vdpauh264dec: improve further

6 files changed, 300 insertions, 92 deletions

diff --git a/sys/vdpau/h264/gsth264dpb.c b/sys/vdpau/h264/gsth264dpb.c
index a2d695686..3f61b61c4 100644
--- a/sys/vdpau/h264/gsth264dpb.c
+++ b/sys/vdpau/h264/gsth264dpb.c
@@ -24,7 +24,8 @@
 enum
 {
   PROP_0,
-  PROP_NUM_REF_FRAMES
+  PROP_NUM_REF_FRAMES,
+  PROP_MAX_LONGTERM_IDX
 };
 
 GST_DEBUG_CATEGORY_STATIC (h264dpb_debug);
@@ -48,15 +49,15 @@ gst_h264_dpb_fill_reference_frames (GstH264DPB * dpb,
     GstVdpH264Frame *frame = frames[i];
 
     reference_frames[i].surface =
-        GST_VDP_VIDEO_BUFFER (GST_VIDEO_FRAME_CAST (frame)->
-        src_buffer)->surface;
+        GST_VDP_VIDEO_BUFFER (GST_VIDEO_FRAME_CAST (frame)->src_buffer)->
+        surface;
 
     reference_frames[i].is_long_term = frame->is_long_term;
     reference_frames[i].top_is_reference = frame->is_reference;
     reference_frames[i].bottom_is_reference = frame->is_reference;
     reference_frames[i].field_order_cnt[0] = frame->poc;
     reference_frames[i].field_order_cnt[1] = frame->poc;
-    reference_frames[i].frame_idx = frame->frame_num;
+    reference_frames[i].frame_idx = frame->frame_idx;
   }
 
   for (i = dpb->n_frames; i < 16; i++) {
@@ -134,8 +135,11 @@ gst_h264_dpb_add (GstH264DPB * dpb, GstVdpH264Frame * h264_frame)
 
   frames = dpb->frames;
 
-  if (h264_frame->is_reference) {
+  if (h264_frame->is_reference && h264_frame->is_long_term &&
+      (h264_frame->frame_idx > dpb->max_longterm_frame_idx))
+    h264_frame->is_reference = FALSE;
 
+  if (h264_frame->is_reference) {
     while (dpb->n_frames == dpb->max_frames) {
       if (!gst_h264_dpb_bump (dpb, G_MAXUINT)) {
         GST_ERROR_OBJECT (dpb, "Couldn't make room in DPB");
@@ -146,14 +150,8 @@ gst_h264_dpb_add (GstH264DPB * dpb, GstVdpH264Frame * h264_frame)
   }
 
   else {
-    if (dpb->n_frames == dpb->max_frames) {
-      while (gst_h264_dpb_bump (dpb, h264_frame->poc));
-
-      dpb->output (dpb, h264_frame);
-    }
-
-    else
-      dpb->frames[dpb->n_frames++] = h264_frame;
+    while (gst_h264_dpb_bump (dpb, h264_frame->poc));
+    dpb->output (dpb, h264_frame);
   }
 
   return TRUE;
@@ -183,7 +181,7 @@ gst_h264_dpb_mark_sliding (GstH264DPB * dpb)
 {
   GstVdpH264Frame **frames;
   guint i;
-  gint mark_idx;
+  gint mark_idx = -1;
 
   if (dpb->n_frames != dpb->max_frames)
     return;
@@ -199,7 +197,7 @@ gst_h264_dpb_mark_sliding (GstH264DPB * dpb)
   if (mark_idx != -1) {
     for (i = mark_idx; i < dpb->n_frames; i++) {
       if (frames[i]->is_reference && !frames[i]->is_long_term &&
-          frames[i]->frame_num < frames[mark_idx]->frame_num)
+          frames[i]->frame_idx < frames[mark_idx]->frame_idx)
         mark_idx = i;
     }
 
@@ -209,6 +207,94 @@ gst_h264_dpb_mark_sliding (GstH264DPB * dpb)
   }
 }
 
+void
+gst_h264_dpb_mark_long_term (GstH264DPB * dpb, guint16 pic_num,
+    guint16 long_term_frame_idx)
+{
+  GstVdpH264Frame **frames;
+  guint i;
+  gint mark_idx = -1;
+
+  frames = dpb->frames;
+  for (i = 0; i < dpb->n_frames; i++) {
+    if (frames[i]->is_reference && !frames[i]->is_long_term &&
+        frames[i]->frame_idx == pic_num) {
+      mark_idx = i;
+      break;
+    }
+  }
+
+  if (mark_idx != -1) {
+    frames[mark_idx]->is_long_term = TRUE;
+    frames[mark_idx]->frame_idx = long_term_frame_idx;
+  }
+}
+
+void
+gst_h264_dpb_mark_short_term_unused (GstH264DPB * dpb, guint16 pic_num)
+{
+  GstVdpH264Frame **frames;
+  guint i;
+  gint mark_idx = -1;
+
+  frames = dpb->frames;
+  for (i = 0; i < dpb->n_frames; i++) {
+    if (frames[i]->is_reference && !frames[i]->is_long_term &&
+        frames[i]->frame_idx == pic_num) {
+      mark_idx = i;
+      break;
+    }
+  }
+
+  if (mark_idx != -1) {
+
+    frames[mark_idx]->is_reference = FALSE;
+    if (!frames[mark_idx]->output_needed)
+      gst_h264_dpb_remove (dpb, mark_idx);
+  }
+}
+
+void
+gst_h264_dpb_mark_long_term_unused (GstH264DPB * dpb, guint16 long_term_pic_num)
+{
+  GstVdpH264Frame **frames;
+  guint i;
+  gint mark_idx = -1;
+
+  frames = dpb->frames;
+  for (i = 0; i < dpb->n_frames; i++) {
+    if (frames[i]->is_reference && frames[i]->is_long_term &&
+        frames[i]->frame_idx == long_term_pic_num) {
+      mark_idx = i;
+      break;
+    }
+  }
+
+  if (mark_idx != -1) {
+
+    frames[mark_idx]->is_reference = FALSE;
+    if (!frames[mark_idx]->output_needed)
+      gst_h264_dpb_remove (dpb, mark_idx);
+  }
+}
+
+void
+gst_h264_dpb_mark_all_unused (GstH264DPB * dpb)
+{
+  GstVdpH264Frame **frames;
+  guint i;
+
+  frames = dpb->frames;
+  for (i = 0; i < dpb->n_frames; i++) {
+    frames[i]->is_reference = FALSE;
+    if (!frames[i]->output_needed) {
+      gst_h264_dpb_remove (dpb, i);
+      i--;
+    }
+  }
+
+}
+
 /* GObject vmethod implementations */
 static void
 gst_h264_dpb_get_property (GObject * object, guint property_id,
@@ -220,6 +306,9 @@ gst_h264_dpb_get_property (GObject * object, guint property_id,
     case PROP_NUM_REF_FRAMES:
       g_value_set_uint (value, dpb->max_frames);
       break;
+    case PROP_MAX_LONGTERM_IDX:
+      g_value_set_int (value, dpb->max_longterm_frame_idx);
+      break;
     default:
       G_OBJECT_WARN_INVALID_PROPERTY_ID (object, property_id, pspec);
       break;
@@ -244,6 +333,27 @@ gst_h264_dpb_set_property (GObject * object, guint property_id,
       break;
     }
 
+    case PROP_MAX_LONGTERM_IDX:
+    {
+      GstVdpH264Frame **frames;
+      guint i;
+
+      dpb->max_longterm_frame_idx = g_value_get_int (value);
+
+      frames = dpb->frames;
+      for (i = dpb->n_frames; i < dpb->n_frames; i++) {
+        if (frames[i]->is_reference && frames[i]->is_long_term &&
+            frames[i]->frame_idx > dpb->max_longterm_frame_idx) {
+          frames[i]->is_reference = FALSE;
+          if (!frames[i]->output_needed) {
+            gst_h264_dpb_remove (dpb, i);
+            i--;
+          }
+        }
+      }
+      break;
+    }
+
     default:
       G_OBJECT_WARN_INVALID_PROPERTY_ID (object, property_id, pspec);
       break;
@@ -262,6 +372,7 @@ static void
 gst_h264_dpb_init (GstH264DPB * dpb)
 {
   dpb->n_frames = 0;
+  dpb->max_longterm_frame_idx = -1;
   dpb->max_frames = MAX_DPB_SIZE;
 }
 
@@ -279,4 +390,9 @@ gst_h264_dpb_class_init (GstH264DPBClass * klass)
           "How many reference frames the DPB should hold ",
           0, 16, 16, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
 
+  g_object_class_install_property (object_class, PROP_MAX_LONGTERM_IDX,
+      g_param_spec_int ("max-longterm-frame-idx", "MaxLongTermFrameIDX",
+          "Maximum long-term frame index",
+          -1, G_MAXINT, -1, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+
 }
diff --git a/sys/vdpau/h264/gsth264dpb.h b/sys/vdpau/h264/gsth264dpb.h
index ff5fe0f55..d8962fe8e 100644
--- a/sys/vdpau/h264/gsth264dpb.h
+++ b/sys/vdpau/h264/gsth264dpb.h
@@ -50,6 +50,7 @@ struct _GstH264DPB
   guint n_frames;
   
   guint max_frames;
+  gint max_longterm_frame_idx;
 
   void (*output) (GstH264DPB *dpb, GstVdpH264Frame *h264_frame);
 };
@@ -67,6 +68,11 @@ void gst_h264_dpb_flush (GstH264DPB *dpb, gboolean output);
 
 void gst_h264_dpb_mark_sliding (GstH264DPB *dpb);
 
+void gst_h264_dpb_mark_long_term_unused (GstH264DPB *dpb, guint16 long_term_pic_num);
+void gst_h264_dpb_mark_short_term_unused (GstH264DPB *dpb, guint16 pic_num);
+void gst_h264_dpb_mark_all_unused (GstH264DPB *dpb);
+void gst_h264_dpb_mark_long_term (GstH264DPB *dpb, guint16 pic_num, guint16 long_term_frame_idx);
+
 GType gst_h264_dpb_get_type (void) G_GNUC_CONST;
 
 G_END_DECLS
diff --git a/sys/vdpau/h264/gsth264parser.c b/sys/vdpau/h264/gsth264parser.c
index 6d442d8d3..8de7c2fdb 100644
--- a/sys/vdpau/h264/gsth264parser.c
+++ b/sys/vdpau/h264/gsth264parser.c
@@ -257,7 +257,7 @@ gst_h264_parser_parse_scaling_list (GstNalReader * reader,
 {
   guint i;
 
-  GST_WARNING ("parsing scaling lists");
+  GST_DEBUG ("parsing scaling lists");
 
   for (i = 0; i < 12; i++) {
     gboolean use_default = FALSE;
@@ -460,6 +460,12 @@ gst_h264_parser_parse_sequence (GstH264Parser * parser, guint8 * data,
       goto error;
   }
 
+  /* calculate ChromaArrayType */
+  if (seq->separate_colour_plane_flag)
+    seq->ChromaArrayType = 0;
+  else
+    seq->ChromaArrayType = seq->chroma_format_idc;
+
   GST_DEBUG ("adding sequence parameter set with id: %d to hash table",
       seq->id);
   g_hash_table_replace (parser->sequences, &seq->id, seq);
@@ -523,6 +529,7 @@ gst_h264_parser_parse_picture (GstH264Parser * parser, guint8 * data,
   GstH264Picture *pic;
   gint seq_parameter_set_id;
   GstH264Sequence *seq;
+  guint8 pic_scaling_matrix_present_flag;
 
   g_return_val_if_fail (GST_IS_H264_PARSER (parser), NULL);
   g_return_val_if_fail (data != NULL, NULL);
@@ -557,27 +564,28 @@ gst_h264_parser_parse_picture (GstH264Parser * parser, guint8 * data,
 
       for (i = 0; i <= pic->num_slice_groups_minus1; i++)
         READ_UE (&reader, pic->run_length_minus1[i]);
-    }
-  } else if (pic->slice_group_map_type == 2) {
-    gint i;
+    } else if (pic->slice_group_map_type == 2) {
+      gint i;
+
+      for (i = 0; i <= pic->num_slice_groups_minus1; i++) {
+        READ_UE (&reader, pic->top_left[i]);
+        READ_UE (&reader, pic->bottom_right[i]);
+      }
+    } else if (pic->slice_group_map_type >= 3 && pic->slice_group_map_type <= 5) {
+      READ_UINT8 (&reader, pic->slice_group_change_direction_flag, 1);
+      READ_UE (&reader, pic->slice_group_change_rate_minus1);
+    } else if (pic->slice_group_map_type == 6) {
+      gint bits;
+      gint i;
+
+      READ_UE (&reader, pic->pic_size_in_map_units_minus1);
+      bits = ceil (log2 (pic->num_slice_groups_minus1 + 1));
 
-    for (i = 0; i <= pic->num_slice_groups_minus1; i++) {
-      READ_UE (&reader, pic->top_left[i]);
-      READ_UE (&reader, pic->bottom_right[i]);
+      pic->slice_group_id =
+          g_new (guint8, pic->pic_size_in_map_units_minus1 + 1);
+      for (i = 0; i <= pic->pic_size_in_map_units_minus1; i++)
+        READ_UINT8 (&reader, pic->slice_group_id[i], bits);
     }
-  } else if (pic->slice_group_map_type >= 3 && pic->slice_group_map_type <= 5) {
-    READ_UINT8 (&reader, pic->slice_group_change_direction_flag, 1);
-    READ_UE (&reader, pic->slice_group_change_rate_minus1);
-  } else if (pic->slice_group_map_type == 6) {
-    gint bits;
-    gint i;
-
-    READ_UE (&reader, pic->pic_size_in_map_units_minus1);
-    bits = ceil (log2 (pic->num_slice_groups_minus1 + 1));
-
-    pic->slice_group_id = g_new (guint8, pic->pic_size_in_map_units_minus1 + 1);
-    for (i = 0; i <= pic->pic_size_in_map_units_minus1; i++)
-      READ_UINT8 (&reader, pic->slice_group_id[i], bits);
   }
 
   READ_UE_ALLOWED (&reader, pic->num_ref_idx_l0_active_minus1, 0, 31);
@@ -597,8 +605,8 @@ gst_h264_parser_parse_picture (GstH264Parser * parser, guint8 * data,
 
   READ_UINT8 (&reader, pic->transform_8x8_mode_flag, 1);
 
-  READ_UINT8 (&reader, pic->scaling_matrix_present_flag, 1);
-  if (pic->scaling_matrix_present_flag) {
+  READ_UINT8 (&reader, pic_scaling_matrix_present_flag, 1);
+  if (pic_scaling_matrix_present_flag) {
     if (seq->scaling_matrix_present_flag) {
       if (!gst_h264_parser_parse_scaling_list (&reader,
               pic->scaling_lists_4x4, pic->scaling_lists_8x8,
@@ -613,6 +621,9 @@ gst_h264_parser_parse_picture (GstH264Parser * parser, guint8 * data,
               default_8x8_inter, default_8x8_intra, seq->chroma_format_idc))
         goto error;
     }
+  } else {
+    memcpy (&pic->scaling_lists_4x4, &seq->scaling_lists_4x4, 96);
+    memcpy (&pic->scaling_lists_8x8, &seq->scaling_lists_8x8, 384);
   }
 
   READ_SE_ALLOWED (&reader, pic->second_chroma_qp_index_offset, -12, 12);
@@ -653,7 +664,7 @@ gst_h264_slice_parse_pred_weight_table (GstH264Slice * slice,
     memset (p->chroma_offset_l0, 0, 64);
   }
 
-  for (i = 0; i <= pic->num_ref_idx_l0_active_minus1; i++) {
+  for (i = 0; i <= slice->num_ref_idx_l0_active_minus1; i++) {
     guint8 luma_weight_l0_flag;
 
     READ_UINT8 (reader, luma_weight_l0_flag, 1);
@@ -666,15 +677,17 @@ gst_h264_slice_parse_pred_weight_table (GstH264Slice * slice,
       gint j;
 
       READ_UINT8 (reader, chroma_weight_l0_flag, 1);
-      for (j = 0; j <= 2; j++) {
-        READ_SE_ALLOWED (reader, p->chroma_weight_l0[i][j], -128, 127);
-        READ_SE_ALLOWED (reader, p->chroma_offset_l0[i][j], -128, 127);
+      if (chroma_weight_l0_flag) {
+        for (j = 0; j < 2; j++) {
+          READ_SE_ALLOWED (reader, p->chroma_weight_l0[i][j], -128, 127);
+          READ_SE_ALLOWED (reader, p->chroma_offset_l0[i][j], -128, 127);
+        }
       }
     }
   }
 
   if (GST_H264_IS_B_SLICE (slice->type)) {
-    for (i = 0; i <= pic->num_ref_idx_l1_active_minus1; i++) {
+    for (i = 0; i <= slice->num_ref_idx_l1_active_minus1; i++) {
       guint8 luma_weight_l1_flag;
 
       READ_UINT8 (reader, luma_weight_l1_flag, 1);
@@ -687,9 +700,11 @@ gst_h264_slice_parse_pred_weight_table (GstH264Slice * slice,
         gint j;
 
         READ_UINT8 (reader, chroma_weight_l1_flag, 1);
-        for (j = 0; j <= 2; j++) {
-          READ_SE_ALLOWED (reader, p->chroma_weight_l1[i][j], -128, 127);
-          READ_SE_ALLOWED (reader, p->chroma_offset_l1[i][j], -128, 127);
+        if (chroma_weight_l1_flag) {
+          for (j = 0; j < 2; j++) {
+            READ_SE_ALLOWED (reader, p->chroma_weight_l1[i][j], -128, 127);
+            READ_SE_ALLOWED (reader, p->chroma_offset_l1[i][j], -128, 127);
+          }
         }
       }
     }
@@ -740,7 +755,7 @@ gst_h264_slice_parse_ref_pic_list_reordering (GstH264Slice * slice,
         if (reordering_of_pic_nums_idc == 0 || reordering_of_pic_nums_idc == 1) {
           guint32 abs_diff_num_minus1;
           READ_UE (reader, abs_diff_num_minus1);
-        } else if (reordering_of_pic_nums_idc == 3) {
+        } else if (reordering_of_pic_nums_idc == 2) {
           guint32 long_term_pic_num;
 
           READ_UE (reader, long_term_pic_num);
@@ -772,32 +787,39 @@ gst_h264_slice_parse_dec_ref_pic_marking (GstH264Slice * slice,
     READ_UINT8 (reader, m->adaptive_ref_pic_marking_mode_flag, 1);
     if (m->adaptive_ref_pic_marking_mode_flag) {
       guint8 memory_management_control_operation;
-      guint i = 0;
 
-      do {
+      m->n_ref_pic_marking = 0;
+      while (1) {
         READ_UE_ALLOWED (reader, memory_management_control_operation, 0, 6);
-        m->ref_pic_marking[i].memory_management_control_operation =
+        if (memory_management_control_operation == 0)
+          break;
+
+        m->ref_pic_marking[m->n_ref_pic_marking].
+            memory_management_control_operation =
             memory_management_control_operation;
 
         if (memory_management_control_operation == 1 ||
             memory_management_control_operation == 3)
-          READ_UE (reader, m->ref_pic_marking[i].difference_of_pic_nums_minus1);
+          READ_UE (reader,
+              m->ref_pic_marking[m->n_ref_pic_marking].
+              difference_of_pic_nums_minus1);
 
         if (memory_management_control_operation == 2)
-          READ_UE (reader, m->ref_pic_marking[i].long_term_pic_num);
+          READ_UE (reader,
+              m->ref_pic_marking[m->n_ref_pic_marking].long_term_pic_num);
 
         if (memory_management_control_operation == 3 ||
             memory_management_control_operation == 6)
-          READ_UE (reader, m->ref_pic_marking[i].long_term_frame_idx);
+          READ_UE (reader,
+              m->ref_pic_marking[m->n_ref_pic_marking].long_term_frame_idx);
 
         if (memory_management_control_operation == 4)
-          READ_UE (reader, m->ref_pic_marking[i].max_long_term_frame_idx_plus1);
+          READ_UE (reader,
+              m->ref_pic_marking[m->n_ref_pic_marking].
+              max_long_term_frame_idx_plus1);
 
-        i++;
+        m->n_ref_pic_marking++;
       }
-      while (memory_management_control_operation != 0);
-
-      m->n_ref_pic_marking = i;
     }
   }
 
diff --git a/sys/vdpau/h264/gsth264parser.h b/sys/vdpau/h264/gsth264parser.h
index 2c18c0532..1c8e6c204 100644
--- a/sys/vdpau/h264/gsth264parser.h
+++ b/sys/vdpau/h264/gsth264parser.h
@@ -71,6 +71,7 @@ typedef struct _GstH264Sequence GstH264Sequence;
 typedef struct _GstH264Picture GstH264Picture;
 
 typedef struct _GstH264DecRefPicMarking GstH264DecRefPicMarking;
+typedef struct _GstH264RefPicMarking GstH264RefPicMarking;
 typedef struct _GstH264PredWeightTable GstH264PredWeightTable;
 typedef struct _GstH264Slice GstH264Slice;
 
@@ -239,33 +240,30 @@ struct _GstH264Picture
 
   guint8 transform_8x8_mode_flag;
 
-  guint8 scaling_matrix_present_flag;
-  /* if scaling_matrix_present_flag == 1 */
   guint8 scaling_lists_4x4[6][16];
   guint8 scaling_lists_8x8[6][64];
 
   guint8 second_chroma_qp_index_offset;
 };
 
+struct _GstH264RefPicMarking
+{
+  guint8 memory_management_control_operation;
+
+  guint32 difference_of_pic_nums_minus1;
+  guint32 long_term_pic_num;
+  guint32 long_term_frame_idx;
+  guint32 max_long_term_frame_idx_plus1;
+};
+
 struct _GstH264DecRefPicMarking
 {
   /* if slice->nal_unit.IdrPicFlag */
   guint8 no_output_of_prior_pics_flag;
   guint8 long_term_reference_flag;
 
-  /* else */
   guint8 adaptive_ref_pic_marking_mode_flag;
-
-  struct {
-    guint8 memory_management_control_operation;
-
-    union {
-      guint32 difference_of_pic_nums_minus1;
-      guint32 long_term_pic_num;
-      guint32 long_term_frame_idx;
-      guint32 max_long_term_frame_idx_plus1;
-    };
-  } ref_pic_marking[10];
+  GstH264RefPicMarking ref_pic_marking[10]; 
   guint8 n_ref_pic_marking;
 };
 
diff --git a/sys/vdpau/h264/gstvdph264dec.c b/sys/vdpau/h264/gstvdph264dec.c
index 93071fb21..0a1adb1b3 100644
--- a/sys/vdpau/h264/gstvdph264dec.c
+++ b/sys/vdpau/h264/gstvdph264dec.c
@@ -236,20 +236,33 @@ gst_vdp_h264_dec_init_frame_info (GstVdpH264Dec * h264_dec,
 
   h264_frame->output_needed = TRUE;
   h264_frame->is_long_term = FALSE;
-  h264_frame->frame_num = slice->frame_num;
+  h264_frame->frame_idx = slice->frame_num;
 
   /* is reference */
   if (slice->nal_unit.ref_idc == 0)
     h264_frame->is_reference = FALSE;
   else if (slice->nal_unit.IdrPicFlag) {
     h264_frame->is_reference = TRUE;
-    h264_frame->is_long_term =
-        slice->dec_ref_pic_marking.long_term_reference_flag;
+    if (slice->dec_ref_pic_marking.long_term_reference_flag) {
+      h264_frame->is_long_term = TRUE;
+      h264_frame->frame_idx = 0;
+    }
   } else {
-    if (slice->dec_ref_pic_marking.adaptive_ref_pic_marking_mode_flag)
-      GST_ERROR ("FIXME: implement adaptive ref pic marking");
-    else
-      h264_frame->is_reference = TRUE;
+    h264_frame->is_reference = TRUE;
+
+    if (slice->dec_ref_pic_marking.adaptive_ref_pic_marking_mode_flag) {
+      GstH264RefPicMarking *marking;
+      guint i;
+
+      marking = slice->dec_ref_pic_marking.ref_pic_marking;
+      for (i = 0; i < slice->dec_ref_pic_marking.n_ref_pic_marking; i++) {
+        if (marking[i].memory_management_control_operation == 6) {
+          h264_frame->is_long_term = TRUE;
+          h264_frame->frame_idx = marking[i].long_term_frame_idx;
+          break;
+        }
+      }
+    }
   }
 
 }
@@ -269,6 +282,10 @@ gst_vdp_h264_dec_idr (GstVdpH264Dec * h264_dec, GstVdpH264Frame * h264_frame)
   else
     gst_h264_dpb_flush (h264_dec->dpb, TRUE);
 
+  if (slice->dec_ref_pic_marking.long_term_reference_flag)
+    g_object_set (h264_dec->dpb, "max-longterm-frame-idx", 0, NULL);
+  else
+    g_object_set (h264_dec->dpb, "max-longterm-frame-idx", -1, NULL);
 
   seq = slice->picture->sequence;
   if (seq != h264_dec->sequence) {
@@ -356,6 +373,8 @@ gst_vdp_h264_dec_fill_info (GstVdpH264Dec * h264_dec,
 
   info.field_pic_flag = slice->field_pic_flag;
   info.bottom_field_flag = slice->bottom_field_flag;
+  info.num_ref_idx_l0_active_minus1 = slice->num_ref_idx_l0_active_minus1;
+  info.num_ref_idx_l1_active_minus1 = slice->num_ref_idx_l1_active_minus1;
 
   info.num_ref_frames = seq->num_ref_frames;
   info.frame_mbs_only_flag = seq->frame_mbs_only_flag;
@@ -375,21 +394,14 @@ gst_vdp_h264_dec_fill_info (GstVdpH264Dec * h264_dec,
   info.chroma_qp_index_offset = pic->chroma_qp_index_offset;
   info.second_chroma_qp_index_offset = pic->second_chroma_qp_index_offset;
   info.pic_init_qp_minus26 = pic->pic_init_qp_minus26;
-  info.num_ref_idx_l0_active_minus1 = pic->num_ref_idx_l0_active_minus1;
-  info.num_ref_idx_l1_active_minus1 = pic->num_ref_idx_l1_active_minus1;
   info.entropy_coding_mode_flag = pic->entropy_coding_mode_flag;
   info.pic_order_present_flag = pic->pic_order_present_flag;
   info.deblocking_filter_control_present_flag =
       pic->deblocking_filter_control_present_flag;
   info.redundant_pic_cnt_present_flag = pic->redundant_pic_cnt_present_flag;
 
-  if (pic->scaling_matrix_present_flag) {
-    memcpy (&info.scaling_lists_4x4, &pic->scaling_lists_4x4, 96);
-    memcpy (&info.scaling_lists_8x8, &pic->scaling_lists_8x8, 128);
-  } else {
-    memcpy (&info.scaling_lists_4x4, &seq->scaling_lists_4x4, 96);
-    memcpy (&info.scaling_lists_8x8, &seq->scaling_lists_8x8, 128);
-  }
+  memcpy (&info.scaling_lists_4x4, &pic->scaling_lists_4x4, 96);
+  memcpy (&info.scaling_lists_8x8, &pic->scaling_lists_8x8, 128);
 
   gst_h264_dpb_fill_reference_frames (h264_dec->dpb, info.referenceFrames);
 
@@ -490,6 +502,8 @@ gst_vdp_h264_dec_handle_frame (GstBaseVideoDecoder * base_video_decoder,
     return GST_FLOW_OK;
   }
 
+
+
   gst_vdp_h264_dec_init_frame_info (h264_dec, h264_frame);
 
 
@@ -515,14 +529,66 @@ gst_vdp_h264_dec_handle_frame (GstBaseVideoDecoder * base_video_decoder,
   frame->src_buffer = GST_BUFFER_CAST (outbuf);
 
 
-
   /* DPB handling */
   if (slice->nal_unit.ref_idc != 0 && !slice->nal_unit.IdrPicFlag) {
-    if (slice->dec_ref_pic_marking.adaptive_ref_pic_marking_mode_flag)
-      GST_ERROR ("FIXME: implement adaptive ref pic marking");
-    else
+    if (slice->dec_ref_pic_marking.adaptive_ref_pic_marking_mode_flag) {
+      GstH264RefPicMarking *marking;
+      guint i;
+
+      marking = slice->dec_ref_pic_marking.ref_pic_marking;
+      for (i = 0; i < slice->dec_ref_pic_marking.n_ref_pic_marking; i++) {
+
+        switch (marking[i].memory_management_control_operation) {
+          case 1:
+          {
+            guint16 pic_num;
+
+            pic_num = slice->frame_num -
+                (marking[i].difference_of_pic_nums_minus1 + 1);
+            gst_h264_dpb_mark_short_term_unused (h264_dec->dpb, pic_num);
+            break;
+          }
+
+          case 2:
+          {
+            gst_h264_dpb_mark_long_term_unused (h264_dec->dpb,
+                marking[i].long_term_pic_num);
+            break;
+          }
+
+          case 3:
+          {
+            guint16 pic_num;
+
+            pic_num = slice->frame_num -
+                (marking[i].difference_of_pic_nums_minus1 + 1);
+            gst_h264_dpb_mark_long_term (h264_dec->dpb, pic_num,
+                marking[i].long_term_frame_idx);
+            break;
+          }
+
+          case 4:
+          {
+            g_object_set (h264_dec->dpb, "max-longterm-frame-idx",
+                marking[i].max_long_term_frame_idx_plus1 - 1, NULL);
+            break;
+          }
+
+          case 5:
+          {
+            gst_h264_dpb_mark_all_unused (h264_dec->dpb);
+            g_object_set (h264_dec->dpb, "max-longterm-frame-idx", -1, NULL);
+            break;
+          }
+
+          default:
+            break;
+        }
+      }
+    } else
       gst_h264_dpb_mark_sliding (h264_dec->dpb);
   }
+
   gst_h264_dpb_add (h264_dec->dpb, h264_frame);
 
   return GST_FLOW_OK;
diff --git a/sys/vdpau/h264/gstvdph264frame.h b/sys/vdpau/h264/gstvdph264frame.h
index adbb1501e..2b8666894 100644
--- a/sys/vdpau/h264/gstvdph264frame.h
+++ b/sys/vdpau/h264/gstvdph264frame.h
@@ -45,7 +45,7 @@ struct _GstVdpH264Frame
   GPtrArray *slices;
 
   guint poc;
-  guint16 frame_num;
+  guint16 frame_idx;
   gboolean is_reference;
   gboolean is_long_term;
   gboolean output_needed;