> ref_field_picture, ref_height-1),
+                             ref_field_picture && ref_field);
+}
+
 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;
+    int mb_xy = h->mb_xy, mb_y = s->mb_y;
     int mb_type_col[2];
     const int16_t (*l1mv0)[2], (*l1mv1)[2];
     const int8_t *l1ref0, *l1ref1;
@@ -157,6 +174,8 @@ static void pred_spatial_direct_motion(H264Context * const h, int *mb_type){
 
     assert(h->ref_list[1][0].reference&3);
 
+    await_reference_mb_row(h, &h->ref_list[1][0], s->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)
 
 
@@ -217,14 +236,17 @@ static void pred_spatial_direct_motion(H264Context * const h, int *mb_type){
 
     if(IS_INTERLACED(h->ref_list[1][0].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;
             b8_stride = 0;
         }else{
-            mb_xy += h->col_fieldoff; // non zero for FL -> FL & differ parity
+            mb_y  += h->col_fieldoff;
+            mb_xy += s->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_type_col[0] = h->ref_list[1][0].mb_type[mb_xy];
             mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy + s->mb_stride];
@@ -260,6 +282,8 @@ single_col:
         }
     }
 
+    await_reference_mb_row(h, &h->ref_list[1][0], mb_y);
+
     l1mv0  = &h->ref_list[1][0].motion_val[0][h->mb2b_xy [mb_xy]];
     l1mv1  = &h->ref_list[1][0].motion_val[1][h->mb2b_xy [mb_xy]];
     l1ref0 = &h->ref_list[1][0].ref_index [0][4*mb_xy];
@@ -384,7 +408,7 @@ 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;
+    int mb_xy = h->mb_xy, mb_y = s->mb_y;
     int mb_type_col[2];
     const int16_t (*l1mv0)[2], (*l1mv1)[2];
     const int8_t *l1ref0, *l1ref1;
@@ -394,16 +418,21 @@ static void pred_temp_direct_motion(H264Context * const h, int *mb_type){
 
     assert(h->ref_list[1][0].reference&3);
 
+    await_reference_mb_row(h, &h->ref_list[1][0], s->mb_y + !!IS_INTERLACED(*mb_type));
+
     if(IS_INTERLACED(h->ref_list[1][0].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;
             b8_stride = 0;
         }else{
-            mb_xy += h->col_fieldoff; // non zero for FL -> FL & differ parity
+            mb_y  += h->col_fieldoff;
+            mb_xy += s->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_type_col[0] = h->ref_list[1][0].mb_type[mb_xy];
             mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy + s->mb_stride];
@@ -440,6 +469,8 @@ single_col:
         }
     }
 
+    await_reference_mb_row(h, &h->ref_list[1][0], mb_y);
+
     l1mv0  = &h->ref_list[1][0].motion_val[0][h->mb2b_xy [mb_xy]];
     l1mv1  = &h->ref_list[1][0].motion_val[1][h->mb2b_xy [mb_xy]];
     l1ref0 = &h->ref_list[1][0].ref_index [0][4*mb_xy];
diff --git a/libavcodec/mdec.c b/libavcodec/mdec.c
index 12d1751..b28ac26 100644
--- a/libavcodec/mdec.c
+++ b/libavcodec/mdec.c
@@ -31,6 +31,7 @@
 #include "dsputil.h"
 #include "mpegvideo.h"
 #include "mpeg12.h"
+#include "thread.h"
 
 typedef struct MDECContext{
     AVCodecContext *avctx;
@@ -162,10 +163,10 @@ static int decode_frame(AVCodecContext *avctx,
     int i;
 
     if(p->data[0])
-        avctx->release_buffer(avctx, p);
+        ff_thread_release_buffer(avctx, p);
 
     p->reference= 0;
-    if(avctx->get_buffer(avctx, p) < 0){
+    if(ff_thread_get_buffer(avctx, p) < 0){
         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
         return -1;
     }
@@ -238,6 +239,18 @@ static av_cold int decode_init(AVCodecContext *avctx){
     return 0;
 }
 
+static av_cold int decode_init_thread_copy(AVCodecContext *avctx){
+    MDECContext * const a = avctx->priv_data;
+    AVFrame *p = (AVFrame*)&a->picture;
+
+    avctx->coded_frame= p;
+    a->avctx= avctx;
+
+    p->qscale_table= av_mallocz( p->qstride * a->mb_height);
+
+    return 0;
+}
+
 static av_cold int decode_end(AVCodecContext *avctx){
     MDECContext * const a = avctx->priv_data;
 
@@ -259,7 +272,8 @@ AVCodec ff_mdec_decoder = {
     NULL,
     decode_end,
     decode_frame,
-    CODEC_CAP_DR1,
+    CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,
     .long_name= NULL_IF_CONFIG_SMALL("Sony PlayStation MDEC (Motion DECoder)"),
+    .init_thread_copy= ONLY_IF_THREADS_ENABLED(decode_init_thread_copy)
 };
 
diff --git a/libavcodec/mimic.c b/libavcodec/mimic.c
index 450dd65..2f3f9d9 100644
--- a/libavcodec/mimic.c
+++ b/libavcodec/mimic.c
@@ -27,6 +27,7 @@
 #include "get_bits.h"
 #include "bytestream.h"
 #include "dsputil.h"
+#include "thread.h"
 
 #define MIMIC_HEADER_SIZE   20
 
@@ -51,6 +52,10 @@ typedef struct {
     ScanTable       scantable;
     DSPContext      dsp;
     VLC             vlc;
+
+    /* Kept in the context so multithreading can have a constant to read from */
+    int             next_cur_index;
+    int             next_prev_index;
 } MimicContext;
 
 static const uint32_t huffcodes[] = {
@@ -121,6 +126,23 @@ static av_cold int mimic_decode_init(AVCodecContext *avctx)
     return 0;
 }
 
+static int mimic_decode_update_thread_context(AVCodecContext *avctx, const AVCodecContext *avctx_from)
+{
+    MimicContext *dst = avctx->priv_data, *src = avctx_from->priv_data;
+
+    if (avctx == avctx_from) return 0;
+
+    dst->cur_index  = src->next_cur_index;
+    dst->prev_index = src->next_prev_index;
+
+    memcpy(dst->buf_ptrs, src->buf_ptrs, sizeof(src->buf_ptrs));
+    memcpy(dst->flipped_ptrs, src->flipped_ptrs, sizeof(src->flipped_ptrs));
+
+    memset(&dst->buf_ptrs[dst->cur_index], 0, sizeof(AVFrame));
+
+    return 0;
+}
+
 static const int8_t vlcdec_lookup[9][64] = {
     {    0, },
     {   -1,   1, },
@@ -205,7 +227,7 @@ static int vlc_decode_block(MimicContext *ctx, int num_coeffs, int qscale)
 static int decode(MimicContext *ctx, int quality, int num_coeffs,
                   int is_iframe)
 {
-    int y, x, plane;
+    int y, x, plane, cur_row = 0;
 
     for(plane = 0; plane < 3; plane++) {
         const int is_chroma = !!plane;
@@ -236,6 +258,7 @@ static int decode(MimicContext *ctx, int quality, int num_coeffs,
                         int index = (ctx->cur_index+backref)&15;
                         uint8_t *p = ctx->flipped_ptrs[index].data[0];
 
+                        ff_thread_await_progress(&ctx->buf_ptrs[index], cur_row, 0);
                         if(p) {
                             p += src -
                                 ctx->flipped_ptrs[ctx->prev_index].data[plane];
@@ -246,6 +269,7 @@ static int decode(MimicContext *ctx, int quality, int num_coeffs,
                         }
                     }
                 } else {
+                    ff_thread_await_progress(&ctx->buf_ptrs[ctx->prev_index], cur_row, 0);
                     ctx->dsp.put_pixels_tab[1][0](dst, src, stride, 8);
                 }
                 src += 8;
@@ -253,6 +277,8 @@ static int decode(MimicContext *ctx, int quality, int num_coeffs,
             }
             src += (stride - ctx->num_hblocks[plane])<<3;
             dst += (stride - ctx->num_hblocks[plane])<<3;
+
+            ff_thread_report_progress(&ctx->buf_ptrs[ctx->cur_index], cur_row++, 0);
         }
     }
 
@@ -326,14 +352,20 @@ static int mimic_decode_frame(AVCodecContext *avctx, void *data,
     }
 
     ctx->buf_ptrs[ctx->cur_index].reference = 1;
-    if(avctx->get_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index])) {
+    ctx->buf_ptrs[ctx->cur_index].pict_type = is_pframe ? FF_P_TYPE:FF_I_TYPE;
+    if(ff_thread_get_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index])) {
         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
         return -1;
     }
 
+    ctx->next_prev_index = ctx->cur_index;
+    ctx->next_cur_index  = (ctx->cur_index - 1) & 15;
+
     prepare_avpic(ctx, &ctx->flipped_ptrs[ctx->cur_index],
                   (AVPicture*) &ctx->buf_ptrs[ctx->cur_index]);
 
+    ff_thread_finish_setup(avctx);
+
     av_fast_malloc(&ctx->swap_buf, &ctx->swap_buf_size,
                                  swap_buf_size + FF_INPUT_BUFFER_PADDING_SIZE);
     if(!ctx->swap_buf)
@@ -345,21 +377,23 @@ static int mimic_decode_frame(AVCodecContext *avctx, void *data,
     init_get_bits(&ctx->gb, ctx->swap_buf, swap_buf_size << 3);
 
     if(!decode(ctx, quality, num_coeffs, !is_pframe)) {
-        avctx->release_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index]);
-        return -1;
+        if (avctx->active_thread_type&FF_THREAD_FRAME)
+            ff_thread_report_progress(&ctx->buf_ptrs[ctx->cur_index], INT_MAX, 0);
+        else {
+            ff_thread_release_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index]);
+            return -1;
+        }
     }
 
-    ctx->buf_ptrs[ctx->cur_index].pict_type = is_pframe ? FF_P_TYPE:FF_I_TYPE;
     *(AVFrame*)data = ctx->buf_ptrs[ctx->cur_index];
     *data_size = sizeof(AVFrame);
 
-    ctx->prev_index = ctx->cur_index;
-    ctx->cur_index--;
-    ctx->cur_index &= 15;
+    ctx->prev_index = ctx->next_prev_index;
+    ctx->cur_index  = ctx->next_cur_index;
 
     /* Only release frames that aren't used for backreferences anymore */
     if(ctx->buf_ptrs[ctx->cur_index].data[0])
-        avctx->release_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index]);
+        ff_thread_release_buffer(avctx, &ctx->buf_ptrs[ctx->cur_index]);
 
     return buf_size;
 }
@@ -370,9 +404,12 @@ static av_cold int mimic_decode_end(AVCodecContext *avctx)
     int i;
 
     av_free(ctx->swap_buf);
+
+    if(avctx->is_copy) return 0;
+
     for(i = 0; i < 16; i++)
         if(ctx->buf_ptrs[i].data[0])
-            avctx->release_buffer(avctx, &ctx->buf_ptrs[i]);
+            ff_thread_release_buffer(avctx, &ctx->buf_ptrs[i]);
     free_vlc(&ctx->vlc);
 
     return 0;
@@ -387,6 +424,7 @@ AVCodec ff_mimic_decoder = {
     NULL,
     mimic_decode_end,
     mimic_decode_frame,
-    CODEC_CAP_DR1,
+    CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,
     .long_name = NULL_IF_CONFIG_SMALL("Mimic"),
+    .update_thread_context = ONLY_IF_THREADS_ENABLED(mimic_decode_update_thread_context)
 };
diff --git a/libavcodec/mpeg12.c b/libavcodec/mpeg12.c
index e33f597..76995ad 100644
--- a/libavcodec/mpeg12.c
+++ b/libavcodec/mpeg12.c
@@ -37,6 +37,7 @@
 #include "bytestream.h"
 #include "vdpau_internal.h"
 #include "xvmc_internal.h"
+#include "thread.h"
 
 //#undef NDEBUG
 //#include <assert.h>
@@ -1179,6 +1180,27 @@ static av_cold int mpeg_decode_init(AVCodecContext *avctx)
     return 0;
 }
 
+static int mpeg_decode_update_thread_context(AVCodecContext *avctx, const AVCodecContext *avctx_from)
+{
+    Mpeg1Context *ctx = avctx->priv_data, *ctx_from = avctx_from->priv_data;
+    MpegEncContext *s = &ctx->mpeg_enc_ctx, *s1 = &ctx_from->mpeg_enc_ctx;
+    int err;
+
+    if(avctx == avctx_from || !ctx_from->mpeg_enc_ctx_allocated || !s1->context_initialized)
+        return 0;
+
+    err = ff_mpeg_update_thread_context(avctx, avctx_from);
+    if(err) return err;
+
+    if(!ctx->mpeg_enc_ctx_allocated)
+        memcpy(s + 1, s1 + 1, sizeof(Mpeg1Context) - sizeof(MpegEncContext));
+
+    if(!(s->pict_type == FF_B_TYPE || s->low_delay))
+        s->picture_number++;
+
+    return 0;
+}
+
 static void quant_matrix_rebuild(uint16_t *matrix, const uint8_t *old_perm,
                                      const uint8_t *new_perm){
     uint16_t temp_matrix[64];
@@ -1595,6 +1617,9 @@ static int mpeg_field_start(MpegEncContext *s, const uint8_t *buf, int buf_size)
         }
 
         *s->current_picture_ptr->pan_scan= s1->pan_scan;
+
+        if (HAVE_PTHREADS && avctx->active_thread_type&FF_THREAD_FRAME)
+            ff_thread_finish_setup(avctx);
     }else{ //second field
             int i;
 
@@ -1769,6 +1794,7 @@ static int mpeg_decode_slice(Mpeg1Context *s1, int mb_y,
             const int mb_size= 16>>s->avctx->lowres;
 
             ff_draw_horiz_band(s, mb_size*(s->mb_y>>field_pic), mb_size);
+            MPV_report_decode_progress(s);
 
             s->mb_x = 0;
             s->mb_y += 1<<field_pic;
@@ -1924,7 +1950,8 @@ static int slice_end(AVCodecContext *avctx, AVFrame *pict)
             *pict= *(AVFrame*)s->current_picture_ptr;
             ff_print_debug_info(s, pict);
         } else {
-            s->picture_number++;
+            if (avctx->active_thread_type&FF_THREAD_FRAME)
+                s->picture_number++;
             /* latency of 1 frame for I- and P-frames */
             /* XXX: use another variable than picture_number */
             if (s->last_picture_ptr != NULL) {
@@ -2260,7 +2287,7 @@ static int decode_chunks(AVCodecContext *avctx,
         buf_ptr = ff_find_start_code(buf_ptr,buf_end, &start_code);
         if (start_code > 0x1ff){
             if(s2->pict_type != FF_B_TYPE || avctx->skip_frame <= AVDISCARD_DEFAULT){
-                if(avctx->thread_count > 1){
+                if(HAVE_THREADS && avctx->active_thread_type&FF_THREAD_SLICE){
                     int i;
 
                     avctx->execute(avctx, slice_decode_thread,  &s2->thread_context[0], NULL, s->slice_count, sizeof(void*));
@@ -2422,7 +2449,7 @@ static int decode_chunks(AVCodecContext *avctx,
                     break;
                 }
 
-                if(avctx->thread_count > 1){
+                if(HAVE_THREADS && avctx->active_thread_type&FF_THREAD_SLICE){
                     int threshold= (s2->mb_height*s->slice_count + avctx->thread_count/2) / avctx->thread_count;
                     if(threshold <= mb_y){
                         MpegEncContext *thread_context= s2->thread_context[s->slice_count];
@@ -2496,6 +2523,7 @@ AVCodec ff_mpeg1video_decoder = {
     .flush= flush,
     .max_lowres= 3,
     .long_name= NULL_IF_CONFIG_SMALL("MPEG-1 video"),
+    .update_thread_context= ONLY_IF_THREADS_ENABLED(mpeg_decode_update_thread_context)
 };
 
 AVCodec ff_mpeg2video_decoder = {
@@ -2532,7 +2560,7 @@ AVCodec ff_mpegvideo_decoder = {
 
 #if CONFIG_MPEG_XVMC_DECODER
 static av_cold int mpeg_mc_decode_init(AVCodecContext *avctx){
-    if( avctx->thread_count > 1)
+    if( avctx->active_thread_type & FF_THREAD_SLICE )
         return -1;
     if( !(avctx->slice_flags & SLICE_FLAG_CODED_ORDER) )
         return -1;
diff --git a/libavcodec/mpeg4videodec.c b/libavcodec/mpeg4videodec.c
index c152b3c..b293bad 100644
--- a/libavcodec/mpeg4videodec.c
+++ b/libavcodec/mpeg4videodec.c
@@ -23,6 +23,7 @@
 #include "mpegvideo.h"
 #include "mpeg4video.h"
 #include "h263.h"
+#include "thread.h"
 
 // The defines below define the number of bits that are read at once for
 // reading vlc values. Changing these may improve speed and data cache needs
@@ -373,7 +374,13 @@ int mpeg4_decode_video_packet_header(MpegEncContext *s)
         return -1;
     }
     if(s->pict_type == FF_B_TYPE){
-        while(s->next_picture.mbskip_table[ s->mb_index2xy[ mb_num ] ]) mb_num++;
+        int mb_x = 0, mb_y = 0;
+
+        while(s->next_picture.mbskip_table[ s->mb_index2xy[ mb_num ] ]) {
+            if (!mb_x) ff_thread_await_progress((AVFrame*)s->next_picture_ptr, mb_y++, 0);
+            mb_num++;
+            if (++mb_x == s->mb_width) mb_x = 0;
+        }
         if(mb_num >= s->mb_num) return -1; // slice contains just skipped MBs which where already decoded
     }
 
@@ -1303,6 +1310,8 @@ static int mpeg4_decode_mb(MpegEncContext *s,
                 s->last_mv[i][1][0]=
                 s->last_mv[i][1][1]= 0;
             }
+
+            ff_thread_await_progress((AVFrame*)s->next_picture_ptr, s->mb_y, 0);
         }
 
         /* if we skipped it in the future P Frame than skip it now too */
@@ -1482,6 +1491,12 @@ end:
     if(s->codec_id==CODEC_ID_MPEG4){
         if(mpeg4_is_resync(s)){
             const int delta= s->mb_x + 1 == s->mb_width ? 2 : 1;
+
+            if(s->pict_type==FF_B_TYPE){
+                ff_thread_await_progress((AVFrame*)s->next_picture_ptr,
+                                        (s->mb_x + delta >= s->mb_width) ? FFMIN(s->mb_y+1, s->mb_height-1) : s->mb_y, 0);
+            }
+
             if(s->pict_type==FF_B_TYPE && s->next_picture.mbskip_table[xy + delta])
                 return SLICE_OK;
             return SLICE_END;
@@ -2239,11 +2254,12 @@ AVCodec ff_mpeg4_decoder = {
     NULL,
     ff_h263_decode_end,
     ff_h263_decode_frame,
-    CODEC_CAP_DRAW_HORIZ_BAND | CODEC_CAP_DR1 | CODEC_CAP_TRUNCATED | CODEC_CAP_DELAY,
+    CODEC_CAP_DRAW_HORIZ_BAND | CODEC_CAP_DR1 | CODEC_CAP_TRUNCATED | CODEC_CAP_DELAY | CODEC_CAP_FRAME_THREADS,
     .flush= ff_mpeg_flush,
     .max_lowres= 3,
     .long_name= NULL_IF_CONFIG_SMALL("MPEG-4 part 2"),
     .pix_fmts= ff_hwaccel_pixfmt_list_420,
+    .update_thread_context= ONLY_IF_THREADS_ENABLED(ff_mpeg_update_thread_context)
 };
 
 
diff --git a/libavcodec/mpegvideo.c b/libavcodec/mpegvideo.c
index 6ba7142..10a488b 100644
--- a/libavcodec/mpegvideo.c
+++ b/libavcodec/mpegvideo.c
@@ -38,6 +38,7 @@
 #include "msmpeg4.h"
 #include "faandct.h"
 #include "xvmc_internal.h"
+#include "thread.h"
 #include <limits.h>
 
 //#undef NDEBUG
@@ -205,7 +206,7 @@ void ff_copy_picture(Picture *dst, Picture *src){
  */
 static void free_frame_buffer(MpegEncContext *s, Picture *pic)
 {
-    s->avctx->release_buffer(s->avctx, (AVFrame*)pic);
+    ff_thread_release_buffer(s->avctx, (AVFrame*)pic);
     av_freep(&pic->hwaccel_picture_private);
 }
 
@@ -227,7 +228,7 @@ static int alloc_frame_buffer(MpegEncContext *s, Picture *pic)
         }
     }
 
-    r = s->avctx->get_buffer(s->avctx, (AVFrame*)pic);
+    r = ff_thread_get_buffer(s->avctx, (AVFrame*)pic);
 
     if (r<0 || !pic->age || !pic->type || !pic->data[0]) {
         av_log(s->avctx, AV_LOG_ERROR, "get_buffer() failed (%d %d %d %p)\n", r, pic->age, pic->type, pic->data[0]);
@@ -458,6 +459,81 @@ void ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src){
 //STOP_TIMER("update_duplicate_context") //about 10k cycles / 0.01 sec for 1000frames on 1ghz with 2 threads
 }
 
+int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src)
+{
+    MpegEncContext *s = dst->priv_data, *s1 = src->priv_data;
+
+    if(dst == src || !s1->context_initialized) return 0;
+
+    //FIXME can parameters change on I-frames? in that case dst may need a reinit
+    if(!s->context_initialized){
+        memcpy(s, s1, sizeof(MpegEncContext));
+
+        s->avctx                 = dst;
+        s->picture_range_start  += MAX_PICTURE_COUNT;
+        s->picture_range_end    += MAX_PICTURE_COUNT;
+        s->bitstream_buffer      = NULL;
+        s->bitstream_buffer_size = s->allocated_bitstream_buffer_size = 0;
+
+        MPV_common_init(s);
+    }
+
+    s->avctx->coded_height  = s1->avctx->coded_height;
+    s->avctx->coded_width   = s1->avctx->coded_width;
+    s->avctx->width         = s1->avctx->width;
+    s->avctx->height        = s1->avctx->height;
+
+    s->coded_picture_number = s1->coded_picture_number;
+    s->picture_number       = s1->picture_number;
+    s->input_picture_number = s1->input_picture_number;
+
+    memcpy(s->picture, s1->picture, s1->picture_count * sizeof(Picture));
+    memcpy(&s->last_picture, &s1->last_picture, (char*)&s1->last_picture_ptr - (char*)&s1->last_picture);
+
+    s->last_picture_ptr     = REBASE_PICTURE(s1->last_picture_ptr,    s, s1);
+    s->current_picture_ptr  = REBASE_PICTURE(s1->current_picture_ptr, s, s1);
+    s->next_picture_ptr     = REBASE_PICTURE(s1->next_picture_ptr,    s, s1);
+
+    memcpy(s->prev_pict_types, s1->prev_pict_types, PREV_PICT_TYPES_BUFFER_SIZE);
+
+    //Error/bug resilience
+    s->next_p_frame_damaged = s1->next_p_frame_damaged;
+    s->workaround_bugs      = s1->workaround_bugs;
+
+    //MPEG4 timing info
+    memcpy(&s->time_increment_bits, &s1->time_increment_bits, (char*)&s1->shape - (char*)&s1->time_increment_bits);
+
+    //B-frame info
+    s->max_b_frames         = s1->max_b_frames;
+    s->low_delay            = s1->low_delay;
+    s->dropable             = s1->dropable;
+
+    //DivX handling (doesn't work)
+    s->divx_packed          = s1->divx_packed;
+
+    if(s1->bitstream_buffer){
+        if (s1->bitstream_buffer_size + FF_INPUT_BUFFER_PADDING_SIZE > s->allocated_bitstream_buffer_size)
+            av_fast_malloc(&s->bitstream_buffer, &s->allocated_bitstream_buffer_size, s1->allocated_bitstream_buffer_size);
+        s->bitstream_buffer_size  = s1->bitstream_buffer_size;
+        memcpy(s->bitstream_buffer, s1->bitstream_buffer, s1->bitstream_buffer_size);
+        memset(s->bitstream_buffer+s->bitstream_buffer_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);
+    }
+
+    //MPEG2/interlacing info
+    memcpy(&s->progressive_sequence, &s1->progressive_sequence, (char*)&s1->rtp_mode - (char*)&s1->progressive_sequence);
+
+    if(!s1->first_field){
+        s->last_pict_type= s1->pict_type;
+        if (s1->current_picture_ptr) s->last_lambda_for[s1->pict_type] = s1->current_picture_ptr->quality;
+
+        if(s1->pict_type!=FF_B_TYPE){
+            s->last_non_b_pict_type= s1->pict_type;
+        }
+    }
+
+    return 0;
+}
+
 /**
  * sets the given MpegEncContext to common defaults (same for encoding and decoding).
  * the changed fields will not depend upon the prior state of the MpegEncContext.
@@ -478,6 +554,9 @@ void MPV_common_defaults(MpegEncContext *s){
 
     s->f_code = 1;
     s->b_code = 1;
+
+    s->picture_range_start = 0;
+    s->picture_range_end = MAX_PICTURE_COUNT;
 }
 
 /**
@@ -506,7 +585,8 @@ av_cold int MPV_common_init(MpegEncContext *s)
         return -1;
     }
 
-    if(s->avctx->thread_count > MAX_THREADS || (s->avctx->thread_count > s->mb_height && s->mb_height)){
+    if(s->avctx->active_thread_type&FF_THREAD_SLICE &&
+       (s->avctx->thread_count > MAX_THREADS || (s->avctx->thread_count > s->mb_height && s->mb_height))){
         av_log(s->avctx, AV_LOG_ERROR, "too many threads\n");
         return -1;
     }
@@ -599,8 +679,9 @@ av_cold int MPV_common_init(MpegEncContext *s)
             FF_ALLOCZ_OR_GOTO(s->avctx, s->dct_offset, 2 * 64 * sizeof(uint16_t), fail)
         }
     }
-    FF_ALLOCZ_OR_GOTO(s->avctx, s->picture, MAX_PICTURE_COUNT * sizeof(Picture), fail)
-    for(i = 0; i < MAX_PICTURE_COUNT; i++) {
+    s->picture_count = MAX_PICTURE_COUNT * FFMAX(1, s->avctx->thread_count);
+    FF_ALLOCZ_OR_GOTO(s->avctx, s->picture, s->picture_count * sizeof(Picture), fail)
+    for(i = 0; i < s->picture_count; i++) {
         avcodec_get_frame_defaults((AVFrame *)&s->picture[i]);
     }
 
@@ -660,20 +741,26 @@ av_cold int MPV_common_init(MpegEncContext *s)
     }
 
     s->context_initialized = 1;
-
     s->thread_context[0]= s;
-    threads = s->avctx->thread_count;
 
-    for(i=1; i<threads; i++){
-        s->thread_context[i]= av_malloc(sizeof(MpegEncContext));
-        memcpy(s->thread_context[i], s, sizeof(MpegEncContext));
-    }
+    if (HAVE_THREADS && s->avctx->active_thread_type&FF_THREAD_SLICE) {
+        threads = s->avctx->thread_count;
 
-    for(i=0; i<threads; i++){
-        if(init_duplicate_context(s->thread_context[i], s) < 0)
-           goto fail;
-        s->thread_context[i]->start_mb_y= (s->mb_height*(i  ) + s->avctx->thread_count/2) / s->avctx->thread_count;
-        s->thread_context[i]->end_mb_y  = (s->mb_height*(i+1) + s->avctx->thread_count/2) / s->avctx->thread_count;
+        for(i=1; i<threads; i++){
+            s->thread_context[i]= av_malloc(sizeof(MpegEncContext));
+            memcpy(s->thread_context[i], s, sizeof(MpegEncContext));
+        }
+
+        for(i=0; i<threads; i++){
+            if(init_duplicate_context(s->thread_context[i], s) < 0)
+               goto fail;
+            s->thread_context[i]->start_mb_y= (s->mb_height*(i  ) + s->avctx->thread_count/2) / s->avctx->thread_count;
+            s->thread_context[i]->end_mb_y  = (s->mb_height*(i+1) + s->avctx->thread_count/2) / s->avctx->thread_count;
+        }
+    } else {
+        if(init_duplicate_context(s, s) < 0) goto fail;
+        s->start_mb_y = 0;
+        s->end_mb_y   = s->mb_height;
     }
 
     return 0;
@@ -687,12 +774,14 @@ void MPV_common_end(MpegEncContext *s)
 {
     int i, j, k;
 
-    for(i=0; i<s->avctx->thread_count; i++){
-        free_duplicate_context(s->thread_context[i]);
-    }
-    for(i=1; i<s->avctx->thread_count; i++){
-        av_freep(&s->thread_context[i]);
-    }
+    if (HAVE_THREADS && s->avctx->active_thread_type&FF_THREAD_SLICE) {
+        for(i=0; i<s->avctx->thread_count; i++){
+            free_duplicate_context(s->thread_context[i]);
+        }
+        for(i=1; i<s->avctx->thread_count; i++){
+            av_freep(&s->thread_context[i]);
+        }
+    } else free_duplicate_context(s);
 
     av_freep(&s->parse_context.buffer);
     s->parse_context.buffer_size=0;
@@ -747,8 +836,8 @@ void MPV_common_end(MpegEncContext *s)
     av_freep(&s->reordered_input_picture);
     av_freep(&s->dct_offset);
 
-    if(s->picture){
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
+    if(s->picture && !s->avctx->is_copy){
+        for(i=0; i<s->picture_count; i++){
             free_picture(s, &s->picture[i]);
         }
     }
@@ -762,7 +851,8 @@ void MPV_common_end(MpegEncContext *s)
     for(i=0; i<3; i++)
         av_freep(&s->visualization_buffer[i]);
 
-    avcodec_default_free_buffers(s->avctx);
+    if(!(s->avctx->active_thread_type&FF_THREAD_FRAME))
+        avcodec_default_free_buffers(s->avctx);
 }
 
 void init_rl(RLTable *rl, uint8_t static_store[2][2*MAX_RUN + MAX_LEVEL + 3])
@@ -860,14 +950,14 @@ int ff_find_unused_picture(MpegEncContext *s, int shared){
     int i;
 
     if(shared){
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
+        for(i=s->picture_range_start; i<s->picture_range_end; i++){
             if(s->picture[i].data[0]==NULL && s->picture[i].type==0) return i;
         }
     }else{
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
+        for(i=s->picture_range_start; i<s->picture_range_end; i++){
             if(s->picture[i].data[0]==NULL && s->picture[i].type!=0) return i; //FIXME
         }
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
+        for(i=s->picture_range_start; i<s->picture_range_end; i++){
             if(s->picture[i].data[0]==NULL) return i;
         }
     }
@@ -924,7 +1014,7 @@ int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx)
         /* release forgotten pictures */
         /* if(mpeg124/h263) */
         if(!s->encoding){
-            for(i=0; i<MAX_PICTURE_COUNT; i++){
+            for(i=0; i<s->picture_count; i++){
                 if(s->picture[i].data[0] && &s->picture[i] != s->next_picture_ptr && s->picture[i].reference){
                     av_log(avctx, AV_LOG_ERROR, "releasing zombie picture\n");
                     free_frame_buffer(s, &s->picture[i]);
@@ -936,7 +1026,7 @@ int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx)
 
     if(!s->encoding){
         /* release non reference frames */
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
+        for(i=0; i<s->picture_count; i++){
             if(s->picture[i].data[0] && !s->picture[i].reference /*&& s->picture[i].type!=FF_BUFFER_TYPE_SHARED*/){
                 free_frame_buffer(s, &s->picture[i]);
             }
@@ -970,6 +1060,7 @@ int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx)
                 s->current_picture_ptr->top_field_first= (s->picture_structure == PICT_TOP_FIELD) == s->first_field;
         }
         s->current_picture_ptr->interlaced_frame= !s->progressive_frame && !s->progressive_sequence;
+        s->current_picture_ptr->field_picture= s->picture_structure != PICT_FRAME;
     }
 
     s->current_picture_ptr->pict_type= s->pict_type;
@@ -998,6 +1089,8 @@ int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx)
             s->last_picture_ptr= &s->picture[i];
             if(ff_alloc_picture(s, s->last_picture_ptr, 0) < 0)
                 return -1;
+            ff_thread_report_progress((AVFrame*)s->last_picture_ptr, INT_MAX, 0);
+            ff_thread_report_progress((AVFrame*)s->last_picture_ptr, INT_MAX, 1);
         }
         if((s->next_picture_ptr==NULL || s->next_picture_ptr->data[0]==NULL) && s->pict_type==FF_B_TYPE){
             /* Allocate a dummy frame */
@@ -1005,6 +1098,8 @@ int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx)
             s->next_picture_ptr= &s->picture[i];
             if(ff_alloc_picture(s, s->next_picture_ptr, 0) < 0)
                 return -1;
+            ff_thread_report_progress((AVFrame*)s->next_picture_ptr, INT_MAX, 0);
+            ff_thread_report_progress((AVFrame*)s->next_picture_ptr, INT_MAX, 1);
         }
     }
 
@@ -1057,7 +1152,7 @@ int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx)
 void MPV_frame_end(MpegEncContext *s)
 {
     int i;
-    /* draw edge for correct motion prediction if outside */
+    /* redraw edges for the frame if decoding didn't complete */
     //just to make sure that all data is rendered.
     if(CONFIG_MPEG_XVMC_DECODER && s->avctx->xvmc_acceleration){
         ff_xvmc_field_end(s);
@@ -1067,10 +1162,14 @@ void MPV_frame_end(MpegEncContext *s)
        && s->current_picture.reference
        && !s->intra_only
        && !(s->flags&CODEC_FLAG_EMU_EDGE)) {
-            s->dsp.draw_edges(s->current_picture.data[0], s->linesize  , s->h_edge_pos   , s->v_edge_pos   , EDGE_WIDTH  );
-            s->dsp.draw_edges(s->current_picture.data[1], s->uvlinesize, s->h_edge_pos>>1, s->v_edge_pos>>1, EDGE_WIDTH/2);
-            s->dsp.draw_edges(s->current_picture.data[2], s->uvlinesize, s->h_edge_pos>>1, s->v_edge_pos>>1, EDGE_WIDTH/2);
+        int edges = EDGE_BOTTOM | EDGE_TOP, h = s->v_edge_pos;
+
+            s->dsp.draw_edges(s->current_picture_ptr->data[0], s->linesize  , s->h_edge_pos   , h   , EDGE_WIDTH  , edges);
+            s->dsp.draw_edges(s->current_picture_ptr->data[1], s->uvlinesize, s->h_edge_pos>>1, h>>1, EDGE_WIDTH/2, edges);
+            s->dsp.draw_edges(s->current_picture_ptr->data[2], s->uvlinesize, s->h_edge_pos>>1, h>>1, EDGE_WIDTH/2, edges);
+
     }
+
     emms_c();
 
     s->last_pict_type    = s->pict_type;
@@ -1091,7 +1190,7 @@ void MPV_frame_end(MpegEncContext *s)
 
     if(s->encoding){
         /* release non-reference frames */
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
+        for(i=0; i<s->picture_count; i++){
             if(s->picture[i].data[0] && !s->picture[i].reference /*&& s->picture[i].type!=FF_BUFFER_TYPE_SHARED*/){
                 free_frame_buffer(s, &s->picture[i]);
             }
@@ -1104,6 +1203,10 @@ void MPV_frame_end(MpegEncContext *s)
     memset(&s->current_picture, 0, sizeof(Picture));
 #endif
     s->avctx->coded_frame= (AVFrame*)s->current_picture_ptr;
+
+    if (s->codec_id != CODEC_ID_H264 && s->current_picture.reference) {
+        ff_thread_report_progress((AVFrame*)s->current_picture_ptr, s->mb_height-1, 0);
+    }
 }
 
 /**
@@ -1768,6 +1871,43 @@ static inline void MPV_motion_lowres(MpegEncContext *s,
     }
 }
 
+/**
+ * find the lowest MB row referenced in the MVs
+ */
+int MPV_lowest_referenced_row(MpegEncContext *s, int dir)
+{
+    int my_max = INT_MIN, my_min = INT_MAX, qpel_shift = !s->quarter_sample;
+    int my, off, i, mvs;
+
+    if (s->picture_structure != PICT_FRAME) goto unhandled;
+
+    switch (s->mv_type) {
+        case MV_TYPE_16X16:
+            mvs = 1;
+            break;
+        case MV_TYPE_16X8:
+            mvs = 2;
+            break;
+        case MV_TYPE_8X8:
+            mvs = 4;
+            break;
+        default:
+            goto unhandled;
+    }
+
+    for (i = 0; i < mvs; i++) {
+        my = s->mv[dir][i][1]<<qpel_shift;
+        my_max = FFMAX(my_max, my);
+        my_min = FFMIN(my_min, my);
+    }
+
+    off = (FFMAX(-my_min, my_max) + 63) >> 6;
+
+    return FFMIN(FFMAX(s->mb_y + off, 0), s->mb_height-1);
+unhandled:
+    return s->mb_height-1;
+}
+
 /* put block[] to dest[] */
 static inline void put_dct(MpegEncContext *s,
                            DCTELEM *block, int i, uint8_t *dest, int line_size, int qscale)
@@ -1932,6 +2072,16 @@ void MPV_decode_mb_internal(MpegEncContext *s, DCTELEM block[12][64],
             /* motion handling */
             /* decoding or more than one mb_type (MC was already done otherwise) */
             if(!s->encoding){
+
+                if(HAVE_PTHREADS && s->avctx->active_thread_type&FF_THREAD_FRAME) {
+                    if (s->mv_dir & MV_DIR_FORWARD) {
+                        ff_thread_await_progress((AVFrame*)s->last_picture_ptr, MPV_lowest_referenced_row(s, 0), 0);
+                    }
+                    if (s->mv_dir & MV_DIR_BACKWARD) {
+                        ff_thread_await_progress((AVFrame*)s->next_picture_ptr, MPV_lowest_referenced_row(s, 1), 0);
+                    }
+                }
+
                 if(lowres_flag){
                     h264_chroma_mc_func *op_pix = s->dsp.put_h264_chroma_pixels_tab;
 
@@ -2096,19 +2246,37 @@ void MPV_decode_mb(MpegEncContext *s, DCTELEM block[12][64]){
  * @param h is the normal height, this will be reduced automatically if needed for the last row
  */
 void ff_draw_horiz_band(MpegEncContext *s, int y, int h){
+	const int field_pic= s->picture_structure != PICT_FRAME;
+    if(field_pic){
+        h <<= 1;
+        y <<= 1;
+    }
+
+    if (!s->avctx->hwaccel
+       && !(s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
+       && s->unrestricted_mv
+       && s->current_picture.reference
+       && !s->intra_only
+       && !(s->flags&CODEC_FLAG_EMU_EDGE)) {
+        int sides = 0, edge_h;
+        if (y==0) sides |= EDGE_TOP;
+        if (y + h >= s->v_edge_pos) sides |= EDGE_BOTTOM;
+
+        edge_h= FFMIN(h, s->v_edge_pos - y);
+
+        s->dsp.draw_edges(s->current_picture_ptr->data[0] +  y    *s->linesize  , s->linesize  , s->h_edge_pos   , edge_h   , EDGE_WIDTH  , sides);
+        s->dsp.draw_edges(s->current_picture_ptr->data[1] + (y>>1)*s->uvlinesize, s->uvlinesize, s->h_edge_pos>>1, edge_h>>1, EDGE_WIDTH/2, sides);
+        s->dsp.draw_edges(s->current_picture_ptr->data[2] + (y>>1)*s->uvlinesize, s->uvlinesize, s->h_edge_pos>>1, edge_h>>1, EDGE_WIDTH/2, sides);
+    }
+
+    h= FFMIN(h, s->avctx->height - y);
+
+    if(field_pic && s->first_field && !(s->avctx->slice_flags&SLICE_FLAG_ALLOW_FIELD)) return;
+
     if (s->avctx->draw_horiz_band) {
         AVFrame *src;
-        const int field_pic= s->picture_structure != PICT_FRAME;
         int offset[4];
 
-        h= FFMIN(h, (s->avctx->height>>field_pic) - y);
-
-        if(field_pic && !(s->avctx->slice_flags&SLICE_FLAG_ALLOW_FIELD)){
-            h <<= 1;
-            y <<= 1;
-            if(s->first_field) return;
-        }
-
         if(s->pict_type==FF_B_TYPE || s->low_delay || (s->avctx->slice_flags&SLICE_FLAG_CODED_ORDER))
             src= (AVFrame*)s->current_picture_ptr;
         else if(s->last_picture_ptr)
@@ -2174,7 +2342,7 @@ void ff_mpeg_flush(AVCodecContext *avctx){
     if(s==NULL || s->picture==NULL)
         return;
 
-    for(i=0; i<MAX_PICTURE_COUNT; i++){
+    for(i=0; i<s->picture_count; i++){
        if(s->picture[i].data[0] && (   s->picture[i].type == FF_BUFFER_TYPE_INTERNAL
                                     || s->picture[i].type == FF_BUFFER_TYPE_USER))
         free_frame_buffer(s, &s->picture[i]);
@@ -2428,3 +2596,9 @@ void ff_set_qscale(MpegEncContext * s, int qscale)
     s->y_dc_scale= s->y_dc_scale_table[ qscale ];
     s->c_dc_scale= s->c_dc_scale_table[ s->chroma_qscale ];
 }
+
+void MPV_report_decode_progress(MpegEncContext *s)
+{
+    if (s->pict_type != FF_B_TYPE && !s->partitioned_frame)
+        ff_thread_report_progress((AVFrame*)s->current_picture_ptr, s->mb_y, 0);
+}
diff --git a/libavcodec/mpegvideo.h b/libavcodec/mpegvideo.h
index 53ab0f4..3e447ab 100644
--- a/libavcodec/mpegvideo.h
+++ b/libavcodec/mpegvideo.h
@@ -123,6 +123,7 @@ typedef struct Picture{
     int ref_poc[2][2][16];      ///< h264 POCs of the frames used as reference (FIXME need per slice)
     int ref_count[2][2];        ///< number of entries in ref_poc              (FIXME need per slice)
     int mbaff;                  ///< h264 1 -> MBAFF frame 0-> not MBAFF
+    int field_picture;          ///< whether or not the picture was encoded in seperate fields
 
     int mb_var_sum;             ///< sum of MB variance for current frame
     int mc_mb_var_sum;          ///< motion compensated MB variance for current frame
@@ -292,6 +293,8 @@ typedef struct MpegEncContext {
     Picture *last_picture_ptr;     ///< pointer to the previous picture.
     Picture *next_picture_ptr;     ///< pointer to the next picture (for bidir pred)
     Picture *current_picture_ptr;  ///< pointer to the current picture
+    int picture_count;             ///< number of allocated pictures (MAX_PICTURE_COUNT * avctx->thread_count)
+    int picture_range_start, picture_range_end; ///< the part of picture that this context can allocate in
     uint8_t *visualization_buffer[3]; //< temporary buffer vor MV visualization
     int last_dc[3];                ///< last DC values for MPEG1
     int16_t *dc_val_base;
@@ -681,6 +684,7 @@ typedef struct MpegEncContext {
     void (*denoise_dct)(struct MpegEncContext *s, DCTELEM *block);
 } MpegEncContext;
 
+#define REBASE_PICTURE(pic, new_ctx, old_ctx) (pic ? &new_ctx->picture[pic - old_ctx->picture] : NULL)
 
 void MPV_decode_defaults(MpegEncContext *s);
 int MPV_common_init(MpegEncContext *s);
@@ -706,6 +710,9 @@ void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix);
 int ff_find_unused_picture(MpegEncContext *s, int shared);
 void ff_denoise_dct(MpegEncContext *s, DCTELEM *block);
 void ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
+int MPV_lowest_referenced_row(MpegEncContext *s, int dir);
+void MPV_report_decode_progress(MpegEncContext *s);
+int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src);
 const uint8_t *ff_find_start_code(const uint8_t *p, const uint8_t *end, uint32_t *state);
 void ff_set_qscale(MpegEncContext * s, int qscale);
 
diff --git a/libavcodec/mpegvideo_enc.c b/libavcodec/mpegvideo_enc.c
index a7f1608..9255fa8 100644
--- a/libavcodec/mpegvideo_enc.c
+++ b/libavcodec/mpegvideo_enc.c
@@ -36,6 +36,7 @@
 #include "mjpegenc.h"
 #include "msmpeg4.h"
 #include "faandct.h"
+#include "thread.h"
 #include "aandcttab.h"
 #include "flv.h"
 #include "mpeg4video.h"
@@ -1238,9 +1239,9 @@ int MPV_encode_picture(AVCodecContext *avctx,
 {
     MpegEncContext *s = avctx->priv_data;
     AVFrame *pic_arg = data;
-    int i, stuffing_count;
+    int i, stuffing_count, context_count = avctx->active_thread_type&FF_THREAD_SLICE ? avctx->thread_count : 1;
 
-    for(i=0; i<avctx->thread_count; i++){
+    for(i=0; i<context_count; i++){
         int start_y= s->thread_context[i]->start_mb_y;
         int   end_y= s->thread_context[i]->  end_mb_y;
         int h= s->mb_height;
@@ -1304,7 +1305,7 @@ vbv_retry:
                     s->last_non_b_time= s->time - s->pp_time;
                 }
 //                av_log(NULL, AV_LOG_ERROR, "R:%d ", s->next_lambda);
-                for(i=0; i<avctx->thread_count; i++){
+                for(i=0; i<context_count; i++){
                     PutBitContext *pb= &s->thread_context[i]->pb;
                     init_put_bits(pb, pb->buf, pb->buf_end - pb->buf);
                 }
@@ -2771,6 +2772,7 @@ static int encode_picture(MpegEncContext *s, int picture_number)
 {
     int i;
     int bits;
+    int context_count = s->avctx->active_thread_type&FF_THREAD_SLICE ? s->avctx->thread_count : 1;
 
     s->picture_number = picture_number;
 
@@ -2810,7 +2812,7 @@ static int encode_picture(MpegEncContext *s, int picture_number)
     }
 
     s->mb_intra=0; //for the rate distortion & bit compare functions
-    for(i=1; i<s->avctx->thread_count; i++){
+    for(i=1; i<context_count; i++){
         ff_update_duplicate_context(s->thread_context[i], s);
     }
 
@@ -2823,11 +2825,11 @@ static int encode_picture(MpegEncContext *s, int picture_number)
         s->lambda2= (s->lambda2* (int64_t)s->avctx->me_penalty_compensation + 128)>>8;
         if(s->pict_type != FF_B_TYPE && s->avctx->me_threshold==0){
             if((s->avctx->pre_me && s->last_non_b_pict_type==FF_I_TYPE) || s->avctx->pre_me==2){
-                s->avctx->execute(s->avctx, pre_estimate_motion_thread, &s->thread_context[0], NULL, s->avctx->thread_count, sizeof(void*));
+                s->avctx->execute(s->avctx, pre_estimate_motion_thread, &s->thread_context[0], NULL, context_count, sizeof(void*));
             }
         }
 
-        s->avctx->execute(s->avctx, estimate_motion_thread, &s->thread_context[0], NULL, s->avctx->thread_count, sizeof(void*));
+        s->avctx->execute(s->avctx, estimate_motion_thread, &s->thread_context[0], NULL, context_count, sizeof(void*));
     }else /* if(s->pict_type == FF_I_TYPE) */{
         /* I-Frame */
         for(i=0; i<s->mb_stride*s->mb_height; i++)
@@ -2835,10 +2837,10 @@ static int encode_picture(MpegEncContext *s, int picture_number)
 
         if(!s->fixed_qscale){
             /* finding spatial complexity for I-frame rate control */
-            s->avctx->execute(s->avctx, mb_var_thread, &s->thread_context[0], NULL, s->avctx->thread_count, sizeof(void*));
+            s->avctx->execute(s->avctx, mb_var_thread, &s->thread_context[0], NULL, context_count, sizeof(void*));
         }
     }
-    for(i=1; i<s->avctx->thread_count; i++){
+    for(i=1; i<context_count; i++){
         merge_context_after_me(s, s->thread_context[i]);
     }
     s->current_picture.mc_mb_var_sum= s->current_picture_ptr->mc_mb_var_sum= s->me.mc_mb_var_sum_temp;
@@ -2974,11 +2976,11 @@ static int encode_picture(MpegEncContext *s, int picture_number)
     bits= put_bits_count(&s->pb);
     s->header_bits= bits - s->last_bits;
 
-    for(i=1; i<s->avctx->thread_count; i++){
+    for(i=1; i<context_count; i++){
         update_duplicate_context_after_me(s->thread_context[i], s);
     }
-    s->avctx->execute(s->avctx, encode_thread, &s->thread_context[0], NULL, s->avctx->thread_count, sizeof(void*));
-    for(i=1; i<s->avctx->thread_count; i++){
+    s->avctx->execute(s->avctx, encode_thread, &s->thread_context[0], NULL, context_count, sizeof(void*));
+    for(i=1; i<context_count; i++){
         merge_context_after_encode(s, s->thread_context[i]);
     }
     emms_c();
diff --git a/libavcodec/snow.c b/libavcodec/snow.c
index 9ab9fed..f330fe3 100644
--- a/libavcodec/snow.c
+++ b/libavcodec/snow.c
@@ -1976,9 +1976,9 @@ static int frame_start(SnowContext *s){
    int h= s->avctx->height;
 
     if(s->current_picture.data[0]){
-        s->dsp.draw_edges(s->current_picture.data[0], s->current_picture.linesize[0], w   , h   , EDGE_WIDTH  );
-        s->dsp.draw_edges(s->current_picture.data[1], s->current_picture.linesize[1], w>>1, h>>1, EDGE_WIDTH/2);
-        s->dsp.draw_edges(s->current_picture.data[2], s->current_picture.linesize[2], w>>1, h>>1, EDGE_WIDTH/2);
+        s->dsp.draw_edges(s->current_picture.data[0], s->current_picture.linesize[0], w   , h   , EDGE_WIDTH  , EDGE_TOP|EDGE_BOTTOM);
+        s->dsp.draw_edges(s->current_picture.data[1], s->current_picture.linesize[1], w>>1, h>>1, EDGE_WIDTH/2, EDGE_TOP|EDGE_BOTTOM);
+        s->dsp.draw_edges(s->current_picture.data[2], s->current_picture.linesize[2], w>>1, h>>1, EDGE_WIDTH/2, EDGE_TOP|EDGE_BOTTOM);
     }
 
     release_buffer(s->avctx);
diff --git a/libavcodec/utils.c b/libavcodec/utils.c
index 40d689e..2c737d9 100644
--- a/libavcodec/utils.c
+++ b/libavcodec/utils.c
@@ -366,6 +366,7 @@ void avcodec_default_release_buffer(AVCodecContext *s, AVFrame *pic){
     assert(pic->type==FF_BUFFER_TYPE_INTERNAL);
     assert(s->internal_buffer_count);
 
+    if(s->internal_buffer){
     buf = NULL; /* avoids warning */
     for(i=0; i<s->internal_buffer_count; i++){ //just 3-5 checks so is not worth to optimize
         buf= &((InternalBuffer*)s->internal_buffer)[i];
@@ -377,6 +378,7 @@ void avcodec_default_release_buffer(AVCodecContext *s, AVFrame *pic){
     last = &((InternalBuffer*)s->internal_buffer)[s->internal_buffer_count];
 
     FFSWAP(InternalBuffer, *buf, *last);
+    }
 
     for(i=0; i<4; i++){
         pic->data[i]=NULL;
@@ -572,6 +574,7 @@ int attribute_align_arg avcodec_open(AVCodecContext *avctx, AVCodec *codec)
             goto free_and_end;
         }
     }
+
     ret=0;
 end:
     entangled_thread_counter--;
@@ -1298,9 +1301,9 @@ int av_lockmgr_register(int (*cb)(void **mutex, enum AVLockOp op))
 unsigned int ff_toupper4(unsigned int x)
 {
     return     toupper( x     &0xFF)
-            + (toupper((x>>8 )&0xFF)<<8 )
-            + (toupper((x>>16)&0xFF)<<16)
-            + (toupper((x>>24)&0xFF)<<24);
+    + (toupper((x>>8 )&0xFF)<<8 )
+    + (toupper((x>>16)&0xFF)<<16)
+    + (toupper((x>>24)&0xFF)<<24);
 }
 
 #if !HAVE_PTHREADS
diff --git a/libavcodec/vp8.c b/libavcodec/vp8.c
index c067f4c..073de8e 100644
--- a/libavcodec/vp8.c
+++ b/libavcodec/vp8.c
@@ -471,7 +471,7 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y)
                                   mb - 1 /* left */,
                                   mb + 1 /* top-left */ };
     enum { CNT_ZERO, CNT_NEAREST, CNT_NEAR, CNT_SPLITMV };
-    enum { EDGE_TOP, EDGE_LEFT, EDGE_TOPLEFT };
+    enum { VP8_EDGE_TOP, VP8_EDGE_LEFT, VP8_EDGE_TOPLEFT };
     int idx = CNT_ZERO;
     int cur_sign_bias = s->sign_bias[mb->ref_frame];
     int8_t *sign_bias = s->sign_bias;
@@ -512,7 +512,7 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y)
         mb->mode = VP8_MVMODE_MV;
 
         /* If we have three distinct MVs, merge first and last if they're the same */
-        if (cnt[CNT_SPLITMV] && AV_RN32A(&near_mv[1+EDGE_TOP]) == AV_RN32A(&near_mv[1+EDGE_TOPLEFT]))
+        if (cnt[CNT_SPLITMV] && AV_RN32A(&near_mv[1+VP8_EDGE_TOP]) == AV_RN32A(&near_mv[1+VP8_EDGE_TOPLEFT]))
             cnt[CNT_NEAREST] += 1;
 
         /* Swap near and nearest if necessary */
@@ -526,9 +526,9 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y)
 
                 /* Choose the best mv out of 0,0 and the nearest mv */
                 clamp_mv(s, &mb->mv, &near_mv[CNT_ZERO + (cnt[CNT_NEAREST] >= cnt[CNT_ZERO])]);
-                cnt[CNT_SPLITMV] = ((mb_edge[EDGE_LEFT]->mode    == VP8_MVMODE_SPLIT) +
-                                    (mb_edge[EDGE_TOP]->mode     == VP8_MVMODE_SPLIT)) * 2 +
-                                    (mb_edge[EDGE_TOPLEFT]->mode == VP8_MVMODE_SPLIT);
+                cnt[CNT_SPLITMV] = ((mb_edge[VP8_EDGE_LEFT]->mode    == VP8_MVMODE_SPLIT) +
+                                    (mb_edge[VP8_EDGE_TOP]->mode     == VP8_MVMODE_SPLIT)) * 2 +
+                                    (mb_edge[VP8_EDGE_TOPLEFT]->mode == VP8_MVMODE_SPLIT);
 
                 if (vp56_rac_get_prob_branchy(c, vp8_mode_contexts[cnt[CNT_SPLITMV]][3])) {
                     mb->mode = VP8_MVMODE_SPLIT;
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index a5b0b3a..3272556 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -783,7 +783,7 @@ static void h263_h_loop_filter_mmx(uint8_t *src, int stride, int qscale){
 
 /* draw the edges of width 'w' of an image of size width, height
    this mmx version can only handle w==8 || w==16 */
-static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w)
+static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w, int sides)
 {
     uint8_t *ptr, *last_line;
     int i;
@@ -838,34 +838,39 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w)
 
     for(i=0;i<w;i+=4) {
         /* top and bottom (and hopefully also the corners) */
-        ptr= buf - (i + 1) * wrap - w;
-        __asm__ volatile(
-                "1:                             \n\t"
-                "movq (%1, %0), %%mm0           \n\t"
-                "movq %%mm0, (%0)               \n\t"
-                "movq %%mm0, (%0, %2)           \n\t"
-                "movq %%mm0, (%0, %2, 2)        \n\t"
-                "movq %%mm0, (%0, %3)           \n\t"
-                "add $8, %0                     \n\t"
-                "cmp %4, %0                     \n\t"
-                " jb 1b                         \n\t"
-                : "+r" (ptr)
-                : "r" ((x86_reg)buf - (x86_reg)ptr - w), "r" ((x86_reg)-wrap), "r" ((x86_reg)-wrap*3), "r" (ptr+width+2*w)
-        );
-        ptr= last_line + (i + 1) * wrap - w;
-        __asm__ volatile(
-                "1:                             \n\t"
-                "movq (%1, %0), %%mm0           \n\t"
-                "movq %%mm0, (%0)               \n\t"
-                "movq %%mm0, (%0, %2)           \n\t"
-                "movq %%mm0, (%0, %2, 2)        \n\t"
-                "movq %%mm0, (%0, %3)           \n\t"
-                "add $8, %0                     \n\t"
-                "cmp %4, %0                     \n\t"
-                " jb 1b                         \n\t"
-                : "+r" (ptr)
-                : "r" ((x86_reg)last_line - (x86_reg)ptr - w), "r" ((x86_reg)wrap), "r" ((x86_reg)wrap*3), "r" (ptr+width+2*w)
-        );
+        if (sides&EDGE_TOP) {
+            ptr= buf - (i + 1) * wrap - w;
+            __asm__ volatile(
+                    "1:                             \n\t"
+                    "movq (%1, %0), %%mm0           \n\t"
+                    "movq %%mm0, (%0)               \n\t"
+                    "movq %%mm0, (%0, %2)           \n\t"
+                    "movq %%mm0, (%0, %2, 2)        \n\t"
+                    "movq %%mm0, (%0, %3)           \n\t"
+                    "add $8, %0                     \n\t"
+                    "cmp %4, %0                     \n\t"
+                    " jb 1b                         \n\t"
+                    : "+r" (ptr)
+                    : "r" ((x86_reg)buf - (x86_reg)ptr - w), "r" ((x86_reg)-wrap), "r" ((x86_reg)-wrap*3), "r" (ptr+width+2*w)
+            );
+        }
+
+        if (sides&EDGE_BOTTOM) {
+            ptr= last_line + (i + 1) * wrap - w;
+            __asm__ volatile(
+                    "1:                             \n\t"
+                    "movq (%1, %0), %%mm0           \n\t"
+                    "movq %%mm0, (%0)               \n\t"
+                    "movq %%mm0, (%0, %2)           \n\t"
+                    "movq %%mm0, (%0, %2, 2)        \n\t"
+                    "movq %%mm0, (%0, %3)           \n\t"
+                    "add $8, %0                     \n\t"
+                    "cmp %4, %0                     \n\t"
+                    " jb 1b                         \n\t"
+                    : "+r" (ptr)
+                    : "r" ((x86_reg)last_line - (x86_reg)ptr - w), "r" ((x86_reg)wrap), "r" ((x86_reg)wrap*3), "r" (ptr+width+2*w)
+            );
+        }
     }
 }