]> git.sesse.net Git - ffmpeg/blobdiff - libavcodec/mpegvideo.c
mpeg12: propagate more real return values through chunk decode error return and fix...
[ffmpeg] / libavcodec / mpegvideo.c
index 168fbacddd3af07574a593612b736be049991ecb..8d404923461371e8c6810531883205711e3752a0 100644 (file)
@@ -5,36 +5,40 @@
  *
  * 4MV & hq & B-frame encoding stuff by Michael Niedermayer <michaelni@gmx.at>
  *
- * This file is part of FFmpeg.
+ * This file is part of Libav.
  *
- * FFmpeg is free software; you can redistribute it and/or
+ * Libav is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
  * License as published by the Free Software Foundation; either
  * version 2.1 of the License, or (at your option) any later version.
  *
- * FFmpeg is distributed in the hope that it will be useful,
+ * Libav is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  * Lesser General Public License for more details.
  *
  * You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
+ * License along with Libav; if not, write to the Free Software
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
 /**
- * @file libavcodec/mpegvideo.c
+ * @file
  * The simplest mpeg encoder (well, it was the simplest!).
  */
 
+#include "libavutil/intmath.h"
+#include "libavutil/imgutils.h"
 #include "avcodec.h"
 #include "dsputil.h"
+#include "internal.h"
 #include "mpegvideo.h"
 #include "mpegvideo_common.h"
 #include "mjpegenc.h"
 #include "msmpeg4.h"
 #include "faandct.h"
 #include "xvmc_internal.h"
+#include "thread.h"
 #include <limits.h>
 
 //#undef NDEBUG
@@ -75,6 +79,48 @@ const uint8_t ff_mpeg1_dc_scale_table[128]={
     8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
 };
 
+static const uint8_t mpeg2_dc_scale_table1[128]={
+//  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
+    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+};
+
+static const uint8_t mpeg2_dc_scale_table2[128]={
+//  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
+    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+};
+
+static const uint8_t mpeg2_dc_scale_table3[128]={
+//  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+};
+
+const uint8_t * const ff_mpeg2_dc_scale_table[4]={
+    ff_mpeg1_dc_scale_table,
+    mpeg2_dc_scale_table1,
+    mpeg2_dc_scale_table2,
+    mpeg2_dc_scale_table3,
+};
+
+const enum PixelFormat ff_pixfmt_list_420[] = {
+    PIX_FMT_YUV420P,
+    PIX_FMT_NONE
+};
+
+const enum PixelFormat ff_hwaccel_pixfmt_list_420[] = {
+    PIX_FMT_DXVA2_VLD,
+    PIX_FMT_VAAPI_VLD,
+    PIX_FMT_YUV420P,
+    PIX_FMT_NONE
+};
 
 const uint8_t *ff_find_start_code(const uint8_t * restrict p, const uint8_t *end, uint32_t * restrict state){
     int i;
@@ -109,6 +155,8 @@ const uint8_t *ff_find_start_code(const uint8_t * restrict p, const uint8_t *end
 /* init common dct for both encoder and decoder */
 av_cold int ff_dct_common_init(MpegEncContext *s)
 {
+    dsputil_init(&s->dsp, s->avctx);
+
     s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_c;
     s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_c;
     s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_c;
@@ -152,14 +200,73 @@ av_cold int ff_dct_common_init(MpegEncContext *s)
 
 void ff_copy_picture(Picture *dst, Picture *src){
     *dst = *src;
-    dst->type= FF_BUFFER_TYPE_COPY;
+    dst->f.type= FF_BUFFER_TYPE_COPY;
+}
+
+/**
+ * Release a frame buffer
+ */
+static void free_frame_buffer(MpegEncContext *s, Picture *pic)
+{
+    /* Windows Media Image codecs allocate internal buffers with different
+       dimensions; ignore user defined callbacks for these */
+    if (s->codec_id != CODEC_ID_WMV3IMAGE && s->codec_id != CODEC_ID_VC1IMAGE)
+        ff_thread_release_buffer(s->avctx, (AVFrame*)pic);
+    else
+        avcodec_default_release_buffer(s->avctx, (AVFrame*)pic);
+    av_freep(&pic->f.hwaccel_picture_private);
+}
+
+/**
+ * Allocate a frame buffer
+ */
+static int alloc_frame_buffer(MpegEncContext *s, Picture *pic)
+{
+    int r;
+
+    if (s->avctx->hwaccel) {
+        assert(!pic->hwaccel_picture_private);
+        if (s->avctx->hwaccel->priv_data_size) {
+            pic->f.hwaccel_picture_private = av_mallocz(s->avctx->hwaccel->priv_data_size);
+            if (!pic->f.hwaccel_picture_private) {
+                av_log(s->avctx, AV_LOG_ERROR, "alloc_frame_buffer() failed (hwaccel private data allocation)\n");
+                return -1;
+            }
+        }
+    }
+
+    if (s->codec_id != CODEC_ID_WMV3IMAGE && s->codec_id != CODEC_ID_VC1IMAGE)
+        r = ff_thread_get_buffer(s->avctx, (AVFrame*)pic);
+    else
+        r = avcodec_default_get_buffer(s->avctx, (AVFrame*)pic);
+
+    if (r < 0 || !pic->f.age || !pic->f.type || !pic->f.data[0]) {
+        av_log(s->avctx, AV_LOG_ERROR, "get_buffer() failed (%d %d %d %p)\n",
+               r, pic->f.age, pic->f.type, pic->f.data[0]);
+        av_freep(&pic->f.hwaccel_picture_private);
+        return -1;
+    }
+
+    if (s->linesize && (s->linesize != pic->f.linesize[0] || s->uvlinesize != pic->f.linesize[1])) {
+        av_log(s->avctx, AV_LOG_ERROR, "get_buffer() failed (stride changed)\n");
+        free_frame_buffer(s, pic);
+        return -1;
+    }
+
+    if (pic->f.linesize[1] != pic->f.linesize[2]) {
+        av_log(s->avctx, AV_LOG_ERROR, "get_buffer() failed (uv stride mismatch)\n");
+        free_frame_buffer(s, pic);
+        return -1;
+    }
+
+    return 0;
 }
 
 /**
  * allocates a Picture
  * The pixels are allocated/set by calling get_buffer() if shared=0
  */
-int alloc_picture(MpegEncContext *s, Picture *pic, int shared){
+int ff_alloc_picture(MpegEncContext *s, Picture *pic, int shared){
     const int big_mb_num= s->mb_stride*(s->mb_height+1) + 1; //the +1 is needed so memset(,,stride*height) does not sig11
     const int mb_array_size= s->mb_stride*s->mb_height;
     const int b8_array_size= s->b8_stride*s->mb_height*2;
@@ -168,79 +275,65 @@ int alloc_picture(MpegEncContext *s, Picture *pic, int shared){
     int r= -1;
 
     if(shared){
-        assert(pic->data[0]);
+        assert(pic->f.data[0]);
         assert(pic->type == 0 || pic->type == FF_BUFFER_TYPE_SHARED);
-        pic->type= FF_BUFFER_TYPE_SHARED;
+        pic->f.type = FF_BUFFER_TYPE_SHARED;
     }else{
-        assert(!pic->data[0]);
-
-        r= s->avctx->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]);
-            return -1;
-        }
-
-        if(s->linesize && (s->linesize != pic->linesize[0] || s->uvlinesize != pic->linesize[1])){
-            av_log(s->avctx, AV_LOG_ERROR, "get_buffer() failed (stride changed)\n");
-            s->avctx->release_buffer(s->avctx, (AVFrame*)pic);
-            return -1;
-        }
+        assert(!pic->f.data[0]);
 
-        if(pic->linesize[1] != pic->linesize[2]){
-            av_log(s->avctx, AV_LOG_ERROR, "get_buffer() failed (uv stride mismatch)\n");
-            s->avctx->release_buffer(s->avctx, (AVFrame*)pic);
+        if (alloc_frame_buffer(s, pic) < 0)
             return -1;
-        }
 
-        s->linesize  = pic->linesize[0];
-        s->uvlinesize= pic->linesize[1];
+        s->linesize   = pic->f.linesize[0];
+        s->uvlinesize = pic->f.linesize[1];
     }
 
-    if(pic->qscale_table==NULL){
+    if (pic->f.qscale_table == NULL) {
         if (s->encoding) {
-            CHECKED_ALLOCZ(pic->mb_var   , mb_array_size * sizeof(int16_t))
-            CHECKED_ALLOCZ(pic->mc_mb_var, mb_array_size * sizeof(int16_t))
-            CHECKED_ALLOCZ(pic->mb_mean  , mb_array_size * sizeof(int8_t))
+            FF_ALLOCZ_OR_GOTO(s->avctx, pic->mb_var   , mb_array_size * sizeof(int16_t)  , fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, pic->mc_mb_var, mb_array_size * sizeof(int16_t)  , fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, pic->mb_mean  , mb_array_size * sizeof(int8_t )  , fail)
         }
 
-        CHECKED_ALLOCZ(pic->mbskip_table , mb_array_size * sizeof(uint8_t)+2) //the +2 is for the slice end check
-        CHECKED_ALLOCZ(pic->qscale_table , mb_array_size * sizeof(uint8_t))
-        CHECKED_ALLOCZ(pic->mb_type_base , (big_mb_num + s->mb_stride) * sizeof(uint32_t))
-        pic->mb_type= pic->mb_type_base + 2*s->mb_stride+1;
+        FF_ALLOCZ_OR_GOTO(s->avctx, pic->f.mbskip_table, mb_array_size * sizeof(uint8_t) + 2, fail) //the +2 is for the slice end check
+        FF_ALLOCZ_OR_GOTO(s->avctx, pic->qscale_table_base , (big_mb_num + s->mb_stride) * sizeof(uint8_t)  , fail)
+        FF_ALLOCZ_OR_GOTO(s->avctx, pic->mb_type_base , (big_mb_num + s->mb_stride) * sizeof(uint32_t), fail)
+        pic->f.mb_type = pic->mb_type_base + 2*s->mb_stride + 1;
+        pic->f.qscale_table = pic->qscale_table_base + 2*s->mb_stride + 1;
         if(s->out_format == FMT_H264){
             for(i=0; i<2; i++){
-                CHECKED_ALLOCZ(pic->motion_val_base[i], 2 * (b4_array_size+4)  * sizeof(int16_t))
-                pic->motion_val[i]= pic->motion_val_base[i]+4;
-                CHECKED_ALLOCZ(pic->ref_index[i], b8_array_size * sizeof(uint8_t))
+                FF_ALLOCZ_OR_GOTO(s->avctx, pic->motion_val_base[i], 2 * (b4_array_size+4)  * sizeof(int16_t), fail)
+                pic->f.motion_val[i] = pic->motion_val_base[i] + 4;
+                FF_ALLOCZ_OR_GOTO(s->avctx, pic->f.ref_index[i], 4*mb_array_size * sizeof(uint8_t), fail)
             }
-            pic->motion_subsample_log2= 2;
+            pic->f.motion_subsample_log2 = 2;
         }else if(s->out_format == FMT_H263 || s->encoding || (s->avctx->debug&FF_DEBUG_MV) || (s->avctx->debug_mv)){
             for(i=0; i<2; i++){
-                CHECKED_ALLOCZ(pic->motion_val_base[i], 2 * (b8_array_size+4) * sizeof(int16_t))
-                pic->motion_val[i]= pic->motion_val_base[i]+4;
-                CHECKED_ALLOCZ(pic->ref_index[i], b8_array_size * sizeof(uint8_t))
+                FF_ALLOCZ_OR_GOTO(s->avctx, pic->motion_val_base[i], 2 * (b8_array_size+4) * sizeof(int16_t), fail)
+                pic->f.motion_val[i] = pic->motion_val_base[i] + 4;
+                FF_ALLOCZ_OR_GOTO(s->avctx, pic->f.ref_index[i], 4*mb_array_size * sizeof(uint8_t), fail)
             }
-            pic->motion_subsample_log2= 3;
+            pic->f.motion_subsample_log2 = 3;
         }
         if(s->avctx->debug&FF_DEBUG_DCT_COEFF) {
-            CHECKED_ALLOCZ(pic->dct_coeff, 64 * mb_array_size * sizeof(DCTELEM)*6)
+            FF_ALLOCZ_OR_GOTO(s->avctx, pic->f.dct_coeff, 64 * mb_array_size * sizeof(DCTELEM) * 6, fail)
         }
-        pic->qstride= s->mb_stride;
-        CHECKED_ALLOCZ(pic->pan_scan , 1 * sizeof(AVPanScan))
+        pic->f.qstride = s->mb_stride;
+        FF_ALLOCZ_OR_GOTO(s->avctx, pic->f.pan_scan , 1 * sizeof(AVPanScan), fail)
     }
 
     /* It might be nicer if the application would keep track of these
      * but it would require an API change. */
     memmove(s->prev_pict_types+1, s->prev_pict_types, PREV_PICT_TYPES_BUFFER_SIZE-1);
-    s->prev_pict_types[0]= s->dropable ? FF_B_TYPE : s->pict_type;
-    if(pic->age < PREV_PICT_TYPES_BUFFER_SIZE && s->prev_pict_types[pic->age] == FF_B_TYPE)
-        pic->age= INT_MAX; // Skipped MBs in B-frames are quite rare in MPEG-1/2 and it is a bit tricky to skip them anyway.
+    s->prev_pict_types[0]= s->dropable ? AV_PICTURE_TYPE_B : s->pict_type;
+    if (pic->f.age < PREV_PICT_TYPES_BUFFER_SIZE && s->prev_pict_types[pic->f.age] == AV_PICTURE_TYPE_B)
+        pic->f.age = INT_MAX; // Skipped MBs in B-frames are quite rare in MPEG-1/2 and it is a bit tricky to skip them anyway.
+    pic->owner2 = s;
 
     return 0;
-fail: //for the CHECKED_ALLOCZ macro
+fail: //for the FF_ALLOCZ_OR_GOTO macro
     if(r>=0)
-        s->avctx->release_buffer(s->avctx, (AVFrame*)pic);
+        free_frame_buffer(s, pic);
     return -1;
 }
 
@@ -250,59 +343,71 @@ fail: //for the CHECKED_ALLOCZ macro
 static void free_picture(MpegEncContext *s, Picture *pic){
     int i;
 
-    if(pic->data[0] && pic->type!=FF_BUFFER_TYPE_SHARED){
-        s->avctx->release_buffer(s->avctx, (AVFrame*)pic);
+    if (pic->f.data[0] && pic->f.type != FF_BUFFER_TYPE_SHARED) {
+        free_frame_buffer(s, pic);
     }
 
     av_freep(&pic->mb_var);
     av_freep(&pic->mc_mb_var);
     av_freep(&pic->mb_mean);
-    av_freep(&pic->mbskip_table);
-    av_freep(&pic->qscale_table);
+    av_freep(&pic->f.mbskip_table);
+    av_freep(&pic->qscale_table_base);
     av_freep(&pic->mb_type_base);
-    av_freep(&pic->dct_coeff);
-    av_freep(&pic->pan_scan);
-    pic->mb_type= NULL;
+    av_freep(&pic->f.dct_coeff);
+    av_freep(&pic->f.pan_scan);
+    pic->f.mb_type = NULL;
     for(i=0; i<2; i++){
         av_freep(&pic->motion_val_base[i]);
-        av_freep(&pic->ref_index[i]);
+        av_freep(&pic->f.ref_index[i]);
     }
 
-    if(pic->type == FF_BUFFER_TYPE_SHARED){
+    if (pic->f.type == FF_BUFFER_TYPE_SHARED) {
         for(i=0; i<4; i++){
-            pic->base[i]=
-            pic->data[i]= NULL;
+            pic->f.base[i] =
+            pic->f.data[i] = NULL;
         }
-        pic->type= 0;
+        pic->f.type = 0;
     }
 }
 
 static int init_duplicate_context(MpegEncContext *s, MpegEncContext *base){
+    int y_size = s->b8_stride * (2 * s->mb_height + 1);
+    int c_size = s->mb_stride * (s->mb_height + 1);
+    int yc_size = y_size + 2 * c_size;
     int i;
 
     // edge emu needs blocksize + filter length - 1 (=17x17 for halfpel / 21x21 for h264)
-    CHECKED_ALLOCZ(s->allocated_edge_emu_buffer, (s->width+64)*2*21*2); //(width + edge + align)*interlaced*MBsize*tolerance
+    FF_ALLOCZ_OR_GOTO(s->avctx, s->allocated_edge_emu_buffer, (s->width+64)*2*21*2, fail); //(width + edge + align)*interlaced*MBsize*tolerance
     s->edge_emu_buffer= s->allocated_edge_emu_buffer + (s->width+64)*2*21;
 
      //FIXME should be linesize instead of s->width*2 but that is not known before get_buffer()
-    CHECKED_ALLOCZ(s->me.scratchpad,  (s->width+64)*4*16*2*sizeof(uint8_t))
+    FF_ALLOCZ_OR_GOTO(s->avctx, s->me.scratchpad,  (s->width+64)*4*16*2*sizeof(uint8_t), fail)
     s->me.temp=         s->me.scratchpad;
     s->rd_scratchpad=   s->me.scratchpad;
     s->b_scratchpad=    s->me.scratchpad;
     s->obmc_scratchpad= s->me.scratchpad + 16;
     if (s->encoding) {
-        CHECKED_ALLOCZ(s->me.map      , ME_MAP_SIZE*sizeof(uint32_t))
-        CHECKED_ALLOCZ(s->me.score_map, ME_MAP_SIZE*sizeof(uint32_t))
+        FF_ALLOCZ_OR_GOTO(s->avctx, s->me.map      , ME_MAP_SIZE*sizeof(uint32_t), fail)
+        FF_ALLOCZ_OR_GOTO(s->avctx, s->me.score_map, ME_MAP_SIZE*sizeof(uint32_t), fail)
         if(s->avctx->noise_reduction){
-            CHECKED_ALLOCZ(s->dct_error_sum, 2 * 64 * sizeof(int))
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->dct_error_sum, 2 * 64 * sizeof(int), fail)
         }
     }
-    CHECKED_ALLOCZ(s->blocks, 64*12*2 * sizeof(DCTELEM))
+    FF_ALLOCZ_OR_GOTO(s->avctx, s->blocks, 64*12*2 * sizeof(DCTELEM), fail)
     s->block= s->blocks[0];
 
     for(i=0;i<12;i++){
         s->pblocks[i] = &s->block[i];
     }
+
+    if (s->out_format == FMT_H263) {
+        /* ac values */
+        FF_ALLOCZ_OR_GOTO(s->avctx, s->ac_val_base, yc_size * sizeof(int16_t) * 16, fail);
+        s->ac_val[0] = s->ac_val_base + s->b8_stride + 1;
+        s->ac_val[1] = s->ac_val_base + y_size + s->mb_stride + 1;
+        s->ac_val[2] = s->ac_val[1] + c_size;
+    }
+
     return 0;
 fail:
     return -1; //free() through MPV_common_end()
@@ -322,6 +427,7 @@ static void free_duplicate_context(MpegEncContext *s){
     av_freep(&s->me.map);
     av_freep(&s->me.score_map);
     av_freep(&s->blocks);
+    av_freep(&s->ac_val_base);
     s->block= NULL;
 }
 
@@ -345,6 +451,10 @@ static void backup_duplicate_context(MpegEncContext *bak, MpegEncContext *src){
     COPY(dct_error_sum);
     COPY(dct_count[0]);
     COPY(dct_count[1]);
+    COPY(ac_val_base);
+    COPY(ac_val[0]);
+    COPY(ac_val[1]);
+    COPY(ac_val[2]);
 #undef COPY
 }
 
@@ -362,6 +472,81 @@ 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->f.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.
@@ -382,6 +567,9 @@ 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;
 }
 
 /**
@@ -398,192 +586,204 @@ void MPV_decode_defaults(MpegEncContext *s){
  */
 av_cold int MPV_common_init(MpegEncContext *s)
 {
-    int y_size, c_size, yc_size, i, mb_array_size, mv_table_size, x, y, threads;
+    int y_size, c_size, yc_size, i, mb_array_size, mv_table_size, x, y,
+        threads = (s->encoding ||
+                   (HAVE_THREADS &&
+                    s->avctx->active_thread_type & FF_THREAD_SLICE)) ?
+                  s->avctx->thread_count : 1;
 
-    s->mb_height = (s->height + 15) / 16;
+    if(s->codec_id == CODEC_ID_MPEG2VIDEO && !s->progressive_sequence)
+        s->mb_height = (s->height + 31) / 32 * 2;
+    else if (s->codec_id != CODEC_ID_H264)
+        s->mb_height = (s->height + 15) / 16;
 
     if(s->avctx->pix_fmt == PIX_FMT_NONE){
         av_log(s->avctx, AV_LOG_ERROR, "decoding to PIX_FMT_NONE is not supported.\n");
         return -1;
     }
 
-    if(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;
+    if((s->encoding || (s->avctx->active_thread_type & FF_THREAD_SLICE)) &&
+       (s->avctx->thread_count > MAX_THREADS || (s->avctx->thread_count > s->mb_height && s->mb_height))){
+        int max_threads = FFMIN(MAX_THREADS, s->mb_height);
+        av_log(s->avctx, AV_LOG_WARNING, "too many threads (%d), reducing to %d\n",
+               s->avctx->thread_count, max_threads);
+        threads = max_threads;
     }
 
-    if((s->width || s->height) && avcodec_check_dimensions(s->avctx, s->width, s->height))
+    if((s->width || s->height) && av_image_check_size(s->width, s->height, 0, s->avctx))
         return -1;
 
-    dsputil_init(&s->dsp, s->avctx);
     ff_dct_common_init(s);
 
     s->flags= s->avctx->flags;
     s->flags2= s->avctx->flags2;
 
-    s->mb_width  = (s->width  + 15) / 16;
-    s->mb_stride = s->mb_width + 1;
-    s->b8_stride = s->mb_width*2 + 1;
-    s->b4_stride = s->mb_width*4 + 1;
-    mb_array_size= s->mb_height * s->mb_stride;
-    mv_table_size= (s->mb_height+2) * s->mb_stride + 1;
+    if (s->width && s->height) {
+        s->mb_width  = (s->width  + 15) / 16;
+        s->mb_stride = s->mb_width + 1;
+        s->b8_stride = s->mb_width*2 + 1;
+        s->b4_stride = s->mb_width*4 + 1;
+        mb_array_size= s->mb_height * s->mb_stride;
+        mv_table_size= (s->mb_height+2) * s->mb_stride + 1;
 
-    /* set chroma shifts */
-    avcodec_get_chroma_sub_sample(s->avctx->pix_fmt,&(s->chroma_x_shift),
-                                                    &(s->chroma_y_shift) );
+        /* set chroma shifts */
+        avcodec_get_chroma_sub_sample(s->avctx->pix_fmt,&(s->chroma_x_shift),
+                                      &(s->chroma_y_shift) );
 
-    /* set default edge pos, will be overriden in decode_header if needed */
-    s->h_edge_pos= s->mb_width*16;
-    s->v_edge_pos= s->mb_height*16;
+        /* set default edge pos, will be overriden in decode_header if needed */
+        s->h_edge_pos= s->mb_width*16;
+        s->v_edge_pos= s->mb_height*16;
 
-    s->mb_num = s->mb_width * s->mb_height;
+        s->mb_num = s->mb_width * s->mb_height;
 
-    s->block_wrap[0]=
-    s->block_wrap[1]=
-    s->block_wrap[2]=
-    s->block_wrap[3]= s->b8_stride;
-    s->block_wrap[4]=
-    s->block_wrap[5]= s->mb_stride;
+        s->block_wrap[0]=
+        s->block_wrap[1]=
+        s->block_wrap[2]=
+        s->block_wrap[3]= s->b8_stride;
+        s->block_wrap[4]=
+        s->block_wrap[5]= s->mb_stride;
 
-    y_size = s->b8_stride * (2 * s->mb_height + 1);
-    c_size = s->mb_stride * (s->mb_height + 1);
-    yc_size = y_size + 2 * c_size;
+        y_size = s->b8_stride * (2 * s->mb_height + 1);
+        c_size = s->mb_stride * (s->mb_height + 1);
+        yc_size = y_size + 2 * c_size;
 
-    /* convert fourcc to upper case */
-    s->codec_tag=          toupper( s->avctx->codec_tag     &0xFF)
-                        + (toupper((s->avctx->codec_tag>>8 )&0xFF)<<8 )
-                        + (toupper((s->avctx->codec_tag>>16)&0xFF)<<16)
-                        + (toupper((s->avctx->codec_tag>>24)&0xFF)<<24);
+        /* convert fourcc to upper case */
+        s->codec_tag = ff_toupper4(s->avctx->codec_tag);
 
-    s->stream_codec_tag=          toupper( s->avctx->stream_codec_tag     &0xFF)
-                               + (toupper((s->avctx->stream_codec_tag>>8 )&0xFF)<<8 )
-                               + (toupper((s->avctx->stream_codec_tag>>16)&0xFF)<<16)
-                               + (toupper((s->avctx->stream_codec_tag>>24)&0xFF)<<24);
+        s->stream_codec_tag = ff_toupper4(s->avctx->stream_codec_tag);
 
-    s->avctx->coded_frame= (AVFrame*)&s->current_picture;
+        s->avctx->coded_frame= (AVFrame*)&s->current_picture;
 
-    CHECKED_ALLOCZ(s->mb_index2xy, (s->mb_num+1)*sizeof(int)) //error ressilience code looks cleaner with this
-    for(y=0; y<s->mb_height; y++){
-        for(x=0; x<s->mb_width; x++){
-            s->mb_index2xy[ x + y*s->mb_width ] = x + y*s->mb_stride;
+        FF_ALLOCZ_OR_GOTO(s->avctx, s->mb_index2xy, (s->mb_num+1)*sizeof(int), fail) //error ressilience code looks cleaner with this
+        for(y=0; y<s->mb_height; y++){
+            for(x=0; x<s->mb_width; x++){
+                s->mb_index2xy[ x + y*s->mb_width ] = x + y*s->mb_stride;
+            }
         }
-    }
-    s->mb_index2xy[ s->mb_height*s->mb_width ] = (s->mb_height-1)*s->mb_stride + s->mb_width; //FIXME really needed?
+        s->mb_index2xy[ s->mb_height*s->mb_width ] = (s->mb_height-1)*s->mb_stride + s->mb_width; //FIXME really needed?
 
-    if (s->encoding) {
-        /* Allocate MV tables */
-        CHECKED_ALLOCZ(s->p_mv_table_base            , mv_table_size * 2 * sizeof(int16_t))
-        CHECKED_ALLOCZ(s->b_forw_mv_table_base       , mv_table_size * 2 * sizeof(int16_t))
-        CHECKED_ALLOCZ(s->b_back_mv_table_base       , mv_table_size * 2 * sizeof(int16_t))
-        CHECKED_ALLOCZ(s->b_bidir_forw_mv_table_base , mv_table_size * 2 * sizeof(int16_t))
-        CHECKED_ALLOCZ(s->b_bidir_back_mv_table_base , mv_table_size * 2 * sizeof(int16_t))
-        CHECKED_ALLOCZ(s->b_direct_mv_table_base     , mv_table_size * 2 * sizeof(int16_t))
-        s->p_mv_table           = s->p_mv_table_base            + s->mb_stride + 1;
-        s->b_forw_mv_table      = s->b_forw_mv_table_base       + s->mb_stride + 1;
-        s->b_back_mv_table      = s->b_back_mv_table_base       + s->mb_stride + 1;
-        s->b_bidir_forw_mv_table= s->b_bidir_forw_mv_table_base + s->mb_stride + 1;
-        s->b_bidir_back_mv_table= s->b_bidir_back_mv_table_base + s->mb_stride + 1;
-        s->b_direct_mv_table    = s->b_direct_mv_table_base     + s->mb_stride + 1;
-
-        if(s->msmpeg4_version){
-            CHECKED_ALLOCZ(s->ac_stats, 2*2*(MAX_LEVEL+1)*(MAX_RUN+1)*2*sizeof(int));
-        }
-        CHECKED_ALLOCZ(s->avctx->stats_out, 256);
-
-        /* Allocate MB type table */
-        CHECKED_ALLOCZ(s->mb_type  , mb_array_size * sizeof(uint16_t)) //needed for encoding
-
-        CHECKED_ALLOCZ(s->lambda_table, mb_array_size * sizeof(int))
-
-        CHECKED_ALLOCZ(s->q_intra_matrix, 64*32 * sizeof(int))
-        CHECKED_ALLOCZ(s->q_inter_matrix, 64*32 * sizeof(int))
-        CHECKED_ALLOCZ(s->q_intra_matrix16, 64*32*2 * sizeof(uint16_t))
-        CHECKED_ALLOCZ(s->q_inter_matrix16, 64*32*2 * sizeof(uint16_t))
-        CHECKED_ALLOCZ(s->input_picture, MAX_PICTURE_COUNT * sizeof(Picture*))
-        CHECKED_ALLOCZ(s->reordered_input_picture, MAX_PICTURE_COUNT * sizeof(Picture*))
+        if (s->encoding) {
+            /* Allocate MV tables */
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->p_mv_table_base            , mv_table_size * 2 * sizeof(int16_t), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->b_forw_mv_table_base       , mv_table_size * 2 * sizeof(int16_t), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->b_back_mv_table_base       , mv_table_size * 2 * sizeof(int16_t), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->b_bidir_forw_mv_table_base , mv_table_size * 2 * sizeof(int16_t), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->b_bidir_back_mv_table_base , mv_table_size * 2 * sizeof(int16_t), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->b_direct_mv_table_base     , mv_table_size * 2 * sizeof(int16_t), fail)
+            s->p_mv_table           = s->p_mv_table_base            + s->mb_stride + 1;
+            s->b_forw_mv_table      = s->b_forw_mv_table_base       + s->mb_stride + 1;
+            s->b_back_mv_table      = s->b_back_mv_table_base       + s->mb_stride + 1;
+            s->b_bidir_forw_mv_table= s->b_bidir_forw_mv_table_base + s->mb_stride + 1;
+            s->b_bidir_back_mv_table= s->b_bidir_back_mv_table_base + s->mb_stride + 1;
+            s->b_direct_mv_table    = s->b_direct_mv_table_base     + s->mb_stride + 1;
+
+            if(s->msmpeg4_version){
+                FF_ALLOCZ_OR_GOTO(s->avctx, s->ac_stats, 2*2*(MAX_LEVEL+1)*(MAX_RUN+1)*2*sizeof(int), fail);
+            }
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->avctx->stats_out, 256, fail);
 
-        if(s->avctx->noise_reduction){
-            CHECKED_ALLOCZ(s->dct_offset, 2 * 64 * sizeof(uint16_t))
+            /* Allocate MB type table */
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->mb_type  , mb_array_size * sizeof(uint16_t), fail) //needed for encoding
+
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->lambda_table, mb_array_size * sizeof(int), fail)
+
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->q_intra_matrix  , 64*32   * sizeof(int), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->q_inter_matrix  , 64*32   * sizeof(int), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->q_intra_matrix16, 64*32*2 * sizeof(uint16_t), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->q_inter_matrix16, 64*32*2 * sizeof(uint16_t), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->input_picture, MAX_PICTURE_COUNT * sizeof(Picture*), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->reordered_input_picture, MAX_PICTURE_COUNT * sizeof(Picture*), fail)
+
+            if(s->avctx->noise_reduction){
+                FF_ALLOCZ_OR_GOTO(s->avctx, s->dct_offset, 2 * 64 * sizeof(uint16_t), fail)
+            }
         }
     }
-    CHECKED_ALLOCZ(s->picture, MAX_PICTURE_COUNT * sizeof(Picture))
 
-    CHECKED_ALLOCZ(s->error_status_table, mb_array_size*sizeof(uint8_t))
+    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]);
+    }
+
+    if (s->width && s->height) {
+        FF_ALLOCZ_OR_GOTO(s->avctx, s->error_status_table, mb_array_size*sizeof(uint8_t), fail)
 
-    if(s->codec_id==CODEC_ID_MPEG4 || (s->flags & CODEC_FLAG_INTERLACED_ME)){
-        /* interlaced direct mode decoding tables */
+        if(s->codec_id==CODEC_ID_MPEG4 || (s->flags & CODEC_FLAG_INTERLACED_ME)){
+            /* interlaced direct mode decoding tables */
             for(i=0; i<2; i++){
                 int j, k;
                 for(j=0; j<2; j++){
                     for(k=0; k<2; k++){
-                        CHECKED_ALLOCZ(s->b_field_mv_table_base[i][j][k]     , mv_table_size * 2 * sizeof(int16_t))
-                        s->b_field_mv_table[i][j][k]    = s->b_field_mv_table_base[i][j][k]     + s->mb_stride + 1;
+                        FF_ALLOCZ_OR_GOTO(s->avctx,    s->b_field_mv_table_base[i][j][k], mv_table_size * 2 * sizeof(int16_t), fail)
+                        s->b_field_mv_table[i][j][k] = s->b_field_mv_table_base[i][j][k] + s->mb_stride + 1;
                     }
-                    CHECKED_ALLOCZ(s->b_field_select_table[i][j]     , mb_array_size * 2 * sizeof(uint8_t))
-                    CHECKED_ALLOCZ(s->p_field_mv_table_base[i][j]     , mv_table_size * 2 * sizeof(int16_t))
-                    s->p_field_mv_table[i][j]    = s->p_field_mv_table_base[i][j]     + s->mb_stride + 1;
+                    FF_ALLOCZ_OR_GOTO(s->avctx, s->b_field_select_table [i][j], mb_array_size * 2 * sizeof(uint8_t), fail)
+                    FF_ALLOCZ_OR_GOTO(s->avctx, s->p_field_mv_table_base[i][j], mv_table_size * 2 * sizeof(int16_t), fail)
+                    s->p_field_mv_table[i][j] = s->p_field_mv_table_base[i][j]+ s->mb_stride + 1;
                 }
-                CHECKED_ALLOCZ(s->p_field_select_table[i]      , mb_array_size * 2 * sizeof(uint8_t))
+                FF_ALLOCZ_OR_GOTO(s->avctx, s->p_field_select_table[i], mb_array_size * 2 * sizeof(uint8_t), fail)
             }
-    }
-    if (s->out_format == FMT_H263) {
-        /* ac values */
-        CHECKED_ALLOCZ(s->ac_val_base, yc_size * sizeof(int16_t) * 16);
-        s->ac_val[0] = s->ac_val_base + s->b8_stride + 1;
-        s->ac_val[1] = s->ac_val_base + y_size + s->mb_stride + 1;
-        s->ac_val[2] = s->ac_val[1] + c_size;
-
-        /* cbp values */
-        CHECKED_ALLOCZ(s->coded_block_base, y_size);
-        s->coded_block= s->coded_block_base + s->b8_stride + 1;
-
-        /* cbp, ac_pred, pred_dir */
-        CHECKED_ALLOCZ(s->cbp_table  , mb_array_size * sizeof(uint8_t))
-        CHECKED_ALLOCZ(s->pred_dir_table, mb_array_size * sizeof(uint8_t))
-    }
+        }
+        if (s->out_format == FMT_H263) {
+            /* cbp values */
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->coded_block_base, y_size, fail);
+            s->coded_block= s->coded_block_base + s->b8_stride + 1;
+
+            /* cbp, ac_pred, pred_dir */
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->cbp_table     , mb_array_size * sizeof(uint8_t), fail)
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->pred_dir_table, mb_array_size * sizeof(uint8_t), fail)
+        }
 
-    if (s->h263_pred || s->h263_plus || !s->encoding) {
-        /* dc values */
-        //MN: we need these for error resilience of intra-frames
-        CHECKED_ALLOCZ(s->dc_val_base, yc_size * sizeof(int16_t));
-        s->dc_val[0] = s->dc_val_base + s->b8_stride + 1;
-        s->dc_val[1] = s->dc_val_base + y_size + s->mb_stride + 1;
-        s->dc_val[2] = s->dc_val[1] + c_size;
-        for(i=0;i<yc_size;i++)
-            s->dc_val_base[i] = 1024;
-    }
+        if (s->h263_pred || s->h263_plus || !s->encoding) {
+            /* dc values */
+            //MN: we need these for error resilience of intra-frames
+            FF_ALLOCZ_OR_GOTO(s->avctx, s->dc_val_base, yc_size * sizeof(int16_t), fail);
+            s->dc_val[0] = s->dc_val_base + s->b8_stride + 1;
+            s->dc_val[1] = s->dc_val_base + y_size + s->mb_stride + 1;
+            s->dc_val[2] = s->dc_val[1] + c_size;
+            for(i=0;i<yc_size;i++)
+                s->dc_val_base[i] = 1024;
+        }
 
-    /* which mb is a intra block */
-    CHECKED_ALLOCZ(s->mbintra_table, mb_array_size);
-    memset(s->mbintra_table, 1, mb_array_size);
+        /* which mb is a intra block */
+        FF_ALLOCZ_OR_GOTO(s->avctx, s->mbintra_table, mb_array_size, fail);
+        memset(s->mbintra_table, 1, mb_array_size);
 
-    /* init macroblock skip table */
-    CHECKED_ALLOCZ(s->mbskip_table, mb_array_size+2);
-    //Note the +1 is for a quicker mpeg4 slice_end detection
-    CHECKED_ALLOCZ(s->prev_pict_types, PREV_PICT_TYPES_BUFFER_SIZE);
+        /* init macroblock skip table */
+        FF_ALLOCZ_OR_GOTO(s->avctx, s->mbskip_table, mb_array_size+2, fail);
+        //Note the +1 is for a quicker mpeg4 slice_end detection
+        FF_ALLOCZ_OR_GOTO(s->avctx, s->prev_pict_types, PREV_PICT_TYPES_BUFFER_SIZE, fail);
 
-    s->parse_context.state= -1;
-    if((s->avctx->debug&(FF_DEBUG_VIS_QP|FF_DEBUG_VIS_MB_TYPE)) || (s->avctx->debug_mv)){
-       s->visualization_buffer[0] = av_malloc((s->mb_width*16 + 2*EDGE_WIDTH) * s->mb_height*16 + 2*EDGE_WIDTH);
-       s->visualization_buffer[1] = av_malloc((s->mb_width*16 + 2*EDGE_WIDTH) * s->mb_height*16 + 2*EDGE_WIDTH);
-       s->visualization_buffer[2] = av_malloc((s->mb_width*16 + 2*EDGE_WIDTH) * s->mb_height*16 + 2*EDGE_WIDTH);
+        s->parse_context.state= -1;
+        if((s->avctx->debug&(FF_DEBUG_VIS_QP|FF_DEBUG_VIS_MB_TYPE)) || (s->avctx->debug_mv)){
+            s->visualization_buffer[0] = av_malloc((s->mb_width*16 + 2*EDGE_WIDTH) * s->mb_height*16 + 2*EDGE_WIDTH);
+            s->visualization_buffer[1] = av_malloc((s->mb_width*16 + 2*EDGE_WIDTH) * s->mb_height*16 + 2*EDGE_WIDTH);
+            s->visualization_buffer[2] = av_malloc((s->mb_width*16 + 2*EDGE_WIDTH) * s->mb_height*16 + 2*EDGE_WIDTH);
+        }
     }
 
     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 (s->width && s->height) {
+    if (s->encoding || (HAVE_THREADS && s->avctx->active_thread_type&FF_THREAD_SLICE)) {
+        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;
+        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;
@@ -597,12 +797,14 @@ 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 (s->encoding || (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;
@@ -634,7 +836,6 @@ void MPV_common_end(MpegEncContext *s)
     }
 
     av_freep(&s->dc_val_base);
-    av_freep(&s->ac_val_base);
     av_freep(&s->coded_block_base);
     av_freep(&s->mbintra_table);
     av_freep(&s->cbp_table);
@@ -658,8 +859,8 @@ 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]);
         }
     }
@@ -673,7 +874,8 @@ 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])
@@ -767,19 +969,37 @@ void init_vlc_rl(RLTable *rl)
     }
 }
 
+void ff_release_unused_pictures(MpegEncContext *s, int remove_current)
+{
+    int i;
+
+    /* release non reference frames */
+    for(i=0; i<s->picture_count; i++){
+        if (s->picture[i].f.data[0] && !s->picture[i].f.reference
+           && (!s->picture[i].owner2 || s->picture[i].owner2 == s)
+           && (remove_current || &s->picture[i] != s->current_picture_ptr)
+           /*&& s->picture[i].type!=FF_BUFFER_TYPE_SHARED*/){
+            free_frame_buffer(s, &s->picture[i]);
+        }
+    }
+}
+
 int ff_find_unused_picture(MpegEncContext *s, int shared){
     int i;
 
     if(shared){
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
-            if(s->picture[i].data[0]==NULL && s->picture[i].type==0) return i;
+        for(i=s->picture_range_start; i<s->picture_range_end; i++){
+            if (s->picture[i].f.data[0] == NULL && s->picture[i].f.type == 0)
+                return i;
         }
     }else{
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
-            if(s->picture[i].data[0]==NULL && s->picture[i].type!=0) return i; //FIXME
+        for(i=s->picture_range_start; i<s->picture_range_end; i++){
+            if (s->picture[i].f.data[0] == NULL && s->picture[i].f.type != 0)
+                return i; //FIXME
         }
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
-            if(s->picture[i].data[0]==NULL) return i;
+        for(i=s->picture_range_start; i<s->picture_range_end; i++){
+            if (s->picture[i].f.data[0] == NULL)
+                return i;
         }
     }
 
@@ -822,104 +1042,127 @@ static void update_noise_reduction(MpegEncContext *s){
 int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx)
 {
     int i;
-    AVFrame *pic;
+    Picture *pic;
     s->mb_skipped = 0;
 
     assert(s->last_picture_ptr==NULL || s->out_format != FMT_H264 || s->codec_id == CODEC_ID_SVQ3);
 
     /* mark&release old frames */
-    if (s->pict_type != FF_B_TYPE && s->last_picture_ptr && s->last_picture_ptr != s->next_picture_ptr && s->last_picture_ptr->data[0]) {
+    if (s->pict_type != AV_PICTURE_TYPE_B && s->last_picture_ptr && s->last_picture_ptr != s->next_picture_ptr && s->last_picture_ptr->f.data[0]) {
       if(s->out_format != FMT_H264 || s->codec_id == CODEC_ID_SVQ3){
-        avctx->release_buffer(avctx, (AVFrame*)s->last_picture_ptr);
+          if (s->last_picture_ptr->owner2 == s)
+              free_frame_buffer(s, s->last_picture_ptr);
 
         /* release forgotten pictures */
         /* if(mpeg124/h263) */
         if(!s->encoding){
-            for(i=0; i<MAX_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");
-                    avctx->release_buffer(avctx, (AVFrame*)&s->picture[i]);
+            for(i=0; i<s->picture_count; i++){
+                if (s->picture[i].owner2 == s && s->picture[i].f.data[0] && &s->picture[i] != s->next_picture_ptr && s->picture[i].f.reference) {
+                    if (!(avctx->active_thread_type & FF_THREAD_FRAME))
+                        av_log(avctx, AV_LOG_ERROR, "releasing zombie picture\n");
+                    free_frame_buffer(s, &s->picture[i]);
                 }
             }
         }
       }
     }
-alloc:
+
     if(!s->encoding){
-        /* release non reference frames */
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
-            if(s->picture[i].data[0] && !s->picture[i].reference /*&& s->picture[i].type!=FF_BUFFER_TYPE_SHARED*/){
-                s->avctx->release_buffer(s->avctx, (AVFrame*)&s->picture[i]);
-            }
-        }
+        ff_release_unused_pictures(s, 1);
 
-        if(s->current_picture_ptr && s->current_picture_ptr->data[0]==NULL)
-            pic= (AVFrame*)s->current_picture_ptr; //we already have a unused image (maybe it was set before reading the header)
+        if (s->current_picture_ptr && s->current_picture_ptr->f.data[0] == NULL)
+            pic= s->current_picture_ptr; //we already have a unused image (maybe it was set before reading the header)
         else{
             i= ff_find_unused_picture(s, 0);
-            pic= (AVFrame*)&s->picture[i];
+            pic= &s->picture[i];
         }
 
-        pic->reference= 0;
+        pic->f.reference = 0;
         if (!s->dropable){
             if (s->codec_id == CODEC_ID_H264)
-                pic->reference = s->picture_structure;
-            else if (s->pict_type != FF_B_TYPE)
-                pic->reference = 3;
+                pic->f.reference = s->picture_structure;
+            else if (s->pict_type != AV_PICTURE_TYPE_B)
+                pic->f.reference = 3;
         }
 
-        pic->coded_picture_number= s->coded_picture_number++;
+        pic->f.coded_picture_number = s->coded_picture_number++;
 
-        if( alloc_picture(s, (Picture*)pic, 0) < 0)
+        if(ff_alloc_picture(s, pic, 0) < 0)
             return -1;
 
-        s->current_picture_ptr= (Picture*)pic;
-        s->current_picture_ptr->top_field_first= s->top_field_first; //FIXME use only the vars from current_pic
-        s->current_picture_ptr->interlaced_frame= !s->progressive_frame && !s->progressive_sequence;
+        s->current_picture_ptr= pic;
+        //FIXME use only the vars from current_pic
+        s->current_picture_ptr->f.top_field_first = s->top_field_first;
+        if(s->codec_id == CODEC_ID_MPEG1VIDEO || s->codec_id == CODEC_ID_MPEG2VIDEO) {
+            if(s->picture_structure != PICT_FRAME)
+                s->current_picture_ptr->f.top_field_first = (s->picture_structure == PICT_TOP_FIELD) == s->first_field;
+        }
+        s->current_picture_ptr->f.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;
+    s->current_picture_ptr->f.pict_type = s->pict_type;
 //    if(s->flags && CODEC_FLAG_QSCALE)
   //      s->current_picture_ptr->quality= s->new_picture_ptr->quality;
-    s->current_picture_ptr->key_frame= s->pict_type == FF_I_TYPE;
+    s->current_picture_ptr->f.key_frame = s->pict_type == AV_PICTURE_TYPE_I;
 
     ff_copy_picture(&s->current_picture, s->current_picture_ptr);
 
-    if (s->pict_type != FF_B_TYPE) {
+    if (s->pict_type != AV_PICTURE_TYPE_B) {
         s->last_picture_ptr= s->next_picture_ptr;
         if(!s->dropable)
             s->next_picture_ptr= s->current_picture_ptr;
     }
 /*    av_log(s->avctx, AV_LOG_DEBUG, "L%p N%p C%p L%p N%p C%p type:%d drop:%d\n", s->last_picture_ptr, s->next_picture_ptr,s->current_picture_ptr,
-        s->last_picture_ptr    ? s->last_picture_ptr->data[0] : NULL,
-        s->next_picture_ptr    ? s->next_picture_ptr->data[0] : NULL,
-        s->current_picture_ptr ? s->current_picture_ptr->data[0] : NULL,
+        s->last_picture_ptr    ? s->last_picture_ptr->f.data[0]    : NULL,
+        s->next_picture_ptr    ? s->next_picture_ptr->f.data[0]    : NULL,
+        s->current_picture_ptr ? s->current_picture_ptr->f.data[0] : NULL,
         s->pict_type, s->dropable);*/
 
-    if(s->last_picture_ptr) ff_copy_picture(&s->last_picture, s->last_picture_ptr);
-    if(s->next_picture_ptr) ff_copy_picture(&s->next_picture, s->next_picture_ptr);
+    if(s->codec_id != CODEC_ID_H264){
+        if ((s->last_picture_ptr == NULL || s->last_picture_ptr->f.data[0] == NULL) &&
+           (s->pict_type!=AV_PICTURE_TYPE_I || s->picture_structure != PICT_FRAME)){
+            if (s->pict_type != AV_PICTURE_TYPE_I)
+                av_log(avctx, AV_LOG_ERROR, "warning: first frame is no keyframe\n");
+            else if (s->picture_structure != PICT_FRAME)
+                av_log(avctx, AV_LOG_INFO, "allocate dummy last picture for field based first keyframe\n");
 
-    if(s->pict_type != FF_I_TYPE && (s->last_picture_ptr==NULL || s->last_picture_ptr->data[0]==NULL) && !s->dropable && s->codec_id != CODEC_ID_H264){
-        av_log(avctx, AV_LOG_ERROR, "warning: first frame is no keyframe\n");
-        assert(s->pict_type != FF_B_TYPE); //these should have been dropped if we don't have a reference
-        goto alloc;
+            /* Allocate a dummy frame */
+            i= ff_find_unused_picture(s, 0);
+            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->f.data[0] == NULL) && s->pict_type == AV_PICTURE_TYPE_B) {
+            /* Allocate a dummy frame */
+            i= ff_find_unused_picture(s, 0);
+            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);
+        }
     }
 
-    assert(s->pict_type == FF_I_TYPE || (s->last_picture_ptr && s->last_picture_ptr->data[0]));
+    if(s->last_picture_ptr) ff_copy_picture(&s->last_picture, s->last_picture_ptr);
+    if(s->next_picture_ptr) ff_copy_picture(&s->next_picture, s->next_picture_ptr);
+
+    assert(s->pict_type == AV_PICTURE_TYPE_I || (s->last_picture_ptr && s->last_picture_ptr->f.data[0]));
 
     if(s->picture_structure!=PICT_FRAME && s->out_format != FMT_H264){
         int i;
         for(i=0; i<4; i++){
             if(s->picture_structure == PICT_BOTTOM_FIELD){
-                 s->current_picture.data[i] += s->current_picture.linesize[i];
+                 s->current_picture.f.data[i] += s->current_picture.f.linesize[i];
             }
-            s->current_picture.linesize[i] *= 2;
-            s->last_picture.linesize[i] *=2;
-            s->next_picture.linesize[i] *=2;
+            s->current_picture.f.linesize[i] *= 2;
+            s->last_picture.f.linesize[i]    *= 2;
+            s->next_picture.f.linesize[i]    *= 2;
         }
     }
 
-    s->hurry_up= s->avctx->hurry_up;
     s->error_recognition= avctx->error_recognition;
 
     /* set dequantizer, we can't do it during init as it might change for mpeg4
@@ -951,31 +1194,41 @@ alloc:
 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);
-    }else if(!s->avctx->hwaccel
+   }else if((s->error_count || s->encoding)
+       && !s->avctx->hwaccel
        && !(s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
        && s->unrestricted_mv
-       && s->current_picture.reference
+       && s->current_picture.f.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 hshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_w;
+            int vshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_h;
+            s->dsp.draw_edges(s->current_picture.f.data[0], s->linesize,
+                              s->h_edge_pos             , s->v_edge_pos,
+                              EDGE_WIDTH        , EDGE_WIDTH        , EDGE_TOP | EDGE_BOTTOM);
+            s->dsp.draw_edges(s->current_picture.f.data[1], s->uvlinesize,
+                              s->h_edge_pos>>hshift, s->v_edge_pos>>vshift,
+                              EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, EDGE_TOP | EDGE_BOTTOM);
+            s->dsp.draw_edges(s->current_picture.f.data[2], s->uvlinesize,
+                              s->h_edge_pos>>hshift, s->v_edge_pos>>vshift,
+                              EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, EDGE_TOP | EDGE_BOTTOM);
     }
+
     emms_c();
 
     s->last_pict_type    = s->pict_type;
-    s->last_lambda_for[s->pict_type]= s->current_picture_ptr->quality;
-    if(s->pict_type!=FF_B_TYPE){
+    s->last_lambda_for[s->pict_type] = s->current_picture_ptr->f.quality;
+    if(s->pict_type!=AV_PICTURE_TYPE_B){
         s->last_non_b_pict_type= s->pict_type;
     }
 #if 0
         /* copy back current_picture variables */
     for(i=0; i<MAX_PICTURE_COUNT; i++){
-        if(s->picture[i].data[0] == s->current_picture.data[0]){
+        if(s->picture[i].f.data[0] == s->current_picture.f.data[0]){
             s->picture[i]= s->current_picture;
             break;
         }
@@ -985,9 +1238,9 @@ void MPV_frame_end(MpegEncContext *s)
 
     if(s->encoding){
         /* release non-reference frames */
-        for(i=0; i<MAX_PICTURE_COUNT; i++){
-            if(s->picture[i].data[0] && !s->picture[i].reference /*&& s->picture[i].type!=FF_BUFFER_TYPE_SHARED*/){
-                s->avctx->release_buffer(s->avctx, (AVFrame*)&s->picture[i]);
+        for(i=0; i<s->picture_count; i++){
+            if (s->picture[i].f.data[0] && !s->picture[i].f.reference /*&& s->picture[i].type != FF_BUFFER_TYPE_SHARED*/) {
+                free_frame_buffer(s, &s->picture[i]);
             }
         }
     }
@@ -998,6 +1251,10 @@ 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.f.reference) {
+        ff_thread_report_progress((AVFrame*)s->current_picture_ptr, s->mb_height-1, 0);
+    }
 }
 
 /**
@@ -1087,19 +1344,19 @@ static void draw_arrow(uint8_t *buf, int sx, int sy, int ex, int ey, int w, int
  */
 void ff_print_debug_info(MpegEncContext *s, AVFrame *pict){
 
-    if(!pict || !pict->mb_type) return;
+    if(s->avctx->hwaccel || !pict || !pict->mb_type) return;
 
     if(s->avctx->debug&(FF_DEBUG_SKIP | FF_DEBUG_QP | FF_DEBUG_MB_TYPE)){
         int x,y;
 
         av_log(s->avctx,AV_LOG_DEBUG,"New frame, type: ");
         switch (pict->pict_type) {
-            case FF_I_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"I\n"); break;
-            case FF_P_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"P\n"); break;
-            case FF_B_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"B\n"); break;
-            case FF_S_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"S\n"); break;
-            case FF_SI_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"SI\n"); break;
-            case FF_SP_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"SP\n"); break;
+            case AV_PICTURE_TYPE_I: av_log(s->avctx,AV_LOG_DEBUG,"I\n"); break;
+            case AV_PICTURE_TYPE_P: av_log(s->avctx,AV_LOG_DEBUG,"P\n"); break;
+            case AV_PICTURE_TYPE_B: av_log(s->avctx,AV_LOG_DEBUG,"B\n"); break;
+            case AV_PICTURE_TYPE_S: av_log(s->avctx,AV_LOG_DEBUG,"S\n"); break;
+            case AV_PICTURE_TYPE_SI: av_log(s->avctx,AV_LOG_DEBUG,"SI\n"); break;
+            case AV_PICTURE_TYPE_SP: av_log(s->avctx,AV_LOG_DEBUG,"SP\n"); break;
         }
         for(y=0; y<s->mb_height; y++){
             for(x=0; x<s->mb_width; x++){
@@ -1154,7 +1411,7 @@ void ff_print_debug_info(MpegEncContext *s, AVFrame *pict){
                         av_log(s->avctx, AV_LOG_DEBUG, "?");
 
 
-                    if(IS_INTERLACED(mb_type) && s->codec_id == CODEC_ID_H264)
+                    if(IS_INTERLACED(mb_type))
                         av_log(s->avctx, AV_LOG_DEBUG, "=");
                     else
                         av_log(s->avctx, AV_LOG_DEBUG, " ");
@@ -1195,15 +1452,15 @@ void ff_print_debug_info(MpegEncContext *s, AVFrame *pict){
                   for(type=0; type<3; type++){
                     int direction = 0;
                     switch (type) {
-                      case 0: if ((!(s->avctx->debug_mv&FF_DEBUG_VIS_MV_P_FOR)) || (pict->pict_type!=FF_P_TYPE))
+                      case 0: if ((!(s->avctx->debug_mv&FF_DEBUG_VIS_MV_P_FOR)) || (pict->pict_type!=AV_PICTURE_TYPE_P))
                                 continue;
                               direction = 0;
                               break;
-                      case 1: if ((!(s->avctx->debug_mv&FF_DEBUG_VIS_MV_B_FOR)) || (pict->pict_type!=FF_B_TYPE))
+                      case 1: if ((!(s->avctx->debug_mv&FF_DEBUG_VIS_MV_B_FOR)) || (pict->pict_type!=AV_PICTURE_TYPE_B))
                                 continue;
                               direction = 0;
                               break;
-                      case 2: if ((!(s->avctx->debug_mv&FF_DEBUG_VIS_MV_B_BACK)) || (pict->pict_type!=FF_B_TYPE))
+                      case 2: if ((!(s->avctx->debug_mv&FF_DEBUG_VIS_MV_B_BACK)) || (pict->pict_type!=AV_PICTURE_TYPE_B))
                                 continue;
                               direction = 1;
                               break;
@@ -1354,6 +1611,7 @@ static inline int hpel_motion_lowres(MpegEncContext *s,
                                   int motion_x, int motion_y)
 {
     const int lowres= s->avctx->lowres;
+    const int op_index= FFMIN(lowres, 2);
     const int s_mask= (2<<lowres)-1;
     int emu=0;
     int sx, sy;
@@ -1372,17 +1630,17 @@ static inline int hpel_motion_lowres(MpegEncContext *s,
 
     if(   (unsigned)src_x > h_edge_pos                 - (!!sx) - w
        || (unsigned)src_y >(v_edge_pos >> field_based) - (!!sy) - h){
-        ff_emulated_edge_mc(s->edge_emu_buffer, src, s->linesize, w+1, (h+1)<<field_based,
+        s->dsp.emulated_edge_mc(s->edge_emu_buffer, src, s->linesize, w+1, (h+1)<<field_based,
                             src_x, src_y<<field_based, h_edge_pos, v_edge_pos);
         src= s->edge_emu_buffer;
         emu=1;
     }
 
-    sx <<= 2 - lowres;
-    sy <<= 2 - lowres;
+    sx= (sx << 2) >> lowres;
+    sy= (sy << 2) >> lowres;
     if(field_select)
         src += s->linesize;
-    pix_op[lowres](dest, src, stride, h, sx, sy);
+    pix_op[op_index](dest, src, stride, h, sx, sy);
     return emu;
 }
 
@@ -1391,17 +1649,18 @@ static av_always_inline void mpeg_motion_lowres(MpegEncContext *s,
                                uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                                int field_based, int bottom_field, int field_select,
                                uint8_t **ref_picture, h264_chroma_mc_func *pix_op,
-                               int motion_x, int motion_y, int h)
+                               int motion_x, int motion_y, int h, int mb_y)
 {
     uint8_t *ptr_y, *ptr_cb, *ptr_cr;
     int mx, my, src_x, src_y, uvsrc_x, uvsrc_y, uvlinesize, linesize, sx, sy, uvsx, uvsy;
     const int lowres= s->avctx->lowres;
+    const int op_index= FFMIN(lowres, 2);
     const int block_s= 8>>lowres;
     const int s_mask= (2<<lowres)-1;
     const int h_edge_pos = s->h_edge_pos >> lowres;
     const int v_edge_pos = s->v_edge_pos >> lowres;
-    linesize   = s->current_picture.linesize[0] << field_based;
-    uvlinesize = s->current_picture.linesize[1] << field_based;
+    linesize   = s->current_picture.f.linesize[0] << field_based;
+    uvlinesize = s->current_picture.f.linesize[1] << field_based;
 
     if(s->quarter_sample){ //FIXME obviously not perfect but qpel will not work in lowres anyway
         motion_x/=2;
@@ -1415,7 +1674,7 @@ static av_always_inline void mpeg_motion_lowres(MpegEncContext *s,
     sx= motion_x & s_mask;
     sy= motion_y & s_mask;
     src_x = s->mb_x*2*block_s               + (motion_x >> (lowres+1));
-    src_y =(s->mb_y*2*block_s>>field_based) + (motion_y >> (lowres+1));
+    src_y =(   mb_y*2*block_s>>field_based) + (motion_y >> (lowres+1));
 
     if (s->out_format == FMT_H263) {
         uvsx = ((motion_x>>1) & s_mask) | (sx&1);
@@ -1428,14 +1687,14 @@ static av_always_inline void mpeg_motion_lowres(MpegEncContext *s,
         uvsx = (2*mx) & s_mask;
         uvsy = (2*my) & s_mask;
         uvsrc_x = s->mb_x*block_s               + (mx >> lowres);
-        uvsrc_y = s->mb_y*block_s               + (my >> lowres);
+        uvsrc_y =    mb_y*block_s               + (my >> lowres);
     } else {
         mx = motion_x / 2;
         my = motion_y / 2;
         uvsx = mx & s_mask;
         uvsy = my & s_mask;
         uvsrc_x = s->mb_x*block_s               + (mx >> (lowres+1));
-        uvsrc_y =(s->mb_y*block_s>>field_based) + (my >> (lowres+1));
+        uvsrc_y =(   mb_y*block_s>>field_based) + (my >> (lowres+1));
     }
 
     ptr_y  = ref_picture[0] + src_y * linesize + src_x;
@@ -1444,21 +1703,21 @@ static av_always_inline void mpeg_motion_lowres(MpegEncContext *s,
 
     if(   (unsigned)src_x > h_edge_pos                 - (!!sx) - 2*block_s
        || (unsigned)src_y >(v_edge_pos >> field_based) - (!!sy) - h){
-            ff_emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize, 17, 17+field_based,
+            s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize, 17, 17+field_based,
                              src_x, src_y<<field_based, h_edge_pos, v_edge_pos);
             ptr_y = s->edge_emu_buffer;
             if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
                 uint8_t *uvbuf= s->edge_emu_buffer+18*s->linesize;
-                ff_emulated_edge_mc(uvbuf  , ptr_cb, s->uvlinesize, 9, 9+field_based,
+                s->dsp.emulated_edge_mc(uvbuf  , ptr_cb, s->uvlinesize, 9, 9+field_based,
                                  uvsrc_x, uvsrc_y<<field_based, h_edge_pos>>1, v_edge_pos>>1);
-                ff_emulated_edge_mc(uvbuf+16, ptr_cr, s->uvlinesize, 9, 9+field_based,
+                s->dsp.emulated_edge_mc(uvbuf+16, ptr_cr, s->uvlinesize, 9, 9+field_based,
                                  uvsrc_x, uvsrc_y<<field_based, h_edge_pos>>1, v_edge_pos>>1);
                 ptr_cb= uvbuf;
                 ptr_cr= uvbuf+16;
             }
     }
 
-    if(bottom_field){ //FIXME use this for field pix too instead of the obnoxious hack which changes picture.data
+    if(bottom_field){ //FIXME use this for field pix too instead of the obnoxious hack which changes picture.f.data
         dest_y += s->linesize;
         dest_cb+= s->uvlinesize;
         dest_cr+= s->uvlinesize;
@@ -1470,15 +1729,15 @@ static av_always_inline void mpeg_motion_lowres(MpegEncContext *s,
         ptr_cr+= s->uvlinesize;
     }
 
-    sx <<= 2 - lowres;
-    sy <<= 2 - lowres;
+    sx= (sx << 2) >> lowres;
+    sy= (sy << 2) >> lowres;
     pix_op[lowres-1](dest_y, ptr_y, linesize, h, sx, sy);
 
     if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
-        uvsx <<= 2 - lowres;
-        uvsy <<= 2 - lowres;
-        pix_op[lowres](dest_cb, ptr_cb, uvlinesize, h >> s->chroma_y_shift, uvsx, uvsy);
-        pix_op[lowres](dest_cr, ptr_cr, uvlinesize, h >> s->chroma_y_shift, uvsx, uvsy);
+        uvsx= (uvsx << 2) >> lowres;
+        uvsy= (uvsy << 2) >> lowres;
+        pix_op[op_index](dest_cb, ptr_cb, uvlinesize, h >> s->chroma_y_shift, uvsx, uvsy);
+        pix_op[op_index](dest_cr, ptr_cr, uvlinesize, h >> s->chroma_y_shift, uvsx, uvsy);
     }
     //FIXME h261 lowres loop filter
 }
@@ -1489,6 +1748,7 @@ static inline void chroma_4mv_motion_lowres(MpegEncContext *s,
                                      h264_chroma_mc_func *pix_op,
                                      int mx, int my){
     const int lowres= s->avctx->lowres;
+    const int op_index= FFMIN(lowres, 2);
     const int block_s= 8>>lowres;
     const int s_mask= (2<<lowres)-1;
     const int h_edge_pos = s->h_edge_pos >> (lowres+1);
@@ -1516,21 +1776,21 @@ static inline void chroma_4mv_motion_lowres(MpegEncContext *s,
     if(s->flags&CODEC_FLAG_EMU_EDGE){
         if(   (unsigned)src_x > h_edge_pos - (!!sx) - block_s
            || (unsigned)src_y > v_edge_pos - (!!sy) - block_s){
-            ff_emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 9, 9, src_x, src_y, h_edge_pos, v_edge_pos);
+            s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 9, 9, src_x, src_y, h_edge_pos, v_edge_pos);
             ptr= s->edge_emu_buffer;
             emu=1;
         }
     }
-    sx <<= 2 - lowres;
-    sy <<= 2 - lowres;
-    pix_op[lowres](dest_cb, ptr, s->uvlinesize, block_s, sx, sy);
+    sx= (sx << 2) >> lowres;
+    sy= (sy << 2) >> lowres;
+    pix_op[op_index](dest_cb, ptr, s->uvlinesize, block_s, sx, sy);
 
     ptr = ref_picture[2] + offset;
     if(emu){
-        ff_emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 9, 9, src_x, src_y, h_edge_pos, v_edge_pos);
+        s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 9, 9, src_x, src_y, h_edge_pos, v_edge_pos);
         ptr= s->edge_emu_buffer;
     }
-    pix_op[lowres](dest_cr, ptr, s->uvlinesize, block_s, sx, sy);
+    pix_op[op_index](dest_cr, ptr, s->uvlinesize, block_s, sx, sy);
 }
 
 /**
@@ -1541,7 +1801,7 @@ static inline void chroma_4mv_motion_lowres(MpegEncContext *s,
  * @param dest_cr chroma cr/v destination pointer
  * @param dir direction (0->forward, 1->backward)
  * @param ref_picture array[3] of pointers to the 3 planes of the reference picture
- * @param pic_op halfpel motion compensation function (average or put normally)
+ * @param pix_op halfpel motion compensation function (average or put normally)
  * the motion vectors are taken from s->mv and the MV type from s->mv_type
  */
 static inline void MPV_motion_lowres(MpegEncContext *s,
@@ -1562,7 +1822,7 @@ static inline void MPV_motion_lowres(MpegEncContext *s,
         mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                     0, 0, 0,
                     ref_picture, pix_op,
-                    s->mv[dir][0][0], s->mv[dir][0][1], 2*block_s);
+                    s->mv[dir][0][0], s->mv[dir][0][1], 2*block_s, mb_y);
         break;
     case MV_TYPE_8X8:
         mx = 0;
@@ -1589,37 +1849,37 @@ static inline void MPV_motion_lowres(MpegEncContext *s,
             mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                         1, 0, s->field_select[dir][0],
                         ref_picture, pix_op,
-                        s->mv[dir][0][0], s->mv[dir][0][1], block_s);
+                        s->mv[dir][0][0], s->mv[dir][0][1], block_s, mb_y);
             /* bottom field */
             mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                         1, 1, s->field_select[dir][1],
                         ref_picture, pix_op,
-                        s->mv[dir][1][0], s->mv[dir][1][1], block_s);
+                        s->mv[dir][1][0], s->mv[dir][1][1], block_s, mb_y);
         } else {
-            if(s->picture_structure != s->field_select[dir][0] + 1 && s->pict_type != FF_B_TYPE && !s->first_field){
-                ref_picture= s->current_picture_ptr->data;
+            if(s->picture_structure != s->field_select[dir][0] + 1 && s->pict_type != AV_PICTURE_TYPE_B && !s->first_field){
+                ref_picture = s->current_picture_ptr->f.data;
             }
 
             mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                         0, 0, s->field_select[dir][0],
                         ref_picture, pix_op,
-                        s->mv[dir][0][0], s->mv[dir][0][1], 2*block_s);
+                        s->mv[dir][0][0], s->mv[dir][0][1], 2*block_s, mb_y>>1);
         }
         break;
     case MV_TYPE_16X8:
         for(i=0; i<2; i++){
             uint8_t ** ref2picture;
 
-            if(s->picture_structure == s->field_select[dir][i] + 1 || s->pict_type == FF_B_TYPE || s->first_field){
+            if(s->picture_structure == s->field_select[dir][i] + 1 || s->pict_type == AV_PICTURE_TYPE_B || s->first_field){
                 ref2picture= ref_picture;
             }else{
-                ref2picture= s->current_picture_ptr->data;
+                ref2picture = s->current_picture_ptr->f.data;
             }
 
             mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                         0, 0, s->field_select[dir][i],
                         ref2picture, pix_op,
-                        s->mv[dir][i][0], s->mv[dir][i][1] + 2*block_s*i, block_s);
+                        s->mv[dir][i][0], s->mv[dir][i][1] + 2*block_s*i, block_s, mb_y>>1);
 
             dest_y += 2*block_s*s->linesize;
             dest_cb+= (2*block_s>>s->chroma_y_shift)*s->uvlinesize;
@@ -1634,7 +1894,7 @@ static inline void MPV_motion_lowres(MpegEncContext *s,
                     mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                                 1, j, j^i,
                                 ref_picture, pix_op,
-                                s->mv[dir][2*i + j][0], s->mv[dir][2*i + j][1], block_s);
+                                s->mv[dir][2*i + j][0], s->mv[dir][2*i + j][1], block_s, mb_y);
                 }
                 pix_op = s->dsp.avg_h264_chroma_pixels_tab;
             }
@@ -1643,14 +1903,14 @@ static inline void MPV_motion_lowres(MpegEncContext *s,
                 mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                             0, 0, s->picture_structure != i+1,
                             ref_picture, pix_op,
-                            s->mv[dir][2*i][0],s->mv[dir][2*i][1],2*block_s);
+                            s->mv[dir][2*i][0],s->mv[dir][2*i][1],2*block_s, mb_y>>1);
 
                 // after put we make avg of the same block
                 pix_op = s->dsp.avg_h264_chroma_pixels_tab;
 
                 //opposite parity is always in the same frame if this is second field
                 if(!s->first_field){
-                    ref_picture = s->current_picture_ptr->data;
+                    ref_picture = s->current_picture_ptr->f.data;
                 }
             }
         }
@@ -1659,6 +1919,43 @@ 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)
@@ -1733,26 +2030,27 @@ static av_always_inline
 void MPV_decode_mb_internal(MpegEncContext *s, DCTELEM block[12][64],
                             int lowres_flag, int is_mpeg12)
 {
-    int mb_x, mb_y;
     const int mb_xy = s->mb_y * s->mb_stride + s->mb_x;
     if(CONFIG_MPEG_XVMC_DECODER && s->avctx->xvmc_acceleration){
         ff_xvmc_decode_mb(s);//xvmc uses pblocks
         return;
     }
 
-    mb_x = s->mb_x;
-    mb_y = s->mb_y;
-
     if(s->avctx->debug&FF_DEBUG_DCT_COEFF) {
        /* save DCT coefficients */
        int i,j;
-       DCTELEM *dct = &s->current_picture.dct_coeff[mb_xy*64*6];
-       for(i=0; i<6; i++)
-           for(j=0; j<64; j++)
+       DCTELEM *dct = &s->current_picture.f.dct_coeff[mb_xy * 64 * 6];
+       av_log(s->avctx, AV_LOG_DEBUG, "DCT coeffs of MB at %dx%d:\n", s->mb_x, s->mb_y);
+       for(i=0; i<6; i++){
+           for(j=0; j<64; j++){
                *dct++ = block[i][s->dsp.idct_permutation[j]];
+               av_log(s->avctx, AV_LOG_DEBUG, "%5d", dct[-1]);
+           }
+           av_log(s->avctx, AV_LOG_DEBUG, "\n");
+       }
     }
 
-    s->current_picture.qscale_table[mb_xy]= s->qscale;
+    s->current_picture.f.qscale_table[mb_xy] = s->qscale;
 
     /* update DC predictors for P macroblocks */
     if (!s->mb_intra) {
@@ -1768,36 +2066,36 @@ void MPV_decode_mb_internal(MpegEncContext *s, DCTELEM block[12][64],
     else if (!is_mpeg12 && (s->h263_pred || s->h263_aic))
         s->mbintra_table[mb_xy]=1;
 
-    if ((s->flags&CODEC_FLAG_PSNR) || !(s->encoding && (s->intra_only || s->pict_type==FF_B_TYPE) && s->avctx->mb_decision != FF_MB_DECISION_RD)) { //FIXME precalc
+    if ((s->flags&CODEC_FLAG_PSNR) || !(s->encoding && (s->intra_only || s->pict_type==AV_PICTURE_TYPE_B) && s->avctx->mb_decision != FF_MB_DECISION_RD)) { //FIXME precalc
         uint8_t *dest_y, *dest_cb, *dest_cr;
         int dct_linesize, dct_offset;
         op_pixels_func (*op_pix)[4];
         qpel_mc_func (*op_qpix)[16];
-        const int linesize= s->current_picture.linesize[0]; //not s->linesize as this would be wrong for field pics
-        const int uvlinesize= s->current_picture.linesize[1];
-        const int readable= s->pict_type != FF_B_TYPE || s->encoding || s->avctx->draw_horiz_band || lowres_flag;
+        const int linesize   = s->current_picture.f.linesize[0]; //not s->linesize as this would be wrong for field pics
+        const int uvlinesize = s->current_picture.f.linesize[1];
+        const int readable= s->pict_type != AV_PICTURE_TYPE_B || s->encoding || s->avctx->draw_horiz_band || lowres_flag;
         const int block_size= lowres_flag ? 8>>s->avctx->lowres : 8;
 
         /* avoid copy if macroblock skipped in last frame too */
         /* skip only during decoding as we might trash the buffers during encoding a bit */
         if(!s->encoding){
             uint8_t *mbskip_ptr = &s->mbskip_table[mb_xy];
-            const int age= s->current_picture.age;
+            const int age = s->current_picture.f.age;
 
             assert(age);
 
             if (s->mb_skipped) {
                 s->mb_skipped= 0;
-                assert(s->pict_type!=FF_I_TYPE);
+                assert(s->pict_type!=AV_PICTURE_TYPE_I);
 
                 (*mbskip_ptr) ++; /* indicate that this time we skipped it */
                 if(*mbskip_ptr >99) *mbskip_ptr= 99;
 
                 /* if previous was skipped too, then nothing to do !  */
-                if (*mbskip_ptr >= age && s->current_picture.reference){
+                if (*mbskip_ptr >= age && s->current_picture.f.reference){
                     return;
                 }
-            } else if(!s->current_picture.reference){
+            } else if(!s->current_picture.f.reference) {
                 (*mbskip_ptr) ++; /* increase counter so the age can be compared cleanly */
                 if(*mbskip_ptr >99) *mbskip_ptr= 99;
             } else{
@@ -1822,45 +2120,54 @@ 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;
 
                     if (s->mv_dir & MV_DIR_FORWARD) {
-                        MPV_motion_lowres(s, dest_y, dest_cb, dest_cr, 0, s->last_picture.data, op_pix);
+                        MPV_motion_lowres(s, dest_y, dest_cb, dest_cr, 0, s->last_picture.f.data, op_pix);
                         op_pix = s->dsp.avg_h264_chroma_pixels_tab;
                     }
                     if (s->mv_dir & MV_DIR_BACKWARD) {
-                        MPV_motion_lowres(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.data, op_pix);
+                        MPV_motion_lowres(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.f.data, op_pix);
                     }
                 }else{
                     op_qpix= s->me.qpel_put;
-                    if ((!s->no_rounding) || s->pict_type==FF_B_TYPE){
+                    if ((!s->no_rounding) || s->pict_type==AV_PICTURE_TYPE_B){
                         op_pix = s->dsp.put_pixels_tab;
                     }else{
                         op_pix = s->dsp.put_no_rnd_pixels_tab;
                     }
                     if (s->mv_dir & MV_DIR_FORWARD) {
-                        MPV_motion(s, dest_y, dest_cb, dest_cr, 0, s->last_picture.data, op_pix, op_qpix);
+                        MPV_motion(s, dest_y, dest_cb, dest_cr, 0, s->last_picture.f.data, op_pix, op_qpix);
                         op_pix = s->dsp.avg_pixels_tab;
                         op_qpix= s->me.qpel_avg;
                     }
                     if (s->mv_dir & MV_DIR_BACKWARD) {
-                        MPV_motion(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.data, op_pix, op_qpix);
+                        MPV_motion(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.f.data, op_pix, op_qpix);
                     }
                 }
             }
 
             /* skip dequant / idct if we are really late ;) */
-            if(s->hurry_up>1) goto skip_idct;
             if(s->avctx->skip_idct){
-                if(  (s->avctx->skip_idct >= AVDISCARD_NONREF && s->pict_type == FF_B_TYPE)
-                   ||(s->avctx->skip_idct >= AVDISCARD_NONKEY && s->pict_type != FF_I_TYPE)
+                if(  (s->avctx->skip_idct >= AVDISCARD_NONREF && s->pict_type == AV_PICTURE_TYPE_B)
+                   ||(s->avctx->skip_idct >= AVDISCARD_NONKEY && s->pict_type != AV_PICTURE_TYPE_I)
                    || s->avctx->skip_idct >= AVDISCARD_ALL)
                     goto skip_idct;
             }
 
             /* add dct residue */
-            if(s->encoding || !(   s->h263_msmpeg4 || s->codec_id==CODEC_ID_MPEG1VIDEO || s->codec_id==CODEC_ID_MPEG2VIDEO
+            if(s->encoding || !(   s->msmpeg4_version || s->codec_id==CODEC_ID_MPEG1VIDEO || s->codec_id==CODEC_ID_MPEG2VIDEO
                                 || (s->codec_id==CODEC_ID_MPEG4 && !s->mpeg_quant))){
                 add_dequant_dct(s, block[0], 0, dest_y                          , dct_linesize, s->qscale);
                 add_dequant_dct(s, block[1], 1, dest_y              + block_size, dct_linesize, s->qscale);
@@ -1908,7 +2215,7 @@ void MPV_decode_mb_internal(MpegEncContext *s, DCTELEM block[12][64],
                     }
                 }//fi gray
             }
-            else if (CONFIG_WMV2) {
+            else if (CONFIG_WMV2_DECODER || CONFIG_WMV2_ENCODER) {
                 ff_wmv2_add_mb(s, block, dest_y, dest_cb, dest_cr);
             }
         } else {
@@ -1986,26 +2293,50 @@ 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.f.reference
+       && !s->intra_only
+       && !(s->flags&CODEC_FLAG_EMU_EDGE)) {
+        int sides = 0, edge_h;
+        int hshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_w;
+        int vshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_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->f.data[0] +  y         *s->linesize  , s->linesize,
+                          s->h_edge_pos        , edge_h        , EDGE_WIDTH        , EDGE_WIDTH        , sides);
+        s->dsp.draw_edges(s->current_picture_ptr->f.data[1] + (y>>vshift)*s->uvlinesize, s->uvlinesize,
+                          s->h_edge_pos>>hshift, edge_h>>hshift, EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, sides);
+        s->dsp.draw_edges(s->current_picture_ptr->f.data[2] + (y>>vshift)*s->uvlinesize, s->uvlinesize,
+                          s->h_edge_pos>>hshift, edge_h>>hshift, EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, 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;
         int offset[4];
 
-        if(s->picture_structure != PICT_FRAME){
-            h <<= 1;
-            y <<= 1;
-            if(s->first_field  && !(s->avctx->slice_flags&SLICE_FLAG_ALLOW_FIELD)) return;
-        }
-
-        h= FFMIN(h, s->avctx->height - y);
-
-        if(s->pict_type==FF_B_TYPE || s->low_delay || (s->avctx->slice_flags&SLICE_FLAG_CODED_ORDER))
+        if(s->pict_type==AV_PICTURE_TYPE_B || s->low_delay || (s->avctx->slice_flags&SLICE_FLAG_CODED_ORDER))
             src= (AVFrame*)s->current_picture_ptr;
         else if(s->last_picture_ptr)
             src= (AVFrame*)s->last_picture_ptr;
         else
             return;
 
-        if(s->pict_type==FF_B_TYPE && s->picture_structure == PICT_FRAME && s->out_format != FMT_H264){
+        if(s->pict_type==AV_PICTURE_TYPE_B && s->picture_structure == PICT_FRAME && s->out_format != FMT_H264){
             offset[0]=
             offset[1]=
             offset[2]=
@@ -2025,8 +2356,8 @@ void ff_draw_horiz_band(MpegEncContext *s, int y, int h){
 }
 
 void ff_init_block_index(MpegEncContext *s){ //FIXME maybe rename
-    const int linesize= s->current_picture.linesize[0]; //not s->linesize as this would be wrong for field pics
-    const int uvlinesize= s->current_picture.linesize[1];
+    const int linesize   = s->current_picture.f.linesize[0]; //not s->linesize as this would be wrong for field pics
+    const int uvlinesize = s->current_picture.f.linesize[1];
     const int mb_size= 4 - s->avctx->lowres;
 
     s->block_index[0]= s->b8_stride*(s->mb_y*2    ) - 2 + s->mb_x*2;
@@ -2037,15 +2368,22 @@ void ff_init_block_index(MpegEncContext *s){ //FIXME maybe rename
     s->block_index[5]= s->mb_stride*(s->mb_y + s->mb_height + 2) + s->b8_stride*s->mb_height*2 + s->mb_x - 1;
     //block_index is not used by mpeg2, so it is not affected by chroma_format
 
-    s->dest[0] = s->current_picture.data[0] + ((s->mb_x - 1) << mb_size);
-    s->dest[1] = s->current_picture.data[1] + ((s->mb_x - 1) << (mb_size - s->chroma_x_shift));
-    s->dest[2] = s->current_picture.data[2] + ((s->mb_x - 1) << (mb_size - s->chroma_x_shift));
+    s->dest[0] = s->current_picture.f.data[0] + ((s->mb_x - 1) <<  mb_size);
+    s->dest[1] = s->current_picture.f.data[1] + ((s->mb_x - 1) << (mb_size - s->chroma_x_shift));
+    s->dest[2] = s->current_picture.f.data[2] + ((s->mb_x - 1) << (mb_size - s->chroma_x_shift));
 
-    if(!(s->pict_type==FF_B_TYPE && s->avctx->draw_horiz_band && s->picture_structure==PICT_FRAME))
+    if(!(s->pict_type==AV_PICTURE_TYPE_B && s->avctx->draw_horiz_band && s->picture_structure==PICT_FRAME))
     {
+        if(s->picture_structure==PICT_FRAME){
         s->dest[0] += s->mb_y *   linesize << mb_size;
         s->dest[1] += s->mb_y * uvlinesize << (mb_size - s->chroma_y_shift);
         s->dest[2] += s->mb_y * uvlinesize << (mb_size - s->chroma_y_shift);
+        }else{
+            s->dest[0] += (s->mb_y>>1) *   linesize << mb_size;
+            s->dest[1] += (s->mb_y>>1) * uvlinesize << (mb_size - s->chroma_y_shift);
+            s->dest[2] += (s->mb_y>>1) * uvlinesize << (mb_size - s->chroma_y_shift);
+            assert((s->mb_y&1) == (s->picture_structure == PICT_BOTTOM_FIELD));
+        }
     }
 }
 
@@ -2056,14 +2394,16 @@ void ff_mpeg_flush(AVCodecContext *avctx){
     if(s==NULL || s->picture==NULL)
         return;
 
-    for(i=0; i<MAX_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))
-        avctx->release_buffer(avctx, (AVFrame*)&s->picture[i]);
+    for(i=0; i<s->picture_count; i++){
+       if (s->picture[i].f.data[0] &&
+           (s->picture[i].f.type == FF_BUFFER_TYPE_INTERNAL ||
+            s->picture[i].f.type == FF_BUFFER_TYPE_USER))
+        free_frame_buffer(s, &s->picture[i]);
     }
     s->current_picture_ptr = s->last_picture_ptr = s->next_picture_ptr = NULL;
 
     s->mb_x= s->mb_y= 0;
+    s->closed_gop= 0;
 
     s->parse_context.state= -1;
     s->parse_context.frame_start_found= 0;
@@ -2309,3 +2649,9 @@ 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 && !s->error_occurred)
+        ff_thread_report_progress((AVFrame*)s->current_picture_ptr, s->mb_y, 0);
+}