]> git.sesse.net Git - ffmpeg/blobdiff - libavcodec/vp3.c
make state transition tables global as they are constant and the code is slightly...
[ffmpeg] / libavcodec / vp3.c
index b22af02b495d77cf84f3937e12ce641bdace39b2..b7c2dcd5f3038f01eb4e8b5106ba40230a8fc858 100644 (file)
@@ -1,28 +1,33 @@
 /*
+ * Copyright (C) 2003-2004 the ffmpeg project
  *
- * Copyright (C) 2003 the ffmpeg project
+ * This file is part of FFmpeg.
  *
- * This library is free software; you can redistribute it and/or
+ * FFmpeg 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 of the License, or (at your option) any later version.
+ * version 2.1 of the License, or (at your option) any later version.
  *
- * This library is distributed in the hope that it will be useful,
+ * FFmpeg 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 this library; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
- *
- * VP3 Video Decoder by Mike Melanson (melanson@pcisys.net)
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  *
  */
 
 /**
  * @file vp3.c
  * On2 VP3 Video Decoder
+ *
+ * VP3 Video Decoder by Mike Melanson (mike at multimedia.cx)
+ * For more information about the VP3 coding process, visit:
+ *   http://multimedia.cx/
+ *
+ * Theora decoder by Alex Beregszaszi
  */
 
 #include <stdio.h>
 #include "avcodec.h"
 #include "dsputil.h"
 #include "mpegvideo.h"
-#include "dsputil.h"
-#include "bswap.h"
 
 #include "vp3data.h"
 
 #define FRAGMENT_PIXELS 8
 
-/* 
+/*
  * Debugging Variables
- * 
+ *
  * Define one or more of the following compile-time variables to 1 to obtain
  * elaborate information about certain aspects of the decoding process.
  *
+ * KEYFRAMES_ONLY: set this to 1 to only see keyframes (VP3 slideshow mode)
  * DEBUG_VP3: high-level decoding flow
  * DEBUG_INIT: initialization parameters
  * DEBUG_DEQUANTIZERS: display how the dequanization tables are built
@@ -59,6 +63,8 @@
  * DEBUG_IDCT: show every detail of the IDCT process
  */
 
+#define KEYFRAMES_ONLY 0
+
 #define DEBUG_VP3 0
 #define DEBUG_INIT 0
 #define DEBUG_DEQUANTIZERS 0
 #define DEBUG_IDCT 0
 
 #if DEBUG_VP3
-#define debug_vp3 printf
+#define debug_vp3(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
 #else
 static inline void debug_vp3(const char *format, ...) { }
 #endif
 
 #if DEBUG_INIT
-#define debug_init printf
+#define debug_init(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
 #else
 static inline void debug_init(const char *format, ...) { }
 #endif
 
 #if DEBUG_DEQUANTIZERS
-#define debug_dequantizers printf 
+#define debug_dequantizers(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
 #else
-static inline void debug_dequantizers(const char *format, ...) { } 
+static inline void debug_dequantizers(const char *format, ...) { }
 #endif
 
 #if DEBUG_BLOCK_CODING
-#define debug_block_coding printf 
+#define debug_block_coding(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
 #else
-static inline void debug_block_coding(const char *format, ...) { } 
+static inline void debug_block_coding(const char *format, ...) { }
 #endif
 
 #if DEBUG_MODES
-#define debug_modes printf 
+#define debug_modes(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
 #else
-static inline void debug_modes(const char *format, ...) { } 
+static inline void debug_modes(const char *format, ...) { }
 #endif
 
 #if DEBUG_VECTORS
-#define debug_vectors printf 
+#define debug_vectors(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
 #else
-static inline void debug_vectors(const char *format, ...) { } 
+static inline void debug_vectors(const char *format, ...) { }
 #endif
 
-#if DEBUG_TOKEN 
-#define debug_token printf 
+#if DEBUG_TOKEN
+#define debug_token(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
 #else
-static inline void debug_token(const char *format, ...) { } 
+static inline void debug_token(const char *format, ...) { }
 #endif
 
 #if DEBUG_VLC
-#define debug_vlc printf 
+#define debug_vlc(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
 #else
-static inline void debug_vlc(const char *format, ...) { } 
+static inline void debug_vlc(const char *format, ...) { }
 #endif
 
 #if DEBUG_DC_PRED
-#define debug_dc_pred printf 
+#define debug_dc_pred(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
 #else
-static inline void debug_dc_pred(const char *format, ...) { } 
+static inline void debug_dc_pred(const char *format, ...) { }
 #endif
 
 #if DEBUG_IDCT
-#define debug_idct printf 
+#define debug_idct(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
 #else
-static inline void debug_idct(const char *format, ...) { } 
+static inline void debug_idct(const char *format, ...) { }
 #endif
 
+typedef struct Coeff {
+    struct Coeff *next;
+    DCTELEM coeff;
+    uint8_t index;
+} Coeff;
+
+//FIXME split things out into their own arrays
 typedef struct Vp3Fragment {
-    DCTELEM coeffs[64];
-    int coding_method;
-    int coeff_count;
-    int last_coeff;
-    int motion_x;
-    int motion_y;
+    Coeff *next_coeff;
     /* address of first pixel taking into account which plane the fragment
      * lives on as well as the plane stride */
     int first_pixel;
     /* this is the macroblock that the fragment belongs to */
-    int macroblock;
+    uint16_t macroblock;
+    uint8_t coding_method;
+    uint8_t coeff_count;
+    int8_t motion_x;
+    int8_t motion_y;
 } Vp3Fragment;
 
 #define SB_NOT_CODED        0
@@ -168,39 +180,39 @@ static int ModeAlphabet[7][CODING_MODE_COUNT] =
     { 0, 0, 0, 0, 0, 0, 0, 0 },
 
     /* scheme 1: Last motion vector dominates */
-    {    MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,  
+    {    MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,
          MODE_INTER_PLUS_MV,    MODE_INTER_NO_MV,
-         MODE_INTRA,            MODE_USING_GOLDEN,      
+         MODE_INTRA,            MODE_USING_GOLDEN,
          MODE_GOLDEN_MV,        MODE_INTER_FOURMV },
 
     /* scheme 2 */
-    {    MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,  
+    {    MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,
          MODE_INTER_NO_MV,      MODE_INTER_PLUS_MV,
-         MODE_INTRA,            MODE_USING_GOLDEN,      
+         MODE_INTRA,            MODE_USING_GOLDEN,
          MODE_GOLDEN_MV,        MODE_INTER_FOURMV },
 
     /* scheme 3 */
-    {    MODE_INTER_LAST_MV,    MODE_INTER_PLUS_MV,     
+    {    MODE_INTER_LAST_MV,    MODE_INTER_PLUS_MV,
          MODE_INTER_PRIOR_LAST, MODE_INTER_NO_MV,
-         MODE_INTRA,            MODE_USING_GOLDEN,      
+         MODE_INTRA,            MODE_USING_GOLDEN,
          MODE_GOLDEN_MV,        MODE_INTER_FOURMV },
 
     /* scheme 4 */
-    {    MODE_INTER_LAST_MV,    MODE_INTER_PLUS_MV,     
+    {    MODE_INTER_LAST_MV,    MODE_INTER_PLUS_MV,
          MODE_INTER_NO_MV,      MODE_INTER_PRIOR_LAST,
-         MODE_INTRA,            MODE_USING_GOLDEN,      
+         MODE_INTRA,            MODE_USING_GOLDEN,
          MODE_GOLDEN_MV,        MODE_INTER_FOURMV },
 
     /* scheme 5: No motion vector dominates */
-    {    MODE_INTER_NO_MV,      MODE_INTER_LAST_MV,     
+    {    MODE_INTER_NO_MV,      MODE_INTER_LAST_MV,
          MODE_INTER_PRIOR_LAST, MODE_INTER_PLUS_MV,
-         MODE_INTRA,            MODE_USING_GOLDEN,      
+         MODE_INTRA,            MODE_USING_GOLDEN,
          MODE_GOLDEN_MV,        MODE_INTER_FOURMV },
 
     /* scheme 6 */
-    {    MODE_INTER_NO_MV,      MODE_USING_GOLDEN,      
+    {    MODE_INTER_NO_MV,      MODE_USING_GOLDEN,
          MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,
-         MODE_INTER_PLUS_MV,    MODE_INTRA,             
+         MODE_INTER_PLUS_MV,    MODE_INTRA,
          MODE_GOLDEN_MV,        MODE_INTER_FOURMV },
 
 };
@@ -209,19 +221,28 @@ static int ModeAlphabet[7][CODING_MODE_COUNT] =
 
 typedef struct Vp3DecodeContext {
     AVCodecContext *avctx;
+    int theora, theora_tables;
+    int version;
     int width, height;
     AVFrame golden_frame;
     AVFrame last_frame;
     AVFrame current_frame;
     int keyframe;
     DSPContext dsp;
+    int flipped_image;
 
+    int qis[3];
+    int nqis;
     int quality_index;
     int last_quality_index;
 
     int superblock_count;
     int superblock_width;
     int superblock_height;
+    int y_superblock_width;
+    int y_superblock_height;
+    int c_superblock_width;
+    int c_superblock_height;
     int u_superblock_start;
     int v_superblock_start;
     unsigned char *superblock_coding;
@@ -235,8 +256,19 @@ typedef struct Vp3DecodeContext {
     int fragment_height;
 
     Vp3Fragment *all_fragments;
-    int u_fragment_start;
-    int v_fragment_start;
+    Coeff *coeffs;
+    Coeff *next_coeff;
+    int fragment_start[3];
+
+    ScanTable scantable;
+
+    /* tables */
+    uint16_t coded_dc_scale_factor[64];
+    uint32_t coded_ac_scale_factor[64];
+    uint8_t base_matrix[384][64];
+    uint8_t qr_count[2][3];
+    uint8_t qr_size [2][3][64];
+    uint16_t qr_base[2][3][64];
 
     /* this is a list of indices into the all_fragments array indicating
      * which of the fragments are coded */
@@ -250,9 +282,14 @@ typedef struct Vp3DecodeContext {
     VLC ac_vlc_3[16];
     VLC ac_vlc_4[16];
 
-    int16_t intra_y_dequant[64];
-    int16_t intra_c_dequant[64];
-    int16_t inter_dequant[64];
+    VLC superblock_run_length_vlc;
+    VLC fragment_run_length_vlc;
+    VLC mode_code_vlc;
+    VLC motion_vector_vlc;
+
+    /* these arrays need to be on 16-byte boundaries since SSE2 operations
+     * index into them */
+    DECLARE_ALIGNED_16(int16_t, qmat[2][4][64]);        //<qmat[is_inter][plane]
 
     /* This table contains superblock_count * 16 entries. Each set of 16
      * numbers corresponds to the fragment indices 0..15 of the superblock.
@@ -270,12 +307,31 @@ typedef struct Vp3DecodeContext {
      * numbers corresponds to the fragment indices 0..5 which comprise
      * the macroblock (4 Y fragments and 2 C fragments). */
     int *macroblock_fragments;
-    /* This is an array of flags indicating whether a particular 
-     * macroblock is coded. */
-    unsigned char *macroblock_coded;
-
+    /* This is an array that indicates how a particular macroblock
+     * is coded. */
+    unsigned char *macroblock_coding;
+
+    int first_coded_y_fragment;
+    int first_coded_c_fragment;
+    int last_coded_y_fragment;
+    int last_coded_c_fragment;
+
+    uint8_t edge_emu_buffer[9*2048]; //FIXME dynamic alloc
+    int8_t qscale_table[2048]; //FIXME dynamic alloc (width+15)/16
+
+    /* Huffman decode */
+    int hti;
+    unsigned int hbits;
+    int entries;
+    int huff_code_size;
+    uint16_t huffman_table[80][32][2];
+
+    uint32_t filter_limit_values[64];
+    int bounding_values_array[256];
 } Vp3DecodeContext;
 
+static int theora_decode_tables(AVCodecContext *avctx, GetBitContext *gb);
+
 /************************************************************************
  * VP3 specific functions
  ************************************************************************/
@@ -284,12 +340,12 @@ typedef struct Vp3DecodeContext {
  * This function sets up all of the various blocks mappings:
  * superblocks <-> fragments, macroblocks <-> fragments,
  * superblocks <-> macroblocks
+ *
+ * Returns 0 is successful; returns 1 if *anything* went wrong.
  */
-static void init_block_mapping(Vp3DecodeContext *s) 
+static int init_block_mapping(Vp3DecodeContext *s)
 {
     int i, j;
-    signed int hilbert_walk_y[16];
-    signed int hilbert_walk_c[16];
     signed int hilbert_walk_mb[4];
 
     int current_fragment = 0;
@@ -305,7 +361,7 @@ static void init_block_mapping(Vp3DecodeContext *s)
     int c_fragment;
 
     signed char travel_width[16] = {
-         1,  1,  0, -1, 
+         1,  1,  0, -1,
          0,  0,  1,  0,
          1,  0,  1,  0,
          0, -1,  0,  1
@@ -328,41 +384,6 @@ static void init_block_mapping(Vp3DecodeContext *s)
 
     debug_vp3("  vp3: initialize block mapping tables\n");
 
-    /* figure out hilbert pattern per these frame dimensions */
-    hilbert_walk_y[0]  = 1;
-    hilbert_walk_y[1]  = 1;
-    hilbert_walk_y[2]  = s->fragment_width;
-    hilbert_walk_y[3]  = -1;
-    hilbert_walk_y[4]  = s->fragment_width;
-    hilbert_walk_y[5]  = s->fragment_width;
-    hilbert_walk_y[6]  = 1;
-    hilbert_walk_y[7]  = -s->fragment_width;
-    hilbert_walk_y[8]  = 1;
-    hilbert_walk_y[9]  = s->fragment_width;
-    hilbert_walk_y[10]  = 1;
-    hilbert_walk_y[11] = -s->fragment_width;
-    hilbert_walk_y[12] = -s->fragment_width;
-    hilbert_walk_y[13] = -1;
-    hilbert_walk_y[14] = -s->fragment_width;
-    hilbert_walk_y[15] = 1;
-
-    hilbert_walk_c[0]  = 1;
-    hilbert_walk_c[1]  = 1;
-    hilbert_walk_c[2]  = s->fragment_width / 2;
-    hilbert_walk_c[3]  = -1;
-    hilbert_walk_c[4]  = s->fragment_width / 2;
-    hilbert_walk_c[5]  = s->fragment_width / 2;
-    hilbert_walk_c[6]  = 1;
-    hilbert_walk_c[7]  = -s->fragment_width / 2;
-    hilbert_walk_c[8]  = 1;
-    hilbert_walk_c[9]  = s->fragment_width / 2;
-    hilbert_walk_c[10]  = 1;
-    hilbert_walk_c[11] = -s->fragment_width / 2;
-    hilbert_walk_c[12] = -s->fragment_width / 2;
-    hilbert_walk_c[13] = -1;
-    hilbert_walk_c[14] = -s->fragment_width / 2;
-    hilbert_walk_c[15] = 1;
-
     hilbert_walk_mb[0] = 1;
     hilbert_walk_mb[1] = s->macroblock_width;
     hilbert_walk_mb[2] = 1;
@@ -379,10 +400,10 @@ static void init_block_mapping(Vp3DecodeContext *s)
             /* start of Y superblocks */
             right_edge = s->fragment_width;
             bottom_edge = s->fragment_height;
-            current_width = 0;
+            current_width = -1;
             current_height = 0;
-            superblock_row_inc = 3 * s->fragment_width;
-            hilbert = hilbert_walk_y;
+            superblock_row_inc = 3 * s->fragment_width -
+                (s->y_superblock_width * 4 - s->fragment_width);
 
             /* the first operation for this variable is to advance by 1 */
             current_fragment = -1;
@@ -392,32 +413,32 @@ static void init_block_mapping(Vp3DecodeContext *s)
             /* start of U superblocks */
             right_edge = s->fragment_width / 2;
             bottom_edge = s->fragment_height / 2;
-            current_width = 0;
+            current_width = -1;
             current_height = 0;
-            superblock_row_inc = 3 * (s->fragment_width / 2);
-            hilbert = hilbert_walk_c;
+            superblock_row_inc = 3 * (s->fragment_width / 2) -
+                (s->c_superblock_width * 4 - s->fragment_width / 2);
 
             /* the first operation for this variable is to advance by 1 */
-            current_fragment = s->u_fragment_start - 1;
+            current_fragment = s->fragment_start[1] - 1;
 
         } else if (i == s->v_superblock_start) {
 
             /* start of V superblocks */
             right_edge = s->fragment_width / 2;
             bottom_edge = s->fragment_height / 2;
-            current_width = 0;
+            current_width = -1;
             current_height = 0;
-            superblock_row_inc = 3 * (s->fragment_width / 2);
-            hilbert = hilbert_walk_c;
+            superblock_row_inc = 3 * (s->fragment_width / 2) -
+                (s->c_superblock_width * 4 - s->fragment_width / 2);
 
             /* the first operation for this variable is to advance by 1 */
-            current_fragment = s->v_fragment_start - 1;
+            current_fragment = s->fragment_start[2] - 1;
 
         }
 
-        if (current_width >= right_edge) {
+        if (current_width >= right_edge - 1) {
             /* reset width and move to next superblock row */
-            current_width = 0;
+            current_width = -1;
             current_height += 4;
 
             /* fragment is now at the start of a new superblock row */
@@ -426,22 +447,24 @@ static void init_block_mapping(Vp3DecodeContext *s)
 
         /* iterate through all 16 fragments in a superblock */
         for (j = 0; j < 16; j++) {
-            current_fragment += hilbert[j];
+            current_fragment += travel_width[j] + right_edge * travel_height[j];
+            current_width += travel_width[j];
             current_height += travel_height[j];
 
             /* check if the fragment is in bounds */
-            if ((current_width <= right_edge) &&
+            if ((current_width < right_edge) &&
                 (current_height < bottom_edge)) {
                 s->superblock_fragments[mapping_index] = current_fragment;
-                debug_init("    mapping fragment %d to superblock %d, position %d\n", 
-                    s->superblock_fragments[mapping_index], i, j);
+                debug_init("    mapping fragment %d to superblock %d, position %d (%d/%d x %d/%d)\n",
+                    s->superblock_fragments[mapping_index], i, j,
+                    current_width, right_edge, current_height, bottom_edge);
             } else {
                 s->superblock_fragments[mapping_index] = -1;
-                debug_init("    superblock %d, position %d has no fragment\n", 
-                    i, j);
+                debug_init("    superblock %d, position %d has no fragment (%d/%d x %d/%d)\n",
+                    i, j,
+                    current_width, right_edge, current_height, bottom_edge);
             }
 
-            current_width += travel_width[j];
             mapping_index++;
         }
     }
@@ -450,17 +473,18 @@ static void init_block_mapping(Vp3DecodeContext *s)
      * all of the Y plane superblocks to build this mapping */
     right_edge = s->macroblock_width;
     bottom_edge = s->macroblock_height;
-    current_width = 0;
+    current_width = -1;
     current_height = 0;
-    superblock_row_inc = s->macroblock_width;
+    superblock_row_inc = s->macroblock_width -
+        (s->y_superblock_width * 2 - s->macroblock_width);;
     hilbert = hilbert_walk_mb;
     mapping_index = 0;
     current_macroblock = -1;
     for (i = 0; i < s->u_superblock_start; i++) {
 
-        if (current_width >= right_edge) {
+        if (current_width >= right_edge - 1) {
             /* reset width and move to next superblock row */
-            current_width = 0;
+            current_width = -1;
             current_height += 2;
 
             /* macroblock is now at the start of a new superblock row */
@@ -470,21 +494,23 @@ static void init_block_mapping(Vp3DecodeContext *s)
         /* iterate through each potential macroblock in the superblock */
         for (j = 0; j < 4; j++) {
             current_macroblock += hilbert_walk_mb[j];
+            current_width += travel_width_mb[j];
             current_height += travel_height_mb[j];
 
             /* check if the macroblock is in bounds */
-            if ((current_width <= right_edge) &&
+            if ((current_width < right_edge) &&
                 (current_height < bottom_edge)) {
                 s->superblock_macroblocks[mapping_index] = current_macroblock;
-                debug_init("    mapping macroblock %d to superblock %d, position %d\n",
-                    s->superblock_macroblocks[mapping_index], i, j);
+                debug_init("    mapping macroblock %d to superblock %d, position %d (%d/%d x %d/%d)\n",
+                    s->superblock_macroblocks[mapping_index], i, j,
+                    current_width, right_edge, current_height, bottom_edge);
             } else {
                 s->superblock_macroblocks[mapping_index] = -1;
-                debug_init("    superblock %d, position %d has no macroblock\n",
-                    i, j);
+                debug_init("    superblock %d, position %d has no macroblock (%d/%d x %d/%d)\n",
+                    i, j,
+                    current_width, right_edge, current_height, bottom_edge);
             }
 
-            current_width += travel_width_mb[j];
             mapping_index++;
         }
     }
@@ -510,33 +536,33 @@ static void init_block_mapping(Vp3DecodeContext *s)
                 s->macroblock_fragments[mapping_index++] = -1;
 
             if (i + 1 < s->fragment_height) {
-                s->all_fragments[current_fragment + s->fragment_width].macroblock = 
+                s->all_fragments[current_fragment + s->fragment_width].macroblock =
                     current_macroblock;
-                s->macroblock_fragments[mapping_index++] = 
+                s->macroblock_fragments[mapping_index++] =
                     current_fragment + s->fragment_width;
                 debug_init("%d ", current_fragment + s->fragment_width);
             } else
                 s->macroblock_fragments[mapping_index++] = -1;
 
             if ((j + 1 < s->fragment_width) && (i + 1 < s->fragment_height)) {
-                s->all_fragments[current_fragment + s->fragment_width + 1].macroblock = 
+                s->all_fragments[current_fragment + s->fragment_width + 1].macroblock =
                     current_macroblock;
-                s->macroblock_fragments[mapping_index++] = 
+                s->macroblock_fragments[mapping_index++] =
                     current_fragment + s->fragment_width + 1;
                 debug_init("%d ", current_fragment + s->fragment_width + 1);
             } else
                 s->macroblock_fragments[mapping_index++] = -1;
 
             /* C planes */
-            c_fragment = s->u_fragment_start + 
+            c_fragment = s->fragment_start[1] +
                 (i * s->fragment_width / 4) + (j / 2);
-        s->all_fragments[c_fragment].macroblock = s->macroblock_count;
+            s->all_fragments[c_fragment].macroblock = s->macroblock_count;
             s->macroblock_fragments[mapping_index++] = c_fragment;
             debug_init("%d ", c_fragment);
 
-            c_fragment = s->v_fragment_start + 
+            c_fragment = s->fragment_start[2] +
                 (i * s->fragment_width / 4) + (j / 2);
-        s->all_fragments[c_fragment].macroblock = s->macroblock_count;
+            s->all_fragments[c_fragment].macroblock = s->macroblock_count;
             s->macroblock_fragments[mapping_index++] = c_fragment;
             debug_init("%d ", c_fragment);
 
@@ -544,237 +570,15 @@ static void init_block_mapping(Vp3DecodeContext *s)
 
             if (j + 2 <= s->fragment_width)
                 current_fragment += 2;
-            else 
+            else
                 current_fragment++;
             current_macroblock++;
         }
 
         current_fragment += s->fragment_width;
     }
-}
-
-/*
- * This function unpacks a single token (which should be in the range 0..31)
- * and returns a zero run (number of zero coefficients in current DCT matrix
- * before next non-zero coefficient), the next DCT coefficient, and the
- * number of consecutive, non-EOB'd DCT blocks to EOB.
- */
-static void unpack_token(GetBitContext *gb, int token, int *zero_run,
-                         DCTELEM *coeff, int *eob_run) 
-{
-    int sign;
-
-    *zero_run = 0;
-    *eob_run = 0;
-    *coeff = 0;
-
-    debug_token("    vp3 token %d: ", token);
-    switch (token) {
-
-    case 0:
-        debug_token("DCT_EOB_TOKEN, EOB next block\n");
-        *eob_run = 1;
-        break;
-
-    case 1:
-        debug_token("DCT_EOB_PAIR_TOKEN, EOB next 2 blocks\n");
-        *eob_run = 2;
-        break;
-
-    case 2:
-        debug_token("DCT_EOB_TRIPLE_TOKEN, EOB next 3 blocks\n");
-        *eob_run = 3;
-        break;
-
-    case 3:
-        debug_token("DCT_REPEAT_RUN_TOKEN, ");
-        *eob_run = get_bits(gb, 2) + 4;
-        debug_token("EOB the next %d blocks\n", *eob_run);
-        break;
-
-    case 4:
-        debug_token("DCT_REPEAT_RUN2_TOKEN, ");
-        *eob_run = get_bits(gb, 3) + 8;
-        debug_token("EOB the next %d blocks\n", *eob_run);
-        break;
-
-    case 5:
-        debug_token("DCT_REPEAT_RUN3_TOKEN, ");
-        *eob_run = get_bits(gb, 4) + 16;
-        debug_token("EOB the next %d blocks\n", *eob_run);
-        break;
-
-    case 6:
-        debug_token("DCT_REPEAT_RUN4_TOKEN, ");
-        *eob_run = get_bits(gb, 12);
-        debug_token("EOB the next %d blocks\n", *eob_run);
-        break;
-
-    case 7:
-        debug_token("DCT_SHORT_ZRL_TOKEN, ");
-        /* note that this token actually indicates that (3 extra bits) + 1 0s
-         * should be output; this case specifies a run of (3 EBs) 0s and a
-         * coefficient of 0. */
-        *zero_run = get_bits(gb, 3);
-        *coeff = 0;
-        debug_token("skip the next %d positions in output matrix\n", *zero_run + 1);
-        break;
-
-    case 8:
-        debug_token("DCT_ZRL_TOKEN, ");
-        /* note that this token actually indicates that (6 extra bits) + 1 0s
-         * should be output; this case specifies a run of (6 EBs) 0s and a
-         * coefficient of 0. */
-        *zero_run = get_bits(gb, 6);
-        *coeff = 0;
-        debug_token("skip the next %d positions in output matrix\n", *zero_run + 1);
-        break;
-
-    case 9:
-        debug_token("ONE_TOKEN, output 1\n");
-        *coeff = 1;
-        break;
-
-    case 10:
-        debug_token("MINUS_ONE_TOKEN, output -1\n");
-        *coeff = -1;
-        break;
-
-    case 11:
-        debug_token("TWO_TOKEN, output 2\n");
-        *coeff = 2;
-        break;
-
-    case 12:
-        debug_token("MINUS_TWO_TOKEN, output -2\n");
-        *coeff = -2;
-        break;
-
-    case 13:
-    case 14:
-    case 15:
-    case 16:
-        debug_token("LOW_VAL_TOKENS, ");
-        if (get_bits(gb, 1))
-            *coeff = -(3 + (token - 13));
-        else
-            *coeff = 3 + (token - 13);
-        debug_token("output %d\n", *coeff);
-        break;
-
-    case 17:
-        debug_token("DCT_VAL_CATEGORY3, ");
-        sign = get_bits(gb, 1);
-        *coeff = 7 + get_bits(gb, 1);
-        if (sign)
-            *coeff = -(*coeff);
-        debug_token("output %d\n", *coeff);
-        break;
-
-    case 18:
-        debug_token("DCT_VAL_CATEGORY4, ");
-        sign = get_bits(gb, 1);
-        *coeff = 9 + get_bits(gb, 2);
-        if (sign)
-            *coeff = -(*coeff);
-        debug_token("output %d\n", *coeff);
-        break;
-
-    case 19:
-        debug_token("DCT_VAL_CATEGORY5, ");
-        sign = get_bits(gb, 1);
-        *coeff = 13 + get_bits(gb, 3);
-        if (sign)
-            *coeff = -(*coeff);
-        debug_token("output %d\n", *coeff);
-        break;
-
-    case 20:
-        debug_token("DCT_VAL_CATEGORY6, ");
-        sign = get_bits(gb, 1);
-        *coeff = 21 + get_bits(gb, 4);
-        if (sign)
-            *coeff = -(*coeff);
-        debug_token("output %d\n", *coeff);
-        break;
-
-    case 21:
-        debug_token("DCT_VAL_CATEGORY7, ");
-        sign = get_bits(gb, 1);
-        *coeff = 37 + get_bits(gb, 5);
-        if (sign)
-            *coeff = -(*coeff);
-        debug_token("output %d\n", *coeff);
-        break;
-
-    case 22:
-        debug_token("DCT_VAL_CATEGORY8, ");
-        sign = get_bits(gb, 1);
-        *coeff = 69 + get_bits(gb, 9);
-        if (sign)
-            *coeff = -(*coeff);
-        debug_token("output %d\n", *coeff);
-        break;
-
-    case 23:
-    case 24:
-    case 25:
-    case 26:
-    case 27:
-        debug_token("DCT_RUN_CATEGORY1, ");
-        *zero_run = token - 22;
-        if (get_bits(gb, 1))
-            *coeff = -1;
-        else
-            *coeff = 1;
-        debug_token("output %d 0s, then %d\n", *zero_run, *coeff);
-        break;
-
-    case 28:
-        debug_token("DCT_RUN_CATEGORY1B, ");
-        if (get_bits(gb, 1))
-            *coeff = -1;
-        else
-            *coeff = 1;
-        *zero_run = 6 + get_bits(gb, 2);
-        debug_token("output %d 0s, then %d\n", *zero_run, *coeff);
-        break;
-
-    case 29:
-        debug_token("DCT_RUN_CATEGORY1C, ");
-        if (get_bits(gb, 1))
-            *coeff = -1;
-        else
-            *coeff = 1;
-        *zero_run = 10 + get_bits(gb, 3);
-        debug_token("output %d 0s, then %d\n", *zero_run, *coeff);
-        break;
-
-    case 30:
-        debug_token("DCT_RUN_CATEGORY2, ");
-        sign = get_bits(gb, 1);
-        *coeff = 2 + get_bits(gb, 1);
-        if (sign)
-            *coeff = -(*coeff);
-        *zero_run = 1;
-        debug_token("output %d 0s, then %d\n", *zero_run, *coeff);
-        break;
-
-    case 31:
-        debug_token("DCT_RUN_CATEGORY2, ");
-        sign = get_bits(gb, 1);
-        *coeff = 2 + get_bits(gb, 1);
-        if (sign)
-            *coeff = -(*coeff);
-        *zero_run = 2 + get_bits(gb, 1);
-        debug_token("output %d 0s, then %d\n", *zero_run, *coeff);
-        break;
-
-    default:
-        printf ("  vp3: help! Got a bad token: %d > 31\n", token);
-        break;
 
-  }
+    return 0;  /* successful path out */
 }
 
 /*
@@ -787,324 +591,90 @@ static void init_frame(Vp3DecodeContext *s, GetBitContext *gb)
     /* zero out all of the fragment information */
     s->coded_fragment_list_index = 0;
     for (i = 0; i < s->fragment_count; i++) {
-        memset(s->all_fragments[i].coeffs, 0, 64 * sizeof(DCTELEM));
         s->all_fragments[i].coeff_count = 0;
-        s->all_fragments[i].last_coeff = 0;
+        s->all_fragments[i].motion_x = 127;
+        s->all_fragments[i].motion_y = 127;
+        s->all_fragments[i].next_coeff= NULL;
+        s->coeffs[i].index=
+        s->coeffs[i].coeff=0;
+        s->coeffs[i].next= NULL;
     }
 }
 
 /*
- * This function sets of the dequantization tables used for a particular
+ * This function sets up the dequantization tables used for a particular
  * frame.
  */
 static void init_dequantizer(Vp3DecodeContext *s)
 {
-
-    int quality_scale = vp31_quality_threshold[s->quality_index];
-    int dc_scale_factor = vp31_dc_scale_factor[s->quality_index];
-    int i, j;
+    int ac_scale_factor = s->coded_ac_scale_factor[s->quality_index];
+    int dc_scale_factor = s->coded_dc_scale_factor[s->quality_index];
+    int i, plane, inter, qri, bmi, bmj, qistart;
 
     debug_vp3("  vp3: initializing dequantization tables\n");
 
-    /* 
-     * Scale dequantizers:
-     *
-     *   quantizer * sf
-     *   --------------
-     *        100
-     *
-     * where sf = dc_scale_factor for DC quantizer
-     *           or quality_scale for AC quantizer
-     *
-     * Then, saturate the result to a lower limit of MIN_DEQUANT_VAL.
-     */
-#define SCALER 1
-
-    /* scale DC quantizers */
-    s->intra_y_dequant[0] = vp31_intra_y_dequant[0] * dc_scale_factor / 100;
-    if (s->intra_y_dequant[0] < MIN_DEQUANT_VAL * 2)
-        s->intra_y_dequant[0] = MIN_DEQUANT_VAL * 2;
-    s->intra_y_dequant[0] *= SCALER;
-
-    s->intra_c_dequant[0] = vp31_intra_c_dequant[0] * dc_scale_factor / 100;
-    if (s->intra_c_dequant[0] < MIN_DEQUANT_VAL * 2)
-        s->intra_c_dequant[0] = MIN_DEQUANT_VAL * 2;
-    s->intra_c_dequant[0] *= SCALER;
-
-    s->inter_dequant[0] = vp31_inter_dequant[0] * dc_scale_factor / 100;
-    if (s->inter_dequant[0] < MIN_DEQUANT_VAL * 4)
-        s->inter_dequant[0] = MIN_DEQUANT_VAL * 4;
-    s->inter_dequant[0] *= SCALER;
-
-    /* scale AC quantizers, zigzag at the same time in preparation for
-     * the dequantization phase */
-    for (i = 1; i < 64; i++) {
-
-        j = quant_index[i];
-
-        s->intra_y_dequant[j] = vp31_intra_y_dequant[i] * quality_scale / 100;
-        if (s->intra_y_dequant[j] < MIN_DEQUANT_VAL)
-            s->intra_y_dequant[j] = MIN_DEQUANT_VAL;
-        s->intra_y_dequant[j] *= SCALER;
-
-        s->intra_c_dequant[j] = vp31_intra_c_dequant[i] * quality_scale / 100;
-        if (s->intra_c_dequant[j] < MIN_DEQUANT_VAL)
-            s->intra_c_dequant[j] = MIN_DEQUANT_VAL;
-        s->intra_c_dequant[j] *= SCALER;
-
-        s->inter_dequant[j] = vp31_inter_dequant[i] * quality_scale / 100;
-        if (s->inter_dequant[j] < MIN_DEQUANT_VAL * 2)
-            s->inter_dequant[j] = MIN_DEQUANT_VAL * 2;
-        s->inter_dequant[j] *= SCALER;
-    }
-
-    /* print debug information as requested */
-    debug_dequantizers("intra Y dequantizers:\n");
-    for (i = 0; i < 8; i++) {
-      for (j = i * 8; j < i * 8 + 8; j++) {
-        debug_dequantizers(" %4d,", s->intra_y_dequant[j]);
-      }
-      debug_dequantizers("\n");
-    }
-    debug_dequantizers("\n");
-
-    debug_dequantizers("intra C dequantizers:\n");
-    for (i = 0; i < 8; i++) {
-      for (j = i * 8; j < i * 8 + 8; j++) {
-        debug_dequantizers(" %4d,", s->intra_c_dequant[j]);
-      }
-      debug_dequantizers("\n");
-    }
-    debug_dequantizers("\n");
-
-    debug_dequantizers("interframe dequantizers:\n");
-    for (i = 0; i < 8; i++) {
-      for (j = i * 8; j < i * 8 + 8; j++) {
-        debug_dequantizers(" %4d,", s->inter_dequant[j]);
-      }
-      debug_dequantizers("\n");
+    for(inter=0; inter<2; inter++){
+        for(plane=0; plane<3; plane++){
+            int sum=0;
+            for(qri=0; qri<s->qr_count[inter][plane]; qri++){
+                sum+= s->qr_size[inter][plane][qri];
+                if(s->quality_index <= sum)
+                    break;
+            }
+            qistart= sum - s->qr_size[inter][plane][qri];
+            bmi= s->qr_base[inter][plane][qri  ];
+            bmj= s->qr_base[inter][plane][qri+1];
+            for(i=0; i<64; i++){
+                int coeff= (  2*(sum    -s->quality_index)*s->base_matrix[bmi][i]
+                            - 2*(qistart-s->quality_index)*s->base_matrix[bmj][i]
+                            + s->qr_size[inter][plane][qri])
+                           / (2*s->qr_size[inter][plane][qri]);
+
+                int qmin= 8<<(inter + !i);
+                int qscale= i ? ac_scale_factor : dc_scale_factor;
+
+                s->qmat[inter][plane][i]= clip((qscale * coeff)/100 * 4, qmin, 4096);
+            }
+        }
     }
-    debug_dequantizers("\n");
-}
-
-/*
- * This function is used to fetch runs of 1s or 0s from the bitstream for
- * use in determining which superblocks are fully and partially coded.
- *
- *  Codeword                RunLength
- *  0                       1
- *  10x                     2-3
- *  110x                    4-5
- *  1110xx                  6-9
- *  11110xxx                10-17
- *  111110xxxx              18-33
- *  111111xxxxxxxxxxxx      34-4129
- */
-static int get_superblock_run_length(GetBitContext *gb)
-{
-
-    if (get_bits(gb, 1) == 0)
-        return 1;
-
-    else if (get_bits(gb, 1) == 0)
-        return (2 + get_bits(gb, 1));
-
-    else if (get_bits(gb, 1) == 0)
-        return (4 + get_bits(gb, 1));
-
-    else if (get_bits(gb, 1) == 0)
-        return (6 + get_bits(gb, 2));
-
-    else if (get_bits(gb, 1) == 0)
-        return (10 + get_bits(gb, 3));
-
-    else if (get_bits(gb, 1) == 0)
-        return (18 + get_bits(gb, 4));
-
-    else
-        return (34 + get_bits(gb, 12));
-
-}
-
-/*
- * This function is used to fetch runs of 1s or 0s from the bitstream for
- * use in determining which particular fragments are coded.
- *
- * Codeword                RunLength
- * 0x                      1-2
- * 10x                     3-4
- * 110x                    5-6
- * 1110xx                  7-10
- * 11110xx                 11-14
- * 11111xxxx               15-30
- */
-static int get_fragment_run_length(GetBitContext *gb)
-{
-
-    if (get_bits(gb, 1) == 0)
-        return (1 + get_bits(gb, 1));
-
-    else if (get_bits(gb, 1) == 0)
-        return (3 + get_bits(gb, 1));
-
-    else if (get_bits(gb, 1) == 0)
-        return (5 + get_bits(gb, 1));
-
-    else if (get_bits(gb, 1) == 0)
-        return (7 + get_bits(gb, 2));
-
-    else if (get_bits(gb, 1) == 0)
-        return (11 + get_bits(gb, 2));
-
-    else
-        return (15 + get_bits(gb, 4));
-
-}
-
-/*
- * This function decodes a VLC from the bitstream and returns a number
- * that ranges from 0..7. The number indicates which of the 8 coding
- * modes to use.
- *
- *  VLC       Number
- *  0            0
- *  10           1
- *  110          2
- *  1110         3
- *  11110        4
- *  111110       5
- *  1111110      6
- *  1111111      7
- *
- */
-static int get_mode_code(GetBitContext *gb)
-{
-
-    if (get_bits(gb, 1) == 0)
-        return 0;
-
-    else if (get_bits(gb, 1) == 0)
-        return 1;
-
-    else if (get_bits(gb, 1) == 0)
-        return 2;
-
-    else if (get_bits(gb, 1) == 0)
-        return 3;
-
-    else if (get_bits(gb, 1) == 0)
-        return 4;
-
-    else if (get_bits(gb, 1) == 0)
-        return 5;
-
-    else if (get_bits(gb, 1) == 0)
-        return 6;
-
-    else
-        return 7;
 
+    memset(s->qscale_table, (FFMAX(s->qmat[0][0][1], s->qmat[0][1][1])+8)/16, 512); //FIXME finetune
 }
 
 /*
- * This function extracts a motion vector from the bitstream using a VLC
- * scheme. 3 bits are fetched from the bitstream and 1 of 8 actions is
- * taken depending on the value on those 3 bits:
- *
- *  0: return 0
- *  1: return 1
- *  2: return -1
- *  3: if (next bit is 1) return -2, else return 2
- *  4: if (next bit is 1) return -3, else return 3
- *  5: return 4 + (next 2 bits), next bit is sign
- *  6: return 8 + (next 3 bits), next bit is sign
- *  7: return 16 + (next 4 bits), next bit is sign
+ * This function initializes the loop filter boundary limits if the frame's
+ * quality index is different from the previous frame's.
  */
-static int get_motion_vector_vlc(GetBitContext *gb)
+static void init_loop_filter(Vp3DecodeContext *s)
 {
-    int bits;
-
-    bits = get_bits(gb, 3);
-
-    switch(bits) {
-
-    case 0:
-        bits = 0;
-        break;
-
-    case 1:
-        bits = 1;
-        break;
-
-    case 2:
-        bits = -1;
-        break;
-
-    case 3:
-        if (get_bits(gb, 1) == 0)
-            bits = 2;
-        else
-            bits = -2;
-        break;
-
-    case 4:
-        if (get_bits(gb, 1) == 0)
-            bits = 3;
-        else
-            bits = -3;
-        break;
-
-    case 5:
-        bits = 4 + get_bits(gb, 2);
-        if (get_bits(gb, 1) == 1)
-            bits = -bits;
-        break;
-
-    case 6:
-        bits = 8 + get_bits(gb, 3);
-        if (get_bits(gb, 1) == 1)
-            bits = -bits;
-        break;
-
-    case 7:
-        bits = 16 + get_bits(gb, 4);
-        if (get_bits(gb, 1) == 1)
-            bits = -bits;
-        break;
-
+    int *bounding_values= s->bounding_values_array+127;
+    int filter_limit;
+    int x;
+
+    filter_limit = s->filter_limit_values[s->quality_index];
+
+    /* set up the bounding values */
+    memset(s->bounding_values_array, 0, 256 * sizeof(int));
+    for (x = 0; x < filter_limit; x++) {
+        bounding_values[-x - filter_limit] = -filter_limit + x;
+        bounding_values[-x] = -x;
+        bounding_values[x] = x;
+        bounding_values[x + filter_limit] = filter_limit - x;
     }
-
-    return bits;
 }
 
 /*
- * This function fetches a 5-bit number from the stream followed by
- * a sign and calls it a motion vector.
- */
-static int get_motion_vector_fixed(GetBitContext *gb)
-{
-
-    int bits;
-
-    bits = get_bits(gb, 5);
-
-    if (get_bits(gb, 1) == 1)
-        bits = -bits;
-
-    return bits;
-}
-
-/*
- * This function unpacks all of the superblock/macroblock/fragment coding 
+ * This function unpacks all of the superblock/macroblock/fragment coding
  * information from the bitstream.
  */
-static void unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
+static int unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
 {
     int bit = 0;
     int current_superblock = 0;
     int current_run = 0;
     int decode_fully_flags = 0;
     int decode_partial_blocks = 0;
+    int first_c_fragment_seen;
 
     int i, j;
     int current_fragment;
@@ -1120,13 +690,16 @@ static void unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
 
         /* unpack the list of partially-coded superblocks */
         bit = get_bits(gb, 1);
-        /* toggle the bit because as soon as the first run length is 
+        /* toggle the bit because as soon as the first run length is
          * fetched the bit will be toggled again */
         bit ^= 1;
         while (current_superblock < s->superblock_count) {
-            if (current_run == 0) {
+            if (current_run-- == 0) {
                 bit ^= 1;
-                current_run = get_superblock_run_length(gb);
+                current_run = get_vlc2(gb,
+                    s->superblock_run_length_vlc.table, 6, 2);
+                if (current_run == 33)
+                    current_run += get_bits(gb, 12);
                 debug_block_coding("      setting superblocks %d..%d to %s\n",
                     current_superblock,
                     current_superblock + current_run - 1,
@@ -1134,18 +707,16 @@ static void unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
 
                 /* if any of the superblocks are not partially coded, flag
                  * a boolean to decode the list of fully-coded superblocks */
-                if (bit == 0)
+                if (bit == 0) {
                     decode_fully_flags = 1;
-            } else {
-
-                /* make a note of the fact that there are partially coded
-                 * superblocks */
-                decode_partial_blocks = 1;
+                } else {
 
+                    /* make a note of the fact that there are partially coded
+                     * superblocks */
+                    decode_partial_blocks = 1;
+                }
             }
-            s->superblock_coding[current_superblock++] = 
-                (bit) ? SB_PARTIALLY_CODED : SB_NOT_CODED;
-            current_run--;
+            s->superblock_coding[current_superblock++] = bit;
         }
 
         /* unpack the list of fully coded superblocks if any of the blocks were
@@ -1155,7 +726,7 @@ static void unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
             current_superblock = 0;
             current_run = 0;
             bit = get_bits(gb, 1);
-            /* toggle the bit because as soon as the first run length is 
+            /* toggle the bit because as soon as the first run length is
              * fetched the bit will be toggled again */
             bit ^= 1;
             while (current_superblock < s->superblock_count) {
@@ -1163,17 +734,18 @@ static void unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
                 /* skip any superblocks already marked as partially coded */
                 if (s->superblock_coding[current_superblock] == SB_NOT_CODED) {
 
-                    if (current_run == 0) {
+                    if (current_run-- == 0) {
                         bit ^= 1;
-                        current_run = get_superblock_run_length(gb);
+                        current_run = get_vlc2(gb,
+                            s->superblock_run_length_vlc.table, 6, 2);
+                        if (current_run == 33)
+                            current_run += get_bits(gb, 12);
                     }
 
                     debug_block_coding("      setting superblock %d to %s\n",
                         current_superblock,
                         (bit) ? "fully coded" : "not coded");
-                    s->superblock_coding[current_superblock] = 
-                        (bit) ? SB_FULLY_CODED : SB_NOT_CODED;
-                    current_run--;
+                    s->superblock_coding[current_superblock] = 2*bit;
                 }
                 current_superblock++;
             }
@@ -1185,7 +757,7 @@ static void unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
 
             current_run = 0;
             bit = get_bits(gb, 1);
-            /* toggle the bit because as soon as the first run length is 
+            /* toggle the bit because as soon as the first run length is
              * fetched the bit will be toggled again */
             bit ^= 1;
         }
@@ -1194,7 +766,11 @@ static void unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
     /* figure out which fragments are coded; iterate through each
      * superblock (all planes) */
     s->coded_fragment_list_index = 0;
-    memset(s->macroblock_coded, 0, s->macroblock_count);
+    s->next_coeff= s->coeffs + s->fragment_count;
+    s->first_coded_y_fragment = s->first_coded_c_fragment = 0;
+    s->last_coded_y_fragment = s->last_coded_c_fragment = -1;
+    first_c_fragment_seen = 0;
+    memset(s->macroblock_coding, MODE_COPY, s->macroblock_count);
     for (i = 0; i < s->superblock_count; i++) {
 
         /* iterate through all 16 fragments in a superblock */
@@ -1202,29 +778,45 @@ static void unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
 
             /* if the fragment is in bounds, check its coding status */
             current_fragment = s->superblock_fragments[i * 16 + j];
+            if (current_fragment >= s->fragment_count) {
+                av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_superblocks(): bad fragment number (%d >= %d)\n",
+                    current_fragment, s->fragment_count);
+                return 1;
+            }
             if (current_fragment != -1) {
                 if (s->superblock_coding[i] == SB_NOT_CODED) {
 
                     /* copy all the fragments from the prior frame */
-                    s->all_fragments[current_fragment].coding_method = 
+                    s->all_fragments[current_fragment].coding_method =
                         MODE_COPY;
 
                 } else if (s->superblock_coding[i] == SB_PARTIALLY_CODED) {
 
                     /* fragment may or may not be coded; this is the case
                      * that cares about the fragment coding runs */
-                    if (current_run == 0) {
+                    if (current_run-- == 0) {
                         bit ^= 1;
-                        current_run = get_fragment_run_length(gb);
+                        current_run = get_vlc2(gb,
+                            s->fragment_run_length_vlc.table, 5, 2);
                     }
 
                     if (bit) {
-                        /* mode will be decoded in the next phase */
-                        s->all_fragments[current_fragment].coding_method = 
+                        /* default mode; actual mode will be decoded in
+                         * the next phase */
+                        s->all_fragments[current_fragment].coding_method =
                             MODE_INTER_NO_MV;
-                        s->coded_fragment_list[s->coded_fragment_list_index++] = 
+                        s->all_fragments[current_fragment].next_coeff= s->coeffs + current_fragment;
+                        s->coded_fragment_list[s->coded_fragment_list_index] =
                             current_fragment;
-                        s->macroblock_coded[s->all_fragments[current_fragment].macroblock] = 1;
+                        if ((current_fragment >= s->fragment_start[1]) &&
+                            (s->last_coded_y_fragment == -1) &&
+                            (!first_c_fragment_seen)) {
+                            s->first_coded_c_fragment = s->coded_fragment_list_index;
+                            s->last_coded_y_fragment = s->first_coded_c_fragment - 1;
+                            first_c_fragment_seen = 1;
+                        }
+                        s->coded_fragment_list_index++;
+                        s->macroblock_coding[s->all_fragments[current_fragment].macroblock] = MODE_INTER_NO_MV;
                         debug_block_coding("      superblock %d is partially coded, fragment %d is coded\n",
                             i, current_fragment);
                     } else {
@@ -1235,30 +827,53 @@ static void unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
                             i, current_fragment);
                     }
 
-                    current_run--;
-
                 } else {
 
                     /* fragments are fully coded in this superblock; actual
                      * coding will be determined in next step */
-                    s->all_fragments[current_fragment].coding_method = 
+                    s->all_fragments[current_fragment].coding_method =
                         MODE_INTER_NO_MV;
-                    s->coded_fragment_list[s->coded_fragment_list_index++] = 
+                    s->all_fragments[current_fragment].next_coeff= s->coeffs + current_fragment;
+                    s->coded_fragment_list[s->coded_fragment_list_index] =
                         current_fragment;
-                    s->macroblock_coded[s->all_fragments[current_fragment].macroblock] = 1;
+                    if ((current_fragment >= s->fragment_start[1]) &&
+                        (s->last_coded_y_fragment == -1) &&
+                        (!first_c_fragment_seen)) {
+                        s->first_coded_c_fragment = s->coded_fragment_list_index;
+                        s->last_coded_y_fragment = s->first_coded_c_fragment - 1;
+                        first_c_fragment_seen = 1;
+                    }
+                    s->coded_fragment_list_index++;
+                    s->macroblock_coding[s->all_fragments[current_fragment].macroblock] = MODE_INTER_NO_MV;
                     debug_block_coding("      superblock %d is fully coded, fragment %d is coded\n",
                         i, current_fragment);
                 }
             }
         }
     }
+
+    if (!first_c_fragment_seen)
+        /* only Y fragments coded in this frame */
+        s->last_coded_y_fragment = s->coded_fragment_list_index - 1;
+    else
+        /* end the list of coded C fragments */
+        s->last_coded_c_fragment = s->coded_fragment_list_index - 1;
+
+    debug_block_coding("    %d total coded fragments, y: %d -> %d, c: %d -> %d\n",
+        s->coded_fragment_list_index,
+        s->first_coded_y_fragment,
+        s->last_coded_y_fragment,
+        s->first_coded_c_fragment,
+        s->last_coded_c_fragment);
+
+    return 0;
 }
 
 /*
  * This function unpacks all the coding mode data for individual macroblocks
  * from the bitstream.
  */
-static void unpack_modes(Vp3DecodeContext *s, GetBitContext *gb)
+static int unpack_modes(Vp3DecodeContext *s, GetBitContext *gb)
 {
     int i, j, k;
     int scheme;
@@ -1284,11 +899,11 @@ static void unpack_modes(Vp3DecodeContext *s, GetBitContext *gb)
         if (scheme == 0) {
             debug_modes("    custom mode alphabet ahead:\n");
             for (i = 0; i < 8; i++)
-                ModeAlphabet[0][i] = get_bits(gb, 3);
+                ModeAlphabet[scheme][get_bits(gb, 3)] = i;
         }
 
         for (i = 0; i < 8; i++)
-            debug_modes("      mode[%d][%d] = %d\n", scheme, i, 
+            debug_modes("      mode[%d][%d] = %d\n", scheme, i,
                 ModeAlphabet[scheme][i]);
 
         /* iterate through all of the macroblocks that contain 1 or more
@@ -1298,19 +913,33 @@ static void unpack_modes(Vp3DecodeContext *s, GetBitContext *gb)
             for (j = 0; j < 4; j++) {
                 current_macroblock = s->superblock_macroblocks[i * 4 + j];
                 if ((current_macroblock == -1) ||
-                    (!s->macroblock_coded[current_macroblock]))
+                    (s->macroblock_coding[current_macroblock] == MODE_COPY))
                     continue;
+                if (current_macroblock >= s->macroblock_count) {
+                    av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_modes(): bad macroblock number (%d >= %d)\n",
+                        current_macroblock, s->macroblock_count);
+                    return 1;
+                }
 
                 /* mode 7 means get 3 bits for each coding mode */
                 if (scheme == 7)
                     coding_mode = get_bits(gb, 3);
                 else
-                    coding_mode = ModeAlphabet[scheme][get_mode_code(gb)];
+                    coding_mode = ModeAlphabet[scheme]
+                        [get_vlc2(gb, s->mode_code_vlc.table, 3, 3)];
 
+                s->macroblock_coding[current_macroblock] = coding_mode;
                 for (k = 0; k < 6; k++) {
-                    current_fragment = 
+                    current_fragment =
                         s->macroblock_fragments[current_macroblock * 6 + k];
-                    if (s->all_fragments[current_fragment].coding_method != 
+                    if (current_fragment == -1)
+                        continue;
+                    if (current_fragment >= s->fragment_count) {
+                        av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_modes(): bad fragment number (%d >= %d)\n",
+                            current_fragment, s->fragment_count);
+                        return 1;
+                    }
+                    if (s->all_fragments[current_fragment].coding_method !=
                         MODE_COPY)
                         s->all_fragments[current_fragment].coding_method =
                             coding_mode;
@@ -1322,13 +951,14 @@ static void unpack_modes(Vp3DecodeContext *s, GetBitContext *gb)
         }
     }
 
+    return 0;
 }
 
 /*
  * This function unpacks all the motion vectors for the individual
  * macroblocks from the bitstream.
  */
-static void unpack_vectors(Vp3DecodeContext *s, GetBitContext *gb)
+static int unpack_vectors(Vp3DecodeContext *s, GetBitContext *gb)
 {
     int i, j, k;
     int coding_mode;
@@ -1342,7 +972,6 @@ static void unpack_vectors(Vp3DecodeContext *s, GetBitContext *gb)
     int current_fragment;
 
     debug_vp3("  vp3: unpacking motion vectors\n");
-
     if (s->keyframe) {
 
         debug_vp3("    keyframe-- there are no motion vectors\n");
@@ -1364,29 +993,40 @@ static void unpack_vectors(Vp3DecodeContext *s, GetBitContext *gb)
             for (j = 0; j < 4; j++) {
                 current_macroblock = s->superblock_macroblocks[i * 4 + j];
                 if ((current_macroblock == -1) ||
-                    (!s->macroblock_coded[current_macroblock]))
+                    (s->macroblock_coding[current_macroblock] == MODE_COPY))
                     continue;
+                if (current_macroblock >= s->macroblock_count) {
+                    av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_vectors(): bad macroblock number (%d >= %d)\n",
+                        current_macroblock, s->macroblock_count);
+                    return 1;
+                }
 
                 current_fragment = s->macroblock_fragments[current_macroblock * 6];
-                switch (s->all_fragments[current_fragment].coding_method) {
+                if (current_fragment >= s->fragment_count) {
+                    av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_vectors(): bad fragment number (%d >= %d\n",
+                        current_fragment, s->fragment_count);
+                    return 1;
+                }
+                switch (s->macroblock_coding[current_macroblock]) {
 
                 case MODE_INTER_PLUS_MV:
                 case MODE_GOLDEN_MV:
                     /* all 6 fragments use the same motion vector */
                     if (coding_mode == 0) {
-                        motion_x[0] = get_motion_vector_vlc(gb);
-                        motion_y[0] = get_motion_vector_vlc(gb);
+                        motion_x[0] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
+                        motion_y[0] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
                     } else {
-                        motion_x[0] = get_motion_vector_fixed(gb);
-                        motion_y[0] = get_motion_vector_fixed(gb);
+                        motion_x[0] = fixed_motion_vector_table[get_bits(gb, 6)];
+                        motion_y[0] = fixed_motion_vector_table[get_bits(gb, 6)];
                     }
+
                     for (k = 1; k < 6; k++) {
                         motion_x[k] = motion_x[0];
                         motion_y[k] = motion_y[0];
                     }
 
                     /* vector maintenance, only on MODE_INTER_PLUS_MV */
-                    if (s->all_fragments[current_fragment].coding_method ==
+                    if (s->macroblock_coding[current_macroblock] ==
                         MODE_INTER_PLUS_MV) {
                         prior_last_motion_x = last_motion_x;
                         prior_last_motion_y = last_motion_y;
@@ -1401,27 +1041,20 @@ static void unpack_vectors(Vp3DecodeContext *s, GetBitContext *gb)
                     motion_x[4] = motion_y[4] = 0;
                     for (k = 0; k < 4; k++) {
                         if (coding_mode == 0) {
-                            motion_x[k] = get_motion_vector_vlc(gb);
-                            motion_y[k] = get_motion_vector_vlc(gb);
+                            motion_x[k] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
+                            motion_y[k] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
                         } else {
-                            motion_x[k] = get_motion_vector_fixed(gb);
-                            motion_y[k] = get_motion_vector_fixed(gb);
+                            motion_x[k] = fixed_motion_vector_table[get_bits(gb, 6)];
+                            motion_y[k] = fixed_motion_vector_table[get_bits(gb, 6)];
                         }
                         motion_x[4] += motion_x[k];
                         motion_y[4] += motion_y[k];
                     }
 
-                    if (motion_x[4] >= 0) 
-                        motion_x[4] = (motion_x[4] + 2) / 4;
-                    else
-                        motion_x[4] = (motion_x[4] - 2) / 4;
-                    motion_x[5] = motion_x[4];
-
-                    if (motion_y[4] >= 0) 
-                        motion_y[4] = (motion_y[4] + 2) / 4;
-                    else
-                        motion_y[4] = (motion_y[4] - 2) / 4;
-                    motion_y[5] = motion_y[4];
+                    motion_x[5]=
+                    motion_x[4]= RSHIFT(motion_x[4], 2);
+                    motion_y[5]=
+                    motion_y[4]= RSHIFT(motion_y[4], 2);
 
                     /* vector maintenance; vector[3] is treated as the
                      * last vector in this case */
@@ -1460,26 +1093,43 @@ static void unpack_vectors(Vp3DecodeContext *s, GetBitContext *gb)
                     last_motion_x = motion_x[0];
                     last_motion_y = motion_y[0];
                     break;
+
+                default:
+                    /* covers intra, inter without MV, golden without MV */
+                    memset(motion_x, 0, 6 * sizeof(int));
+                    memset(motion_y, 0, 6 * sizeof(int));
+
+                    /* no vector maintenance */
+                    break;
                 }
 
                 /* assign the motion vectors to the correct fragments */
                 debug_vectors("    vectors for macroblock starting @ fragment %d (coding method %d):\n",
                     current_fragment,
-                    s->all_fragments[current_fragment].coding_method);
+                    s->macroblock_coding[current_macroblock]);
                 for (k = 0; k < 6; k++) {
-                    current_fragment = 
+                    current_fragment =
                         s->macroblock_fragments[current_macroblock * 6 + k];
+                    if (current_fragment == -1)
+                        continue;
+                    if (current_fragment >= s->fragment_count) {
+                        av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_vectors(): bad fragment number (%d >= %d)\n",
+                            current_fragment, s->fragment_count);
+                        return 1;
+                    }
                     s->all_fragments[current_fragment].motion_x = motion_x[k];
-                    s->all_fragments[current_fragment].motion_x = motion_y[k];
+                    s->all_fragments[current_fragment].motion_y = motion_y[k];
                     debug_vectors("    vector %d: fragment %d = (%d, %d)\n",
                         k, current_fragment, motion_x[k], motion_y[k]);
                 }
             }
         }
     }
+
+    return 0;
 }
 
-/* 
+/*
  * This function is called by unpack_dct_coeffs() to extract the VLCs from
  * the bitstream. The VLCs encode tokens which are used to unpack DCT
  * data. This function unpacks all the VLCs for either the Y plane or both
@@ -1498,11 +1148,21 @@ static int unpack_vlcs(Vp3DecodeContext *s, GetBitContext *gb,
 {
     int i;
     int token;
-    int zero_run;
-    DCTELEM coeff;
+    int zero_run = 0;
+    DCTELEM coeff = 0;
     Vp3Fragment *fragment;
+    uint8_t *perm= s->scantable.permutated;
+    int bits_to_get;
+
+    if ((first_fragment >= s->fragment_count) ||
+        (last_fragment >= s->fragment_count)) {
+
+        av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_vlcs(): bad fragment number (%d -> %d ?)\n",
+            first_fragment, last_fragment);
+        return 0;
+    }
 
-    for (i = first_fragment; i < last_fragment; i++) {
+    for (i = first_fragment; i <= last_fragment; i++) {
 
         fragment = &s->all_fragments[s->coded_fragment_list[i]];
         if (fragment->coeff_count > coeff_index)
@@ -1513,20 +1173,39 @@ static int unpack_vlcs(Vp3DecodeContext *s, GetBitContext *gb,
             token = get_vlc2(gb, table->table, 5, 3);
             debug_vlc(" token = %2d, ", token);
             /* use the token to get a zero run, a coefficient, and an eob run */
-            unpack_token(gb, token, &zero_run, &coeff, &eob_run);
+            if (token <= 6) {
+                eob_run = eob_run_base[token];
+                if (eob_run_get_bits[token])
+                    eob_run += get_bits(gb, eob_run_get_bits[token]);
+                coeff = zero_run = 0;
+            } else {
+                bits_to_get = coeff_get_bits[token];
+                if (!bits_to_get)
+                    coeff = coeff_tables[token][0];
+                else
+                    coeff = coeff_tables[token][get_bits(gb, bits_to_get)];
+
+                zero_run = zero_run_base[token];
+                if (zero_run_get_bits[token])
+                    zero_run += get_bits(gb, zero_run_get_bits[token]);
+            }
         }
 
         if (!eob_run) {
             fragment->coeff_count += zero_run;
-            if (fragment->coeff_count < 64)
-                fragment->coeffs[fragment->coeff_count++] = coeff;
+            if (fragment->coeff_count < 64){
+                fragment->next_coeff->coeff= coeff;
+                fragment->next_coeff->index= perm[fragment->coeff_count++]; //FIXME perm here already?
+                fragment->next_coeff->next= s->next_coeff;
+                s->next_coeff->next=NULL;
+                fragment->next_coeff= s->next_coeff++;
+            }
             debug_vlc(" fragment %d coeff = %d\n",
-                s->coded_fragment_list[i], fragment->coeffs[coeff_index]);
+                s->coded_fragment_list[i], fragment->next_coeff[coeff_index]);
         } else {
-            fragment->last_coeff = fragment->coeff_count;
-            fragment->coeff_count = 64;
-            debug_vlc(" fragment %d eob with %d coefficients\n", 
-                s->coded_fragment_list[i], fragment->last_coeff);
+            fragment->coeff_count |= 128;
+            debug_vlc(" fragment %d eob with %d coefficients\n",
+                s->coded_fragment_list[i], fragment->coeff_count&127);
             eob_run--;
         }
     }
@@ -1538,7 +1217,7 @@ static int unpack_vlcs(Vp3DecodeContext *s, GetBitContext *gb,
  * This function unpacks all of the DCT coefficient data from the
  * bitstream.
  */
-static void unpack_dct_coeffs(Vp3DecodeContext *s, GetBitContext *gb)
+static int unpack_dct_coeffs(Vp3DecodeContext *s, GetBitContext *gb)
 {
     int i;
     int dc_y_table;
@@ -1547,42 +1226,6 @@ static void unpack_dct_coeffs(Vp3DecodeContext *s, GetBitContext *gb)
     int ac_c_table;
     int residual_eob_run = 0;
 
-    /* for the binary search */
-    int left, middle, right, found;
-    /* this indicates the first fragment of the color plane data */
-    int plane_split = 0;
-
-    debug_vp3("  vp3: unpacking DCT coefficients\n");
-
-    /* find the plane split (the first color plane fragment) using a binary 
-     * search; test the boundaries first */
-    if (s->coded_fragment_list_index == 0)
-        return;
-    if (s->u_fragment_start <= s->coded_fragment_list[0])
-        plane_split = 0;  /* this means no Y fragments */
-    else if (s->coded_fragment_list[s->coded_fragment_list_index - 1] >
-        s->u_fragment_start) {
-
-        left = 0;
-        right = s->coded_fragment_list_index - 1;
-        found = 0;
-        do {
-            middle = (left + right + 1) / 2;
-            if ((s->coded_fragment_list[middle] >= s->u_fragment_start) &&
-                (s->coded_fragment_list[middle - 1] < s->u_fragment_start))
-                found = 1;
-            else if (s->coded_fragment_list[middle] < s->u_fragment_start)
-                left = middle;
-            else
-                right = middle;
-        } while (!found);
-
-        plane_split = middle;
-    }
-
-    debug_vp3("  plane split @ index %d (fragment %d)\n", plane_split,
-        s->coded_fragment_list[plane_split]);
-
     /* fetch the DC table indices */
     dc_y_table = get_bits(gb, 4);
     dc_c_table = get_bits(gb, 4);
@@ -1590,91 +1233,92 @@ static void unpack_dct_coeffs(Vp3DecodeContext *s, GetBitContext *gb)
     /* unpack the Y plane DC coefficients */
     debug_vp3("  vp3: unpacking Y plane DC coefficients using table %d\n",
         dc_y_table);
-    residual_eob_run = unpack_vlcs(s, gb, &s->dc_vlc[dc_y_table], 0, 
-        0, plane_split, residual_eob_run);
+    residual_eob_run = unpack_vlcs(s, gb, &s->dc_vlc[dc_y_table], 0,
+        s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
 
     /* unpack the C plane DC coefficients */
     debug_vp3("  vp3: unpacking C plane DC coefficients using table %d\n",
         dc_c_table);
     residual_eob_run = unpack_vlcs(s, gb, &s->dc_vlc[dc_c_table], 0,
-        plane_split, s->coded_fragment_list_index, residual_eob_run);
+        s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
 
-    /* fetch the level 1 AC table indices */
+    /* fetch the AC table indices */
     ac_y_table = get_bits(gb, 4);
     ac_c_table = get_bits(gb, 4);
 
-    /* unpack the level 1 AC coefficients (coeffs 1-5) */
+    /* unpack the group 1 AC coefficients (coeffs 1-5) */
     for (i = 1; i <= 5; i++) {
 
         debug_vp3("  vp3: unpacking level %d Y plane AC coefficients using table %d\n",
             i, ac_y_table);
-        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_1[ac_y_table], i, 
-            0, plane_split, residual_eob_run);
+        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_1[ac_y_table], i,
+            s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
 
         debug_vp3("  vp3: unpacking level %d C plane AC coefficients using table %d\n",
             i, ac_c_table);
-        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_1[ac_c_table], i, 
-            plane_split, s->coded_fragment_list_index, residual_eob_run);
+        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_1[ac_c_table], i,
+            s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
     }
 
-    /* unpack the level 2 AC coefficients (coeffs 6-14) */
+    /* unpack the group 2 AC coefficients (coeffs 6-14) */
     for (i = 6; i <= 14; i++) {
 
         debug_vp3("  vp3: unpacking level %d Y plane AC coefficients using table %d\n",
             i, ac_y_table);
-        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_2[ac_y_table], i, 
-            0, plane_split, residual_eob_run);
+        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_2[ac_y_table], i,
+            s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
 
         debug_vp3("  vp3: unpacking level %d C plane AC coefficients using table %d\n",
             i, ac_c_table);
-        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_2[ac_c_table], i, 
-            plane_split, s->coded_fragment_list_index, residual_eob_run);
+        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_2[ac_c_table], i,
+            s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
     }
 
-    /* unpack the level 3 AC coefficients (coeffs 15-27) */
+    /* unpack the group 3 AC coefficients (coeffs 15-27) */
     for (i = 15; i <= 27; i++) {
 
         debug_vp3("  vp3: unpacking level %d Y plane AC coefficients using table %d\n",
             i, ac_y_table);
-        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_3[ac_y_table], i, 
-            0, plane_split, residual_eob_run);
+        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_3[ac_y_table], i,
+            s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
 
         debug_vp3("  vp3: unpacking level %d C plane AC coefficients using table %d\n",
             i, ac_c_table);
-        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_3[ac_c_table], i, 
-            plane_split, s->coded_fragment_list_index, residual_eob_run);
+        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_3[ac_c_table], i,
+            s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
     }
 
-    /* unpack the level 4 AC coefficients (coeffs 28-63) */
+    /* unpack the group 4 AC coefficients (coeffs 28-63) */
     for (i = 28; i <= 63; i++) {
 
         debug_vp3("  vp3: unpacking level %d Y plane AC coefficients using table %d\n",
             i, ac_y_table);
-        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_4[ac_y_table], i, 
-            0, plane_split, residual_eob_run);
+        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_4[ac_y_table], i,
+            s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
 
         debug_vp3("  vp3: unpacking level %d C plane AC coefficients using table %d\n",
             i, ac_c_table);
-        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_4[ac_c_table], i, 
-            plane_split, s->coded_fragment_list_index, residual_eob_run);
+        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_4[ac_c_table], i,
+            s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
     }
+
+    return 0;
 }
 
 /*
  * This function reverses the DC prediction for each coded fragment in
- * the frame. Much of this function is adapted directly from the original 
+ * the frame. Much of this function is adapted directly from the original
  * VP3 source code.
  */
 #define COMPATIBLE_FRAME(x) \
   (compatible_frame[s->all_fragments[x].coding_method] == current_frame_type)
 #define FRAME_CODED(x) (s->all_fragments[x].coding_method != MODE_COPY)
-#define HIGHBITDUPPED(X) (((signed short) X)  >> 15)
-static inline int iabs (int x) { return ((x < 0) ? -x : x); }
+#define DC_COEFF(u) (s->coeffs[u].index ? 0 : s->coeffs[u].coeff) //FIXME do somethin to simplify this
 
 static void reverse_dc_prediction(Vp3DecodeContext *s,
                                   int first_fragment,
                                   int fragment_width,
-                                  int fragment_height) 
+                                  int fragment_height)
 {
 
 #define PUL 8
@@ -1685,23 +1329,7 @@ static void reverse_dc_prediction(Vp3DecodeContext *s,
     int x, y;
     int i = first_fragment;
 
-    /*
-     * Fragment prediction groups:
-     *
-     * 32222222226
-     * 10000000004
-     * 10000000004
-     * 10000000004
-     * 10000000004
-     *
-     * Note: Groups 5 and 7 do not exist as it would mean that the 
-     * fragment's x coordinate is both 0 and (width - 1) at the same time.
-     */
-    int predictor_group;
-    short predicted_dc;
-
-    /* validity flags for the left, up-left, up, and up-right fragments */
-    int fl, ful, fu, fur;
+    int predicted_dc;
 
     /* DC values for the left, up-left, up, and up-right fragments */
     int vl, vul, vu, vur;
@@ -1709,38 +1337,36 @@ static void reverse_dc_prediction(Vp3DecodeContext *s,
     /* indices for the left, up-left, up, and up-right fragments */
     int l, ul, u, ur;
 
-    /* 
+    /*
      * The 6 fields mean:
      *   0: up-left multiplier
      *   1: up multiplier
      *   2: up-right multiplier
      *   3: left multiplier
-     *   4: mask
-     *   5: right bit shift divisor (e.g., 7 means >>=7, a.k.a. div by 128)
      */
-    int predictor_transform[16][6] = {
-        {  0,  0,  0,  0,   0,  0 },
-        {  0,  0,  0,  1,   0,  0 },        // PL
-        {  0,  0,  1,  0,   0,  0 },        // PUR
-        {  0,  0, 53, 75, 127,  7 },        // PUR|PL
-        {  0,  1,  0,  0,   0,  0 },        // PU
-        {  0,  1,  0,  1,   1,  1 },        // PU|PL
-        {  0,  1,  0,  0,   0,  0 },        // PU|PUR
-        {  0,  0, 53, 75, 127,  7 },        // PU|PUR|PL
-        {  1,  0,  0,  0,   0,  0 },        // PUL
-        {  0,  0,  0,  1,   0,  0 },        // PUL|PL
-        {  1,  0,  1,  0,   1,  1 },        // PUL|PUR
-        {  0,  0, 53, 75, 127,  7 },        // PUL|PUR|PL
-        {  0,  1,  0,  0,   0,  0 },        // PUL|PU
-        {-26, 29,  0, 29,  31,  5 },        // PUL|PU|PL
-        {  3, 10,  3,  0,  15,  4 },        // PUL|PU|PUR
-        {-26, 29,  0, 29,  31,  5 }         // PUL|PU|PUR|PL
+    int predictor_transform[16][4] = {
+        {  0,  0,  0,  0},
+        {  0,  0,  0,128},        // PL
+        {  0,  0,128,  0},        // PUR
+        {  0,  0, 53, 75},        // PUR|PL
+        {  0,128,  0,  0},        // PU
+        {  0, 64,  0, 64},        // PU|PL
+        {  0,128,  0,  0},        // PU|PUR
+        {  0,  0, 53, 75},        // PU|PUR|PL
+        {128,  0,  0,  0},        // PUL
+        {  0,  0,  0,128},        // PUL|PL
+        { 64,  0, 64,  0},        // PUL|PUR
+        {  0,  0, 53, 75},        // PUL|PUR|PL
+        {  0,128,  0,  0},        // PUL|PU
+       {-104,116,  0,116},        // PUL|PU|PL
+        { 24, 80, 24,  0},        // PUL|PU|PUR
+       {-104,116,  0,116}         // PUL|PU|PUR|PL
     };
 
     /* This table shows which types of blocks can use other blocks for
      * prediction. For example, INTRA is the only mode in this table to
      * have a frame number of 0. That means INTRA blocks can only predict
-     * from other INTRA blocks. There are 2 golden frame coding types; 
+     * from other INTRA blocks. There are 2 golden frame coding types;
      * blocks encoding in these modes can only predict from other blocks
      * that were encoded with these 1 of these 2 modes. */
     unsigned char compatible_frame[8] = {
@@ -1774,115 +1400,35 @@ static void reverse_dc_prediction(Vp3DecodeContext *s,
             /* reverse prediction if this block was coded */
             if (s->all_fragments[i].coding_method != MODE_COPY) {
 
-                current_frame_type = 
+                current_frame_type =
                     compatible_frame[s->all_fragments[i].coding_method];
-                predictor_group = (x == 0) + ((y == 0) << 1) +
-                    ((x + 1 == fragment_width) << 2);
-                debug_dc_pred(" frag %d: group %d, orig DC = %d, ",
-                    i, predictor_group, s->all_fragments[i].coeffs[0]);
-
-                switch (predictor_group) {
-
-                case 0:
-                    /* main body of fragments; consider all 4 possible
-                     * fragments for prediction */
-
-                    /* calculate the indices of the predicting fragments */
-                    ul = i - fragment_width - 1;
-                    u = i - fragment_width;
-                    ur = i - fragment_width + 1;
-                    l = i - 1;
-
-                    /* fetch the DC values for the predicting fragments */
-                    vul = s->all_fragments[ul].coeffs[0];
-                    vu = s->all_fragments[u].coeffs[0];
-                    vur = s->all_fragments[ur].coeffs[0];
-                    vl = s->all_fragments[l].coeffs[0];
-
-                    /* figure out which fragments are valid */
-                    ful = FRAME_CODED(ul) && COMPATIBLE_FRAME(ul);
-                    fu = FRAME_CODED(u) && COMPATIBLE_FRAME(u);
-                    fur = FRAME_CODED(ur) && COMPATIBLE_FRAME(ur);
-                    fl = FRAME_CODED(l) && COMPATIBLE_FRAME(l);
-
-                    /* decide which predictor transform to use */
-                    transform = (fl*PL) | (fu*PU) | (ful*PUL) | (fur*PUR);
-
-                    break;
-
-                case 1:
-                    /* left column of fragments, not including top corner;
-                     * only consider up and up-right fragments */
-
-                    /* calculate the indices of the predicting fragments */
-                    u = i - fragment_width;
-                    ur = i - fragment_width + 1;
-
-                    /* fetch the DC values for the predicting fragments */
-                    vu = s->all_fragments[u].coeffs[0];
-                    vur = s->all_fragments[ur].coeffs[0];
-
-                    /* figure out which fragments are valid */
-                    fur = FRAME_CODED(ur) && COMPATIBLE_FRAME(ur);
-                    fu = FRAME_CODED(u) && COMPATIBLE_FRAME(u);
-
-                    /* decide which predictor transform to use */
-                    transform = (fu*PU) | (fur*PUR);
-
-                    break;
-
-                case 2:
-                case 6:
-                    /* top row of fragments, not including top-left frag;
-                     * only consider the left fragment for prediction */
-
-                    /* calculate the indices of the predicting fragments */
-                    l = i - 1;
-
-                    /* fetch the DC values for the predicting fragments */
-                    vl = s->all_fragments[l].coeffs[0];
-
-                    /* figure out which fragments are valid */
-                    fl = FRAME_CODED(l) && COMPATIBLE_FRAME(l);
-
-                    /* decide which predictor transform to use */
-                    transform = (fl*PL);
-
-                    break;
-
-                case 3:
-                    /* top-left fragment */
-
-                    /* nothing to predict from in this case */
-                    transform = 0;
-
-                    break;
-
-                case 4:
-                    /* right column of fragments, not including top corner;
-                     * consider up-left, up, and left fragments for
-                     * prediction */
-
-                    /* calculate the indices of the predicting fragments */
-                    ul = i - fragment_width - 1;
-                    u = i - fragment_width;
-                    l = i - 1;
-
-                    /* fetch the DC values for the predicting fragments */
-                    vul = s->all_fragments[ul].coeffs[0];
-                    vu = s->all_fragments[u].coeffs[0];
-                    vl = s->all_fragments[l].coeffs[0];
-
-                    /* figure out which fragments are valid */
-                    ful = FRAME_CODED(ul) && COMPATIBLE_FRAME(ul);
-                    fu = FRAME_CODED(u) && COMPATIBLE_FRAME(u);
-                    fl = FRAME_CODED(l) && COMPATIBLE_FRAME(l);
-
-                    /* decide which predictor transform to use */
-                    transform = (fl*PL) | (fu*PU) | (ful*PUL);
-
-                    break;
-
+                debug_dc_pred(" frag %d: orig DC = %d, ",
+                    i, DC_COEFF(i));
+
+                transform= 0;
+                if(x){
+                    l= i-1;
+                    vl = DC_COEFF(l);
+                    if(FRAME_CODED(l) && COMPATIBLE_FRAME(l))
+                        transform |= PL;
+                }
+                if(y){
+                    u= i-fragment_width;
+                    vu = DC_COEFF(u);
+                    if(FRAME_CODED(u) && COMPATIBLE_FRAME(u))
+                        transform |= PU;
+                    if(x){
+                        ul= i-fragment_width-1;
+                        vul = DC_COEFF(ul);
+                        if(FRAME_CODED(ul) && COMPATIBLE_FRAME(ul))
+                            transform |= PUL;
+                    }
+                    if(x + 1 < fragment_width){
+                        ur= i-fragment_width+1;
+                        vur = DC_COEFF(ur);
+                        if(FRAME_CODED(ur) && COMPATIBLE_FRAME(ur))
+                            transform |= PUR;
+                    }
                 }
 
                 debug_dc_pred("transform = %d, ", transform);
@@ -1891,9 +1437,9 @@ static void reverse_dc_prediction(Vp3DecodeContext *s,
 
                     /* if there were no fragments to predict from, use last
                      * DC saved */
-                    s->all_fragments[i].coeffs[0] += last_dc[current_frame_type];
-                    debug_dc_pred("from last DC (%d) = %d\n", 
-                        current_frame_type, s->all_fragments[i].coeffs[0]);
+                    predicted_dc = last_dc[current_frame_type];
+                    debug_dc_pred("from last DC (%d) = %d\n",
+                        current_frame_type, DC_COEFF(i));
 
                 } else {
 
@@ -1904,150 +1450,397 @@ static void reverse_dc_prediction(Vp3DecodeContext *s,
                         (predictor_transform[transform][2] * vur) +
                         (predictor_transform[transform][3] * vl);
 
-                    /* if there is a shift value in the transform, add
-                     * the sign bit before the shift */
-                    if (predictor_transform[transform][5] != 0) {
-                        predicted_dc += ((predicted_dc >> 15) & 
-                            predictor_transform[transform][4]);
-                        predicted_dc >>= predictor_transform[transform][5];
-                    }
+                    predicted_dc /= 128;
 
                     /* check for outranging on the [ul u l] and
                      * [ul u ur l] predictors */
                     if ((transform == 13) || (transform == 15)) {
-                        if (iabs(predicted_dc - vu) > 128)
+                        if (ABS(predicted_dc - vu) > 128)
                             predicted_dc = vu;
-                        else if (iabs(predicted_dc - vl) > 128)
+                        else if (ABS(predicted_dc - vl) > 128)
                             predicted_dc = vl;
-                        else if (iabs(predicted_dc - vul) > 128)
+                        else if (ABS(predicted_dc - vul) > 128)
                             predicted_dc = vul;
                     }
 
-                    /* at long last, apply the predictor */
-                    s->all_fragments[i].coeffs[0] += predicted_dc;
-                    debug_dc_pred("from pred DC = %d\n", 
-                    s->all_fragments[i].coeffs[0]);
+                    debug_dc_pred("from pred DC = %d\n",
+                    DC_COEFF(i));
                 }
 
+                /* at long last, apply the predictor */
+                if(s->coeffs[i].index){
+                    *s->next_coeff= s->coeffs[i];
+                    s->coeffs[i].index=0;
+                    s->coeffs[i].coeff=0;
+                    s->coeffs[i].next= s->next_coeff++;
+                }
+                s->coeffs[i].coeff += predicted_dc;
                 /* save the DC */
-                last_dc[current_frame_type] = s->all_fragments[i].coeffs[0];
+                last_dc[current_frame_type] = DC_COEFF(i);
+                if(DC_COEFF(i) && !(s->all_fragments[i].coeff_count&127)){
+                    s->all_fragments[i].coeff_count= 129;
+//                    s->all_fragments[i].next_coeff= s->next_coeff;
+                    s->coeffs[i].next= s->next_coeff;
+                    (s->next_coeff++)->next=NULL;
+                }
             }
         }
     }
 }
 
+
+static void horizontal_filter(unsigned char *first_pixel, int stride,
+    int *bounding_values);
+static void vertical_filter(unsigned char *first_pixel, int stride,
+    int *bounding_values);
+
 /*
- * This function performs the final rendering of each fragment's data
- * onto the output frame.
+ * Perform the final rendering for a particular slice of data.
+ * The slice number ranges from 0..(macroblock_height - 1).
  */
-static void render_fragments(Vp3DecodeContext *s,
-                             int first_fragment,
-                             int fragment_width,
-                             int fragment_height,
-                             int plane /* 0 = Y, 1 = U, 2 = V */) 
+static void render_slice(Vp3DecodeContext *s, int slice)
 {
-    int x, y;
+    int x;
     int m, n;
-    int i = first_fragment;
-    int j;
     int16_t *dequantizer;
-    DCTELEM dequant_block[64];
-    unsigned char *output_plane;
-    unsigned char *last_plane;
-    unsigned char *golden_plane;
-    int stride;
-
-    debug_vp3("  vp3: rendering final fragments for %s\n",
-        (plane == 0) ? "Y plane" : (plane == 1) ? "U plane" : "V plane");
-
-    /* set up plane-specific parameters */
-    if (plane == 0) {
-        dequantizer = s->intra_y_dequant;
-        output_plane = s->current_frame.data[0];
-        last_plane = s->last_frame.data[0];
-        golden_plane = s->golden_frame.data[0];
-        stride = -s->current_frame.linesize[0];
-    } else if (plane == 1) {
-        dequantizer = s->intra_c_dequant;
-        output_plane = s->current_frame.data[1];
-        last_plane = s->last_frame.data[1];
-        golden_plane = s->golden_frame.data[1];
-        stride = -s->current_frame.linesize[1];
-    } else {
-        dequantizer = s->intra_c_dequant;
-        output_plane = s->current_frame.data[2];
-        last_plane = s->last_frame.data[2];
-        golden_plane = s->golden_frame.data[2];
-        stride = -s->current_frame.linesize[2];
-    }
+    DECLARE_ALIGNED_16(DCTELEM, block[64]);
+    int motion_x = 0xdeadbeef, motion_y = 0xdeadbeef;
+    int motion_halfpel_index;
+    uint8_t *motion_source;
+    int plane;
+    int current_macroblock_entry = slice * s->macroblock_width * 6;
+
+    if (slice >= s->macroblock_height)
+        return;
 
-    /* for each fragment row... */
-    for (y = 0; y < fragment_height; y++) {
+    for (plane = 0; plane < 3; plane++) {
+        uint8_t *output_plane = s->current_frame.data    [plane];
+        uint8_t *  last_plane = s->   last_frame.data    [plane];
+        uint8_t *golden_plane = s-> golden_frame.data    [plane];
+        int stride            = s->current_frame.linesize[plane];
+        int plane_width       = s->width  >> !!plane;
+        int plane_height      = s->height >> !!plane;
+        int y =        slice *  FRAGMENT_PIXELS << !plane ;
+        int slice_height = y + (FRAGMENT_PIXELS << !plane);
+        int i = s->macroblock_fragments[current_macroblock_entry + plane + 3*!!plane];
 
-        /* for each fragment in a row... */
-        for (x = 0; x < fragment_width; x++, i++) {
+        if (!s->flipped_image) stride = -stride;
 
-            /* transform if this block was coded */
-            if (s->all_fragments[i].coding_method == MODE_INTRA) {
-                /* dequantize the DCT coefficients */
-                for (j = 0; j < 64; j++)
-                    dequant_block[dequant_index[j]] =
-                        s->all_fragments[i].coeffs[j] *
-                        dequantizer[j];
-                dequant_block[0] += 1024;
-
-                debug_idct("fragment %d:\n", i);
-                debug_idct("dequantized block:\n");
-                for (m = 0; m < 8; m++) {
-                    for (n = 0; n < 8; n++) {
-                        debug_idct(" %5d", dequant_block[m * 8 + n]);
-                    }
-                    debug_idct("\n");
+
+        if(ABS(stride) > 2048)
+            return; //various tables are fixed size
+
+        /* for each fragment row in the slice (both of them)... */
+        for (; y < slice_height; y += 8) {
+
+            /* for each fragment in a row... */
+            for (x = 0; x < plane_width; x += 8, i++) {
+
+                if ((i < 0) || (i >= s->fragment_count)) {
+                    av_log(s->avctx, AV_LOG_ERROR, "  vp3:render_slice(): bad fragment number (%d)\n", i);
+                    return;
                 }
-                debug_idct("\n");
 
-                /* invert DCT and place in final output */
-                s->dsp.idct_put(
-                    output_plane + s->all_fragments[i].first_pixel,
-                    stride, dequant_block);
+                /* transform if this block was coded */
+                if ((s->all_fragments[i].coding_method != MODE_COPY) &&
+                    !((s->avctx->flags & CODEC_FLAG_GRAY) && plane)) {
 
-/*
-                debug_idct("idct block:\n");
-                for (m = 0; m < 8; m++) {
-                    for (n = 0; n < 8; n++) {
-                        debug_idct(" %3d", pixels[m * 8 + n]);
+                    if ((s->all_fragments[i].coding_method == MODE_USING_GOLDEN) ||
+                        (s->all_fragments[i].coding_method == MODE_GOLDEN_MV))
+                        motion_source= golden_plane;
+                    else
+                        motion_source= last_plane;
+
+                    motion_source += s->all_fragments[i].first_pixel;
+                    motion_halfpel_index = 0;
+
+                    /* sort out the motion vector if this fragment is coded
+                     * using a motion vector method */
+                    if ((s->all_fragments[i].coding_method > MODE_INTRA) &&
+                        (s->all_fragments[i].coding_method != MODE_USING_GOLDEN)) {
+                        int src_x, src_y;
+                        motion_x = s->all_fragments[i].motion_x;
+                        motion_y = s->all_fragments[i].motion_y;
+                        if(plane){
+                            motion_x= (motion_x>>1) | (motion_x&1);
+                            motion_y= (motion_y>>1) | (motion_y&1);
+                        }
+
+                        src_x= (motion_x>>1) + x;
+                        src_y= (motion_y>>1) + y;
+                        if ((motion_x == 127) || (motion_y == 127))
+                            av_log(s->avctx, AV_LOG_ERROR, " help! got invalid motion vector! (%X, %X)\n", motion_x, motion_y);
+
+                        motion_halfpel_index = motion_x & 0x01;
+                        motion_source += (motion_x >> 1);
+
+                        motion_halfpel_index |= (motion_y & 0x01) << 1;
+                        motion_source += ((motion_y >> 1) * stride);
+
+                        if(src_x<0 || src_y<0 || src_x + 9 >= plane_width || src_y + 9 >= plane_height){
+                            uint8_t *temp= s->edge_emu_buffer;
+                            if(stride<0) temp -= 9*stride;
+                            else temp += 9*stride;
+
+                            ff_emulated_edge_mc(temp, motion_source, stride, 9, 9, src_x, src_y, plane_width, plane_height);
+                            motion_source= temp;
+                        }
+                    }
+
+
+                    /* first, take care of copying a block from either the
+                     * previous or the golden frame */
+                    if (s->all_fragments[i].coding_method != MODE_INTRA) {
+                        /* Note, it is possible to implement all MC cases with
+                           put_no_rnd_pixels_l2 which would look more like the
+                           VP3 source but this would be slower as
+                           put_no_rnd_pixels_tab is better optimzed */
+                        if(motion_halfpel_index != 3){
+                            s->dsp.put_no_rnd_pixels_tab[1][motion_halfpel_index](
+                                output_plane + s->all_fragments[i].first_pixel,
+                                motion_source, stride, 8);
+                        }else{
+                            int d= (motion_x ^ motion_y)>>31; // d is 0 if motion_x and _y have the same sign, else -1
+                            s->dsp.put_no_rnd_pixels_l2[1](
+                                output_plane + s->all_fragments[i].first_pixel,
+                                motion_source - d,
+                                motion_source + stride + 1 + d,
+                                stride, 8);
+                        }
+                        dequantizer = s->qmat[1][plane];
+                    }else{
+                        dequantizer = s->qmat[0][plane];
+                    }
+
+                    /* dequantize the DCT coefficients */
+                    debug_idct("fragment %d, coding mode %d, DC = %d, dequant = %d:\n",
+                        i, s->all_fragments[i].coding_method,
+                        DC_COEFF(i), dequantizer[0]);
+
+                    if(s->avctx->idct_algo==FF_IDCT_VP3){
+                        Coeff *coeff= s->coeffs + i;
+                        memset(block, 0, sizeof(block));
+                        while(coeff->next){
+                            block[coeff->index]= coeff->coeff * dequantizer[coeff->index];
+                            coeff= coeff->next;
+                        }
+                    }else{
+                        Coeff *coeff= s->coeffs + i;
+                        memset(block, 0, sizeof(block));
+                        while(coeff->next){
+                            block[coeff->index]= (coeff->coeff * dequantizer[coeff->index] + 2)>>2;
+                            coeff= coeff->next;
+                        }
+                    }
+
+                    /* invert DCT and place (or add) in final output */
+
+                    if (s->all_fragments[i].coding_method == MODE_INTRA) {
+                        if(s->avctx->idct_algo!=FF_IDCT_VP3)
+                            block[0] += 128<<3;
+                        s->dsp.idct_put(
+                            output_plane + s->all_fragments[i].first_pixel,
+                            stride,
+                            block);
+                    } else {
+                        s->dsp.idct_add(
+                            output_plane + s->all_fragments[i].first_pixel,
+                            stride,
+                            block);
+                    }
+
+                    debug_idct("block after idct_%s():\n",
+                        (s->all_fragments[i].coding_method == MODE_INTRA)?
+                        "put" : "add");
+                    for (m = 0; m < 8; m++) {
+                        for (n = 0; n < 8; n++) {
+                            debug_idct(" %3d", *(output_plane +
+                                s->all_fragments[i].first_pixel + (m * stride + n)));
+                        }
+                        debug_idct("\n");
                     }
                     debug_idct("\n");
-                }
-                debug_idct("\n");
-*/
-            } else if (s->all_fragments[i].coding_method == MODE_COPY) {
 
-                /* copy directly from the previous frame */
-                for (m = 0; m < 8; m++)
-                    memcpy(
-                        output_plane + s->all_fragments[i].first_pixel + stride * m,
-                        last_plane + s->all_fragments[i].first_pixel + stride * m,
-                        8);
+                } else {
 
-            } else {
+                    /* copy directly from the previous frame */
+                    s->dsp.put_pixels_tab[1][0](
+                        output_plane + s->all_fragments[i].first_pixel,
+                        last_plane + s->all_fragments[i].first_pixel,
+                        stride, 8);
 
-                /* carry out the motion compensation */
+                }
+#if 0
+                /* perform the left edge filter if:
+                 *   - the fragment is not on the left column
+                 *   - the fragment is coded in this frame
+                 *   - the fragment is not coded in this frame but the left
+                 *     fragment is coded in this frame (this is done instead
+                 *     of a right edge filter when rendering the left fragment
+                 *     since this fragment is not available yet) */
+                if ((x > 0) &&
+                    ((s->all_fragments[i].coding_method != MODE_COPY) ||
+                     ((s->all_fragments[i].coding_method == MODE_COPY) &&
+                      (s->all_fragments[i - 1].coding_method != MODE_COPY)) )) {
+                    horizontal_filter(
+                        output_plane + s->all_fragments[i].first_pixel + 7*stride,
+                        -stride, s->bounding_values_array + 127);
+                }
 
+                /* perform the top edge filter if:
+                 *   - the fragment is not on the top row
+                 *   - the fragment is coded in this frame
+                 *   - the fragment is not coded in this frame but the above
+                 *     fragment is coded in this frame (this is done instead
+                 *     of a bottom edge filter when rendering the above
+                 *     fragment since this fragment is not available yet) */
+                if ((y > 0) &&
+                    ((s->all_fragments[i].coding_method != MODE_COPY) ||
+                     ((s->all_fragments[i].coding_method == MODE_COPY) &&
+                      (s->all_fragments[i - fragment_width].coding_method != MODE_COPY)) )) {
+                    vertical_filter(
+                        output_plane + s->all_fragments[i].first_pixel - stride,
+                        -stride, s->bounding_values_array + 127);
+                }
+#endif
             }
         }
     }
 
+     /* this looks like a good place for slice dispatch... */
+     /* algorithm:
+      *   if (slice == s->macroblock_height - 1)
+      *     dispatch (both last slice & 2nd-to-last slice);
+      *   else if (slice > 0)
+      *     dispatch (slice - 1);
+      */
+
     emms_c();
+}
+
+static void horizontal_filter(unsigned char *first_pixel, int stride,
+    int *bounding_values)
+{
+    unsigned char *end;
+    int filter_value;
+
+    for (end= first_pixel + 8*stride; first_pixel != end; first_pixel += stride) {
+        filter_value =
+            (first_pixel[-2] - first_pixel[ 1])
+         +3*(first_pixel[ 0] - first_pixel[-1]);
+        filter_value = bounding_values[(filter_value + 4) >> 3];
+        first_pixel[-1] = clip_uint8(first_pixel[-1] + filter_value);
+        first_pixel[ 0] = clip_uint8(first_pixel[ 0] - filter_value);
+    }
+}
+
+static void vertical_filter(unsigned char *first_pixel, int stride,
+    int *bounding_values)
+{
+    unsigned char *end;
+    int filter_value;
+    const int nstride= -stride;
+
+    for (end= first_pixel + 8; first_pixel < end; first_pixel++) {
+        filter_value =
+            (first_pixel[2 * nstride] - first_pixel[ stride])
+         +3*(first_pixel[0          ] - first_pixel[nstride]);
+        filter_value = bounding_values[(filter_value + 4) >> 3];
+        first_pixel[nstride] = clip_uint8(first_pixel[nstride] + filter_value);
+        first_pixel[0] = clip_uint8(first_pixel[0] - filter_value);
+    }
+}
+
+static void apply_loop_filter(Vp3DecodeContext *s)
+{
+    int plane;
+    int x, y;
+    int *bounding_values= s->bounding_values_array+127;
+
+#if 0
+    int bounding_values_array[256];
+    int filter_limit;
+
+    /* find the right loop limit value */
+    for (x = 63; x >= 0; x--) {
+        if (vp31_ac_scale_factor[x] >= s->quality_index)
+            break;
+    }
+    filter_limit = vp31_filter_limit_values[s->quality_index];
+
+    /* set up the bounding values */
+    memset(bounding_values_array, 0, 256 * sizeof(int));
+    for (x = 0; x < filter_limit; x++) {
+        bounding_values[-x - filter_limit] = -filter_limit + x;
+        bounding_values[-x] = -x;
+        bounding_values[x] = x;
+        bounding_values[x + filter_limit] = filter_limit - x;
+    }
+#endif
+
+    for (plane = 0; plane < 3; plane++) {
+        int width           = s->fragment_width  >> !!plane;
+        int height          = s->fragment_height >> !!plane;
+        int fragment        = s->fragment_start        [plane];
+        int stride          = s->current_frame.linesize[plane];
+        uint8_t *plane_data = s->current_frame.data    [plane];
+        if (!s->flipped_image) stride = -stride;
+
+        for (y = 0; y < height; y++) {
+
+            for (x = 0; x < width; x++) {
+START_TIMER
+                /* do not perform left edge filter for left columns frags */
+                if ((x > 0) &&
+                    (s->all_fragments[fragment].coding_method != MODE_COPY)) {
+                    horizontal_filter(
+                        plane_data + s->all_fragments[fragment].first_pixel,
+                        stride, bounding_values);
+                }
 
+                /* do not perform top edge filter for top row fragments */
+                if ((y > 0) &&
+                    (s->all_fragments[fragment].coding_method != MODE_COPY)) {
+                    vertical_filter(
+                        plane_data + s->all_fragments[fragment].first_pixel,
+                        stride, bounding_values);
+                }
+
+                /* do not perform right edge filter for right column
+                 * fragments or if right fragment neighbor is also coded
+                 * in this frame (it will be filtered in next iteration) */
+                if ((x < width - 1) &&
+                    (s->all_fragments[fragment].coding_method != MODE_COPY) &&
+                    (s->all_fragments[fragment + 1].coding_method == MODE_COPY)) {
+                    horizontal_filter(
+                        plane_data + s->all_fragments[fragment + 1].first_pixel,
+                        stride, bounding_values);
+                }
+
+                /* do not perform bottom edge filter for bottom row
+                 * fragments or if bottom fragment neighbor is also coded
+                 * in this frame (it will be filtered in the next row) */
+                if ((y < height - 1) &&
+                    (s->all_fragments[fragment].coding_method != MODE_COPY) &&
+                    (s->all_fragments[fragment + width].coding_method == MODE_COPY)) {
+                    vertical_filter(
+                        plane_data + s->all_fragments[fragment + width].first_pixel,
+                        stride, bounding_values);
+                }
+
+                fragment++;
+STOP_TIMER("loop filter")
+            }
+        }
+    }
 }
 
-/* 
+/*
  * This function computes the first pixel addresses for each fragment.
  * This function needs to be invoked after the first frame is allocated
  * so that it has access to the plane strides.
  */
-static void vp3_calculate_pixel_addresses(Vp3DecodeContext *s) 
+static void vp3_calculate_pixel_addresses(Vp3DecodeContext *s)
 {
 
     int i, x, y;
@@ -2057,37 +1850,84 @@ static void vp3_calculate_pixel_addresses(Vp3DecodeContext *s)
     i = 0;
     for (y = s->fragment_height; y > 0; y--) {
         for (x = 0; x < s->fragment_width; x++) {
-            s->all_fragments[i++].first_pixel = 
+            s->all_fragments[i++].first_pixel =
                 s->golden_frame.linesize[0] * y * FRAGMENT_PIXELS -
                     s->golden_frame.linesize[0] +
                     x * FRAGMENT_PIXELS;
-            debug_init("  fragment %d, first pixel @ %d\n", 
+            debug_init("  fragment %d, first pixel @ %d\n",
                 i-1, s->all_fragments[i-1].first_pixel);
         }
     }
 
     /* U plane */
-    i = s->u_fragment_start;
+    i = s->fragment_start[1];
     for (y = s->fragment_height / 2; y > 0; y--) {
         for (x = 0; x < s->fragment_width / 2; x++) {
-            s->all_fragments[i++].first_pixel = 
+            s->all_fragments[i++].first_pixel =
                 s->golden_frame.linesize[1] * y * FRAGMENT_PIXELS -
                     s->golden_frame.linesize[1] +
                     x * FRAGMENT_PIXELS;
-            debug_init("  fragment %d, first pixel @ %d\n", 
+            debug_init("  fragment %d, first pixel @ %d\n",
                 i-1, s->all_fragments[i-1].first_pixel);
         }
     }
 
     /* V plane */
-    i = s->v_fragment_start;
+    i = s->fragment_start[2];
     for (y = s->fragment_height / 2; y > 0; y--) {
         for (x = 0; x < s->fragment_width / 2; x++) {
-            s->all_fragments[i++].first_pixel = 
+            s->all_fragments[i++].first_pixel =
+                s->golden_frame.linesize[2] * y * FRAGMENT_PIXELS -
+                    s->golden_frame.linesize[2] +
+                    x * FRAGMENT_PIXELS;
+            debug_init("  fragment %d, first pixel @ %d\n",
+                i-1, s->all_fragments[i-1].first_pixel);
+        }
+    }
+}
+
+/* FIXME: this should be merged with the above! */
+static void theora_calculate_pixel_addresses(Vp3DecodeContext *s)
+{
+
+    int i, x, y;
+
+    /* figure out the first pixel addresses for each of the fragments */
+    /* Y plane */
+    i = 0;
+    for (y = 1; y <= s->fragment_height; y++) {
+        for (x = 0; x < s->fragment_width; x++) {
+            s->all_fragments[i++].first_pixel =
+                s->golden_frame.linesize[0] * y * FRAGMENT_PIXELS -
+                    s->golden_frame.linesize[0] +
+                    x * FRAGMENT_PIXELS;
+            debug_init("  fragment %d, first pixel @ %d\n",
+                i-1, s->all_fragments[i-1].first_pixel);
+        }
+    }
+
+    /* U plane */
+    i = s->fragment_start[1];
+    for (y = 1; y <= s->fragment_height / 2; y++) {
+        for (x = 0; x < s->fragment_width / 2; x++) {
+            s->all_fragments[i++].first_pixel =
+                s->golden_frame.linesize[1] * y * FRAGMENT_PIXELS -
+                    s->golden_frame.linesize[1] +
+                    x * FRAGMENT_PIXELS;
+            debug_init("  fragment %d, first pixel @ %d\n",
+                i-1, s->all_fragments[i-1].first_pixel);
+        }
+    }
+
+    /* V plane */
+    i = s->fragment_start[2];
+    for (y = 1; y <= s->fragment_height / 2; y++) {
+        for (x = 0; x < s->fragment_width / 2; x++) {
+            s->all_fragments[i++].first_pixel =
                 s->golden_frame.linesize[2] * y * FRAGMENT_PIXELS -
                     s->golden_frame.linesize[2] +
                     x * FRAGMENT_PIXELS;
-            debug_init("  fragment %d, first pixel @ %d\n", 
+            debug_init("  fragment %d, first pixel @ %d\n",
                 i-1, s->all_fragments[i-1].first_pixel);
         }
     }
@@ -2099,24 +1939,46 @@ static void vp3_calculate_pixel_addresses(Vp3DecodeContext *s)
 static int vp3_decode_init(AVCodecContext *avctx)
 {
     Vp3DecodeContext *s = avctx->priv_data;
-    int i;
+    int i, inter, plane;
+    int c_width;
+    int c_height;
+    int y_superblock_count;
+    int c_superblock_count;
+
+    if (avctx->codec_tag == MKTAG('V','P','3','0'))
+        s->version = 0;
+    else
+        s->version = 1;
 
     s->avctx = avctx;
-    s->width = avctx->width;
-    s->height = avctx->height;
+    s->width = (avctx->width + 15) & 0xFFFFFFF0;
+    s->height = (avctx->height + 15) & 0xFFFFFFF0;
     avctx->pix_fmt = PIX_FMT_YUV420P;
     avctx->has_b_frames = 0;
+    if(avctx->idct_algo==FF_IDCT_AUTO)
+        avctx->idct_algo=FF_IDCT_VP3;
     dsputil_init(&s->dsp, avctx);
 
+    ff_init_scantable(s->dsp.idct_permutation, &s->scantable, ff_zigzag_direct);
+
     /* initialize to an impossible value which will force a recalculation
      * in the first frame decode */
     s->quality_index = -1;
 
-    s->superblock_width = (s->width + 31) / 32;
-    s->superblock_height = (s->height + 31) / 32;
-    s->superblock_count = s->superblock_width * s->superblock_height * 3 / 2;
-    s->u_superblock_start = s->superblock_width * s->superblock_height;
-    s->v_superblock_start = s->superblock_width * s->superblock_height * 5 / 4;
+    s->y_superblock_width = (s->width + 31) / 32;
+    s->y_superblock_height = (s->height + 31) / 32;
+    y_superblock_count = s->y_superblock_width * s->y_superblock_height;
+
+    /* work out the dimensions for the C planes */
+    c_width = s->width / 2;
+    c_height = s->height / 2;
+    s->c_superblock_width = (c_width + 31) / 32;
+    s->c_superblock_height = (c_height + 31) / 32;
+    c_superblock_count = s->c_superblock_width * s->c_superblock_height;
+
+    s->superblock_count = y_superblock_count + (c_superblock_count * 2);
+    s->u_superblock_start = y_superblock_count;
+    s->v_superblock_start = s->u_superblock_start + c_superblock_count;
     s->superblock_coding = av_malloc(s->superblock_count);
 
     s->macroblock_width = (s->width + 15) / 16;
@@ -2128,69 +1990,136 @@ static int vp3_decode_init(AVCodecContext *avctx)
 
     /* fragment count covers all 8x8 blocks for all 3 planes */
     s->fragment_count = s->fragment_width * s->fragment_height * 3 / 2;
-    s->u_fragment_start = s->fragment_width * s->fragment_height;
-    s->v_fragment_start = s->fragment_width * s->fragment_height * 5 / 4;
-
-    debug_init("  width: %d x %d\n", s->width, s->height);
-    debug_init("  superblocks: %d x %d, %d total\n",
-        s->superblock_width, s->superblock_height, s->superblock_count);
+    s->fragment_start[1] = s->fragment_width * s->fragment_height;
+    s->fragment_start[2] = s->fragment_width * s->fragment_height * 5 / 4;
+
+    debug_init("  Y plane: %d x %d\n", s->width, s->height);
+    debug_init("  C plane: %d x %d\n", c_width, c_height);
+    debug_init("  Y superblocks: %d x %d, %d total\n",
+        s->y_superblock_width, s->y_superblock_height, y_superblock_count);
+    debug_init("  C superblocks: %d x %d, %d total\n",
+        s->c_superblock_width, s->c_superblock_height, c_superblock_count);
+    debug_init("  total superblocks = %d, U starts @ %d, V starts @ %d\n",
+        s->superblock_count, s->u_superblock_start, s->v_superblock_start);
     debug_init("  macroblocks: %d x %d, %d total\n",
         s->macroblock_width, s->macroblock_height, s->macroblock_count);
     debug_init("  %d fragments, %d x %d, u starts @ %d, v starts @ %d\n",
         s->fragment_count,
         s->fragment_width,
         s->fragment_height,
-        s->u_fragment_start,
-        s->v_fragment_start);
+        s->fragment_start[1],
+        s->fragment_start[2]);
 
     s->all_fragments = av_malloc(s->fragment_count * sizeof(Vp3Fragment));
+    s->coeffs = av_malloc(s->fragment_count * sizeof(Coeff) * 65);
     s->coded_fragment_list = av_malloc(s->fragment_count * sizeof(int));
     s->pixel_addresses_inited = 0;
 
-    /* init VLC tables */
-    for (i = 0; i < 16; i++) {
-
-        /* Dc histograms */
-        init_vlc(&s->dc_vlc[i], 5, 32,
-            &dc_bias[i][0][1], 4, 2,
-            &dc_bias[i][0][0], 4, 2);
-
-        /* level 1 AC histograms */
-        init_vlc(&s->ac_vlc_1[i], 5, 32,
-            &ac_bias_0[i][0][1], 4, 2,
-            &ac_bias_0[i][0][0], 4, 2);
-
-        /* level 2 AC histograms */
-        init_vlc(&s->ac_vlc_2[i], 5, 32,
-            &ac_bias_1[i][0][1], 4, 2,
-            &ac_bias_1[i][0][0], 4, 2);
-
-        /* level 3 AC histograms */
-        init_vlc(&s->ac_vlc_3[i], 5, 32,
-            &ac_bias_2[i][0][1], 4, 2,
-            &ac_bias_2[i][0][0], 4, 2);
-
-        /* level 4 AC histograms */
-        init_vlc(&s->ac_vlc_4[i], 5, 32,
-            &ac_bias_3[i][0][1], 4, 2,
-            &ac_bias_3[i][0][0], 4, 2);
+    if (!s->theora_tables)
+    {
+        for (i = 0; i < 64; i++) {
+            s->coded_dc_scale_factor[i] = vp31_dc_scale_factor[i];
+            s->coded_ac_scale_factor[i] = vp31_ac_scale_factor[i];
+            s->base_matrix[0][i] = vp31_intra_y_dequant[i];
+            s->base_matrix[1][i] = vp31_intra_c_dequant[i];
+            s->base_matrix[2][i] = vp31_inter_dequant[i];
+            s->filter_limit_values[i] = vp31_filter_limit_values[i];
+        }
+
+        for(inter=0; inter<2; inter++){
+            for(plane=0; plane<3; plane++){
+                s->qr_count[inter][plane]= 1;
+                s->qr_size [inter][plane][0]= 63;
+                s->qr_base [inter][plane][0]=
+                s->qr_base [inter][plane][1]= 2*inter + (!!plane)*!inter;
+            }
+        }
+
+        /* init VLC tables */
+        for (i = 0; i < 16; i++) {
+
+            /* DC histograms */
+            init_vlc(&s->dc_vlc[i], 5, 32,
+                &dc_bias[i][0][1], 4, 2,
+                &dc_bias[i][0][0], 4, 2, 0);
+
+            /* group 1 AC histograms */
+            init_vlc(&s->ac_vlc_1[i], 5, 32,
+                &ac_bias_0[i][0][1], 4, 2,
+                &ac_bias_0[i][0][0], 4, 2, 0);
+
+            /* group 2 AC histograms */
+            init_vlc(&s->ac_vlc_2[i], 5, 32,
+                &ac_bias_1[i][0][1], 4, 2,
+                &ac_bias_1[i][0][0], 4, 2, 0);
+
+            /* group 3 AC histograms */
+            init_vlc(&s->ac_vlc_3[i], 5, 32,
+                &ac_bias_2[i][0][1], 4, 2,
+                &ac_bias_2[i][0][0], 4, 2, 0);
+
+            /* group 4 AC histograms */
+            init_vlc(&s->ac_vlc_4[i], 5, 32,
+                &ac_bias_3[i][0][1], 4, 2,
+                &ac_bias_3[i][0][0], 4, 2, 0);
+        }
+    } else {
+        for (i = 0; i < 16; i++) {
+
+            /* DC histograms */
+            init_vlc(&s->dc_vlc[i], 5, 32,
+                &s->huffman_table[i][0][1], 4, 2,
+                &s->huffman_table[i][0][0], 4, 2, 0);
+
+            /* group 1 AC histograms */
+            init_vlc(&s->ac_vlc_1[i], 5, 32,
+                &s->huffman_table[i+16][0][1], 4, 2,
+                &s->huffman_table[i+16][0][0], 4, 2, 0);
+
+            /* group 2 AC histograms */
+            init_vlc(&s->ac_vlc_2[i], 5, 32,
+                &s->huffman_table[i+16*2][0][1], 4, 2,
+                &s->huffman_table[i+16*2][0][0], 4, 2, 0);
+
+            /* group 3 AC histograms */
+            init_vlc(&s->ac_vlc_3[i], 5, 32,
+                &s->huffman_table[i+16*3][0][1], 4, 2,
+                &s->huffman_table[i+16*3][0][0], 4, 2, 0);
+
+            /* group 4 AC histograms */
+            init_vlc(&s->ac_vlc_4[i], 5, 32,
+                &s->huffman_table[i+16*4][0][1], 4, 2,
+                &s->huffman_table[i+16*4][0][0], 4, 2, 0);
+        }
     }
 
-    /* build quantization table */
-    for (i = 0; i < 64; i++)
-        quant_index[dequant_index[i]] = i;
+    init_vlc(&s->superblock_run_length_vlc, 6, 34,
+        &superblock_run_length_vlc_table[0][1], 4, 2,
+        &superblock_run_length_vlc_table[0][0], 4, 2, 0);
+
+    init_vlc(&s->fragment_run_length_vlc, 5, 30,
+        &fragment_run_length_vlc_table[0][1], 4, 2,
+        &fragment_run_length_vlc_table[0][0], 4, 2, 0);
+
+    init_vlc(&s->mode_code_vlc, 3, 8,
+        &mode_code_vlc_table[0][1], 2, 1,
+        &mode_code_vlc_table[0][0], 2, 1, 0);
+
+    init_vlc(&s->motion_vector_vlc, 6, 63,
+        &motion_vector_vlc_table[0][1], 2, 1,
+        &motion_vector_vlc_table[0][0], 2, 1, 0);
 
     /* work out the block mapping tables */
     s->superblock_fragments = av_malloc(s->superblock_count * 16 * sizeof(int));
     s->superblock_macroblocks = av_malloc(s->superblock_count * 4 * sizeof(int));
     s->macroblock_fragments = av_malloc(s->macroblock_count * 6 * sizeof(int));
-    s->macroblock_coded = av_malloc(s->macroblock_count + 1);
+    s->macroblock_coding = av_malloc(s->macroblock_count + 1);
     init_block_mapping(s);
 
-    /* make sure that frames are available to be freed on the first decode */
-    if(avctx->get_buffer(avctx, &s->golden_frame) < 0) {
-        printf("vp3: get_buffer() failed\n");
-        return -1;
+    for (i = 0; i < 3; i++) {
+        s->current_frame.data[i] = NULL;
+        s->last_frame.data[i] = NULL;
+        s->golden_frame.data[i] = NULL;
     }
 
     return 0;
@@ -2199,91 +2128,205 @@ static int vp3_decode_init(AVCodecContext *avctx)
 /*
  * This is the ffmpeg/libavcodec API frame decode function.
  */
-static int vp3_decode_frame(AVCodecContext *avctx, 
+static int vp3_decode_frame(AVCodecContext *avctx,
                             void *data, int *data_size,
                             uint8_t *buf, int buf_size)
 {
     Vp3DecodeContext *s = avctx->priv_data;
     GetBitContext gb;
     static int counter = 0;
-
-    *data_size = 0;
+    int i;
 
     init_get_bits(&gb, buf, buf_size * 8);
 
-    s->keyframe = get_bits(&gb, 1);
-    s->keyframe ^= 1;
-    skip_bits(&gb, 1);
+    if (s->theora && get_bits1(&gb))
+    {
+#if 1
+        av_log(avctx, AV_LOG_ERROR, "Header packet passed to frame decoder, skipping\n");
+        return -1;
+#else
+        int ptype = get_bits(&gb, 7);
+
+        skip_bits(&gb, 6*8); /* "theora" */
+
+        switch(ptype)
+        {
+            case 1:
+                theora_decode_comments(avctx, &gb);
+                break;
+            case 2:
+                theora_decode_tables(avctx, &gb);
+                    init_dequantizer(s);
+                break;
+            default:
+                av_log(avctx, AV_LOG_ERROR, "Unknown Theora config packet: %d\n", ptype);
+        }
+        return buf_size;
+#endif
+    }
+
+    s->keyframe = !get_bits1(&gb);
+    if (!s->theora)
+        skip_bits(&gb, 1);
     s->last_quality_index = s->quality_index;
-    s->quality_index = get_bits(&gb, 6);
-    if (s->quality_index != s->last_quality_index)
-        init_dequantizer(s);
 
-    debug_vp3(" VP3 frame #%d: Q index = %d", counter, s->quality_index);
+    s->nqis=0;
+    do{
+        s->qis[s->nqis++]= get_bits(&gb, 6);
+    } while(s->theora >= 0x030200 && s->nqis<3 && get_bits1(&gb));
+
+    s->quality_index= s->qis[0];
+
+    if (s->avctx->debug & FF_DEBUG_PICT_INFO)
+        av_log(s->avctx, AV_LOG_INFO, " VP3 %sframe #%d: Q index = %d\n",
+            s->keyframe?"key":"", counter, s->quality_index);
     counter++;
 
+    if (s->quality_index != s->last_quality_index) {
+        init_dequantizer(s);
+        init_loop_filter(s);
+    }
+
     if (s->keyframe) {
-        /* release the previous golden frame and get a new one */
-        avctx->release_buffer(avctx, &s->golden_frame);
+        if (!s->theora)
+        {
+            skip_bits(&gb, 4); /* width code */
+            skip_bits(&gb, 4); /* height code */
+            if (s->version)
+            {
+                s->version = get_bits(&gb, 5);
+                if (counter == 1)
+                    av_log(s->avctx, AV_LOG_DEBUG, "VP version: %d\n", s->version);
+            }
+        }
+        if (s->version || s->theora)
+        {
+                if (get_bits1(&gb))
+                    av_log(s->avctx, AV_LOG_ERROR, "Warning, unsupported keyframe coding type?!\n");
+            skip_bits(&gb, 2); /* reserved? */
+        }
+
+        if (s->last_frame.data[0] == s->golden_frame.data[0]) {
+            if (s->golden_frame.data[0])
+                avctx->release_buffer(avctx, &s->golden_frame);
+            s->last_frame= s->golden_frame; /* ensure that we catch any access to this released frame */
+        } else {
+            if (s->golden_frame.data[0])
+                avctx->release_buffer(avctx, &s->golden_frame);
+            if (s->last_frame.data[0])
+                avctx->release_buffer(avctx, &s->last_frame);
+        }
 
-        s->golden_frame.reference = 0;
+        s->golden_frame.reference = 3;
         if(avctx->get_buffer(avctx, &s->golden_frame) < 0) {
-            printf("vp3: get_buffer() failed\n");
+            av_log(s->avctx, AV_LOG_ERROR, "vp3: get_buffer() failed\n");
             return -1;
         }
 
         /* golden frame is also the current frame */
-        memcpy(&s->current_frame, &s->golden_frame, sizeof(AVFrame));
+        s->current_frame= s->golden_frame;
 
         /* time to figure out pixel addresses? */
         if (!s->pixel_addresses_inited)
-            vp3_calculate_pixel_addresses(s);
-
+        {
+            if (!s->flipped_image)
+                vp3_calculate_pixel_addresses(s);
+            else
+                theora_calculate_pixel_addresses(s);
+            s->pixel_addresses_inited = 1;
+        }
     } else {
-
         /* allocate a new current frame */
-        s->current_frame.reference = 0;
+        s->current_frame.reference = 3;
+        if (!s->pixel_addresses_inited) {
+            av_log(s->avctx, AV_LOG_ERROR, "vp3: first frame not a keyframe\n");
+            return -1;
+        }
         if(avctx->get_buffer(avctx, &s->current_frame) < 0) {
-            printf("vp3: get_buffer() failed\n");
+            av_log(s->avctx, AV_LOG_ERROR, "vp3: get_buffer() failed\n");
             return -1;
         }
-
     }
 
-    if (s->keyframe) {
-      debug_vp3(", keyframe\n");
-      /* skip the other 2 header bytes for now */
-      skip_bits(&gb, 16);
-    } else
-      debug_vp3("\n");
+    s->current_frame.qscale_table= s->qscale_table; //FIXME allocate individual tables per AVFrame
+    s->current_frame.qstride= 0;
 
+    {START_TIMER
     init_frame(s, &gb);
+    STOP_TIMER("init_frame")}
+
+#if KEYFRAMES_ONLY
+if (!s->keyframe) {
 
-    unpack_superblocks(s, &gb);
-    unpack_modes(s, &gb);
-    unpack_vectors(s, &gb);
-    unpack_dct_coeffs(s, &gb);
+    memcpy(s->current_frame.data[0], s->golden_frame.data[0],
+        s->current_frame.linesize[0] * s->height);
+    memcpy(s->current_frame.data[1], s->golden_frame.data[1],
+        s->current_frame.linesize[1] * s->height / 2);
+    memcpy(s->current_frame.data[2], s->golden_frame.data[2],
+        s->current_frame.linesize[2] * s->height / 2);
+
+} else {
+#endif
+
+    {START_TIMER
+    if (unpack_superblocks(s, &gb)){
+        av_log(s->avctx, AV_LOG_ERROR, "error in unpack_superblocks\n");
+        return -1;
+    }
+    STOP_TIMER("unpack_superblocks")}
+    {START_TIMER
+    if (unpack_modes(s, &gb)){
+        av_log(s->avctx, AV_LOG_ERROR, "error in unpack_modes\n");
+        return -1;
+    }
+    STOP_TIMER("unpack_modes")}
+    {START_TIMER
+    if (unpack_vectors(s, &gb)){
+        av_log(s->avctx, AV_LOG_ERROR, "error in unpack_vectors\n");
+        return -1;
+    }
+    STOP_TIMER("unpack_vectors")}
+    {START_TIMER
+    if (unpack_dct_coeffs(s, &gb)){
+        av_log(s->avctx, AV_LOG_ERROR, "error in unpack_dct_coeffs\n");
+        return -1;
+    }
+    STOP_TIMER("unpack_dct_coeffs")}
+    {START_TIMER
 
     reverse_dc_prediction(s, 0, s->fragment_width, s->fragment_height);
-    reverse_dc_prediction(s, s->u_fragment_start,
-        s->fragment_width / 2, s->fragment_height / 2);
-    reverse_dc_prediction(s, s->v_fragment_start,
-        s->fragment_width / 2, s->fragment_height / 2);
+    if ((avctx->flags & CODEC_FLAG_GRAY) == 0) {
+        reverse_dc_prediction(s, s->fragment_start[1],
+            s->fragment_width / 2, s->fragment_height / 2);
+        reverse_dc_prediction(s, s->fragment_start[2],
+            s->fragment_width / 2, s->fragment_height / 2);
+    }
+    STOP_TIMER("reverse_dc_prediction")}
+    {START_TIMER
 
-    render_fragments(s, 0, s->fragment_width, s->fragment_height, 0);
-    render_fragments(s, s->u_fragment_start,
-        s->fragment_width / 2, s->fragment_height / 2, 1);
-    render_fragments(s, s->v_fragment_start,
-        s->fragment_width / 2, s->fragment_height / 2, 2);
+    for (i = 0; i < s->macroblock_height; i++)
+        render_slice(s, i);
+    STOP_TIMER("render_fragments")}
+
+    {START_TIMER
+    apply_loop_filter(s);
+    STOP_TIMER("apply_loop_filter")}
+#if KEYFRAMES_ONLY
+}
+#endif
 
     *data_size=sizeof(AVFrame);
     *(AVFrame*)data= s->current_frame;
 
-    /* release the last frame, if it was allocated */
-    avctx->release_buffer(avctx, &s->last_frame);
+    /* release the last frame, if it is allocated and if it is not the
+     * golden frame */
+    if ((s->last_frame.data[0]) &&
+        (s->last_frame.data[0] != s->golden_frame.data[0]))
+        avctx->release_buffer(avctx, &s->last_frame);
 
     /* shuffle frames (last = current) */
-    memcpy(&s->last_frame, &s->current_frame, sizeof(AVFrame));
+    s->last_frame= s->current_frame;
+    s->current_frame.data[0]= NULL; /* ensure that we catch any access to this released frame */
 
     return buf_size;
 }
@@ -2296,17 +2339,294 @@ static int vp3_decode_end(AVCodecContext *avctx)
     Vp3DecodeContext *s = avctx->priv_data;
 
     av_free(s->all_fragments);
+    av_free(s->coeffs);
     av_free(s->coded_fragment_list);
     av_free(s->superblock_fragments);
     av_free(s->superblock_macroblocks);
     av_free(s->macroblock_fragments);
-    av_free(s->macroblock_coded);
+    av_free(s->macroblock_coding);
 
     /* release all frames */
-    avctx->release_buffer(avctx, &s->golden_frame);
-    avctx->release_buffer(avctx, &s->last_frame);
-    avctx->release_buffer(avctx, &s->current_frame);
+    if (s->golden_frame.data[0] && s->golden_frame.data[0] != s->last_frame.data[0])
+        avctx->release_buffer(avctx, &s->golden_frame);
+    if (s->last_frame.data[0])
+        avctx->release_buffer(avctx, &s->last_frame);
+    /* no need to release the current_frame since it will always be pointing
+     * to the same frame as either the golden or last frame */
+
+    return 0;
+}
+
+static int read_huffman_tree(AVCodecContext *avctx, GetBitContext *gb)
+{
+    Vp3DecodeContext *s = avctx->priv_data;
 
+    if (get_bits(gb, 1)) {
+        int token;
+        if (s->entries >= 32) { /* overflow */
+            av_log(avctx, AV_LOG_ERROR, "huffman tree overflow\n");
+            return -1;
+        }
+        token = get_bits(gb, 5);
+        //av_log(avctx, AV_LOG_DEBUG, "hti %d hbits %x token %d entry : %d size %d\n", s->hti, s->hbits, token, s->entries, s->huff_code_size);
+        s->huffman_table[s->hti][token][0] = s->hbits;
+        s->huffman_table[s->hti][token][1] = s->huff_code_size;
+        s->entries++;
+    }
+    else {
+        if (s->huff_code_size >= 32) {/* overflow */
+            av_log(avctx, AV_LOG_ERROR, "huffman tree overflow\n");
+            return -1;
+        }
+        s->huff_code_size++;
+        s->hbits <<= 1;
+        read_huffman_tree(avctx, gb);
+        s->hbits |= 1;
+        read_huffman_tree(avctx, gb);
+        s->hbits >>= 1;
+        s->huff_code_size--;
+    }
+    return 0;
+}
+
+static int theora_decode_header(AVCodecContext *avctx, GetBitContext *gb)
+{
+    Vp3DecodeContext *s = avctx->priv_data;
+
+    s->theora = get_bits_long(gb, 24);
+    av_log(avctx, AV_LOG_INFO, "Theora bitstream version %X\n", s->theora);
+
+    /* 3.2.0 aka alpha3 has the same frame orientation as original vp3 */
+    /* but previous versions have the image flipped relative to vp3 */
+    if (s->theora < 0x030200)
+    {
+        s->flipped_image = 1;
+        av_log(avctx, AV_LOG_DEBUG, "Old (<alpha3) Theora bitstream, flipped image\n");
+    }
+
+    s->width = get_bits(gb, 16) << 4;
+    s->height = get_bits(gb, 16) << 4;
+
+    if(avcodec_check_dimensions(avctx, s->width, s->height)){
+        av_log(avctx, AV_LOG_ERROR, "Invalid dimensions (%dx%d)\n", s->width, s->height);
+        s->width= s->height= 0;
+        return -1;
+    }
+
+    if (s->theora >= 0x030400)
+    {
+        skip_bits(gb, 32); /* total number of superblocks in a frame */
+        // fixme, the next field is 36bits long
+        skip_bits(gb, 32); /* total number of blocks in a frame */
+        skip_bits(gb, 4); /* total number of blocks in a frame */
+        skip_bits(gb, 32); /* total number of macroblocks in a frame */
+
+        skip_bits(gb, 24); /* frame width */
+        skip_bits(gb, 24); /* frame height */
+    }
+    else
+    {
+        skip_bits(gb, 24); /* frame width */
+        skip_bits(gb, 24); /* frame height */
+    }
+
+  if (s->theora >= 0x030200) {
+    skip_bits(gb, 8); /* offset x */
+    skip_bits(gb, 8); /* offset y */
+  }
+
+    skip_bits(gb, 32); /* fps numerator */
+    skip_bits(gb, 32); /* fps denumerator */
+    skip_bits(gb, 24); /* aspect numerator */
+    skip_bits(gb, 24); /* aspect denumerator */
+
+    if (s->theora < 0x030200)
+        skip_bits(gb, 5); /* keyframe frequency force */
+    skip_bits(gb, 8); /* colorspace */
+    if (s->theora >= 0x030400)
+        skip_bits(gb, 2); /* pixel format: 420,res,422,444 */
+    skip_bits(gb, 24); /* bitrate */
+
+    skip_bits(gb, 6); /* quality hint */
+
+    if (s->theora >= 0x030200)
+    {
+        skip_bits(gb, 5); /* keyframe frequency force */
+
+        if (s->theora < 0x030400)
+            skip_bits(gb, 5); /* spare bits */
+    }
+
+//    align_get_bits(gb);
+
+    avctx->width = s->width;
+    avctx->height = s->height;
+
+    return 0;
+}
+
+static int theora_decode_tables(AVCodecContext *avctx, GetBitContext *gb)
+{
+    Vp3DecodeContext *s = avctx->priv_data;
+    int i, n, matrices, inter, plane;
+
+    if (s->theora >= 0x030200) {
+        n = get_bits(gb, 3);
+        /* loop filter limit values table */
+        for (i = 0; i < 64; i++)
+            s->filter_limit_values[i] = get_bits(gb, n);
+    }
+
+    if (s->theora >= 0x030200)
+        n = get_bits(gb, 4) + 1;
+    else
+        n = 16;
+    /* quality threshold table */
+    for (i = 0; i < 64; i++)
+        s->coded_ac_scale_factor[i] = get_bits(gb, n);
+
+    if (s->theora >= 0x030200)
+        n = get_bits(gb, 4) + 1;
+    else
+        n = 16;
+    /* dc scale factor table */
+    for (i = 0; i < 64; i++)
+        s->coded_dc_scale_factor[i] = get_bits(gb, n);
+
+    if (s->theora >= 0x030200)
+        matrices = get_bits(gb, 9) + 1;
+    else
+        matrices = 3;
+
+    if(matrices > 384){
+        av_log(avctx, AV_LOG_ERROR, "invalid number of base matrixes\n");
+        return -1;
+    }
+
+    for(n=0; n<matrices; n++){
+        for (i = 0; i < 64; i++)
+            s->base_matrix[n][i]= get_bits(gb, 8);
+    }
+
+    for (inter = 0; inter <= 1; inter++) {
+        for (plane = 0; plane <= 2; plane++) {
+            int newqr= 1;
+            if (inter || plane > 0)
+                newqr = get_bits(gb, 1);
+            if (!newqr) {
+                int qtj, plj;
+                if(inter && get_bits(gb, 1)){
+                    qtj = 0;
+                    plj = plane;
+                }else{
+                    qtj= (3*inter + plane - 1) / 3;
+                    plj= (plane + 2) % 3;
+                }
+                s->qr_count[inter][plane]= s->qr_count[qtj][plj];
+                memcpy(s->qr_size[inter][plane], s->qr_size[qtj][plj], sizeof(s->qr_size[0][0]));
+                memcpy(s->qr_base[inter][plane], s->qr_base[qtj][plj], sizeof(s->qr_base[0][0]));
+            } else {
+                int qri= 0;
+                int qi = 0;
+
+                for(;;){
+                    i= get_bits(gb, av_log2(matrices-1)+1);
+                    if(i>= matrices){
+                        av_log(avctx, AV_LOG_ERROR, "invalid base matrix index\n");
+                        return -1;
+                    }
+                    s->qr_base[inter][plane][qri]= i;
+                    if(qi >= 63)
+                        break;
+                    i = get_bits(gb, av_log2(63-qi)+1) + 1;
+                    s->qr_size[inter][plane][qri++]= i;
+                    qi += i;
+                }
+
+                if (qi > 63) {
+                    av_log(avctx, AV_LOG_ERROR, "invalid qi %d > 63\n", qi);
+                    return -1;
+                }
+                s->qr_count[inter][plane]= qri;
+            }
+        }
+    }
+
+    /* Huffman tables */
+    for (s->hti = 0; s->hti < 80; s->hti++) {
+        s->entries = 0;
+        s->huff_code_size = 1;
+        if (!get_bits(gb, 1)) {
+            s->hbits = 0;
+            read_huffman_tree(avctx, gb);
+            s->hbits = 1;
+            read_huffman_tree(avctx, gb);
+        }
+    }
+
+    s->theora_tables = 1;
+
+    return 0;
+}
+
+static int theora_decode_init(AVCodecContext *avctx)
+{
+    Vp3DecodeContext *s = avctx->priv_data;
+    GetBitContext gb;
+    int ptype;
+    uint8_t *p= avctx->extradata;
+    int op_bytes, i;
+
+    s->theora = 1;
+
+    if (!avctx->extradata_size)
+    {
+        av_log(avctx, AV_LOG_ERROR, "Missing extradata!\n");
+        return -1;
+    }
+
+  for(i=0;i<3;i++) {
+    op_bytes = *(p++)<<8;
+    op_bytes += *(p++);
+
+    init_get_bits(&gb, p, op_bytes);
+    p += op_bytes;
+
+    ptype = get_bits(&gb, 8);
+    debug_vp3("Theora headerpacket type: %x\n", ptype);
+
+     if (!(ptype & 0x80))
+     {
+        av_log(avctx, AV_LOG_ERROR, "Invalid extradata!\n");
+//        return -1;
+     }
+
+    // FIXME: check for this aswell
+    skip_bits(&gb, 6*8); /* "theora" */
+
+    switch(ptype)
+    {
+        case 0x80:
+            theora_decode_header(avctx, &gb);
+                break;
+        case 0x81:
+// FIXME: is this needed? it breaks sometimes
+//            theora_decode_comments(avctx, gb);
+            break;
+        case 0x82:
+            theora_decode_tables(avctx, &gb);
+            break;
+        default:
+            av_log(avctx, AV_LOG_ERROR, "Unknown Theora config packet: %d\n", ptype&~0x80);
+            break;
+    }
+    if(8*op_bytes != get_bits_count(&gb))
+        av_log(avctx, AV_LOG_ERROR, "%d bits left in packet %X\n", 8*op_bytes - get_bits_count(&gb), ptype);
+    if (s->theora < 0x030200)
+        break;
+  }
+
+    vp3_decode_init(avctx);
     return 0;
 }
 
@@ -2322,3 +2642,18 @@ AVCodec vp3_decoder = {
     0,
     NULL
 };
+
+#ifndef CONFIG_LIBTHEORA
+AVCodec theora_decoder = {
+    "theora",
+    CODEC_TYPE_VIDEO,
+    CODEC_ID_THEORA,
+    sizeof(Vp3DecodeContext),
+    theora_decode_init,
+    NULL,
+    vp3_decode_end,
+    vp3_decode_frame,
+    0,
+    NULL
+};
+#endif