Diff of /trunk/xvidcore/src/decoder.c

-revision 100, Thu Apr  4 13:58:18 2002 UTC
+revision 2178, Wed Nov  6 10:33:01 2019 UTC
 Line 1
- /**************************************************************************
+ /*****************************************************************************
   *
   *      XVID MPEG-4 VIDEO CODEC
-  *      decoder main
+  *  - Decoder Module -
   *
-  *      This program is an implementation of a part of one or more MPEG-4
+  *  Copyright(C) 2002      MinChen <chenm001@163.com>
-  *      Video tools as specified in ISO/IEC 14496-2 standard.  Those intending
+  *               2002-2010 Peter Ross <pross@xvid.org>
-  *      to use this software module in hardware or software products are
-  *      advised that its use may infringe existing patents or copyrights, and
-  *      any such use would be at such party's own risk.  The original
-  *      developer of this software module and his/her company, and subsequent
-  *      editors and their companies, will have no liability for use of this
-  *      software or modifications or derivatives thereof.
   *
-  *      This program is xvid_free software; you can redistribute it and/or modify
+  *  This program is free software ; you can redistribute it and/or modify
   *      it under the terms of the GNU General Public License as published by
-  *      the xvid_free Software Foundation; either version 2 of the License, or
+  *  the Free Software Foundation ; either version 2 of the License, or
   *      (at your option) any later version.
   *
   *      This program is distributed in the hope that it will be useful,
-Line 23
+Line 17
   *      GNU General Public License for more details.
   *
   *      You should have received a copy of the GNU General Public License
-  *      along with this program; if not, write to the xvid_free Software
+  *  along with this program ; if not, write to the Free Software
-  *      Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
   *
-  *************************************************************************/
+  * $Id$
- /**************************************************************************
-  *
-  *      History:
   *
-  *  29.03.2002  interlacing fix - compensated block wasn't being used when
+  ****************************************************************************/
-  *              reconstructing blocks, thus artifacts
-  *              interlacing speedup - used transfers to re-interlace
-  *              interlaced decoding should be as fast as progressive now
-  *  26.03.2002  interlacing support - moved transfers outside decode loop
-  *      26.12.2001      decoder_mbinter: dequant/idct moved within if(coded) block
-  *      22.12.2001      block based interpolation
-  *      01.12.2001      inital version; (c)2001 peter ross <pross@cs.rmit.edu.au>
-  *
-  *************************************************************************/
+ #include <stdio.h>
  #include <stdlib.h>
- #include <string.h>  // memset
+ #include <string.h>
+ #ifdef BFRAMES_DEC_DEBUG
+ #define BFRAMES_DEC
+ #endif
  #include "xvid.h"
  #include "portab.h"
+ #include "global.h"
  #include "decoder.h"
  #include "bitstream/bitstream.h"
  #include "bitstream/mbcoding.h"
- #include "quant/quant_h263.h"
+ #include "quant/quant.h"
- #include "quant/quant_mpeg4.h"
+ #include "quant/quant_matrix.h"
  #include "dct/idct.h"
  #include "dct/fdct.h"
  #include "utils/mem_transfer.h"
  #include "image/interpolate8x8.h"
+ #include "image/font.h"
+ #include "image/qpel.h"
  #include "bitstream/mbcoding.h"
  #include "prediction/mbprediction.h"
  #include "utils/timer.h"
  #include "utils/emms.h"
+ #include "motion/motion.h"
+ #include "motion/gmc.h"
  #include "image/image.h"
  #include "image/colorspace.h"
+ #include "image/postprocessing.h"
  #include "utils/mem_align.h"
- int decoder_create(XVID_DEC_PARAM * param)
+ #define DIV2ROUND(n)  (((n)>>1)|((n)&1))
- {
+ #define DIV2(n)       ((n)>>1)
-         DECODER * dec;
+ #define DIVUVMOV(n) (((n) >> 1) + roundtab_79[(n) & 0x3]) //
-         dec = xvid_malloc(sizeof(DECODER), CACHE_LINE);
+ static int
-         if (dec == NULL)
+ decoder_resize(DECODER * dec)
          {
-                 return XVID_ERR_MEMORY;
+         /* free existing */
-         }
+         image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
-         param->handle = dec;
+         image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
+         image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
+         image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
+         image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
+         image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
+   image_null(&dec->cur);
+   image_null(&dec->refn[0]);
+   image_null(&dec->refn[1]);
+   image_null(&dec->tmp);
+   image_null(&dec->qtmp);
+   image_null(&dec->gmc);
-         dec->width = param->width;
-         dec->height = param->height;
+   xvid_free(dec->last_mbs);
+   xvid_free(dec->mbs);
+   xvid_free(dec->qscale);
+   dec->last_mbs = NULL;
+   dec->mbs = NULL;
+   dec->qscale = NULL;
+         /* realloc */
          dec->mb_width = (dec->width + 15) / 16;
          dec->mb_height = (dec->height + 15) / 16;
          dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE;
          dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE;
-         if (image_create(&dec->cur, dec->edged_width, dec->edged_height))
+         if (   image_create(&dec->cur, dec->edged_width, dec->edged_height)
-         {
+             || image_create(&dec->refn[0], dec->edged_width, dec->edged_height)
+             || image_create(&dec->refn[1], dec->edged_width, dec->edged_height)         /* Support B-frame to reference last 2 frame */
+             || image_create(&dec->tmp, dec->edged_width, dec->edged_height)
+             || image_create(&dec->qtmp, dec->edged_width, dec->edged_height)
+       || image_create(&dec->gmc, dec->edged_width, dec->edged_height) )
+     goto memory_error;
+         dec->mbs =
+                 xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
+                                         CACHE_LINE);
+         if (dec->mbs == NULL)
+           goto memory_error;
+         memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
+         /* For skip MB flag */
+         dec->last_mbs =
+                 xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
+                                         CACHE_LINE);
+         if (dec->last_mbs == NULL)
+           goto memory_error;
+         memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
+         /* nothing happens if that fails */
+         dec->qscale =
+                 xvid_malloc(sizeof(int) * dec->mb_width * dec->mb_height, CACHE_LINE);
+         if (dec->qscale)
+                 memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height);
+         return 0;
+ memory_error:
+         /* Most structures were deallocated / nullifieded, so it should be safe */
+         /* decoder_destroy(dec) minus the write_timer */
+   xvid_free(dec->mbs);
+   image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
+   image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
+   image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
+   image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
+   image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
                  xvid_free(dec);
                  return XVID_ERR_MEMORY;
          }
-         if (image_create(&dec->refn, dec->edged_width, dec->edged_height))
+ int
+ decoder_create(xvid_dec_create_t * create)
          {
-                 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
+   int ret = 0;
-                 xvid_free(dec);
+   DECODER *dec;
+   if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */
+     return XVID_ERR_VERSION;
+   dec = xvid_malloc(sizeof(DECODER), CACHE_LINE);
+   if (dec == NULL) {
                  return XVID_ERR_MEMORY;
          }
-         dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE);
+   memset(dec, 0, sizeof(DECODER));
-         if (dec->mbs == NULL)
-         {
+   dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE);
-                 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
+   if (dec->mpeg_quant_matrices == NULL) {
                  xvid_free(dec);
                  return XVID_ERR_MEMORY;
          }
+   create->handle = dec;
+   dec->width = MAX(0, create->width);
+   dec->height = MAX(0, create->height);
+   dec->num_threads = MAX(0, create->num_threads);
+   image_null(&dec->cur);
+   image_null(&dec->refn[0]);
+   image_null(&dec->refn[1]);
+   image_null(&dec->tmp);
+   image_null(&dec->qtmp);
+   /* image based GMC */
+   image_null(&dec->gmc);
+   dec->mbs = NULL;
+   dec->last_mbs = NULL;
+   dec->qscale = NULL;
          init_timer();
+   init_postproc(&dec->postproc);
+   init_mpeg_matrix(dec->mpeg_quant_matrices);
-         return XVID_ERR_OK;
+   /* For B-frame support (used to save reference frame's time */
+   dec->frames = 0;
+   dec->time = dec->time_base = dec->last_time_base = 0;
+   dec->low_delay = 0;
+   dec->packed_mode = 0;
+   dec->time_inc_resolution = 1; /* until VOL header says otherwise */
+   dec->ver_id = 1;
+   if (create->fourcc == ((int)('X')|((int)('V')<<8)|
+                          ((int)('I')<<16)|((int)('D')<<24))) { /* XVID */
+     dec->bs_version = 0; /* Initially assume oldest xvid version */
+   }
+   else {
+         dec->bs_version = 0xffff; /* Initialize to very high value -> assume bugfree stream */
  }
+   dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
- int decoder_destroy(DECODER * dec)
+   ret = decoder_resize(dec);
+   if (ret == XVID_ERR_MEMORY) create->handle = NULL;
+   return ret;
+ }
+ int
+ decoder_destroy(DECODER * dec)
  {
+   xvid_free(dec->last_mbs);
          xvid_free(dec->mbs);
-         image_destroy(&dec->refn, dec->edged_width, dec->edged_height);
+   xvid_free(dec->qscale);
+   /* image based GMC */
+   image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
+   image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
+   image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
+   image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
+   image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
          image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
+   xvid_free(dec->mpeg_quant_matrices);
          xvid_free(dec);
          write_timer();
-         return XVID_ERR_OK;
+   return 0;
  }
+ static const int32_t dquant_table[4] = {
- static const int32_t dquant_table[4] =
- {
          -1, -2, 1, 2
  };
+ /* decode an intra macroblock */
- // decode an intra macroblock
+ static void
+ decoder_mbintra(DECODER * dec,
- void decoder_mbintra(DECODER * dec,
                       MACROBLOCK * pMB,
                       const uint32_t x_pos,
                       const uint32_t y_pos,
-Line 145
+Line 253
                       const uint32_t cbp,
                       Bitstream * bs,
                       const uint32_t quant,
-                      const uint32_t intra_dc_threshold)
+         const uint32_t intra_dc_threshold,
+         const unsigned int bound)
  {
          DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
-Line 155
+Line 264
          uint32_t stride2 = stride / 2;
          uint32_t next_block = stride * 8;
          uint32_t i;
-         uint32_t iQuant = pMB->quant;
+   uint32_t iQuant = MAX(1, pMB->quant);
          uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
          pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
          pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
          pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
-         memset(block, 0, 6*64*sizeof(int16_t));         // clear
+   memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */
-         for (i = 0; i < 6; i++)
+   for (i = 0; i < 6; i++) {
-         {
                  uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4);
                  int16_t predictors[8];
                  int start_coeff;
                  start_timer();
-                 predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i*64], iQuant, iDcScaler, predictors);
+     predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64],
-                 if (!acpred_flag)
+            iQuant, iDcScaler, predictors, bound);
-                 {
+     if (!acpred_flag) {
                          pMB->acpred_directions[i] = 0;
                  }
                  stop_prediction_timer();
-                 if (quant < intra_dc_threshold)
+     if (quant < intra_dc_threshold) {
-                 {
                          int dc_size;
                          int dc_dif;
                          dc_size = i < 4 ?  get_dc_size_lum(bs) : get_dc_size_chrom(bs);
                          dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ;
-                         if (dc_size > 8)
+       if (dc_size > 8) {
-                         {
+         BitstreamSkip(bs, 1); /* marker */
-                                 BitstreamSkip(bs, 1);           // marker
                          }
                          block[i*64 + 0] = dc_dif;
                          start_coeff = 1;
-                 }
-                 else
+       DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif);
-                 {
+     } else {
                          start_coeff = 0;
                  }
                  start_timer();
-                 if (cbp & (1 << (5-i)))                 // coded
+     if (cbp & (1 << (5 - i))) /* coded */
                  {
-                         get_intra_block(bs, &block[i*64], pMB->acpred_directions[i], start_coeff);
+       int direction = dec->alternate_vertical_scan ?
+: pMB->acpred_directions[i];
+       get_intra_block(bs, &block[i * 64], direction, start_coeff);
                  }
                  stop_coding_timer();
                  start_timer();
-                 add_acdc(pMB, i, &block[i*64], iDcScaler, predictors);
+     add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version);
                  stop_prediction_timer();
                  start_timer();
-                 if (dec->quant_type == 0)
+     if (dec->quant_type == 0) {
-                 {
+       dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
-                         dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler);
+     } else {
-                 }
+       dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
-                 else
-                 {
-                         dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler);
                  }
                  stop_iquant_timer();
                  start_timer();
-                 idct(&data[i*64]);
+     idct((short * const)&data[i * 64]);
                  stop_idct_timer();
          }
-         if (pMB->field_dct)
+   if (dec->interlacing && pMB->field_dct) {
-         {
                  next_block = stride;
                  stride *= 2;
          }
-Line 242
+Line 348
          stop_transfer_timer();
  }
+ static void
+ decoder_mb_decode(DECODER * dec,
+         const uint32_t cbp,
+         Bitstream * bs,
+         uint8_t * pY_Cur,
+         uint8_t * pU_Cur,
+         uint8_t * pV_Cur,
+         const MACROBLOCK * pMB)
+ {
+   DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE);
+   int stride = dec->edged_width;
+   int i;
+   const uint32_t iQuant = MAX(1, pMB->quant);
+   const int direction = dec->alternate_vertical_scan ? 2 : 0;
+   typedef void (*get_inter_block_function_t)(
+       Bitstream * bs,
+       int16_t * block,
+       int direction,
+       const int quant,
+       const uint16_t *matrix);
+   typedef void (*add_residual_function_t)(
+       uint8_t *predicted_block,
+       const int16_t *residual,
+       int stride);
+   const get_inter_block_function_t get_inter_block = (dec->quant_type == 0)
+     ? (get_inter_block_function_t)get_inter_block_h263
+     : (get_inter_block_function_t)get_inter_block_mpeg;
+   uint8_t *dst[6];
+   int strides[6];
+   if (dec->interlacing && pMB->field_dct) {
+     dst[0] = pY_Cur;
+     dst[1] = pY_Cur + 8;
+     dst[2] = pY_Cur + stride;
+     dst[3] = dst[2] + 8;
+     dst[4] = pU_Cur;
+     dst[5] = pV_Cur;
+     strides[0] = strides[1] = strides[2] = strides[3] = stride*2;
+     strides[4] = stride/2;
+     strides[5] = stride/2;
+   } else {
+     dst[0] = pY_Cur;
+     dst[1] = pY_Cur + 8;
+     dst[2] = pY_Cur + 8*stride;
+     dst[3] = dst[2] + 8;
+     dst[4] = pU_Cur;
+     dst[5] = pV_Cur;
+     strides[0] = strides[1] = strides[2] = strides[3] = stride;
+     strides[4] = stride/2;
+     strides[5] = stride/2;
+   }
+   for (i = 0; i < 6; i++) {
+     /* Process only coded blocks */
+     if (cbp & (1 << (5 - i))) {
+       /* Clear the block */
+       memset(&data[0], 0, 64*sizeof(int16_t));
+       /* Decode coeffs and dequantize on the fly */
+       start_timer();
+       get_inter_block(bs, &data[0], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices));
+       stop_coding_timer();
+       /* iDCT */
+       start_timer();
+       idct((short * const)&data[0]);
+       stop_idct_timer();
- #define SIGN(X) (((X)>0)?1:-1)
+       /* Add this residual to the predicted block */
- #define ABS(X) (((X)>0)?(X):-(X))
+       start_timer();
- static const uint32_t roundtab[16] =
+       transfer_16to8add(dst[i], &data[0], strides[i]);
- { 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 };
+       stop_transfer_timer();
+     }
+   }
- // decode an inter macroblock
+ }
- void decoder_mbinter(DECODER * dec,
+ static void __inline
+ validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec)
+ {
+   /* clip a vector to valid range
+      prevents crashes if bitstream is broken
+   */
+   int shift = 5 + dec->quarterpel;
+   int xborder_high = (int)(dec->mb_width - x_pos) << shift;
+   int xborder_low = (-(int)x_pos-1) << shift;
+   int yborder_high = (int)(dec->mb_height - y_pos) << shift;
+   int yborder_low = (-(int)y_pos-1) << shift;
+ #define CHECK_MV(mv) \
+   do { \
+   if ((mv).x > xborder_high) { \
+     DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \
+     (mv).x = xborder_high; \
+   } else if ((mv).x < xborder_low) { \
+     DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", (mv).x, xborder_low, x_pos, y_pos); \
+     (mv).x = xborder_low; \
+   } \
+   if ((mv).y > yborder_high) { \
+     DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", (mv).y, yborder_high, x_pos, y_pos); \
+     (mv).y = yborder_high; \
+   } else if ((mv).y < yborder_low) { \
+     DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \
+     (mv).y = yborder_low; \
+   } \
+   } while (0)
+   CHECK_MV(mv[0]);
+   CHECK_MV(mv[1]);
+   CHECK_MV(mv[2]);
+   CHECK_MV(mv[3]);
+ }
+ /* Up to this version, chroma rounding was wrong with qpel.
+  * So we try to be backward compatible to avoid artifacts */
+ #define BS_VERSION_BUGGY_CHROMA_ROUNDING 1
+ /* decode an inter macroblock */
+ static void
+ decoder_mbinter(DECODER * dec,
                       const MACROBLOCK * pMB,
                       const uint32_t x_pos,
                       const uint32_t y_pos,
-                      const uint32_t acpred_flag,
                       const uint32_t cbp,
                       Bitstream * bs,
-                      const uint32_t quant,
+         const uint32_t rounding,
-                      const uint32_t rounding)
+         const int ref,
+                 const int bvop)
  {
-         DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE);
-         DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
          uint32_t stride = dec->edged_width;
          uint32_t stride2 = stride / 2;
-         uint32_t next_block = stride * 8;
          uint32_t i;
-         uint32_t iQuant = pMB->quant;
          uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
          int uv_dx, uv_dy;
+   VECTOR mv[4]; /* local copy of mvs */
          pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
          pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
          pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
+   for (i = 0; i < 4; i++)
+     mv[i] = pMB->mvs[i];
-         if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)
+   validate_vector(mv, x_pos, y_pos, dec);
-         {
-                 uv_dx = pMB->mvs[0].x;
-                 uv_dy = pMB->mvs[0].y;
-                 uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2;
+   start_timer();
-                 uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2;
-         }
+   if ((pMB->mode != MODE_INTER4V) || (bvop)) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
+     uv_dx = mv[0].x;
+     uv_dy = mv[0].y;
+     if (dec->quarterpel) {
+                         if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
+                                 uv_dx = (uv_dx>>1) | (uv_dx&1);
+                                 uv_dy = (uv_dy>>1) | (uv_dy&1);
+                         }
+                         else {
+         uv_dx /= 2;
+         uv_dy /= 2;
+       }
+     }
+     uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
+     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
+     if (dec->quarterpel)
+       interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64,
+                   dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
+                       mv[0].x, mv[0].y, stride, rounding);
          else
-         {
+       interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos,
-                 int sum;
+                   mv[0].x, mv[0].y, stride, rounding);
-                 sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
-                 uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) );
-                 sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
+   } else {  /* MODE_INTER4V */
-                 uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) );
+     if(dec->quarterpel) {
+                         if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
+                                 int z;
+                                 uv_dx = 0; uv_dy = 0;
+                                 for (z = 0; z < 4; z++) {
+                                   uv_dx += ((mv[z].x>>1) | (mv[z].x&1));
+                                   uv_dy += ((mv[z].y>>1) | (mv[z].y&1));
+                                 }
+                         }
+                         else {
+         uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2);
+         uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2);
+       }
+     } else {
+       uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x;
+       uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y;
+     }
+     uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
+     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
+     if (dec->quarterpel) {
+       interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
+                   dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
+                   mv[0].x, mv[0].y, stride, rounding);
+       interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
+                   dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
+                   mv[1].x, mv[1].y, stride, rounding);
+       interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
+                   dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
+                   mv[2].x, mv[2].y, stride, rounding);
+       interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
+                   dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
+                   mv[3].x, mv[3].y, stride, rounding);
+     } else {
+       interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos,
+                 mv[0].x, mv[0].y, stride, rounding);
+       interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos,
+                 mv[1].x, mv[1].y, stride, rounding);
+       interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8,
+                 mv[2].x, mv[2].y, stride, rounding);
+       interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8,
+                 mv[3].x, mv[3].y, stride, rounding);
+     }
          }
-         start_timer();
+   /* chroma */
-         interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos,     16*y_pos    , pMB->mvs[0].x, pMB->mvs[0].y, stride,  rounding);
+   interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
-         interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos + 8, 16*y_pos    , pMB->mvs[1].x, pMB->mvs[1].y, stride,  rounding);
+               uv_dx, uv_dy, stride2, rounding);
-         interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos,     16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride,  rounding);
+   interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
-         interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride,  rounding);
+               uv_dx, uv_dy, stride2, rounding);
-         interpolate8x8_switch(dec->cur.u, dec->refn.u, 8*x_pos,      8*y_pos,      uv_dx,         uv_dy,         stride2, rounding);
-         interpolate8x8_switch(dec->cur.v, dec->refn.v, 8*x_pos,      8*y_pos,      uv_dx,         uv_dy,         stride2, rounding);
          stop_comp_timer();
-         for (i = 0; i < 6; i++)
+   if (cbp)
-         {
+     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
-                 if (cbp & (1 << (5-i)))                 // coded
+ }
+ /* decode an inter macroblock in field mode */
+ static void
+ decoder_mbinter_field(DECODER * dec,
+         const MACROBLOCK * pMB,
+         const uint32_t x_pos,
+         const uint32_t y_pos,
+         const uint32_t cbp,
+         Bitstream * bs,
+         const uint32_t rounding,
+         const int ref,
+                 const int bvop)
                  {
-                         memset(&block[i*64], 0, 64 * sizeof(int16_t));          // clear
+   uint32_t stride = dec->edged_width;
+   uint32_t stride2 = stride / 2;
-                         start_timer();
+   uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
-                         get_inter_block(bs, &block[i*64]);
-                         stop_coding_timer();
+   int uvtop_dx, uvtop_dy;
+   int uvbot_dx, uvbot_dy;
+   VECTOR mv[4]; /* local copy of mvs */
+   /* Get pointer to memory areas */
+   pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
+   pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
+   pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
+   mv[0] = pMB->mvs[0];
+   mv[1] = pMB->mvs[1];
+   memset(&mv[2],0,2*sizeof(VECTOR));
+   validate_vector(mv, x_pos, y_pos, dec);
                          start_timer();
-                         if (dec->quant_type == 0)
+   if((pMB->mode!=MODE_INTER4V) || (bvop))   /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
                          {
-                                 dequant_inter(&data[i*64], &block[i*64], iQuant);
+     /* Prepare top field vector */
+     uvtop_dx = DIV2ROUND(mv[0].x);
+     uvtop_dy = DIV2ROUND(mv[0].y);
+     /* Prepare bottom field vector */
+     uvbot_dx = DIV2ROUND(mv[1].x);
+     uvbot_dy = DIV2ROUND(mv[1].y);
+     if(dec->quarterpel)
+     {
+       /* NOT supported */
                          }
                          else
                          {
-                                 dequant4_inter(&data[i*64], &block[i*64], iQuant);
+       /* Interpolate top field left part(we use double stride for every 2nd line) */
+       interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride,
+*x_pos,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding);
+       /* top field right part */
+       interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride,
+*x_pos+8,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding);
+       /* Interpolate bottom field left part(we use double stride for every 2nd line) */
+       interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride,
+*x_pos,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding);
+       /* Bottom field right part */
+       interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride,
+*x_pos+8,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding);
+       /* Interpolate field1 U */
+       interpolate8x4_switch(dec->cur.u,dec->refn[ref].u+pMB->field_for_top*stride2,
+*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding);
+       /* Interpolate field1 V */
+       interpolate8x4_switch(dec->cur.v,dec->refn[ref].v+pMB->field_for_top*stride2,
+*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding);
+       /* Interpolate field2 U */
+       interpolate8x4_switch(dec->cur.u+stride2,dec->refn[ref].u+pMB->field_for_bot*stride2,
+*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding);
+       /* Interpolate field2 V */
+       interpolate8x4_switch(dec->cur.v+stride2,dec->refn[ref].v+pMB->field_for_bot*stride2,
+*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding);
                          }
-                         stop_iquant_timer();
-                         start_timer();
-                         idct(&data[i*64]);
-                         stop_idct_timer();
                  }
+   else
+   {
+     /* We don't expect 4 motion vectors in interlaced mode */
          }
-         if (pMB->field_dct)
+   stop_comp_timer();
-         {
-                 next_block = stride;
+   /* Must add error correction? */
-                 stride *= 2;
+   if(cbp)
+    decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
          }
+ static void
+ decoder_mbgmc(DECODER * dec,
+         MACROBLOCK * const pMB,
+         const uint32_t x_pos,
+         const uint32_t y_pos,
+         const uint32_t fcode,
+         const uint32_t cbp,
+         Bitstream * bs,
+         const uint32_t rounding)
+ {
+   const uint32_t stride = dec->edged_width;
+   const uint32_t stride2 = stride / 2;
+   uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
+   uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
+   uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
+   NEW_GMC_DATA * gmc_data = &dec->new_gmc_data;
+   pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
          start_timer();
-         if (cbp & 32)
-                 transfer_16to8add(pY_Cur,                  &data[0*64], stride);
+ /* this is where the calculations are done */
-         if (cbp & 16)
-                 transfer_16to8add(pY_Cur + 8,              &data[1*64], stride);
+   gmc_data->predict_16x16(gmc_data,
-         if (cbp & 8)
+       dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y,
-                 transfer_16to8add(pY_Cur + next_block,     &data[2*64], stride);
+       stride, stride, x_pos, y_pos, rounding);
-         if (cbp & 4)
-                 transfer_16to8add(pY_Cur + 8 + next_block, &data[3*64], stride);
+   gmc_data->predict_8x8(gmc_data,
-         if (cbp & 2)
+       dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u,
-                 transfer_16to8add(pU_Cur,                  &data[4*64], stride2);
+       dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v,
-         if (cbp & 1)
+       stride2, stride2, x_pos, y_pos, rounding);
-                 transfer_16to8add(pV_Cur,                  &data[5*64], stride2);
+   gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel);
+   pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode);
+   pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode);
+   pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
          stop_transfer_timer();
+   if (cbp)
+     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
  }
- void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold)
+ static void
+ decoder_iframe(DECODER * dec,
+         Bitstream * bs,
+         int quant,
+         int intra_dc_threshold)
  {
+   uint32_t bound;
          uint32_t x, y;
+   const uint32_t mb_width = dec->mb_width;
+   const uint32_t mb_height = dec->mb_height;
-         for (y = 0; y < dec->mb_height; y++)
+   bound = 0;
-         {
-                 for (x = 0; x < dec->mb_width; x++)
-                 {
-                         MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x];
+   for (y = 0; y < mb_height; y++) {
+     for (x = 0; x < mb_width; x++) {
+       MACROBLOCK *mb;
                          uint32_t mcbpc;
                          uint32_t cbpc;
                          uint32_t acpred_flag;
                          uint32_t cbpy;
                          uint32_t cbp;
+       while (BitstreamShowBits(bs, 9) == 1)
+         BitstreamSkip(bs, 9);
+       if (check_resync_marker(bs, 0))
+       {
+         bound = read_video_packet_header(bs, dec, 0,
+               &quant, NULL, NULL, &intra_dc_threshold);
+         x = bound % mb_width;
+         y = MIN((bound / mb_width), (mb_height-1));
+       }
+       mb = &dec->mbs[y * dec->mb_width + x];
+       DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
                          mcbpc = get_mcbpc_intra(bs);
                          mb->mode = mcbpc & 7;
                          cbpc = (mcbpc >> 4);
                          acpred_flag = BitstreamGetBit(bs);
-                         if (mb->mode == MODE_STUFFING)
-                         {
-                                 DEBUG("-- STUFFING ?");
-                                 continue;
-                         }
                          cbpy = get_cbpy(bs, 1);
                          cbp = (cbpy << 2) | cbpc;
-                         if (mb->mode == MODE_INTRA_Q)
+       if (mb->mode == MODE_INTRA_Q) {
-                         {
                                  quant += dquant_table[BitstreamGetBits(bs,2)];
-                                 if (quant > 31)
+         if (quant > 31) {
-                                 {
                                          quant = 31;
-                                 }
+         } else if (quant < 1) {
-                                 else if (quant < 1)
-                                 {
                                          quant = 1;
                                  }
                          }
                          mb->quant = quant;
+       mb->mvs[0].x = mb->mvs[0].y =
+       mb->mvs[1].x = mb->mvs[1].y =
+       mb->mvs[2].x = mb->mvs[2].y =
+       mb->mvs[3].x = mb->mvs[3].y =0;
-                         if (dec->interlacing)
+       if (dec->interlacing) {
-                         {
                                  mb->field_dct = BitstreamGetBit(bs);
-                                 DEBUG1("deci: field_dct: ", mb->field_dct);
+         DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct);
                          }
-                         decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold);
+       decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
+               intra_dc_threshold, bound);
                  }
+     if(dec->out_frm)
+       output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
          }
  }
- void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode)
+ static void
+ get_motion_vector(DECODER * dec,
+         Bitstream * bs,
+         int x,
+         int y,
+         int k,
+         VECTOR * ret_mv,
+         int fcode,
+         const int bound)
  {
-         int scale_fac = 1 << (fcode - 1);
+   const int scale_fac = 1 << (fcode - 1);
-         int high = (32 * scale_fac) - 1;
+   const int high = (32 * scale_fac) - 1;
-         int low = ((-32) * scale_fac);
+   const int low = ((-32) * scale_fac);
-         int range = (64 * scale_fac);
+   const int range = (64 * scale_fac);
-         VECTOR pmv[4];
+   const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k);
-         uint32_t psad[4];
+   VECTOR mv;
-         int mv_x, mv_y;
+   mv.x = get_mv(bs, fcode);
-         int pmv_x, pmv_y;
+   mv.y = get_mv(bs, fcode);
+   DPRINTF(XVID_DEBUG_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)\n", mv.x, mv.y, pmv.x, pmv.y, mv.x+pmv.x, mv.y+pmv.y);
-         get_pmvdata(dec->mbs, x, y, dec->mb_width, k, pmv, psad);
+   mv.x += pmv.x;
+   mv.y += pmv.y;
-         pmv_x = pmv[0].x;
+   if (mv.x < low) {
-         pmv_y = pmv[0].y;
+     mv.x += range;
+   } else if (mv.x > high) {
+     mv.x -= range;
+   }
-         mv_x = get_mv(bs, fcode);
+   if (mv.y < low) {
-         mv_y = get_mv(bs, fcode);
+     mv.y += range;
+   } else if (mv.y > high) {
+     mv.y -= range;
+   }
-         mv_x += pmv_x;
+   ret_mv->x = mv.x;
-         mv_y += pmv_y;
+   ret_mv->y = mv.y;
+ }
-         if (mv_x < low)
+ /* We use this when decoder runs interlaced -> different prediction */
+ static void get_motion_vector_interlaced(DECODER * dec,
+         Bitstream * bs,
+         int x,
+         int y,
+         int k,
+         MACROBLOCK *pMB,
+         int fcode,
+         const int bound)
          {
-                 mv_x += range;
+   const int scale_fac = 1 << (fcode - 1);
-         }
+   const int high = (32 * scale_fac) - 1;
-         else if (mv_x > high)
+   const int low = ((-32) * scale_fac);
+   const int range = (64 * scale_fac);
+   /* Get interlaced prediction */
+   const VECTOR pmv=get_pmv2_interlaced(dec->mbs,dec->mb_width,bound,x,y,k);
+   VECTOR mv,mvf1,mvf2;
+   if(!pMB->field_pred)
          {
-                 mv_x -= range;
+     mv.x = get_mv(bs,fcode);
+     mv.y = get_mv(bs,fcode);
+     mv.x += pmv.x;
+     mv.y += pmv.y;
+     if(mv.x<low) {
+       mv.x += range;
+     } else if (mv.x>high) {
+       mv.x-=range;
          }
-         if (mv_y < low)
+     if (mv.y < low) {
-         {
+       mv.y += range;
-                 mv_y += range;
+     } else if (mv.y > high) {
+       mv.y -= range;
          }
-         else if (mv_y > high)
+     pMB->mvs[0]=pMB->mvs[1]=pMB->mvs[2]=pMB->mvs[3]=mv;
+   }
+   else
          {
-                 mv_y -= range;
+     mvf1.x = get_mv(bs, fcode);
+     mvf1.y = get_mv(bs, fcode);
+     mvf1.x += pmv.x;
+     mvf1.y = 2*(mvf1.y+pmv.y/2); /* It's multiple of 2 */
+     if (mvf1.x < low) {
+       mvf1.x += range;
+     } else if (mvf1.x > high) {
+       mvf1.x -= range;
          }
-         mv->x = mv_x;
+     if (mvf1.y < low) {
-         mv->y = mv_y;
+       mvf1.y += range;
+     } else if (mvf1.y > high) {
+       mvf1.y -= range;
+     }
+     mvf2.x = get_mv(bs, fcode);
+     mvf2.y = get_mv(bs, fcode);
+     mvf2.x += pmv.x;
+     mvf2.y = 2*(mvf2.y+pmv.y/2); /* It's multiple of 2 */
+     if (mvf2.x < low) {
+       mvf2.x += range;
+     } else if (mvf2.x > high) {
+       mvf2.x -= range;
  }
+     if (mvf2.y < low) {
+       mvf2.y += range;
+     } else if (mvf2.y > high) {
+       mvf2.y -= range;
+     }
- void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold)
+     pMB->mvs[0]=mvf1;
- {
+     pMB->mvs[1]=mvf2;
+     pMB->mvs[2].x=pMB->mvs[3].x=0;
+     pMB->mvs[2].y=pMB->mvs[3].y=0;
-         uint32_t x, y;
+     /* Calculate average for as it is field predicted */
+     pMB->mvs_avg.x=DIV2ROUND(pMB->mvs[0].x+pMB->mvs[1].x);
+     pMB->mvs_avg.y=DIV2ROUND(pMB->mvs[0].y+pMB->mvs[1].y);
+   }
+ }
-         image_swap(&dec->cur, &dec->refn);
+ /* for P_VOP set gmc_warp to NULL */
+ static void
+ decoder_pframe(DECODER * dec,
+         Bitstream * bs,
+         int rounding,
+         int quant,
+         int fcode,
+         int intra_dc_threshold,
+         const WARPPOINTS *const gmc_warp)
+ {
+   uint32_t x, y;
+   uint32_t bound;
+   int cp_mb, st_mb;
+   const uint32_t mb_width = dec->mb_width;
+   const uint32_t mb_height = dec->mb_height;
+   if (!dec->is_edged[0]) {
          start_timer();
-         image_setedges(&dec->refn, dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing);
+     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
+             dec->width, dec->height, dec->bs_version);
+     dec->is_edged[0] = 1;
          stop_edges_timer();
+   }
-         for (y = 0; y < dec->mb_height; y++)
+   if (gmc_warp) {
-         {
+     /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */
-                 for (x = 0; x < dec->mb_width; x++)
+     generate_GMCparameters( dec->sprite_warping_points,
-                 {
+         dec->sprite_warping_accuracy, gmc_warp,
-                         MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x];
+         dec->width, dec->height, &dec->new_gmc_data);
-                         if (!BitstreamGetBit(bs))                       // not_coded
+     /* image warping is done block-based in decoder_mbgmc(), now */
-                         {
+   }
-                                 uint32_t mcbpc;
-                                 uint32_t cbpc;
+   bound = 0;
-                                 uint32_t acpred_flag;
-                                 uint32_t cbpy;
+   for (y = 0; y < mb_height; y++) {
-                                 uint32_t cbp;
+     cp_mb = st_mb = 0;
-                                 uint32_t intra;
+     for (x = 0; x < mb_width; x++) {
+       MACROBLOCK *mb;
+       /* skip stuffing */
+       while (BitstreamShowBits(bs, 10) == 1)
+         BitstreamSkip(bs, 10);
+       if (check_resync_marker(bs, fcode - 1)) {
+         bound = read_video_packet_header(bs, dec, fcode - 1,
+           &quant, &fcode, NULL, &intra_dc_threshold);
+         x = bound % mb_width;
+         y = MIN((bound / mb_width), (mb_height-1));
+       }
+       mb = &dec->mbs[y * dec->mb_width + x];
+       DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
+       if (!(BitstreamGetBit(bs))) { /* block _is_ coded */
+         uint32_t mcbpc, cbpc, cbpy, cbp;
+         uint32_t intra, acpred_flag = 0;
+         int mcsel = 0;    /* mcsel: '0'=local motion, '1'=GMC */
+         cp_mb++;
                                  mcbpc = get_mcbpc_inter(bs);
                                  mb->mode = mcbpc & 7;
                                  cbpc = (mcbpc >> 4);
-                                 acpred_flag = 0;
+         DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode);
+         DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc);
                                  intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q);
-                                 if (intra)
+         if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q))
-                                 {
+           mcsel = BitstreamGetBit(bs);
+         else if (intra)
                                          acpred_flag = BitstreamGetBit(bs);
-                                 }
-                                 if (mb->mode == MODE_STUFFING)
-                                 {
-                                         DEBUG("-- STUFFING ?");
-                                         continue;
-                                 }
                                  cbpy = get_cbpy(bs, intra);
+         DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel);
                                  cbp = (cbpy << 2) | cbpc;
-                                 if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q)
+         if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) {
-                                 {
+           int dquant = dquant_table[BitstreamGetBits(bs, 2)];
-                                         quant += dquant_table[BitstreamGetBits(bs,2)];
+           DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant);
-                                         if (quant > 31)
+           quant += dquant;
-                                         {
+           if (quant > 31) {
                                                  quant = 31;
-                                         }
+           } else if (quant < 1) {
-                                         else if (mb->quant < 1)
-                                         {
                                                  quant = 1;
                                          }
+           DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant);
                                  }
                                  mb->quant = quant;
-                                 if (dec->interlacing)
+         mb->field_pred=0;
-                                 {
+         if (dec->interlacing) {
+           if (cbp || intra) {
                                          mb->field_dct = BitstreamGetBit(bs);
-                                         DEBUG1("decp: field_dct: ", mb->field_dct);
+             DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
+           }
-                                         if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)
+           if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) {
-                                         {
                                                  mb->field_pred = BitstreamGetBit(bs);
-                                                 DEBUG1("decp: field_pred: ", mb->field_pred);
+             DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
-                                                 if (mb->field_pred)
+             if (mb->field_pred) {
-                                                 {
                                                          mb->field_for_top = BitstreamGetBit(bs);
-                                                         DEBUG1("decp: field_for_top: ", mb->field_for_top);
+               DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
                                                          mb->field_for_bot = BitstreamGetBit(bs);
-                                                         DEBUG1("decp: field_for_bot: ", mb->field_for_bot);
+               DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
                                                  }
                                          }
                                  }
-                                 if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)
+         if (mcsel) {
-                                 {
+           decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding);
-                                         if (dec->interlacing && mb->field_pred)
+           continue;
-                                         {
-                                                 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode);
+         } else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
-                                                 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode);
+           if(dec->interlacing) {
+             /* Get motion vectors interlaced, field_pred is handled there */
+             get_motion_vector_interlaced(dec, bs, x, y, 0, mb, fcode, bound);
+           } else {
+             get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
+             mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
+           }
+         } else if (mb->mode == MODE_INTER4V ) {
+           /* interlaced missing here */
+           get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
+           get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound);
+           get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound);
+           get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound);
+         } else { /* MODE_INTRA, MODE_INTRA_Q */
+           mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
+           mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
+           decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
+                   intra_dc_threshold, bound);
+           continue;
                                          }
+         /* See how to decode */
+         if(!mb->field_pred)
+          decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0, 0);
                                          else
-                                         {
+          decoder_mbinter_field(dec, mb, x, y, cbp, bs, rounding, 0, 0);
-                                                 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode);
-                                                 mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x;
+       } else if (gmc_warp) {  /* a not coded S(GMC)-VOP macroblock */
-                                                 mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y;
+         mb->mode = MODE_NOT_CODED_GMC;
+         mb->quant = quant;
+         decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding);
+         if(dec->out_frm && cp_mb > 0) {
+           output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
+           cp_mb = 0;
+         }
+         st_mb = x+1;
+       } else { /* not coded P_VOP macroblock */
+         mb->mode = MODE_NOT_CODED;
+         mb->quant = quant;
+         mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
+         mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
+         mb->field_pred=0; /* (!) */
+         decoder_mbinter(dec, mb, x, y, 0, bs,
+                                 rounding, 0, 0);
+         if(dec->out_frm && cp_mb > 0) {
+           output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
+           cp_mb = 0;
                                          }
+         st_mb = x+1;
                                  }
-                                 else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */)
+     }
+     if(dec->out_frm && cp_mb > 0)
+       output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
+   }
+ }
+ /* decode B-frame motion vector */
+ static void
+ get_b_motion_vector(Bitstream * bs,
+           VECTOR * mv,
+           int fcode,
+           const VECTOR pmv,
+           const DECODER * const dec,
+           const int x, const int y)
                                  {
-                                         get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode);
+   const int scale_fac = 1 << (fcode - 1);
-                                         get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode);
+   const int high = (32 * scale_fac) - 1;
-                                         get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode);
+   const int low = ((-32) * scale_fac);
-                                         get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode);
+   const int range = (64 * scale_fac);
+   int mv_x = get_mv(bs, fcode);
+   int mv_y = get_mv(bs, fcode);
+   mv_x += pmv.x;
+   mv_y += pmv.y;
+   if (mv_x < low)
+     mv_x += range;
+   else if (mv_x > high)
+     mv_x -= range;
+   if (mv_y < low)
+     mv_y += range;
+   else if (mv_y > high)
+     mv_y -= range;
+   mv->x = mv_x;
+   mv->y = mv_y;
                                  }
-                                 else  // MODE_INTRA, MODE_INTRA_Q
+ /* decode an B-frame direct & interpolate macroblock */
+ static void
+ decoder_bf_interpolate_mbinter(DECODER * dec,
+                 IMAGE forward,
+                 IMAGE backward,
+                 MACROBLOCK * pMB,
+                 const uint32_t x_pos,
+                 const uint32_t y_pos,
+                 Bitstream * bs,
+                 const int direct)
                                  {
-                                         mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
+   uint32_t stride = dec->edged_width;
-                                         mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
+   uint32_t stride2 = stride / 2;
-                                         decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold);
+   int uv_dx, uv_dy;
-                                         continue;
+   int b_uv_dx, b_uv_dy;
+   uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
+   const uint32_t cbp = pMB->cbp;
+   pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
+   pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
+   pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
+   validate_vector(pMB->mvs, x_pos, y_pos, dec);
+   validate_vector(pMB->b_mvs, x_pos, y_pos, dec);
+   if (!direct) {
+     uv_dx = pMB->mvs[0].x;
+     uv_dy = pMB->mvs[0].y;
+     b_uv_dx = pMB->b_mvs[0].x;
+     b_uv_dy = pMB->b_mvs[0].y;
+     if (dec->quarterpel) {
+                         if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
+                                 uv_dx = (uv_dx>>1) | (uv_dx&1);
+                                 uv_dy = (uv_dy>>1) | (uv_dy&1);
+                                 b_uv_dx = (b_uv_dx>>1) | (b_uv_dx&1);
+                                 b_uv_dy = (b_uv_dy>>1) | (b_uv_dy&1);
+                         }
+                         else {
+         uv_dx /= 2;
+         uv_dy /= 2;
+         b_uv_dx /= 2;
+         b_uv_dy /= 2;
+       }
                                  }
-                                 decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding);
+     uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
+     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
+     b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
+     b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
+   } else {
+           if (dec->quarterpel) { /* for qpel the /2 shall be done before summation. We've done it right in the encoder in the past. */
+                                                          /* TODO: figure out if we ever did it wrong on the encoder side. If yes, add some workaround */
+                 if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
+                         int z;
+                         uv_dx = 0; uv_dy = 0;
+                         b_uv_dx = 0; b_uv_dy = 0;
+                         for (z = 0; z < 4; z++) {
+                           uv_dx += ((pMB->mvs[z].x>>1) | (pMB->mvs[z].x&1));
+                           uv_dy += ((pMB->mvs[z].y>>1) | (pMB->mvs[z].y&1));
+                           b_uv_dx += ((pMB->b_mvs[z].x>>1) | (pMB->b_mvs[z].x&1));
+                           b_uv_dy += ((pMB->b_mvs[z].y>>1) | (pMB->b_mvs[z].y&1));
+                         }
+                 }
+                 else {
+                         uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
+                         uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
+                         b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2);
+                         b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2);
+                 }
+         } else {
+       uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
+       uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
+       b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
+       b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
+     }
+     uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
+     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
+     b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf];
+     b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf];
                          }
-                         else    // not coded
+   start_timer();
+   if(dec->quarterpel) {
+     if(!direct) {
+       interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
+                     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
+                     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
+     } else {
+       interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
+                     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
+                     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
+       interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
+                     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
+                     pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
+       interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
+                     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
+                     pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
+       interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
+                     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
+                     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
+     }
+   } else {
+     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
+               pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
+     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
+               pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
+     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
+               pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
+     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8,
+               pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
+   }
+   interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
+             uv_dy, stride2, 0);
+   interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx,
+             uv_dy, stride2, 0);
+   if(dec->quarterpel) {
+     if(!direct) {
+       interpolate16x16_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
+           dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
+           pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
+     } else {
+       interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
+           dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
+           pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
+       interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
+           dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
+           pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
+       interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
+           dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
+           pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
+       interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
+           dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
+           pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
+     }
+   } else {
+     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, 16 * y_pos,
+         pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
+     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8,
+* y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
+     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos,
+* y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
+     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8,
+* y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
+   }
+   interpolate8x8_add_switch(dec->cur.u, backward.u, 8 * x_pos, 8 * y_pos,
+       b_uv_dx, b_uv_dy, stride2, 0);
+   interpolate8x8_add_switch(dec->cur.v, backward.v, 8 * x_pos, 8 * y_pos,
+       b_uv_dx, b_uv_dy, stride2, 0);
+   stop_comp_timer();
+   if (cbp)
+     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
+ }
+ /* for decode B-frame dbquant */
+ static __inline int32_t
+ get_dbquant(Bitstream * bs)
                          {
+   if (!BitstreamGetBit(bs))   /*  '0' */
+     return (0);
+   else if (!BitstreamGetBit(bs))  /* '10' */
+     return (-2);
+   else              /* '11' */
+     return (2);
+ }
-                                 mb->mode = MODE_NOT_CODED;
+ /*
-                                 mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
+  * decode B-frame mb_type
-                                 mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
+  * bit    ret_value
+  * 1    0
+  * 01   1
+  * 001    2
+  * 0001   3
+  */
+ static int32_t __inline
+ get_mbtype(Bitstream * bs)
+ {
+   int32_t mb_type;
+   for (mb_type = 0; mb_type <= 3; mb_type++)
+     if (BitstreamGetBit(bs))
+       return (mb_type);
+   return -1;
+ }
-                                 // copy macroblock directly from ref to cur
+ static int __inline get_resync_len_b(const int fcode_backward,
+                                      const int fcode_forward) {
+   int resync_len = ((fcode_forward>fcode_backward) ? fcode_forward : fcode_backward) - 1;
+   if (resync_len < 1) resync_len = 1;
+   return resync_len;
+ }
+ static void
+ decoder_bframe(DECODER * dec,
+         Bitstream * bs,
+         int quant,
+         int fcode_forward,
+         int fcode_backward)
+ {
+   uint32_t x, y;
+   VECTOR mv;
+   const VECTOR zeromv = {0,0};
+   int i;
+   int resync_len;
+   if (!dec->is_edged[0]) {
                                  start_timer();
+     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
+             dec->width, dec->height, dec->bs_version);
+     dec->is_edged[0] = 1;
+     stop_edges_timer();
+   }
-                                 transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x),
+   if (!dec->is_edged[1]) {
-                                                  dec->refn.y + (16*y)*dec->edged_width + (16*x),
+     start_timer();
-                                                  dec->edged_width);
+     image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height,
+             dec->width, dec->height, dec->bs_version);
+     dec->is_edged[1] = 1;
+     stop_edges_timer();
+   }
-                                 transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8),
+   resync_len = get_resync_len_b(fcode_backward, fcode_forward);
-                                                  dec->refn.y + (16*y)*dec->edged_width + (16*x+8),
+   for (y = 0; y < dec->mb_height; y++) {
-                                                  dec->edged_width);
+     /* Initialize Pred Motion Vector */
+     dec->p_fmv = dec->p_bmv = zeromv;
+     for (x = 0; x < dec->mb_width; x++) {
+       MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x];
+       MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x];
+       int intra_dc_threshold; /* fake variable */
-                                 transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x),
+       mv =
-                                                  dec->refn.y + (16*y+8)*dec->edged_width + (16*x),
+       mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] =
-                                                  dec->edged_width);
+       mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv;
+       mb->quant = quant;
-                                 transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8),
+       /*
-                                                  dec->refn.y + (16*y+8)*dec->edged_width + (16*x+8),
+        * skip if the co-located P_VOP macroblock is not coded
-                                                  dec->edged_width);
+        * if not codec in co-located S_VOP macroblock is _not_
+        * automatically skipped
+        */
+       if (last_mb->mode == MODE_NOT_CODED) {
+         mb->cbp = 0;
+         mb->mode = MODE_FORWARD;
+         decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1);
+         continue;
+       }
-                                 transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x),
+       if (check_resync_marker(bs, resync_len)) {
-                                                  dec->refn.u + (8*y)*dec->edged_width/2 + (8*x),
+         int bound = read_video_packet_header(bs, dec, resync_len, &quant,
-                                                  dec->edged_width/2);
+                            &fcode_forward, &fcode_backward, &intra_dc_threshold);
+                 bound = MAX(0, bound-1); /* valid bound must always be >0 */
+         x = bound % dec->mb_width;
+         y = MIN((bound / dec->mb_width), (dec->mb_height-1));
+         /* reset predicted macroblocks */
+         dec->p_fmv = dec->p_bmv = zeromv;
+         /* update resync len with new fcodes */
+         resync_len = get_resync_len_b(fcode_backward, fcode_forward);
+                 continue; /* re-init loop */
+           }
-                                 transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x),
+       if (!BitstreamGetBit(bs)) { /* modb=='0' */
-                                                  dec->refn.v + (8*y)*dec->edged_width/2 + (8*x),
+         const uint8_t modb2 = BitstreamGetBit(bs);
-                                                  dec->edged_width/2);
-                                 stop_transfer_timer();
+         mb->mode = get_mbtype(bs);
+         if (!modb2)   /* modb=='00' */
+           mb->cbp = BitstreamGetBits(bs, 6);
+         else
+           mb->cbp = 0;
+         if (mb->mode && mb->cbp) {
+           quant += get_dbquant(bs);
+           if (quant > 31)
+             quant = 31;
+           else if (quant < 1)
+             quant = 1;
                          }
+         mb->quant = quant;
+         if (dec->interlacing) {
+           if (mb->cbp) {
+             mb->field_dct = BitstreamGetBit(bs);
+             DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
+           }
+           if (mb->mode) {
+             mb->field_pred = BitstreamGetBit(bs);
+             DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
+             if (mb->field_pred) {
+               mb->field_for_top = BitstreamGetBit(bs);
+               DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
+               mb->field_for_bot = BitstreamGetBit(bs);
+               DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
                  }
          }
  }
- int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame)
+       } else {
- {
+         mb->mode = MODE_DIRECT_NONE_MV;
+         mb->cbp = 0;
+       }
-         Bitstream bs;
+       switch (mb->mode) {
-         uint32_t rounding;
+       case MODE_DIRECT:
-         uint32_t quant;
+         get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y);
-         uint32_t fcode;
-         uint32_t intra_dc_threshold;
-         start_global_timer();
+       case MODE_DIRECT_NONE_MV:
+         for (i = 0; i < 4; i++) {
+           mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x;
+           mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y;
-         BitstreamInit(&bs, frame->bitstream, frame->length);
+           mb->b_mvs[i].x = (mv.x)
+             ?  mb->mvs[i].x - last_mb->mvs[i].x
+             : last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp;
+           mb->b_mvs[i].y = (mv.y)
+             ? mb->mvs[i].y - last_mb->mvs[i].y
+             : last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp;
+         }
-         switch (BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode, &intra_dc_threshold))
+         decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
-         {
+                         mb, x, y, bs, 1);
-         case P_VOP :
-                 decoder_pframe(dec, &bs, rounding, quant, fcode, intra_dc_threshold);
                  break;
-         case I_VOP :
+       case MODE_INTERPOLATE:
-                 //DEBUG1("",intra_dc_threshold);
+         get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
-                 decoder_iframe(dec, &bs, quant, intra_dc_threshold);
+         dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
+         get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
+         dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0];
+         decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
+                       mb, x, y, bs, 0);
                  break;
-         case B_VOP :    // ignore
+       case MODE_BACKWARD:
+         get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
+         dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
+         decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
                  break;
-         case N_VOP :    // vop not coded
+       case MODE_FORWARD:
+         get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
+         dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
+         decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1);
                  break;
          default :
-                 return XVID_ERR_FAIL;
+         DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode);
+       }
+     } /* End of for */
+   }
          }
-         frame->length = BitstreamPos(&bs) / 8;
+ /* perform post processing if necessary, and output the image */
+ static void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs,
+           xvid_dec_frame_t * frame, xvid_dec_stats_t * stats,
+           int coding_type, int quant)
+ {
+   const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0;
-         start_timer();
+   if (dec->cartoon_mode)
-         image_output(&dec->cur, dec->width, dec->height, dec->edged_width,
+     frame->general &= ~XVID_FILMEFFECT;
-                      frame->image, frame->stride, frame->colorspace);
-         stop_conv_timer();
-         emms();
+   if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0)
+     && mbs != NULL) /* post process */
+   {
+     /* note: image is stored to tmp */
+     image_copy(&dec->tmp, img, dec->edged_width, dec->height);
+     image_postproc(&dec->postproc, &dec->tmp, dec->edged_width,
+              mbs, dec->mb_width, dec->mb_height, dec->mb_width,
+              frame->general, brightness, dec->frames, (coding_type == B_VOP), dec->num_threads);
+     img = &dec->tmp;
+   }
+   if ((frame->output.plane[0] != NULL) && (frame->output.stride[0] >= dec->width)) {
+     image_output(img, dec->width, dec->height,
+            dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride,
+            frame->output.csp, dec->interlacing);
+   }
+   if (stats) {
+     stats->type = coding2type(coding_type);
+     stats->data.vop.time_base = (int)dec->time_base;
+     stats->data.vop.time_increment = 0; /* XXX: todo */
+     stats->data.vop.qscale_stride = dec->mb_width;
+     stats->data.vop.qscale = dec->qscale;
+     if (stats->data.vop.qscale != NULL && mbs != NULL) {
+       unsigned int i;
+       for (i = 0; i < dec->mb_width*dec->mb_height; i++)
+         stats->data.vop.qscale[i] = mbs[i].quant;
+     } else
+       stats->data.vop.qscale = NULL;
+   }
+ }
+ int
+ decoder_decode(DECODER * dec,
+         xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
+ {
+   Bitstream bs;
+   uint32_t rounding = 0;
+   uint32_t quant = 2;
+   uint32_t fcode_forward = 0;
+   uint32_t fcode_backward = 0;
+   uint32_t intra_dc_threshold = 0;
+   WARPPOINTS gmc_warp;
+   int coding_type = -1;
+   int success, output, seen_something;
+   if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1))  /* v1.x.x */
+     return XVID_ERR_VERSION;
+   start_global_timer();
+   memset((void *)&gmc_warp, 0, sizeof(WARPPOINTS));
+   dec->low_delay_default = (frame->general & XVID_LOWDELAY);
+   if ((frame->general & XVID_DISCONTINUITY))
+     dec->frames = 0;
+   dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL;
+   if(frame->length<0) {  /* decoder flush */
+     int ret;
+     /* if not decoding "low_delay/packed", and this isn't low_delay and
+       we have a reference frame, then outout the reference frame */
+     if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) {
+       decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
+       dec->frames = 0;
+       ret = 0;
+     } else {
+       if (stats) stats->type = XVID_TYPE_NOTHING;
+       ret = XVID_ERR_END;
+     }
+     emms();
          stop_global_timer();
+     return ret;
+   }
+   BitstreamInit(&bs, frame->bitstream, frame->length);
+   /* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */
+   if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
+   {
+     image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width,
+            (uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing);
+     if (stats) stats->type = XVID_TYPE_NOTHING;
+     emms();
+     return 1; /* one byte consumed */
+   }
+   success = 0;
+   output = 0;
+   seen_something = 0;
+ repeat:
+   coding_type = BitstreamReadHeaders(&bs, dec, &rounding,
+       &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp);
+   DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i,  packed=%i,  time=%"
+ #if defined(_MSC_VER)
+     "I64"
+ #else
+     "ll"
+ #endif
+     "i,  time_pp=%i,  time_bp=%i\n",
+               coding_type,  dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
+   if (coding_type == -1) { /* nothing */
+     if (success) goto done;
+     if (stats) stats->type = XVID_TYPE_NOTHING;
+     emms();
+     return BitstreamPos(&bs)/8;
+   }
+   if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */
+     if (coding_type == -3)
+       if (decoder_resize(dec)) return XVID_ERR_MEMORY;
+     if(stats) {
+       stats->type = XVID_TYPE_VOL;
+       stats->data.vol.general = 0;
+           stats->data.vop.general = 0;
+           if (dec->interlacing) {
+                 stats->data.vol.general |= XVID_VOL_INTERLACING;
+                 if (dec->top_field_first) {
+               stats->data.vop.general |= XVID_VOP_TOPFIELDFIRST;
+                 }
+           }
+       stats->data.vol.width = dec->width;
+       stats->data.vol.height = dec->height;
+       stats->data.vol.par = dec->aspect_ratio;
+       stats->data.vol.par_width = dec->par_width;
+       stats->data.vol.par_height = dec->par_height;
+       emms();
+       return BitstreamPos(&bs)/8; /* number of bytes consumed */
+     }
+     goto repeat;
+   }
+   if((dec->frames == 0 && coding_type != I_VOP) || (!dec->width || !dec->height)) {
+     /* 1st frame is not an i-vop */
+     goto repeat;
+   }
-         return XVID_ERR_OK;
+   dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.x = dec->p_fmv.y = 0;  /* init pred vector to 0 */
+   /* packed_mode: special-N_VOP treament */
+   if (dec->packed_mode && coding_type == N_VOP) {
+     if (dec->low_delay_default && dec->frames > 0) {
+       decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
+       output = 1;
+     }
+     /* ignore otherwise */
+   } else if (coding_type != B_VOP) {
+     switch(coding_type) {
+     case I_VOP :
+       decoder_iframe(dec, &bs, quant, intra_dc_threshold);
+       break;
+     case P_VOP :
+       decoder_pframe(dec, &bs, rounding, quant,
+                         fcode_forward, intra_dc_threshold, NULL);
+       break;
+     case S_VOP :
+       decoder_pframe(dec, &bs, rounding, quant,
+                         fcode_forward, intra_dc_threshold, &gmc_warp);
+       break;
+     case N_VOP :
+       /* XXX: not_coded vops are not used for forward prediction */
+       /* we should not swap(last_mbs,mbs) */
+       image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
+       SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */
+       break;
+     }
+     /* note: for packed_mode, output is performed when the special-N_VOP is decoded */
+     if (!(dec->low_delay_default && dec->packed_mode)) {
+       if(dec->low_delay) {
+         decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
+         output = 1;
+       } else if (dec->frames > 0) { /* is the reference frame valid? */
+         /* output the reference frame */
+         decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
+         output = 1;
+       }
+     }
+     image_swap(&dec->refn[0], &dec->refn[1]);
+     dec->is_edged[1] = dec->is_edged[0];
+     image_swap(&dec->cur, &dec->refn[0]);
+     dec->is_edged[0] = 0;
+     SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs);
+     dec->last_coding_type = coding_type;
+     dec->frames++;
+     seen_something = 1;
+   } else {  /* B_VOP */
+     if (dec->low_delay) {
+       DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n");
+       dec->low_delay = 0;
+     }
+     if (dec->frames < 2) {
+       /* attemping to decode a bvop without atleast 2 reference frames */
+       image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
+             "broken b-frame, mising ref frames");
+       if (stats) stats->type = XVID_TYPE_NOTHING;
+     } else if (dec->time_pp <= dec->time_bp) {
+       /* this occurs when dx50_bvop_compatibility==0 sequences are
+       decoded in vfw. */
+       image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
+             "broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp);
+       if (stats) stats->type = XVID_TYPE_NOTHING;
+     } else {
+       decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward);
+       decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
+     }
+     output = 1;
+     dec->frames++;
+   }
+ #if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */
+    BitstreamByteAlign(&bs);
+ #endif
+   /* low_delay_default mode: repeat in packed_mode */
+   if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) {
+     success = 1;
+     goto repeat;
+   }
+ done :
+   /* if we reach here without outputing anything _and_
+      the calling application has specified low_delay_default,
+      we *must* output something.
+      this always occurs on the first call to decode() call
+      when bframes are present in the bitstream. it may also
+      occur if no vops  were seen in the bitstream
+      if packed_mode is enabled, then we output the recently
+      decoded frame (the very first ivop). otherwise we have
+      nothing to display, and therefore output a black screen.
+   */
+   if (dec->low_delay_default && output == 0) {
+     if (dec->packed_mode && seen_something) {
+       decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
+     } else {
+       image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128);
+       decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant);
+       if (stats) stats->type = XVID_TYPE_NOTHING;
+     }
+   }
+   emms();
+   stop_global_timer();
+   return (BitstreamPos(&bs)+7)/8; /* number of bytes consumed */
  }

 Legend:



Removed from v.100
 


changed lines


 
Added in v.2178
 Legend:



Removed from v.100
 


changed lines


 
Added in v.2178
-Removed from v.100
+Added in v.2178

No admin address has been configured	ViewVC Help
Powered by ViewVC 1.0.4