vc1: Split the decoder in components
authorLuca Barbato <lu_zero@gentoo.org>
Wed, 16 Jul 2014 18:18:20 +0000 (20:18 +0200)
committerLuca Barbato <lu_zero@gentoo.org>
Wed, 8 Oct 2014 16:34:48 +0000 (18:34 +0200)
Speed up the overall compilation time.

libavcodec/Makefile
libavcodec/vc1.h
libavcodec/vc1_block.c [new file with mode: 0644]
libavcodec/vc1_loopfilter.c [new file with mode: 0644]
libavcodec/vc1_mc.c [new file with mode: 0644]
libavcodec/vc1_pred.c [new file with mode: 0644]
libavcodec/vc1_pred.h [new file with mode: 0644]
libavcodec/vc1data.h
libavcodec/vc1dec.c

index 69b92b6..62e41b8 100644 (file)
@@ -390,7 +390,9 @@ OBJS-$(CONFIG_V410_ENCODER)            += v410enc.o
 OBJS-$(CONFIG_V210X_DECODER)           += v210x.o
 OBJS-$(CONFIG_VB_DECODER)              += vb.o
 OBJS-$(CONFIG_VBLE_DECODER)            += vble.o
-OBJS-$(CONFIG_VC1_DECODER)             += vc1dec.o vc1.o vc1data.o vc1dsp.o \
+OBJS-$(CONFIG_VC1_DECODER)             += vc1dec.o vc1_block.o vc1_loopfilter.o \
+                                          vc1_mc.o vc1_pred.o vc1.o vc1data.o \
+                                          vc1dsp.o \
                                           msmpeg4dec.o msmpeg4.o msmpeg4data.o
 OBJS-$(CONFIG_VCR1_DECODER)            += vcr1.o
 OBJS-$(CONFIG_VMDAUDIO_DECODER)        += vmdaudio.o
index 780344a..0c4958c 100644 (file)
@@ -400,4 +400,16 @@ void ff_vc1_init_transposed_scantables(VC1Context *v);
 int  ff_vc1_decode_end(AVCodecContext *avctx);
 void ff_vc1_decode_blocks(VC1Context *v);
 
+void ff_vc1_loop_filter_iblk(VC1Context *v, int pq);
+void ff_vc1_loop_filter_iblk_delayed(VC1Context *v, int pq);
+void ff_vc1_smooth_overlap_filter_iblk(VC1Context *v);
+void ff_vc1_apply_p_loop_filter(VC1Context *v);
+
+void ff_vc1_mc_1mv(VC1Context *v, int dir);
+void ff_vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg);
+void ff_vc1_mc_4mv_chroma(VC1Context *v, int dir);
+void ff_vc1_mc_4mv_chroma4(VC1Context *v, int dir, int dir2, int avg);
+
+void ff_vc1_interp_mc(VC1Context *v);
+
 #endif /* AVCODEC_VC1_H */
diff --git a/libavcodec/vc1_block.c b/libavcodec/vc1_block.c
new file mode 100644 (file)
index 0000000..acd5145
--- /dev/null
@@ -0,0 +1,3054 @@
+/*
+ * VC-1 and WMV3 decoder
+ * Copyright (c) 2011 Mashiat Sarker Shakkhar
+ * Copyright (c) 2006-2007 Konstantin Shishkov
+ * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * VC-1 and WMV3 block decoding routines
+ */
+
+#include "avcodec.h"
+#include "mpegutils.h"
+#include "mpegvideo.h"
+#include "msmpeg4data.h"
+#include "unary.h"
+#include "vc1.h"
+#include "vc1_pred.h"
+#include "vc1acdata.h"
+#include "vc1data.h"
+
+#define MB_INTRA_VLC_BITS 9
+#define DC_VLC_BITS 9
+
+// offset tables for interlaced picture MVDATA decoding
+static const int offset_table1[9] = {  0,  1,  2,  4,  8, 16, 32,  64, 128 };
+static const int offset_table2[9] = {  0,  1,  3,  7, 15, 31, 63, 127, 255 };
+
+/***********************************************************************/
+/**
+ * @name VC-1 Bitplane decoding
+ * @see 8.7, p56
+ * @{
+ */
+
+/**
+ * Imode types
+ * @{
+ */
+enum Imode {
+    IMODE_RAW,
+    IMODE_NORM2,
+    IMODE_DIFF2,
+    IMODE_NORM6,
+    IMODE_DIFF6,
+    IMODE_ROWSKIP,
+    IMODE_COLSKIP
+};
+/** @} */ //imode defines
+
+static void init_block_index(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    ff_init_block_index(s);
+    if (v->field_mode && !(v->second_field ^ v->tff)) {
+        s->dest[0] += s->current_picture_ptr->f->linesize[0];
+        s->dest[1] += s->current_picture_ptr->f->linesize[1];
+        s->dest[2] += s->current_picture_ptr->f->linesize[2];
+    }
+}
+
+/** @} */ //Bitplane group
+
+static void vc1_put_signed_blocks_clamped(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    int topleft_mb_pos, top_mb_pos;
+    int stride_y, fieldtx = 0;
+    int v_dist;
+
+    /* The put pixels loop is always one MB row behind the decoding loop,
+     * because we can only put pixels when overlap filtering is done, and
+     * for filtering of the bottom edge of a MB, we need the next MB row
+     * present as well.
+     * Within the row, the put pixels loop is also one MB col behind the
+     * decoding loop. The reason for this is again, because for filtering
+     * of the right MB edge, we need the next MB present. */
+    if (!s->first_slice_line) {
+        if (s->mb_x) {
+            topleft_mb_pos = (s->mb_y - 1) * s->mb_stride + s->mb_x - 1;
+            if (v->fcm == ILACE_FRAME)
+                fieldtx = v->fieldtx_plane[topleft_mb_pos];
+            stride_y       = s->linesize << fieldtx;
+            v_dist         = (16 - fieldtx) >> (fieldtx == 0);
+            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][0],
+                                              s->dest[0] - 16 * s->linesize - 16,
+                                              stride_y);
+            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][1],
+                                              s->dest[0] - 16 * s->linesize - 8,
+                                              stride_y);
+            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][2],
+                                              s->dest[0] - v_dist * s->linesize - 16,
+                                              stride_y);
+            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][3],
+                                              s->dest[0] - v_dist * s->linesize - 8,
+                                              stride_y);
+            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][4],
+                                              s->dest[1] - 8 * s->uvlinesize - 8,
+                                              s->uvlinesize);
+            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][5],
+                                              s->dest[2] - 8 * s->uvlinesize - 8,
+                                              s->uvlinesize);
+        }
+        if (s->mb_x == s->mb_width - 1) {
+            top_mb_pos = (s->mb_y - 1) * s->mb_stride + s->mb_x;
+            if (v->fcm == ILACE_FRAME)
+                fieldtx = v->fieldtx_plane[top_mb_pos];
+            stride_y   = s->linesize << fieldtx;
+            v_dist     = fieldtx ? 15 : 8;
+            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][0],
+                                              s->dest[0] - 16 * s->linesize,
+                                              stride_y);
+            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][1],
+                                              s->dest[0] - 16 * s->linesize + 8,
+                                              stride_y);
+            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][2],
+                                              s->dest[0] - v_dist * s->linesize,
+                                              stride_y);
+            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][3],
+                                              s->dest[0] - v_dist * s->linesize + 8,
+                                              stride_y);
+            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][4],
+                                              s->dest[1] - 8 * s->uvlinesize,
+                                              s->uvlinesize);
+            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][5],
+                                              s->dest[2] - 8 * s->uvlinesize,
+                                              s->uvlinesize);
+        }
+    }
+
+#define inc_blk_idx(idx) do { \
+        idx++; \
+        if (idx >= v->n_allocated_blks) \
+            idx = 0; \
+    } while (0)
+
+    inc_blk_idx(v->topleft_blk_idx);
+    inc_blk_idx(v->top_blk_idx);
+    inc_blk_idx(v->left_blk_idx);
+    inc_blk_idx(v->cur_blk_idx);
+}
+
+/***********************************************************************/
+/**
+ * @name VC-1 Block-level functions
+ * @see 7.1.4, p91 and 8.1.1.7, p(1)04
+ * @{
+ */
+
+/**
+ * @def GET_MQUANT
+ * @brief Get macroblock-level quantizer scale
+ */
+#define GET_MQUANT()                                           \
+    if (v->dquantfrm) {                                        \
+        int edges = 0;                                         \
+        if (v->dqprofile == DQPROFILE_ALL_MBS) {               \
+            if (v->dqbilevel) {                                \
+                mquant = (get_bits1(gb)) ? v->altpq : v->pq;   \
+            } else {                                           \
+                mqdiff = get_bits(gb, 3);                      \
+                if (mqdiff != 7)                               \
+                    mquant = v->pq + mqdiff;                   \
+                else                                           \
+                    mquant = get_bits(gb, 5);                  \
+            }                                                  \
+        }                                                      \
+        if (v->dqprofile == DQPROFILE_SINGLE_EDGE)             \
+            edges = 1 << v->dqsbedge;                          \
+        else if (v->dqprofile == DQPROFILE_DOUBLE_EDGES)       \
+            edges = (3 << v->dqsbedge) % 15;                   \
+        else if (v->dqprofile == DQPROFILE_FOUR_EDGES)         \
+            edges = 15;                                        \
+        if ((edges&1) && !s->mb_x)                             \
+            mquant = v->altpq;                                 \
+        if ((edges&2) && s->first_slice_line)                  \
+            mquant = v->altpq;                                 \
+        if ((edges&4) && s->mb_x == (s->mb_width - 1))         \
+            mquant = v->altpq;                                 \
+        if ((edges&8) && s->mb_y == (s->mb_height - 1))        \
+            mquant = v->altpq;                                 \
+        if (!mquant || mquant > 31) {                          \
+            av_log(v->s.avctx, AV_LOG_ERROR,                   \
+                   "Overriding invalid mquant %d\n", mquant);  \
+            mquant = 1;                                        \
+        }                                                      \
+    }
+
+/**
+ * @def GET_MVDATA(_dmv_x, _dmv_y)
+ * @brief Get MV differentials
+ * @see MVDATA decoding from 8.3.5.2, p(1)20
+ * @param _dmv_x Horizontal differential for decoded MV
+ * @param _dmv_y Vertical differential for decoded MV
+ */
+#define GET_MVDATA(_dmv_x, _dmv_y)                                      \
+    index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table, \
+                         VC1_MV_DIFF_VLC_BITS, 2);                      \
+    if (index > 36) {                                                   \
+        mb_has_coeffs = 1;                                              \
+        index -= 37;                                                    \
+    } else                                                              \
+        mb_has_coeffs = 0;                                              \
+    s->mb_intra = 0;                                                    \
+    if (!index) {                                                       \
+        _dmv_x = _dmv_y = 0;                                            \
+    } else if (index == 35) {                                           \
+        _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample);          \
+        _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample);          \
+    } else if (index == 36) {                                           \
+        _dmv_x = 0;                                                     \
+        _dmv_y = 0;                                                     \
+        s->mb_intra = 1;                                                \
+    } else {                                                            \
+        index1 = index % 6;                                             \
+        if (!s->quarter_sample && index1 == 5) val = 1;                 \
+        else                                   val = 0;                 \
+        if (size_table[index1] - val > 0)                               \
+            val = get_bits(gb, size_table[index1] - val);               \
+        else                                   val = 0;                 \
+        sign = 0 - (val&1);                                             \
+        _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \
+                                                                        \
+        index1 = index / 6;                                             \
+        if (!s->quarter_sample && index1 == 5) val = 1;                 \
+        else                                   val = 0;                 \
+        if (size_table[index1] - val > 0)                               \
+            val = get_bits(gb, size_table[index1] - val);               \
+        else                                   val = 0;                 \
+        sign = 0 - (val & 1);                                           \
+        _dmv_y = (sign ^ ((val >> 1) + offset_table[index1])) - sign;   \
+    }
+
+static av_always_inline void get_mvdata_interlaced(VC1Context *v, int *dmv_x,
+                                                   int *dmv_y, int *pred_flag)
+{
+    int index, index1;
+    int extend_x = 0, extend_y = 0;
+    GetBitContext *gb = &v->s.gb;
+    int bits, esc;
+    int val, sign;
+    const int* offs_tab;
+
+    if (v->numref) {
+        bits = VC1_2REF_MVDATA_VLC_BITS;
+        esc  = 125;
+    } else {
+        bits = VC1_1REF_MVDATA_VLC_BITS;
+        esc  = 71;
+    }
+    switch (v->dmvrange) {
+    case 1:
+        extend_x = 1;
+        break;
+    case 2:
+        extend_y = 1;
+        break;
+    case 3:
+        extend_x = extend_y = 1;
+        break;
+    }
+    index = get_vlc2(gb, v->imv_vlc->table, bits, 3);
+    if (index == esc) {
+        *dmv_x = get_bits(gb, v->k_x);
+        *dmv_y = get_bits(gb, v->k_y);
+        if (v->numref) {
+            if (pred_flag) {
+                *pred_flag = *dmv_y & 1;
+                *dmv_y     = (*dmv_y + *pred_flag) >> 1;
+            } else {
+                *dmv_y     = (*dmv_y + (*dmv_y & 1)) >> 1;
+            }
+        }
+    }
+    else {
+        if (extend_x)
+            offs_tab = offset_table2;
+        else
+            offs_tab = offset_table1;
+        index1 = (index + 1) % 9;
+        if (index1 != 0) {
+            val    = get_bits(gb, index1 + extend_x);
+            sign   = 0 -(val & 1);
+            *dmv_x = (sign ^ ((val >> 1) + offs_tab[index1])) - sign;
+        } else
+            *dmv_x = 0;
+        if (extend_y)
+            offs_tab = offset_table2;
+        else
+            offs_tab = offset_table1;
+        index1 = (index + 1) / 9;
+        if (index1 > v->numref) {
+            val    = get_bits(gb, (index1 + (extend_y << v->numref)) >> v->numref);
+            sign   = 0 - (val & 1);
+            *dmv_y = (sign ^ ((val >> 1) + offs_tab[index1 >> v->numref])) - sign;
+        } else
+            *dmv_y = 0;
+        if (v->numref && pred_flag)
+            *pred_flag = index1 & 1;
+    }
+}
+
+/** Reconstruct motion vector for B-frame and do motion compensation
+ */
+static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2],
+                            int direct, int mode)
+{
+    if (direct) {
+        ff_vc1_mc_1mv(v, 0);
+        ff_vc1_interp_mc(v);
+        return;
+    }
+    if (mode == BMV_TYPE_INTERPOLATED) {
+        ff_vc1_mc_1mv(v, 0);
+        ff_vc1_interp_mc(v);
+        return;
+    }
+
+    ff_vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));
+}
+
+/** Get predicted DC value for I-frames only
+ * prediction dir: left=0, top=1
+ * @param s MpegEncContext
+ * @param overlap flag indicating that overlap filtering is used
+ * @param pq integer part of picture quantizer
+ * @param[in] n block index in the current MB
+ * @param dc_val_ptr Pointer to DC predictor
+ * @param dir_ptr Prediction direction for use in AC prediction
+ */
+static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
+                                int16_t **dc_val_ptr, int *dir_ptr)
+{
+    int a, b, c, wrap, pred, scale;
+    int16_t *dc_val;
+    static const uint16_t dcpred[32] = {
+        -1, 1024,  512,  341,  256,  205,  171,  146,  128,
+             114,  102,   93,   85,   79,   73,   68,   64,
+              60,   57,   54,   51,   49,   47,   45,   43,
+              41,   39,   38,   37,   35,   34,   33
+    };
+
+    /* find prediction - wmv3_dc_scale always used here in fact */
+    if (n < 4) scale = s->y_dc_scale;
+    else       scale = s->c_dc_scale;
+
+    wrap   = s->block_wrap[n];
+    dc_val = s->dc_val[0] + s->block_index[n];
+
+    /* B A
+     * C X
+     */
+    c = dc_val[ - 1];
+    b = dc_val[ - 1 - wrap];
+    a = dc_val[ - wrap];
+
+    if (pq < 9 || !overlap) {
+        /* Set outer values */
+        if (s->first_slice_line && (n != 2 && n != 3))
+            b = a = dcpred[scale];
+        if (s->mb_x == 0 && (n != 1 && n != 3))
+            b = c = dcpred[scale];
+    } else {
+        /* Set outer values */
+        if (s->first_slice_line && (n != 2 && n != 3))
+            b = a = 0;
+        if (s->mb_x == 0 && (n != 1 && n != 3))
+            b = c = 0;
+    }
+
+    if (abs(a - b) <= abs(b - c)) {
+        pred     = c;
+        *dir_ptr = 1; // left
+    } else {
+        pred     = a;
+        *dir_ptr = 0; // top
+    }
+
+    /* update predictor */
+    *dc_val_ptr = &dc_val[0];
+    return pred;
+}
+
+
+/** Get predicted DC value
+ * prediction dir: left=0, top=1
+ * @param s MpegEncContext
+ * @param overlap flag indicating that overlap filtering is used
+ * @param pq integer part of picture quantizer
+ * @param[in] n block index in the current MB
+ * @param a_avail flag indicating top block availability
+ * @param c_avail flag indicating left block availability
+ * @param dc_val_ptr Pointer to DC predictor
+ * @param dir_ptr Prediction direction for use in AC prediction
+ */
+static inline int ff_vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
+                              int a_avail, int c_avail,
+                              int16_t **dc_val_ptr, int *dir_ptr)
+{
+    int a, b, c, wrap, pred;
+    int16_t *dc_val;
+    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+    int q1, q2 = 0;
+    int dqscale_index;
+
+    wrap = s->block_wrap[n];
+    dc_val = s->dc_val[0] + s->block_index[n];
+
+    /* B A
+     * C X
+     */
+    c = dc_val[ - 1];
+    b = dc_val[ - 1 - wrap];
+    a = dc_val[ - wrap];
+    /* scale predictors if needed */
+    q1 = s->current_picture.qscale_table[mb_pos];
+    dqscale_index = s->y_dc_scale_table[q1] - 1;
+    if (dqscale_index < 0)
+        return 0;
+    if (c_avail && (n != 1 && n != 3)) {
+        q2 = s->current_picture.qscale_table[mb_pos - 1];
+        if (q2 && q2 != q1)
+            c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;
+    }
+    if (a_avail && (n != 2 && n != 3)) {
+        q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
+        if (q2 && q2 != q1)
+            a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;
+    }
+    if (a_avail && c_avail && (n != 3)) {
+        int off = mb_pos;
+        if (n != 1)
+            off--;
+        if (n != 2)
+            off -= s->mb_stride;
+        q2 = s->current_picture.qscale_table[off];
+        if (q2 && q2 != q1)
+            b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;
+    }
+
+    if (a_avail && c_avail) {
+        if (abs(a - b) <= abs(b - c)) {
+            pred     = c;
+            *dir_ptr = 1; // left
+        } else {
+            pred     = a;
+            *dir_ptr = 0; // top
+        }
+    } else if (a_avail) {
+        pred     = a;
+        *dir_ptr = 0; // top
+    } else if (c_avail) {
+        pred     = c;
+        *dir_ptr = 1; // left
+    } else {
+        pred     = 0;
+        *dir_ptr = 1; // left
+    }
+
+    /* update predictor */
+    *dc_val_ptr = &dc_val[0];
+    return pred;
+}
+
+/** @} */ // Block group
+
+/**
+ * @name VC1 Macroblock-level functions in Simple/Main Profiles
+ * @see 7.1.4, p91 and 8.1.1.7, p(1)04
+ * @{
+ */
+
+static inline int vc1_coded_block_pred(MpegEncContext * s, int n,
+                                       uint8_t **coded_block_ptr)
+{
+    int xy, wrap, pred, a, b, c;
+
+    xy   = s->block_index[n];
+    wrap = s->b8_stride;
+
+    /* B C
+     * A X
+     */
+    a = s->coded_block[xy - 1       ];
+    b = s->coded_block[xy - 1 - wrap];
+    c = s->coded_block[xy     - wrap];
+
+    if (b == c) {
+        pred = a;
+    } else {
+        pred = c;
+    }
+
+    /* store value */
+    *coded_block_ptr = &s->coded_block[xy];
+
+    return pred;
+}
+
+/**
+ * Decode one AC coefficient
+ * @param v The VC1 context
+ * @param last Last coefficient
+ * @param skip How much zero coefficients to skip
+ * @param value Decoded AC coefficient value
+ * @param codingset set of VLC to decode data
+ * @see 8.1.3.4
+ */
+static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip,
+                                int *value, int codingset)
+{
+    GetBitContext *gb = &v->s.gb;
+    int index, escape, run = 0, level = 0, lst = 0;
+
+    index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
+    if (index != ff_vc1_ac_sizes[codingset] - 1) {
+        run   = vc1_index_decode_table[codingset][index][0];
+        level = vc1_index_decode_table[codingset][index][1];
+        lst   = index >= vc1_last_decode_table[codingset] || get_bits_left(gb) < 0;
+        if (get_bits1(gb))
+            level = -level;
+    } else {
+        escape = decode210(gb);
+        if (escape != 2) {
+            index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
+            run   = vc1_index_decode_table[codingset][index][0];
+            level = vc1_index_decode_table[codingset][index][1];
+            lst   = index >= vc1_last_decode_table[codingset];
+            if (escape == 0) {
+                if (lst)
+                    level += vc1_last_delta_level_table[codingset][run];
+                else
+                    level += vc1_delta_level_table[codingset][run];
+            } else {
+                if (lst)
+                    run += vc1_last_delta_run_table[codingset][level] + 1;
+                else
+                    run += vc1_delta_run_table[codingset][level] + 1;
+            }
+            if (get_bits1(gb))
+                level = -level;
+        } else {
+            int sign;
+            lst = get_bits1(gb);
+            if (v->s.esc3_level_length == 0) {
+                if (v->pq < 8 || v->dquantfrm) { // table 59
+                    v->s.esc3_level_length = get_bits(gb, 3);
+                    if (!v->s.esc3_level_length)
+                        v->s.esc3_level_length = get_bits(gb, 2) + 8;
+                } else { // table 60
+                    v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;
+                }
+                v->s.esc3_run_length = 3 + get_bits(gb, 2);
+            }
+            run   = get_bits(gb, v->s.esc3_run_length);
+            sign  = get_bits1(gb);
+            level = get_bits(gb, v->s.esc3_level_length);
+            if (sign)
+                level = -level;
+        }
+    }
+
+    *last  = lst;
+    *skip  = run;
+    *value = level;
+}
+
+/** Decode intra block in intra frames - should be faster than decode_intra_block
+ * @param v VC1Context
+ * @param block block to decode
+ * @param[in] n subblock index
+ * @param coded are AC coeffs present or not
+ * @param codingset set of VLC to decode data
+ */
+static int vc1_decode_i_block(VC1Context *v, int16_t block[64], int n,
+                              int coded, int codingset)
+{
+    GetBitContext *gb = &v->s.gb;
+    MpegEncContext *s = &v->s;
+    int dc_pred_dir = 0; /* Direction of the DC prediction used */
+    int i;
+    int16_t *dc_val;
+    int16_t *ac_val, *ac_val2;
+    int dcdiff;
+
+    /* Get DC differential */
+    if (n < 4) {
+        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
+    } else {
+        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
+    }
+    if (dcdiff < 0) {
+        av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
+        return -1;
+    }
+    if (dcdiff) {
+        if (dcdiff == 119 /* ESC index value */) {
+            /* TODO: Optimize */
+            if (v->pq == 1)      dcdiff = get_bits(gb, 10);
+            else if (v->pq == 2) dcdiff = get_bits(gb, 9);
+            else                 dcdiff = get_bits(gb, 8);
+        } else {
+            if (v->pq == 1)
+                dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;
+            else if (v->pq == 2)
+                dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;
+        }
+        if (get_bits1(gb))
+            dcdiff = -dcdiff;
+    }
+
+    /* Prediction */
+    dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);
+    *dc_val = dcdiff;
+
+    /* Store the quantized DC coeff, used for prediction */
+    if (n < 4) {
+        block[0] = dcdiff * s->y_dc_scale;
+    } else {
+        block[0] = dcdiff * s->c_dc_scale;
+    }
+    /* Skip ? */
+    if (!coded) {
+        goto not_coded;
+    }
+
+    // AC Decoding
+    i = 1;
+
+    {
+        int last = 0, skip, value;
+        const uint8_t *zz_table;
+        int scale;
+        int k;
+
+        scale = v->pq * 2 + v->halfpq;
+
+        if (v->s.ac_pred) {
+            if (!dc_pred_dir)
+                zz_table = v->zz_8x8[2];
+            else
+                zz_table = v->zz_8x8[3];
+        } else
+            zz_table = v->zz_8x8[1];
+
+        ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;
+        ac_val2 = ac_val;
+        if (dc_pred_dir) // left
+            ac_val -= 16;
+        else // top
+            ac_val -= 16 * s->block_wrap[n];
+
+        while (!last) {
+            vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
+            i += skip;
+            if (i > 63)
+                break;
+            block[zz_table[i++]] = value;
+        }
+
+        /* apply AC prediction if needed */
+        if (s->ac_pred) {
+            if (dc_pred_dir) { // left
+                for (k = 1; k < 8; k++)
+                    block[k << v->left_blk_sh] += ac_val[k];
+            } else { // top
+                for (k = 1; k < 8; k++)
+                    block[k << v->top_blk_sh] += ac_val[k + 8];
+            }
+        }
+        /* save AC coeffs for further prediction */
+        for (k = 1; k < 8; k++) {
+            ac_val2[k]     = block[k << v->left_blk_sh];
+            ac_val2[k + 8] = block[k << v->top_blk_sh];
+        }
+
+        /* scale AC coeffs */
+        for (k = 1; k < 64; k++)
+            if (block[k]) {
+                block[k] *= scale;
+                if (!v->pquantizer)
+                    block[k] += (block[k] < 0) ? -v->pq : v->pq;
+            }
+
+        if (s->ac_pred) i = 63;
+    }
+
+not_coded:
+    if (!coded) {
+        int k, scale;
+        ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;
+        ac_val2 = ac_val;
+
+        i = 0;
+        scale = v->pq * 2 + v->halfpq;
+        memset(ac_val2, 0, 16 * 2);
+        if (dc_pred_dir) { // left
+            ac_val -= 16;
+            if (s->ac_pred)
+                memcpy(ac_val2, ac_val, 8 * 2);
+        } else { // top
+            ac_val -= 16 * s->block_wrap[n];
+            if (s->ac_pred)
+                memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
+        }
+
+        /* apply AC prediction if needed */
+        if (s->ac_pred) {
+            if (dc_pred_dir) { //left
+                for (k = 1; k < 8; k++) {
+                    block[k << v->left_blk_sh] = ac_val[k] * scale;
+                    if (!v->pquantizer && block[k << v->left_blk_sh])
+                        block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -v->pq : v->pq;
+                }
+            } else { // top
+                for (k = 1; k < 8; k++) {
+                    block[k << v->top_blk_sh] = ac_val[k + 8] * scale;
+                    if (!v->pquantizer && block[k << v->top_blk_sh])
+                        block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -v->pq : v->pq;
+                }
+            }
+            i = 63;
+        }
+    }
+    s->block_last_index[n] = i;
+
+    return 0;
+}
+
+/** Decode intra block in intra frames - should be faster than decode_intra_block
+ * @param v VC1Context
+ * @param block block to decode
+ * @param[in] n subblock number
+ * @param coded are AC coeffs present or not
+ * @param codingset set of VLC to decode data
+ * @param mquant quantizer value for this macroblock
+ */
+static int vc1_decode_i_block_adv(VC1Context *v, int16_t block[64], int n,
+                                  int coded, int codingset, int mquant)
+{
+    GetBitContext *gb = &v->s.gb;
+    MpegEncContext *s = &v->s;
+    int dc_pred_dir = 0; /* Direction of the DC prediction used */
+    int i;
+    int16_t *dc_val;
+    int16_t *ac_val, *ac_val2;
+    int dcdiff;
+    int a_avail = v->a_avail, c_avail = v->c_avail;
+    int use_pred = s->ac_pred;
+    int scale;
+    int q1, q2 = 0;
+    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+
+    /* Get DC differential */
+    if (n < 4) {
+        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
+    } else {
+        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
+    }
+    if (dcdiff < 0) {
+        av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
+        return -1;
+    }
+    if (dcdiff) {
+        if (dcdiff == 119 /* ESC index value */) {
+            /* TODO: Optimize */
+            if (mquant == 1)      dcdiff = get_bits(gb, 10);
+            else if (mquant == 2) dcdiff = get_bits(gb, 9);
+            else                  dcdiff = get_bits(gb, 8);
+        } else {
+            if (mquant == 1)
+                dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;
+            else if (mquant == 2)
+                dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;
+        }
+        if (get_bits1(gb))
+            dcdiff = -dcdiff;
+    }
+
+    /* Prediction */
+    dcdiff += ff_vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);
+    *dc_val = dcdiff;
+
+    /* Store the quantized DC coeff, used for prediction */
+    if (n < 4) {
+        block[0] = dcdiff * s->y_dc_scale;
+    } else {
+        block[0] = dcdiff * s->c_dc_scale;
+    }
+
+    //AC Decoding
+    i = 1;
+
+    /* check if AC is needed at all */
+    if (!a_avail && !c_avail)
+        use_pred = 0;
+    ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;
+    ac_val2 = ac_val;
+
+    scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);
+
+    if (dc_pred_dir) // left
+        ac_val -= 16;
+    else // top
+        ac_val -= 16 * s->block_wrap[n];
+
+    q1 = s->current_picture.qscale_table[mb_pos];
+    if ( dc_pred_dir && c_avail && mb_pos)
+        q2 = s->current_picture.qscale_table[mb_pos - 1];
+    if (!dc_pred_dir && a_avail && mb_pos >= s->mb_stride)
+        q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
+    if ( dc_pred_dir && n == 1)
+        q2 = q1;
+    if (!dc_pred_dir && n == 2)
+        q2 = q1;
+    if (n == 3)
+        q2 = q1;
+
+    if (coded) {
+        int last = 0, skip, value;
+        const uint8_t *zz_table;
+        int k;
+
+        if (v->s.ac_pred) {
+            if (!use_pred && v->fcm == ILACE_FRAME) {
+                zz_table = v->zzi_8x8;
+            } else {
+                if (!dc_pred_dir) // top
+                    zz_table = v->zz_8x8[2];
+                else // left
+                    zz_table = v->zz_8x8[3];
+            }
+        } else {
+            if (v->fcm != ILACE_FRAME)
+                zz_table = v->zz_8x8[1];
+            else
+                zz_table = v->zzi_8x8;
+        }
+
+        while (!last) {
+            vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
+            i += skip;
+            if (i > 63)
+                break;
+            block[zz_table[i++]] = value;
+        }
+
+        /* apply AC prediction if needed */
+        if (use_pred) {
+            /* scale predictors if needed*/
+            if (q2 && q1 != q2) {
+                q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
+                q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
+
+                if (q1 < 1)
+                    return AVERROR_INVALIDDATA;
+                if (dc_pred_dir) { // left
+                    for (k = 1; k < 8; k++)
+                        block[k << v->left_blk_sh] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
+                } else { // top
+                    for (k = 1; k < 8; k++)
+                        block[k << v->top_blk_sh] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
+                }
+            } else {
+                if (dc_pred_dir) { //left
+                    for (k = 1; k < 8; k++)
+                        block[k << v->left_blk_sh] += ac_val[k];
+                } else { //top
+                    for (k = 1; k < 8; k++)
+                        block[k << v->top_blk_sh] += ac_val[k + 8];
+                }
+            }
+        }
+        /* save AC coeffs for further prediction */
+        for (k = 1; k < 8; k++) {
+            ac_val2[k    ] = block[k << v->left_blk_sh];
+            ac_val2[k + 8] = block[k << v->top_blk_sh];
+        }
+
+        /* scale AC coeffs */
+        for (k = 1; k < 64; k++)
+            if (block[k]) {
+                block[k] *= scale;
+                if (!v->pquantizer)
+                    block[k] += (block[k] < 0) ? -mquant : mquant;
+            }
+
+        if (use_pred) i = 63;
+    } else { // no AC coeffs
+        int k;
+
+        memset(ac_val2, 0, 16 * 2);
+        if (dc_pred_dir) { // left
+            if (use_pred) {
+                memcpy(ac_val2, ac_val, 8 * 2);
+                if (q2 && q1 != q2) {
+                    q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
+                    q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
+                    if (q1 < 1)
+                        return AVERROR_INVALIDDATA;
+                    for (k = 1; k < 8; k++)
+                        ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
+                }
+            }
+        } else { // top
+            if (use_pred) {
+                memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
+                if (q2 && q1 != q2) {
+                    q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
+                    q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
+                    if (q1 < 1)
+                        return AVERROR_INVALIDDATA;
+                    for (k = 1; k < 8; k++)
+                        ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
+                }
+            }
+        }
+
+        /* apply AC prediction if needed */
+        if (use_pred) {
+            if (dc_pred_dir) { // left
+                for (k = 1; k < 8; k++) {
+                    block[k << v->left_blk_sh] = ac_val2[k] * scale;
+                    if (!v->pquantizer && block[k << v->left_blk_sh])
+                        block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -mquant : mquant;
+                }
+            } else { // top
+                for (k = 1; k < 8; k++) {
+                    block[k << v->top_blk_sh] = ac_val2[k + 8] * scale;
+                    if (!v->pquantizer && block[k << v->top_blk_sh])
+                        block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -mquant : mquant;
+                }
+            }
+            i = 63;
+        }
+    }
+    s->block_last_index[n] = i;
+
+    return 0;
+}
+
+/** Decode intra block in inter frames - more generic version than vc1_decode_i_block
+ * @param v VC1Context
+ * @param block block to decode
+ * @param[in] n subblock index
+ * @param coded are AC coeffs present or not
+ * @param mquant block quantizer
+ * @param codingset set of VLC to decode data
+ */
+static int vc1_decode_intra_block(VC1Context *v, int16_t block[64], int n,
+                                  int coded, int mquant, int codingset)
+{
+    GetBitContext *gb = &v->s.gb;
+    MpegEncContext *s = &v->s;
+    int dc_pred_dir = 0; /* Direction of the DC prediction used */
+    int i;
+    int16_t *dc_val;
+    int16_t *ac_val, *ac_val2;
+    int dcdiff;
+    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+    int a_avail = v->a_avail, c_avail = v->c_avail;
+    int use_pred = s->ac_pred;
+    int scale;
+    int q1, q2 = 0;
+
+    s->bdsp.clear_block(block);
+
+    /* XXX: Guard against dumb values of mquant */
+    mquant = (mquant < 1) ? 0 : ((mquant > 31) ? 31 : mquant);
+
+    /* Set DC scale - y and c use the same */
+    s->y_dc_scale = s->y_dc_scale_table[mquant];
+    s->c_dc_scale = s->c_dc_scale_table[mquant];
+
+    /* Get DC differential */
+    if (n < 4) {
+        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
+    } else {
+        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
+    }
+    if (dcdiff < 0) {
+        av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
+        return -1;
+    }
+    if (dcdiff) {
+        if (dcdiff == 119 /* ESC index value */) {
+            /* TODO: Optimize */
+            if (mquant == 1)      dcdiff = get_bits(gb, 10);
+            else if (mquant == 2) dcdiff = get_bits(gb, 9);
+            else                  dcdiff = get_bits(gb, 8);
+        } else {
+            if (mquant == 1)
+                dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;
+            else if (mquant == 2)
+                dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;
+        }
+        if (get_bits1(gb))
+            dcdiff = -dcdiff;
+    }
+
+    /* Prediction */
+    dcdiff += ff_vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);
+    *dc_val = dcdiff;
+
+    /* Store the quantized DC coeff, used for prediction */
+
+    if (n < 4) {
+        block[0] = dcdiff * s->y_dc_scale;
+    } else {
+        block[0] = dcdiff * s->c_dc_scale;
+    }
+
+    //AC Decoding
+    i = 1;
+
+    /* check if AC is needed at all and adjust direction if needed */
+    if (!a_avail) dc_pred_dir = 1;
+    if (!c_avail) dc_pred_dir = 0;
+    if (!a_avail && !c_avail) use_pred = 0;
+    ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
+    ac_val2 = ac_val;
+
+    scale = mquant * 2 + v->halfpq;
+
+    if (dc_pred_dir) //left
+        ac_val -= 16;
+    else //top
+        ac_val -= 16 * s->block_wrap[n];
+
+    q1 = s->current_picture.qscale_table[mb_pos];
+    if (dc_pred_dir && c_avail && mb_pos)
+        q2 = s->current_picture.qscale_table[mb_pos - 1];
+    if (!dc_pred_dir && a_avail && mb_pos >= s->mb_stride)
+        q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
+    if ( dc_pred_dir && n == 1)
+        q2 = q1;
+    if (!dc_pred_dir && n == 2)
+        q2 = q1;
+    if (n == 3) q2 = q1;
+
+    if (coded) {
+        int last = 0, skip, value;
+        int k;
+
+        while (!last) {
+            vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
+            i += skip;
+            if (i > 63)
+                break;
+            if (v->fcm == PROGRESSIVE)
+                block[v->zz_8x8[0][i++]] = value;
+            else {
+                if (use_pred && (v->fcm == ILACE_FRAME)) {
+                    if (!dc_pred_dir) // top
+                        block[v->zz_8x8[2][i++]] = value;
+                    else // left
+                        block[v->zz_8x8[3][i++]] = value;
+                } else {
+                    block[v->zzi_8x8[i++]] = value;
+                }
+            }
+        }
+
+        /* apply AC prediction if needed */
+        if (use_pred) {
+            /* scale predictors if needed*/
+            if (q2 && q1 != q2) {
+                q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
+                q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
+
+                if (q1 < 1)
+                    return AVERROR_INVALIDDATA;
+                if (dc_pred_dir) { // left
+                    for (k = 1; k < 8; k++)
+                        block[k << v->left_blk_sh] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
+                } else { //top
+                    for (k = 1; k < 8; k++)
+                        block[k << v->top_blk_sh] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
+                }
+            } else {
+                if (dc_pred_dir) { // left
+                    for (k = 1; k < 8; k++)
+                        block[k << v->left_blk_sh] += ac_val[k];
+                } else { // top
+                    for (k = 1; k < 8; k++)
+                        block[k << v->top_blk_sh] += ac_val[k + 8];
+                }
+            }
+        }
+        /* save AC coeffs for further prediction */
+        for (k = 1; k < 8; k++) {
+            ac_val2[k    ] = block[k << v->left_blk_sh];
+            ac_val2[k + 8] = block[k << v->top_blk_sh];
+        }
+
+        /* scale AC coeffs */
+        for (k = 1; k < 64; k++)
+            if (block[k]) {
+                block[k] *= scale;
+                if (!v->pquantizer)
+                    block[k] += (block[k] < 0) ? -mquant : mquant;
+            }
+
+        if (use_pred) i = 63;
+    } else { // no AC coeffs
+        int k;
+
+        memset(ac_val2, 0, 16 * 2);
+        if (dc_pred_dir) { // left
+            if (use_pred) {
+                memcpy(ac_val2, ac_val, 8 * 2);
+                if (q2 && q1 != q2) {
+                    q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
+                    q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
+                    if (q1 < 1)
+                        return AVERROR_INVALIDDATA;
+                    for (k = 1; k < 8; k++)
+                        ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
+                }
+            }
+        } else { // top
+            if (use_pred) {
+                memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
+                if (q2 && q1 != q2) {
+                    q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
+                    q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
+                    if (q1 < 1)
+                        return AVERROR_INVALIDDATA;
+                    for (k = 1; k < 8; k++)
+                        ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
+                }
+            }
+        }
+
+        /* apply AC prediction if needed */
+        if (use_pred) {
+            if (dc_pred_dir) { // left
+                for (k = 1; k < 8; k++) {
+                    block[k << v->left_blk_sh] = ac_val2[k] * scale;
+                    if (!v->pquantizer && block[k << v->left_blk_sh])
+                        block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -mquant : mquant;
+                }
+            } else { // top
+                for (k = 1; k < 8; k++) {
+                    block[k << v->top_blk_sh] = ac_val2[k + 8] * scale;
+                    if (!v->pquantizer && block[k << v->top_blk_sh])
+                        block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -mquant : mquant;
+                }
+            }
+            i = 63;
+        }
+    }
+    s->block_last_index[n] = i;
+
+    return 0;
+}
+
+/** Decode P block
+ */
+static int vc1_decode_p_block(VC1Context *v, int16_t block[64], int n,
+                              int mquant, int ttmb, int first_block,
+                              uint8_t *dst, int linesize, int skip_block,
+                              int *ttmb_out)
+{
+    MpegEncContext *s = &v->s;
+    GetBitContext *gb = &s->gb;
+    int i, j;
+    int subblkpat = 0;
+    int scale, off, idx, last, skip, value;
+    int ttblk = ttmb & 7;
+    int pat = 0;
+
+    s->bdsp.clear_block(block);
+
+    if (ttmb == -1) {
+        ttblk = ff_vc1_ttblk_to_tt[v->tt_index][get_vlc2(gb, ff_vc1_ttblk_vlc[v->tt_index].table, VC1_TTBLK_VLC_BITS, 1)];
+    }
+    if (ttblk == TT_4X4) {
+        subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);
+    }
+    if ((ttblk != TT_8X8 && ttblk != TT_4X4)
+        && ((v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))
+            || (!v->res_rtm_flag && !first_block))) {
+        subblkpat = decode012(gb);
+        if (subblkpat)
+            subblkpat ^= 3; // swap decoded pattern bits
+        if (ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM)
+            ttblk = TT_8X4;
+        if (ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT)
+            ttblk = TT_4X8;
+    }
+    scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);
+
+    // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT
+    if (ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {
+        subblkpat = 2 - (ttblk == TT_8X4_TOP);
+        ttblk     = TT_8X4;
+    }
+    if (ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {
+        subblkpat = 2 - (ttblk == TT_4X8_LEFT);
+        ttblk     = TT_4X8;
+    }
+    switch (ttblk) {
+    case TT_8X8:
+        pat  = 0xF;
+        i    = 0;
+        last = 0;
+        while (!last) {
+            vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
+            i += skip;
+            if (i > 63)
+                break;
+            if (!v->fcm)
+                idx = v->zz_8x8[0][i++];
+            else
+                idx = v->zzi_8x8[i++];
+            block[idx] = value * scale;
+            if (!v->pquantizer)
+                block[idx] += (block[idx] < 0) ? -mquant : mquant;
+        }
+        if (!skip_block) {
+            if (i == 1)
+                v->vc1dsp.vc1_inv_trans_8x8_dc(dst, linesize, block);
+            else {
+                v->vc1dsp.vc1_inv_trans_8x8(block);
+                s->idsp.add_pixels_clamped(block, dst, linesize);
+            }
+        }
+        break;
+    case TT_4X4:
+        pat = ~subblkpat & 0xF;
+        for (j = 0; j < 4; j++) {
+            last = subblkpat & (1 << (3 - j));
+            i    = 0;
+            off  = (j & 1) * 4 + (j & 2) * 16;
+            while (!last) {
+                vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
+                i += skip;
+                if (i > 15)
+                    break;
+                if (!v->fcm)
+                    idx = ff_vc1_simple_progressive_4x4_zz[i++];
+                else
+                    idx = ff_vc1_adv_interlaced_4x4_zz[i++];
+                block[idx + off] = value * scale;
+                if (!v->pquantizer)
+                    block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;
+            }
+            if (!(subblkpat & (1 << (3 - j))) && !skip_block) {
+                if (i == 1)
+                    v->vc1dsp.vc1_inv_trans_4x4_dc(dst + (j & 1) * 4 + (j & 2) * 2 * linesize, linesize, block + off);
+                else
+                    v->vc1dsp.vc1_inv_trans_4x4(dst + (j & 1) * 4 + (j & 2) *  2 * linesize, linesize, block + off);
+            }
+        }
+        break;
+    case TT_8X4:
+        pat = ~((subblkpat & 2) * 6 + (subblkpat & 1) * 3) & 0xF;
+        for (j = 0; j < 2; j++) {
+            last = subblkpat & (1 << (1 - j));
+            i    = 0;
+            off  = j * 32;
+            while (!last) {
+                vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
+                i += skip;
+                if (i > 31)
+                    break;
+                if (!v->fcm)
+                    idx = v->zz_8x4[i++] + off;
+                else
+                    idx = ff_vc1_adv_interlaced_8x4_zz[i++] + off;
+                block[idx] = value * scale;
+                if (!v->pquantizer)
+                    block[idx] += (block[idx] < 0) ? -mquant : mquant;
+            }
+            if (!(subblkpat & (1 << (1 - j))) && !skip_block) {
+                if (i == 1)
+                    v->vc1dsp.vc1_inv_trans_8x4_dc(dst + j * 4 * linesize, linesize, block + off);
+                else
+                    v->vc1dsp.vc1_inv_trans_8x4(dst + j * 4 * linesize, linesize, block + off);
+            }
+        }
+        break;
+    case TT_4X8:
+        pat = ~(subblkpat * 5) & 0xF;
+        for (j = 0; j < 2; j++) {
+            last = subblkpat & (1 << (1 - j));
+            i    = 0;
+            off  = j * 4;
+            while (!last) {
+                vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
+                i += skip;
+                if (i > 31)
+                    break;
+                if (!v->fcm)
+                    idx = v->zz_4x8[i++] + off;
+                else
+                    idx = ff_vc1_adv_interlaced_4x8_zz[i++] + off;
+                block[idx] = value * scale;
+                if (!v->pquantizer)
+                    block[idx] += (block[idx] < 0) ? -mquant : mquant;
+            }
+            if (!(subblkpat & (1 << (1 - j))) && !skip_block) {
+                if (i == 1)
+                    v->vc1dsp.vc1_inv_trans_4x8_dc(dst + j * 4, linesize, block + off);
+                else
+                    v->vc1dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);
+            }
+        }
+        break;
+    }
+    if (ttmb_out)
+        *ttmb_out |= ttblk << (n * 4);
+    return pat;
+}
+
+/** @} */ // Macroblock group
+
+static const int size_table  [6] = { 0, 2, 3, 4,  5,  8 };
+static const int offset_table[6] = { 0, 1, 3, 7, 15, 31 };
+
+/** Decode one P-frame MB
+ */
+static int vc1_decode_p_mb(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    GetBitContext *gb = &s->gb;
+    int i, j;
+    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+    int cbp; /* cbp decoding stuff */
+    int mqdiff, mquant; /* MB quantization */
+    int ttmb = v->ttfrm; /* MB Transform type */
+
+    int mb_has_coeffs = 1; /* last_flag */
+    int dmv_x, dmv_y; /* Differential MV components */
+    int index, index1; /* LUT indexes */
+    int val, sign; /* temp values */
+    int first_block = 1;
+    int dst_idx, off;
+    int skipped, fourmv;
+    int block_cbp = 0, pat, block_tt = 0, block_intra = 0;
+
+    mquant = v->pq; /* lossy initialization */
+
+    if (v->mv_type_is_raw)
+        fourmv = get_bits1(gb);
+    else
+        fourmv = v->mv_type_mb_plane[mb_pos];
+    if (v->skip_is_raw)
+        skipped = get_bits1(gb);
+    else
+        skipped = v->s.mbskip_table[mb_pos];
+
+    if (!fourmv) { /* 1MV mode */
+        if (!skipped) {
+            GET_MVDATA(dmv_x, dmv_y);
+
+            if (s->mb_intra) {
+                s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
+                s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
+            }
+            s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
+            ff_vc1_pred_mv(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0, 0);
+
+            /* FIXME Set DC val for inter block ? */
+            if (s->mb_intra && !mb_has_coeffs) {
+                GET_MQUANT();
+                s->ac_pred = get_bits1(gb);
+                cbp        = 0;
+            } else if (mb_has_coeffs) {
+                if (s->mb_intra)
+                    s->ac_pred = get_bits1(gb);
+                cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
+                GET_MQUANT();
+            } else {
+                mquant = v->pq;
+                cbp    = 0;
+            }
+            s->current_picture.qscale_table[mb_pos] = mquant;
+
+            if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)
+                ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,
+                                VC1_TTMB_VLC_BITS, 2);
+            if (!s->mb_intra) ff_vc1_mc_1mv(v, 0);
+            dst_idx = 0;
+            for (i = 0; i < 6; i++) {
+                s->dc_val[0][s->block_index[i]] = 0;
+                dst_idx += i >> 2;
+                val = ((cbp >> (5 - i)) & 1);
+                off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
+                v->mb_type[0][s->block_index[i]] = s->mb_intra;
+                if (s->mb_intra) {
+                    /* check if prediction blocks A and C are available */
+                    v->a_avail = v->c_avail = 0;
+                    if (i == 2 || i == 3 || !s->first_slice_line)
+                        v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
+                    if (i == 1 || i == 3 || s->mb_x)
+                        v->c_avail = v->mb_type[0][s->block_index[i] - 1];
+
+                    vc1_decode_intra_block(v, s->block[i], i, val, mquant,
+                                           (i & 4) ? v->codingset2 : v->codingset);
+                    if ((i>3) && (s->flags & CODEC_FLAG_GRAY))
+                        continue;
+                    v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);
+                    if (v->rangeredfrm)
+                        for (j = 0; j < 64; j++)
+                            s->block[i][j] <<= 1;
+                    s->idsp.put_signed_pixels_clamped(s->block[i],
+                                                      s->dest[dst_idx] + off,
+                                                      i & 4 ? s->uvlinesize
+                                                            : s->linesize);
+                    if (v->pq >= 9 && v->overlap) {
+                        if (v->c_avail)
+                            v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);
+                        if (v->a_avail)
+                            v->vc1dsp.vc1_v_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);
+                    }
+                    block_cbp   |= 0xF << (i << 2);
+                    block_intra |= 1 << i;
+                } else if (val) {
+                    pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block,
+                                             s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize,
+                                             (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);
+                    block_cbp |= pat << (i << 2);
+                    if (!v->ttmbf && ttmb < 8)
+                        ttmb = -1;
+                    first_block = 0;
+                }
+            }
+        } else { // skipped
+            s->mb_intra = 0;
+            for (i = 0; i < 6; i++) {
+                v->mb_type[0][s->block_index[i]] = 0;
+                s->dc_val[0][s->block_index[i]]  = 0;
+            }
+            s->current_picture.mb_type[mb_pos]      = MB_TYPE_SKIP;
+            s->current_picture.qscale_table[mb_pos] = 0;
+            ff_vc1_pred_mv(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0, 0);
+            ff_vc1_mc_1mv(v, 0);
+        }
+    } else { // 4MV mode
+        if (!skipped /* unskipped MB */) {
+            int intra_count = 0, coded_inter = 0;
+            int is_intra[6], is_coded[6];
+            /* Get CBPCY */
+            cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
+            for (i = 0; i < 6; i++) {
+                val = ((cbp >> (5 - i)) & 1);
+                s->dc_val[0][s->block_index[i]] = 0;
+                s->mb_intra                     = 0;
+                if (i < 4) {
+                    dmv_x = dmv_y = 0;
+                    s->mb_intra   = 0;
+                    mb_has_coeffs = 0;
+                    if (val) {
+                        GET_MVDATA(dmv_x, dmv_y);
+                    }
+                    ff_vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], 0, 0);
+                    if (!s->mb_intra)
+                        ff_vc1_mc_4mv_luma(v, i, 0, 0);
+                    intra_count += s->mb_intra;
+                    is_intra[i]  = s->mb_intra;
+                    is_coded[i]  = mb_has_coeffs;
+                }
+                if (i & 4) {
+                    is_intra[i] = (intra_count >= 3);
+                    is_coded[i] = val;
+                }
+                if (i == 4)
+                    ff_vc1_mc_4mv_chroma(v, 0);
+                v->mb_type[0][s->block_index[i]] = is_intra[i];
+                if (!coded_inter)
+                    coded_inter = !is_intra[i] & is_coded[i];
+            }
+            // if there are no coded blocks then don't do anything more
+            dst_idx = 0;
+            if (!intra_count && !coded_inter)
+                goto end;
+            GET_MQUANT();
+            s->current_picture.qscale_table[mb_pos] = mquant;
+            /* test if block is intra and has pred */
+            {
+                int intrapred = 0;
+                for (i = 0; i < 6; i++)
+                    if (is_intra[i]) {
+                        if (((!s->first_slice_line || (i == 2 || i == 3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])
+                            || ((s->mb_x || (i == 1 || i == 3)) && v->mb_type[0][s->block_index[i] - 1])) {
+                            intrapred = 1;
+                            break;
+                        }
+                    }
+                if (intrapred)
+                    s->ac_pred = get_bits1(gb);
+                else
+                    s->ac_pred = 0;
+            }
+            if (!v->ttmbf && coded_inter)
+                ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
+            for (i = 0; i < 6; i++) {
+                dst_idx    += i >> 2;
+                off         = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
+                s->mb_intra = is_intra[i];
+                if (is_intra[i]) {
+                    /* check if prediction blocks A and C are available */
+                    v->a_avail = v->c_avail = 0;
+                    if (i == 2 || i == 3 || !s->first_slice_line)
+                        v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
+                    if (i == 1 || i == 3 || s->mb_x)
+                        v->c_avail = v->mb_type[0][s->block_index[i] - 1];
+
+                    vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant,
+                                           (i & 4) ? v->codingset2 : v->codingset);
+                    if ((i>3) && (s->flags & CODEC_FLAG_GRAY))
+                        continue;
+                    v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);
+                    if (v->rangeredfrm)
+                        for (j = 0; j < 64; j++)
+                            s->block[i][j] <<= 1;
+                    s->idsp.put_signed_pixels_clamped(s->block[i],
+                                                      s->dest[dst_idx] + off,
+                                                      (i & 4) ? s->uvlinesize
+                                                              : s->linesize);
+                    if (v->pq >= 9 && v->overlap) {
+                        if (v->c_avail)
+                            v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);
+                        if (v->a_avail)
+                            v->vc1dsp.vc1_v_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);
+                    }
+                    block_cbp   |= 0xF << (i << 2);
+                    block_intra |= 1 << i;
+                } else if (is_coded[i]) {
+                    pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,
+                                             first_block, s->dest[dst_idx] + off,
+                                             (i & 4) ? s->uvlinesize : s->linesize,
+                                             (i & 4) && (s->flags & CODEC_FLAG_GRAY),
+                                             &block_tt);
+                    block_cbp |= pat << (i << 2);
+                    if (!v->ttmbf && ttmb < 8)
+                        ttmb = -1;
+                    first_block = 0;
+                }
+            }
+        } else { // skipped MB
+            s->mb_intra                               = 0;
+            s->current_picture.qscale_table[mb_pos] = 0;
+            for (i = 0; i < 6; i++) {
+                v->mb_type[0][s->block_index[i]] = 0;
+                s->dc_val[0][s->block_index[i]]  = 0;
+            }
+            for (i = 0; i < 4; i++) {
+                ff_vc1_pred_mv(v, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0], 0, 0);
+                ff_vc1_mc_4mv_luma(v, i, 0, 0);
+            }
+            ff_vc1_mc_4mv_chroma(v, 0);
+            s->current_picture.qscale_table[mb_pos] = 0;
+        }
+    }
+end:
+    v->cbp[s->mb_x]      = block_cbp;
+    v->ttblk[s->mb_x]    = block_tt;
+    v->is_intra[s->mb_x] = block_intra;
+
+    return 0;
+}
+
+/* Decode one macroblock in an interlaced frame p picture */
+
+static int vc1_decode_p_mb_intfr(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    GetBitContext *gb = &s->gb;
+    int i;
+    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+    int cbp = 0; /* cbp decoding stuff */
+    int mqdiff, mquant; /* MB quantization */
+    int ttmb = v->ttfrm; /* MB Transform type */
+
+    int mb_has_coeffs = 1; /* last_flag */
+    int dmv_x, dmv_y; /* Differential MV components */
+    int val; /* temp value */
+    int first_block = 1;
+    int dst_idx, off;
+    int skipped, fourmv = 0, twomv = 0;
+    int block_cbp = 0, pat, block_tt = 0;
+    int idx_mbmode = 0, mvbp;
+    int stride_y, fieldtx;
+
+    mquant = v->pq; /* Loosy initialization */
+
+    if (v->skip_is_raw)
+        skipped = get_bits1(gb);
+    else
+        skipped = v->s.mbskip_table[mb_pos];
+    if (!skipped) {
+        if (v->fourmvswitch)
+            idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_4MV_MBMODE_VLC_BITS, 2); // try getting this done
+        else
+            idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_NON4MV_MBMODE_VLC_BITS, 2); // in a single line
+        switch (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0]) {
+        /* store the motion vector type in a flag (useful later) */
+        case MV_PMODE_INTFR_4MV:
+            fourmv = 1;
+            v->blk_mv_type[s->block_index[0]] = 0;
+            v->blk_mv_type[s->block_index[1]] = 0;
+            v->blk_mv_type[s->block_index[2]] = 0;
+            v->blk_mv_type[s->block_index[3]] = 0;
+            break;
+        case MV_PMODE_INTFR_4MV_FIELD:
+            fourmv = 1;
+            v->blk_mv_type[s->block_index[0]] = 1;
+            v->blk_mv_type[s->block_index[1]] = 1;
+            v->blk_mv_type[s->block_index[2]] = 1;
+            v->blk_mv_type[s->block_index[3]] = 1;
+            break;
+        case MV_PMODE_INTFR_2MV_FIELD:
+            twomv = 1;
+            v->blk_mv_type[s->block_index[0]] = 1;
+            v->blk_mv_type[s->block_index[1]] = 1;
+            v->blk_mv_type[s->block_index[2]] = 1;
+            v->blk_mv_type[s->block_index[3]] = 1;
+            break;
+        case MV_PMODE_INTFR_1MV:
+            v->blk_mv_type[s->block_index[0]] = 0;
+            v->blk_mv_type[s->block_index[1]] = 0;
+            v->blk_mv_type[s->block_index[2]] = 0;
+            v->blk_mv_type[s->block_index[3]] = 0;
+            break;
+        }
+        if (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_INTRA) { // intra MB
+            for (i = 0; i < 4; i++) {
+                s->current_picture.motion_val[1][s->block_index[i]][0] = 0;
+                s->current_picture.motion_val[1][s->block_index[i]][1] = 0;
+            }
+            s->current_picture.mb_type[mb_pos]                     = MB_TYPE_INTRA;
+            s->mb_intra = v->is_intra[s->mb_x] = 1;
+            for (i = 0; i < 6; i++)
+                v->mb_type[0][s->block_index[i]] = 1;
+            fieldtx = v->fieldtx_plane[mb_pos] = get_bits1(gb);
+            mb_has_coeffs = get_bits1(gb);
+            if (mb_has_coeffs)
+                cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
+            v->s.ac_pred = v->acpred_plane[mb_pos] = get_bits1(gb);
+            GET_MQUANT();
+            s->current_picture.qscale_table[mb_pos] = mquant;
+            /* Set DC scale - y and c use the same (not sure if necessary here) */
+            s->y_dc_scale = s->y_dc_scale_table[mquant];
+            s->c_dc_scale = s->c_dc_scale_table[mquant];
+            dst_idx = 0;
+            for (i = 0; i < 6; i++) {
+                s->dc_val[0][s->block_index[i]] = 0;
+                dst_idx += i >> 2;
+                val = ((cbp >> (5 - i)) & 1);
+                v->mb_type[0][s->block_index[i]] = s->mb_intra;
+                v->a_avail = v->c_avail = 0;
+                if (i == 2 || i == 3 || !s->first_slice_line)
+                    v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
+                if (i == 1 || i == 3 || s->mb_x)
+                    v->c_avail = v->mb_type[0][s->block_index[i] - 1];
+
+                vc1_decode_intra_block(v, s->block[i], i, val, mquant,
+                                       (i & 4) ? v->codingset2 : v->codingset);
+                if ((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
+                v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);
+                if (i < 4) {
+                    stride_y = s->linesize << fieldtx;
+                    off = (fieldtx) ? ((i & 1) * 8) + ((i & 2) >> 1) * s->linesize : (i & 1) * 8 + 4 * (i & 2) * s->linesize;
+                } else {
+                    stride_y = s->uvlinesize;
+                    off = 0;
+                }
+                s->idsp.put_signed_pixels_clamped(s->block[i],
+                                                  s->dest[dst_idx] + off,
+                                                  stride_y);
+                //TODO: loop filter
+            }
+
+        } else { // inter MB
+            mb_has_coeffs = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][3];
+            if (mb_has_coeffs)
+                cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
+            if (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_2MV_FIELD) {
+                v->twomvbp = get_vlc2(gb, v->twomvbp_vlc->table, VC1_2MV_BLOCK_PATTERN_VLC_BITS, 1);
+            } else {
+                if ((ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_4MV)
+                    || (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_4MV_FIELD)) {
+                    v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);
+                }
+            }
+            s->mb_intra = v->is_intra[s->mb_x] = 0;
+            for (i = 0; i < 6; i++)
+                v->mb_type[0][s->block_index[i]] = 0;
+            fieldtx = v->fieldtx_plane[mb_pos] = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][1];
+            /* for all motion vector read MVDATA and motion compensate each block */
+            dst_idx = 0;
+            if (fourmv) {
+                mvbp = v->fourmvbp;
+                for (i = 0; i < 6; i++) {
+                    if (i < 4) {
+                        dmv_x = dmv_y = 0;
+                        val   = ((mvbp >> (3 - i)) & 1);
+                        if (val) {
+                            get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);
+                        }
+                        ff_vc1_pred_mv_intfr(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], 0);
+                        ff_vc1_mc_4mv_luma(v, i, 0, 0);
+                    } else if (i == 4) {
+                        ff_vc1_mc_4mv_chroma4(v, 0, 0, 0);
+                    }
+                }
+            } else if (twomv) {
+                mvbp  = v->twomvbp;
+                dmv_x = dmv_y = 0;
+                if (mvbp & 2) {
+                    get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);
+                }
+                ff_vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], 0);
+                ff_vc1_mc_4mv_luma(v, 0, 0, 0);
+                ff_vc1_mc_4mv_luma(v, 1, 0, 0);
+                dmv_x = dmv_y = 0;
+                if (mvbp & 1) {
+                    get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);
+                }
+                ff_vc1_pred_mv_intfr(v, 2, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], 0);
+                ff_vc1_mc_4mv_luma(v, 2, 0, 0);
+                ff_vc1_mc_4mv_luma(v, 3, 0, 0);
+                ff_vc1_mc_4mv_chroma4(v, 0, 0, 0);
+            } else {
+                mvbp = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][2];
+                dmv_x = dmv_y = 0;
+                if (mvbp) {
+                    get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);
+                }
+                ff_vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0);
+                ff_vc1_mc_1mv(v, 0);
+            }
+            if (cbp)
+                GET_MQUANT();  // p. 227
+            s->current_picture.qscale_table[mb_pos] = mquant;
+            if (!v->ttmbf && cbp)
+                ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
+            for (i = 0; i < 6; i++) {
+                s->dc_val[0][s->block_index[i]] = 0;
+                dst_idx += i >> 2;
+                val = ((cbp >> (5 - i)) & 1);
+                if (!fieldtx)
+                    off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
+                else
+                    off = (i & 4) ? 0 : ((i & 1) * 8 + ((i > 1) * s->linesize));
+                if (val) {
+                    pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,
+                                             first_block, s->dest[dst_idx] + off,
+                                             (i & 4) ? s->uvlinesize : (s->linesize << fieldtx),
+                                             (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);
+                    block_cbp |= pat << (i << 2);
+                    if (!v->ttmbf && ttmb < 8)
+                        ttmb = -1;
+                    first_block = 0;
+                }
+            }
+        }
+    } else { // skipped
+        s->mb_intra = v->is_intra[s->mb_x] = 0;
+        for (i = 0; i < 6; i++) {
+            v->mb_type[0][s->block_index[i]] = 0;
+            s->dc_val[0][s->block_index[i]] = 0;
+        }
+        s->current_picture.mb_type[mb_pos]      = MB_TYPE_SKIP;
+        s->current_picture.qscale_table[mb_pos] = 0;
+        v->blk_mv_type[s->block_index[0]] = 0;
+        v->blk_mv_type[s->block_index[1]] = 0;
+        v->blk_mv_type[s->block_index[2]] = 0;
+        v->blk_mv_type[s->block_index[3]] = 0;
+        ff_vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0);
+        ff_vc1_mc_1mv(v, 0);
+    }
+    if (s->mb_x == s->mb_width - 1)
+        memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0])*s->mb_stride);
+    return 0;
+}
+
+static int vc1_decode_p_mb_intfi(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    GetBitContext *gb = &s->gb;
+    int i;
+    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+    int cbp = 0; /* cbp decoding stuff */
+    int mqdiff, mquant; /* MB quantization */
+    int ttmb = v->ttfrm; /* MB Transform type */
+
+    int mb_has_coeffs = 1; /* last_flag */
+    int dmv_x, dmv_y; /* Differential MV components */
+    int val; /* temp values */
+    int first_block = 1;
+    int dst_idx, off;
+    int pred_flag;
+    int block_cbp = 0, pat, block_tt = 0;
+    int idx_mbmode = 0;
+
+    mquant = v->pq; /* Loosy initialization */
+
+    idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_IF_MBMODE_VLC_BITS, 2);
+    if (idx_mbmode <= 1) { // intra MB
+        s->mb_intra = v->is_intra[s->mb_x] = 1;
+        s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;
+        s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;
+        s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_INTRA;
+        GET_MQUANT();
+        s->current_picture.qscale_table[mb_pos] = mquant;
+        /* Set DC scale - y and c use the same (not sure if necessary here) */
+        s->y_dc_scale = s->y_dc_scale_table[mquant];
+        s->c_dc_scale = s->c_dc_scale_table[mquant];
+        v->s.ac_pred  = v->acpred_plane[mb_pos] = get_bits1(gb);
+        mb_has_coeffs = idx_mbmode & 1;
+        if (mb_has_coeffs)
+            cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_ICBPCY_VLC_BITS, 2);
+        dst_idx = 0;
+        for (i = 0; i < 6; i++) {
+            s->dc_val[0][s->block_index[i]]  = 0;
+            v->mb_type[0][s->block_index[i]] = 1;
+            dst_idx += i >> 2;
+            val = ((cbp >> (5 - i)) & 1);
+            v->a_avail = v->c_avail = 0;
+            if (i == 2 || i == 3 || !s->first_slice_line)
+                v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
+            if (i == 1 || i == 3 || s->mb_x)
+                v->c_avail = v->mb_type[0][s->block_index[i] - 1];
+
+            vc1_decode_intra_block(v, s->block[i], i, val, mquant,
+                                   (i & 4) ? v->codingset2 : v->codingset);
+            if ((i>3) && (s->flags & CODEC_FLAG_GRAY))
+                continue;
+            v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);
+            off  = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
+            s->idsp.put_signed_pixels_clamped(s->block[i],
+                                              s->dest[dst_idx] + off,
+                                              (i & 4) ? s->uvlinesize
+                                                      : s->linesize);
+            // TODO: loop filter
+        }
+    } else {
+        s->mb_intra = v->is_intra[s->mb_x] = 0;
+        s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16;
+        for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 0;
+        if (idx_mbmode <= 5) { // 1-MV
+            dmv_x = dmv_y = pred_flag = 0;
+            if (idx_mbmode & 1) {
+                get_mvdata_interlaced(v, &dmv_x, &dmv_y, &pred_flag);
+            }
+            ff_vc1_pred_mv(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], pred_flag, 0);
+            ff_vc1_mc_1mv(v, 0);
+            mb_has_coeffs = !(idx_mbmode & 2);
+        } else { // 4-MV
+            v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);
+            for (i = 0; i < 6; i++) {
+                if (i < 4) {
+                    dmv_x = dmv_y = pred_flag = 0;
+                    val   = ((v->fourmvbp >> (3 - i)) & 1);
+                    if (val) {
+                        get_mvdata_interlaced(v, &dmv_x, &dmv_y, &pred_flag);
+                    }
+                    ff_vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], pred_flag, 0);
+                    ff_vc1_mc_4mv_luma(v, i, 0, 0);
+                } else if (i == 4)
+                    ff_vc1_mc_4mv_chroma(v, 0);
+            }
+            mb_has_coeffs = idx_mbmode & 1;
+        }
+        if (mb_has_coeffs)
+            cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
+        if (cbp) {
+            GET_MQUANT();
+        }
+        s->current_picture.qscale_table[mb_pos] = mquant;
+        if (!v->ttmbf && cbp) {
+            ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
+        }
+        dst_idx = 0;
+        for (i = 0; i < 6; i++) {
+            s->dc_val[0][s->block_index[i]] = 0;
+            dst_idx += i >> 2;
+            val = ((cbp >> (5 - i)) & 1);
+            off = (i & 4) ? 0 : (i & 1) * 8 + (i & 2) * 4 * s->linesize;
+            if (val) {
+                pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,
+                                         first_block, s->dest[dst_idx] + off,
+                                         (i & 4) ? s->uvlinesize : s->linesize,
+                                         (i & 4) && (s->flags & CODEC_FLAG_GRAY),
+                                         &block_tt);
+                block_cbp |= pat << (i << 2);
+                if (!v->ttmbf && ttmb < 8) ttmb = -1;
+                first_block = 0;
+            }
+        }
+    }
+    if (s->mb_x == s->mb_width - 1)
+        memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride);
+    return 0;
+}
+
+/** Decode one B-frame MB (in Main profile)
+ */
+static void vc1_decode_b_mb(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    GetBitContext *gb = &s->gb;
+    int i, j;
+    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+    int cbp = 0; /* cbp decoding stuff */
+    int mqdiff, mquant; /* MB quantization */
+    int ttmb = v->ttfrm; /* MB Transform type */
+    int mb_has_coeffs = 0; /* last_flag */
+    int index, index1; /* LUT indexes */
+    int val, sign; /* temp values */
+    int first_block = 1;
+    int dst_idx, off;
+    int skipped, direct;
+    int dmv_x[2], dmv_y[2];
+    int bmvtype = BMV_TYPE_BACKWARD;
+
+    mquant      = v->pq; /* lossy initialization */
+    s->mb_intra = 0;
+
+    if (v->dmb_is_raw)
+        direct = get_bits1(gb);
+    else
+        direct = v->direct_mb_plane[mb_pos];
+    if (v->skip_is_raw)
+        skipped = get_bits1(gb);
+    else
+        skipped = v->s.mbskip_table[mb_pos];
+
+    dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;
+    for (i = 0; i < 6; i++) {
+        v->mb_type[0][s->block_index[i]] = 0;
+        s->dc_val[0][s->block_index[i]]  = 0;
+    }
+    s->current_picture.qscale_table[mb_pos] = 0;
+
+    if (!direct) {
+        if (!skipped) {
+            GET_MVDATA(dmv_x[0], dmv_y[0]);
+            dmv_x[1] = dmv_x[0];
+            dmv_y[1] = dmv_y[0];
+        }
+        if (skipped || !s->mb_intra) {
+            bmvtype = decode012(gb);
+            switch (bmvtype) {
+            case 0:
+                bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;
+                break;
+            case 1:
+                bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;
+                break;
+            case 2:
+                bmvtype  = BMV_TYPE_INTERPOLATED;
+                dmv_x[0] = dmv_y[0] = 0;
+            }
+        }
+    }
+    for (i = 0; i < 6; i++)
+        v->mb_type[0][s->block_index[i]] = s->mb_intra;
+
+    if (skipped) {
+        if (direct)
+            bmvtype = BMV_TYPE_INTERPOLATED;
+        ff_vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
+        vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
+        return;
+    }
+    if (direct) {
+        cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
+        GET_MQUANT();
+        s->mb_intra = 0;
+        s->current_picture.qscale_table[mb_pos] = mquant;
+        if (!v->ttmbf)
+            ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
+        dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;
+        ff_vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
+        vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
+    } else {
+        if (!mb_has_coeffs && !s->mb_intra) {
+            /* no coded blocks - effectively skipped */
+            ff_vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
+            vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
+            return;
+        }
+        if (s->mb_intra && !mb_has_coeffs) {
+            GET_MQUANT();
+            s->current_picture.qscale_table[mb_pos] = mquant;
+            s->ac_pred = get_bits1(gb);
+            cbp = 0;
+            ff_vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
+        } else {
+            if (bmvtype == BMV_TYPE_INTERPOLATED) {
+                GET_MVDATA(dmv_x[0], dmv_y[0]);
+                if (!mb_has_coeffs) {
+                    /* interpolated skipped block */
+                    ff_vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
+                    vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
+                    return;
+                }
+            }
+            ff_vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
+            if (!s->mb_intra) {
+                vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
+            }
+            if (s->mb_intra)
+                s->ac_pred = get_bits1(gb);
+            cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
+            GET_MQUANT();
+            s->current_picture.qscale_table[mb_pos] = mquant;
+            if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)
+                ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
+        }
+    }
+    dst_idx = 0;
+    for (i = 0; i < 6; i++) {
+        s->dc_val[0][s->block_index[i]] = 0;
+        dst_idx += i >> 2;
+        val = ((cbp >> (5 - i)) & 1);
+        off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
+        v->mb_type[0][s->block_index[i]] = s->mb_intra;
+        if (s->mb_intra) {
+            /* check if prediction blocks A and C are available */
+            v->a_avail = v->c_avail = 0;
+            if (i == 2 || i == 3 || !s->first_slice_line)
+                v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
+            if (i == 1 || i == 3 || s->mb_x)
+                v->c_avail = v->mb_type[0][s->block_index[i] - 1];
+
+            vc1_decode_intra_block(v, s->block[i], i, val, mquant,
+                                   (i & 4) ? v->codingset2 : v->codingset);
+            if ((i>3) && (s->flags & CODEC_FLAG_GRAY))
+                continue;
+            v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);
+            if (v->rangeredfrm)
+                for (j = 0; j < 64; j++)
+                    s->block[i][j] <<= 1;
+            s->idsp.put_signed_pixels_clamped(s->block[i],
+                                              s->dest[dst_idx] + off,
+                                              i & 4 ? s->uvlinesize
+                                                    : s->linesize);
+        } else if (val) {
+            vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,
+                               first_block, s->dest[dst_idx] + off,
+                               (i & 4) ? s->uvlinesize : s->linesize,
+                               (i & 4) && (s->flags & CODEC_FLAG_GRAY), NULL);
+            if (!v->ttmbf && ttmb < 8)
+                ttmb = -1;
+            first_block = 0;
+        }
+    }
+}
+
+/** Decode one B-frame MB (in interlaced field B picture)
+ */
+static void vc1_decode_b_mb_intfi(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    GetBitContext *gb = &s->gb;
+    int i, j;
+    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+    int cbp = 0; /* cbp decoding stuff */
+    int mqdiff, mquant; /* MB quantization */
+    int ttmb = v->ttfrm; /* MB Transform type */
+    int mb_has_coeffs = 0; /* last_flag */
+    int val; /* temp value */
+    int first_block = 1;
+    int dst_idx, off;
+    int fwd;
+    int dmv_x[2], dmv_y[2], pred_flag[2];
+    int bmvtype = BMV_TYPE_BACKWARD;
+    int idx_mbmode, interpmvp;
+
+    mquant      = v->pq; /* Loosy initialization */
+    s->mb_intra = 0;
+
+    idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_IF_MBMODE_VLC_BITS, 2);
+    if (idx_mbmode <= 1) { // intra MB
+        s->mb_intra = v->is_intra[s->mb_x] = 1;
+        s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
+        s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
+        s->current_picture.mb_type[mb_pos + v->mb_off]         = MB_TYPE_INTRA;
+        GET_MQUANT();
+        s->current_picture.qscale_table[mb_pos] = mquant;
+        /* Set DC scale - y and c use the same (not sure if necessary here) */
+        s->y_dc_scale = s->y_dc_scale_table[mquant];
+        s->c_dc_scale = s->c_dc_scale_table[mquant];
+        v->s.ac_pred  = v->acpred_plane[mb_pos] = get_bits1(gb);
+        mb_has_coeffs = idx_mbmode & 1;
+        if (mb_has_coeffs)
+            cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_ICBPCY_VLC_BITS, 2);
+        dst_idx = 0;
+        for (i = 0; i < 6; i++) {
+            s->dc_val[0][s->block_index[i]] = 0;
+            dst_idx += i >> 2;
+            val = ((cbp >> (5 - i)) & 1);
+            v->mb_type[0][s->block_index[i]] = s->mb_intra;
+            v->a_avail                       = v->c_avail = 0;
+            if (i == 2 || i == 3 || !s->first_slice_line)
+                v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
+            if (i == 1 || i == 3 || s->mb_x)
+                v->c_avail = v->mb_type[0][s->block_index[i] - 1];
+
+            vc1_decode_intra_block(v, s->block[i], i, val, mquant,
+                                   (i & 4) ? v->codingset2 : v->codingset);
+            if ((i>3) && (s->flags & CODEC_FLAG_GRAY))
+                continue;
+            v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);
+            if (v->rangeredfrm)
+                for (j = 0; j < 64; j++)
+                    s->block[i][j] <<= 1;
+            off  = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
+            s->idsp.put_signed_pixels_clamped(s->block[i],
+                                              s->dest[dst_idx] + off,
+                                              (i & 4) ? s->uvlinesize
+                                                      : s->linesize);
+            // TODO: yet to perform loop filter
+        }
+    } else {
+        s->mb_intra = v->is_intra[s->mb_x] = 0;
+        s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16;
+        for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 0;
+        if (v->fmb_is_raw)
+            fwd = v->forward_mb_plane[mb_pos] = get_bits1(gb);
+        else
+            fwd = v->forward_mb_plane[mb_pos];
+        if (idx_mbmode <= 5) { // 1-MV
+            dmv_x[0]     = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;
+            pred_flag[0] = pred_flag[1] = 0;
+            if (fwd)
+                bmvtype = BMV_TYPE_FORWARD;
+            else {
+                bmvtype = decode012(gb);
+                switch (bmvtype) {
+                case 0:
+                    bmvtype = BMV_TYPE_BACKWARD;
+                    break;
+                case 1:
+                    bmvtype = BMV_TYPE_DIRECT;
+                    break;
+                case 2:
+                    bmvtype   = BMV_TYPE_INTERPOLATED;
+                    interpmvp = get_bits1(gb);
+                }
+            }
+            v->bmvtype = bmvtype;
+            if (bmvtype != BMV_TYPE_DIRECT && idx_mbmode & 1) {
+                get_mvdata_interlaced(v, &dmv_x[bmvtype == BMV_TYPE_BACKWARD], &dmv_y[bmvtype == BMV_TYPE_BACKWARD], &pred_flag[bmvtype == BMV_TYPE_BACKWARD]);
+            }
+            if (bmvtype == BMV_TYPE_INTERPOLATED && interpmvp) {
+                get_mvdata_interlaced(v, &dmv_x[1], &dmv_y[1], &pred_flag[1]);
+            }
+            if (bmvtype == BMV_TYPE_DIRECT) {
+                dmv_x[0] = dmv_y[0] = pred_flag[0] = 0;
+                dmv_x[1] = dmv_y[1] = pred_flag[0] = 0;
+            }
+            ff_vc1_pred_b_mv_intfi(v, 0, dmv_x, dmv_y, 1, pred_flag);
+            vc1_b_mc(v, dmv_x, dmv_y, (bmvtype == BMV_TYPE_DIRECT), bmvtype);
+            mb_has_coeffs = !(idx_mbmode & 2);
+        } else { // 4-MV
+            if (fwd)
+                bmvtype = BMV_TYPE_FORWARD;
+            v->bmvtype  = bmvtype;
+            v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);
+            for (i = 0; i < 6; i++) {
+                if (i < 4) {
+                    dmv_x[0] = dmv_y[0] = pred_flag[0] = 0;
+                    dmv_x[1] = dmv_y[1] = pred_flag[1] = 0;
+                    val = ((v->fourmvbp >> (3 - i)) & 1);
+                    if (val) {
+                        get_mvdata_interlaced(v, &dmv_x[bmvtype == BMV_TYPE_BACKWARD],
+                                                 &dmv_y[bmvtype == BMV_TYPE_BACKWARD],
+                                             &pred_flag[bmvtype == BMV_TYPE_BACKWARD]);
+                    }
+                    ff_vc1_pred_b_mv_intfi(v, i, dmv_x, dmv_y, 0, pred_flag);
+                    ff_vc1_mc_4mv_luma(v, i, bmvtype == BMV_TYPE_BACKWARD, 0);
+                } else if (i == 4)
+                    ff_vc1_mc_4mv_chroma(v, bmvtype == BMV_TYPE_BACKWARD);
+            }
+            mb_has_coeffs = idx_mbmode & 1;
+        }
+        if (mb_has_coeffs)
+            cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
+        if (cbp) {
+            GET_MQUANT();
+        }
+        s->current_picture.qscale_table[mb_pos] = mquant;
+        if (!v->ttmbf && cbp) {
+            ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
+        }
+        dst_idx = 0;
+        for (i = 0; i < 6; i++) {
+            s->dc_val[0][s->block_index[i]] = 0;
+            dst_idx += i >> 2;
+            val = ((cbp >> (5 - i)) & 1);
+            off = (i & 4) ? 0 : (i & 1) * 8 + (i & 2) * 4 * s->linesize;
+            if (val) {
+                vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,
+                                   first_block, s->dest[dst_idx] + off,
+                                   (i & 4) ? s->uvlinesize : s->linesize,
+                                   (i & 4) && (s->flags & CODEC_FLAG_GRAY), NULL);
+                if (!v->ttmbf && ttmb < 8)
+                    ttmb = -1;
+                first_block = 0;
+            }
+        }
+    }
+}
+
+/** Decode one B-frame MB (in interlaced frame B picture)
+ */
+static int vc1_decode_b_mb_intfr(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    GetBitContext *gb = &s->gb;
+    int i, j;
+    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+    int cbp = 0; /* cbp decoding stuff */
+    int mqdiff, mquant; /* MB quantization */
+    int ttmb = v->ttfrm; /* MB Transform type */
+    int mvsw = 0; /* motion vector switch */
+    int mb_has_coeffs = 1; /* last_flag */
+    int dmv_x, dmv_y; /* Differential MV components */
+    int val; /* temp value */
+    int first_block = 1;
+    int dst_idx, off;
+    int skipped, direct, twomv = 0;
+    int block_cbp = 0, pat, block_tt = 0;
+    int idx_mbmode = 0, mvbp;
+    int stride_y, fieldtx;
+    int bmvtype = BMV_TYPE_BACKWARD;
+    int dir, dir2;
+
+    mquant = v->pq; /* Lossy initialization */
+    s->mb_intra = 0;
+    if (v->skip_is_raw)
+        skipped = get_bits1(gb);
+    else
+        skipped = v->s.mbskip_table[mb_pos];
+
+    if (!skipped) {
+        idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_NON4MV_MBMODE_VLC_BITS, 2);
+        if (ff_vc1_mbmode_intfrp[0][idx_mbmode][0] == MV_PMODE_INTFR_2MV_FIELD) {
+            twomv = 1;
+            v->blk_mv_type[s->block_index[0]] = 1;
+            v->blk_mv_type[s->block_index[1]] = 1;
+            v->blk_mv_type[s->block_index[2]] = 1;
+            v->blk_mv_type[s->block_index[3]] = 1;
+        } else {
+            v->blk_mv_type[s->block_index[0]] = 0;
+            v->blk_mv_type[s->block_index[1]] = 0;
+            v->blk_mv_type[s->block_index[2]] = 0;
+            v->blk_mv_type[s->block_index[3]] = 0;
+        }
+    }
+
+    if (v->dmb_is_raw)
+        direct = get_bits1(gb);
+    else
+        direct = v->direct_mb_plane[mb_pos];
+
+    if (direct) {
+        s->mv[0][0][0] = s->current_picture.motion_val[0][s->block_index[0]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][0], v->bfraction, 0, s->quarter_sample);
+        s->mv[0][0][1] = s->current_picture.motion_val[0][s->block_index[0]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][1], v->bfraction, 0, s->quarter_sample);
+        s->mv[1][0][0] = s->current_picture.motion_val[1][s->block_index[0]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][0], v->bfraction, 1, s->quarter_sample);
+        s->mv[1][0][1] = s->current_picture.motion_val[1][s->block_index[0]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][1], v->bfraction, 1, s->quarter_sample);
+
+        if (twomv) {
+            s->mv[0][2][0] = s->current_picture.motion_val[0][s->block_index[2]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][0], v->bfraction, 0, s->quarter_sample);
+            s->mv[0][2][1] = s->current_picture.motion_val[0][s->block_index[2]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][1], v->bfraction, 0, s->quarter_sample);
+            s->mv[1][2][0] = s->current_picture.motion_val[1][s->block_index[2]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][0], v->bfraction, 1, s->quarter_sample);
+            s->mv[1][2][1] = s->current_picture.motion_val[1][s->block_index[2]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][1], v->bfraction, 1, s->quarter_sample);
+
+            for (i = 1; i < 4; i += 2) {
+                s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0] = s->mv[0][i-1][0];
+                s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1] = s->mv[0][i-1][1];
+                s->mv[1][i][0] = s->current_picture.motion_val[1][s->block_index[i]][0] = s->mv[1][i-1][0];
+                s->mv[1][i][1] = s->current_picture.motion_val[1][s->block_index[i]][1] = s->mv[1][i-1][1];
+            }
+        } else {
+            for (i = 1; i < 4; i++) {
+                s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0] = s->mv[0][0][0];
+                s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1] = s->mv[0][0][1];
+                s->mv[1][i][0] = s->current_picture.motion_val[1][s->block_index[i]][0] = s->mv[1][0][0];
+                s->mv[1][i][1] = s->current_picture.motion_val[1][s->block_index[i]][1] = s->mv[1][0][1];
+            }
+        }
+    }
+
+    if (ff_vc1_mbmode_intfrp[0][idx_mbmode][0] == MV_PMODE_INTFR_INTRA) { // intra MB
+        for (i = 0; i < 4; i++) {
+            s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0] = 0;
+            s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1] = 0;
+            s->mv[1][i][0] = s->current_picture.motion_val[1][s->block_index[i]][0] = 0;
+            s->mv[1][i][1] = s->current_picture.motion_val[1][s->block_index[i]][1] = 0;
+        }
+        s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
+        s->mb_intra = v->is_intra[s->mb_x] = 1;
+        for (i = 0; i < 6; i++)
+            v->mb_type[0][s->block_index[i]] = 1;
+        fieldtx = v->fieldtx_plane[mb_pos] = get_bits1(gb);
+        mb_has_coeffs = get_bits1(gb);
+        if (mb_has_coeffs)
+            cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
+        v->s.ac_pred = v->acpred_plane[mb_pos] = get_bits1(gb);
+        GET_MQUANT();
+        s->current_picture.qscale_table[mb_pos] = mquant;
+        /* Set DC scale - y and c use the same (not sure if necessary here) */
+        s->y_dc_scale = s->y_dc_scale_table[mquant];
+        s->c_dc_scale = s->c_dc_scale_table[mquant];
+        dst_idx = 0;
+        for (i = 0; i < 6; i++) {
+            s->dc_val[0][s->block_index[i]] = 0;
+            dst_idx += i >> 2;
+            val = ((cbp >> (5 - i)) & 1);
+            v->mb_type[0][s->block_index[i]] = s->mb_intra;
+            v->a_avail = v->c_avail = 0;
+            if (i == 2 || i == 3 || !s->first_slice_line)
+                v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
+            if (i == 1 || i == 3 || s->mb_x)
+                v->c_avail = v->mb_type[0][s->block_index[i] - 1];
+
+            vc1_decode_intra_block(v, s->block[i], i, val, mquant,
+                                   (i & 4) ? v->codingset2 : v->codingset);
+            if (i > 3 && (s->flags & CODEC_FLAG_GRAY))
+                continue;
+            v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);
+            if (i < 4) {
+                stride_y = s->linesize << fieldtx;
+                off = (fieldtx) ? ((i & 1) * 8) + ((i & 2) >> 1) * s->linesize : (i & 1) * 8 + 4 * (i & 2) * s->linesize;
+            } else {
+                stride_y = s->uvlinesize;
+                off = 0;
+            }
+            s->idsp.put_signed_pixels_clamped(s->block[i],
+                                              s->dest[dst_idx] + off,
+                                              stride_y);
+        }
+    } else {
+        s->mb_intra = v->is_intra[s->mb_x] = 0;
+        if (!direct) {
+            if (skipped || !s->mb_intra) {
+                bmvtype = decode012(gb);
+                switch (bmvtype) {
+                case 0:
+                    bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;
+                    break;
+                case 1:
+                    bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;
+                    break;
+                case 2:
+                    bmvtype  = BMV_TYPE_INTERPOLATED;
+                }
+            }
+
+            if (twomv && bmvtype != BMV_TYPE_INTERPOLATED)
+                mvsw = get_bits1(gb);
+        }
+
+        if (!skipped) { // inter MB
+            mb_has_coeffs = ff_vc1_mbmode_intfrp[0][idx_mbmode][3];
+            if (mb_has_coeffs)
+                cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
+            if (!direct) {
+                if (bmvtype == BMV_TYPE_INTERPOLATED && twomv) {
+                    v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);
+                } else if (bmvtype == BMV_TYPE_INTERPOLATED || twomv) {
+                    v->twomvbp = get_vlc2(gb, v->twomvbp_vlc->table, VC1_2MV_BLOCK_PATTERN_VLC_BITS, 1);
+                }
+            }
+
+            for (i = 0; i < 6; i++)
+                v->mb_type[0][s->block_index[i]] = 0;
+            fieldtx = v->fieldtx_plane[mb_pos] = ff_vc1_mbmode_intfrp[0][idx_mbmode][1];
+            /* for all motion vector read MVDATA and motion compensate each block */
+            dst_idx = 0;
+            if (direct) {
+                if (twomv) {
+                    for (i = 0; i < 4; i++) {
+                        ff_vc1_mc_4mv_luma(v, i, 0, 0);
+                        ff_vc1_mc_4mv_luma(v, i, 1, 1);
+                    }
+                    ff_vc1_mc_4mv_chroma4(v, 0, 0, 0);
+                    ff_vc1_mc_4mv_chroma4(v, 1, 1, 1);
+                } else {
+                    ff_vc1_mc_1mv(v, 0);
+                    ff_vc1_interp_mc(v);
+                }
+            } else if (twomv && bmvtype == BMV_TYPE_INTERPOLATED) {
+                mvbp = v->fourmvbp;
+                for (i = 0; i < 4; i++) {
+                    dir = i==1 || i==3;
+                    dmv_x = dmv_y = 0;
+                    val = ((mvbp >> (3 - i)) & 1);
+                    if (val)
+                        get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);
+                    j = i > 1 ? 2 : 0;
+                    ff_vc1_pred_mv_intfr(v, j, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], dir);
+                    ff_vc1_mc_4mv_luma(v, j, dir, dir);
+                    ff_vc1_mc_4mv_luma(v, j+1, dir, dir);
+                }
+
+                ff_vc1_mc_4mv_chroma4(v, 0, 0, 0);
+                ff_vc1_mc_4mv_chroma4(v, 1, 1, 1);
+            } else if (bmvtype == BMV_TYPE_INTERPOLATED) {
+                mvbp = v->twomvbp;
+                dmv_x = dmv_y = 0;
+                if (mvbp & 2)
+                    get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);
+
+                ff_vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0);
+                ff_vc1_mc_1mv(v, 0);
+
+                dmv_x = dmv_y = 0;
+                if (mvbp & 1)
+                    get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);
+
+                ff_vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 1);
+                ff_vc1_interp_mc(v);
+            } else if (twomv) {
+                dir = bmvtype == BMV_TYPE_BACKWARD;
+                dir2 = dir;
+                if (mvsw)
+                    dir2 = !dir;
+                mvbp = v->twomvbp;
+                dmv_x = dmv_y = 0;
+                if (mvbp & 2)
+                    get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);
+                ff_vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], dir);
+
+                dmv_x = dmv_y = 0;
+                if (mvbp & 1)
+                    get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);
+                ff_vc1_pred_mv_intfr(v, 2, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], dir2);
+
+                if (mvsw) {
+                    for (i = 0; i < 2; i++) {
+                        s->mv[dir][i+2][0] = s->mv[dir][i][0] = s->current_picture.motion_val[dir][s->block_index[i+2]][0] = s->current_picture.motion_val[dir][s->block_index[i]][0];
+                        s->mv[dir][i+2][1] = s->mv[dir][i][1] = s->current_picture.motion_val[dir][s->block_index[i+2]][1] = s->current_picture.motion_val[dir][s->block_index[i]][1];
+                        s->mv[dir2][i+2][0] = s->mv[dir2][i][0] = s->current_picture.motion_val[dir2][s->block_index[i]][0] = s->current_picture.motion_val[dir2][s->block_index[i+2]][0];
+                        s->mv[dir2][i+2][1] = s->mv[dir2][i][1] = s->current_picture.motion_val[dir2][s->block_index[i]][1] = s->current_picture.motion_val[dir2][s->block_index[i+2]][1];
+                    }
+                } else {
+                    ff_vc1_pred_mv_intfr(v, 0, 0, 0, 2, v->range_x, v->range_y, v->mb_type[0], !dir);
+                    ff_vc1_pred_mv_intfr(v, 2, 0, 0, 2, v->range_x, v->range_y, v->mb_type[0], !dir);
+                }
+
+                ff_vc1_mc_4mv_luma(v, 0, dir, 0);
+                ff_vc1_mc_4mv_luma(v, 1, dir, 0);
+                ff_vc1_mc_4mv_luma(v, 2, dir2, 0);
+                ff_vc1_mc_4mv_luma(v, 3, dir2, 0);
+                ff_vc1_mc_4mv_chroma4(v, dir, dir2, 0);
+            } else {
+                dir = bmvtype == BMV_TYPE_BACKWARD;
+
+                mvbp = ff_vc1_mbmode_intfrp[0][idx_mbmode][2];
+                dmv_x = dmv_y = 0;
+                if (mvbp)
+                    get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);
+
+                ff_vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], dir);
+                v->blk_mv_type[s->block_index[0]] = 1;
+                v->blk_mv_type[s->block_index[1]] = 1;
+                v->blk_mv_type[s->block_index[2]] = 1;
+                v->blk_mv_type[s->block_index[3]] = 1;
+                ff_vc1_pred_mv_intfr(v, 0, 0, 0, 2, v->range_x, v->range_y, 0, !dir);
+                for (i = 0; i < 2; i++) {
+                    s->mv[!dir][i+2][0] = s->mv[!dir][i][0] = s->current_picture.motion_val[!dir][s->block_index[i+2]][0] = s->current_picture.motion_val[!dir][s->block_index[i]][0];
+                    s->mv[!dir][i+2][1] = s->mv[!dir][i][1] = s->current_picture.motion_val[!dir][s->block_index[i+2]][1] = s->current_picture.motion_val[!dir][s->block_index[i]][1];
+                }
+                ff_vc1_mc_1mv(v, dir);
+            }
+
+            if (cbp)
+                GET_MQUANT();  // p. 227
+            s->current_picture.qscale_table[mb_pos] = mquant;
+            if (!v->ttmbf && cbp)
+                ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
+            for (i = 0; i < 6; i++) {
+                s->dc_val[0][s->block_index[i]] = 0;
+                dst_idx += i >> 2;
+                val = ((cbp >> (5 - i)) & 1);
+                if (!fieldtx)
+                    off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
+                else
+                    off = (i & 4) ? 0 : ((i & 1) * 8 + ((i > 1) * s->linesize));
+                if (val) {
+                    pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,
+                                             first_block, s->dest[dst_idx] + off,
+                                             (i & 4) ? s->uvlinesize : (s->linesize << fieldtx),
+                                             (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);
+                    block_cbp |= pat << (i << 2);
+                    if (!v->ttmbf && ttmb < 8)
+                        ttmb = -1;
+                    first_block = 0;
+                }
+            }
+
+        } else { // skipped
+            dir = 0;
+            for (i = 0; i < 6; i++) {
+                v->mb_type[0][s->block_index[i]] = 0;
+                s->dc_val[0][s->block_index[i]] = 0;
+            }
+            s->current_picture.mb_type[mb_pos]      = MB_TYPE_SKIP;
+            s->current_picture.qscale_table[mb_pos] = 0;
+            v->blk_mv_type[s->block_index[0]] = 0;
+            v->blk_mv_type[s->block_index[1]] = 0;
+            v->blk_mv_type[s->block_index[2]] = 0;
+            v->blk_mv_type[s->block_index[3]] = 0;
+
+            if (!direct) {
+                if (bmvtype == BMV_TYPE_INTERPOLATED) {
+                    ff_vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0);
+                    ff_vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 1);
+                } else {
+                    dir = bmvtype == BMV_TYPE_BACKWARD;
+                    ff_vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], dir);
+                    if (mvsw) {
+                        int dir2 = dir;
+                        if (mvsw)
+                            dir2 = !dir;
+                        for (i = 0; i < 2; i++) {
+                            s->mv[dir][i+2][0] = s->mv[dir][i][0] = s->current_picture.motion_val[dir][s->block_index[i+2]][0] = s->current_picture.motion_val[dir][s->block_index[i]][0];
+                            s->mv[dir][i+2][1] = s->mv[dir][i][1] = s->current_picture.motion_val[dir][s->block_index[i+2]][1] = s->current_picture.motion_val[dir][s->block_index[i]][1];
+                            s->mv[dir2][i+2][0] = s->mv[dir2][i][0] = s->current_picture.motion_val[dir2][s->block_index[i]][0] = s->current_picture.motion_val[dir2][s->block_index[i+2]][0];
+                            s->mv[dir2][i+2][1] = s->mv[dir2][i][1] = s->current_picture.motion_val[dir2][s->block_index[i]][1] = s->current_picture.motion_val[dir2][s->block_index[i+2]][1];
+                        }
+                    } else {
+                        v->blk_mv_type[s->block_index[0]] = 1;
+                        v->blk_mv_type[s->block_index[1]] = 1;
+                        v->blk_mv_type[s->block_index[2]] = 1;
+                        v->blk_mv_type[s->block_index[3]] = 1;
+                        ff_vc1_pred_mv_intfr(v, 0, 0, 0, 2, v->range_x, v->range_y, 0, !dir);
+                        for (i = 0; i < 2; i++) {
+                            s->mv[!dir][i+2][0] = s->mv[!dir][i][0] = s->current_picture.motion_val[!dir][s->block_index[i+2]][0] = s->current_picture.motion_val[!dir][s->block_index[i]][0];
+                            s->mv[!dir][i+2][1] = s->mv[!dir][i][1] = s->current_picture.motion_val[!dir][s->block_index[i+2]][1] = s->current_picture.motion_val[!dir][s->block_index[i]][1];
+                        }
+                    }
+                }
+            }
+
+            ff_vc1_mc_1mv(v, dir);
+            if (direct || bmvtype == BMV_TYPE_INTERPOLATED) {
+                ff_vc1_interp_mc(v);
+            }
+        }
+    }
+    if (s->mb_x == s->mb_width - 1)
+        memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride);
+    v->cbp[s->mb_x]      = block_cbp;
+    v->ttblk[s->mb_x]    = block_tt;
+    return 0;
+}
+
+/** Decode blocks of I-frame
+ */
+static void vc1_decode_i_blocks(VC1Context *v)
+{
+    int k, j;
+    MpegEncContext *s = &v->s;
+    int cbp, val;
+    uint8_t *coded_val;
+    int mb_pos;
+
+    /* select codingmode used for VLC tables selection */
+    switch (v->y_ac_table_index) {
+    case 0:
+        v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
+        break;
+    case 1:
+        v->codingset = CS_HIGH_MOT_INTRA;
+        break;
+    case 2:
+        v->codingset = CS_MID_RATE_INTRA;
+        break;
+    }
+
+    switch (v->c_ac_table_index) {
+    case 0:
+        v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
+        break;
+    case 1:
+        v->codingset2 = CS_HIGH_MOT_INTER;
+        break;
+    case 2:
+        v->codingset2 = CS_MID_RATE_INTER;
+        break;
+    }
+
+    /* Set DC scale - y and c use the same */
+    s->y_dc_scale = s->y_dc_scale_table[v->pq];
+    s->c_dc_scale = s->c_dc_scale_table[v->pq];
+
+    //do frame decode
+    s->mb_x = s->mb_y = 0;
+    s->mb_intra         = 1;
+    s->first_slice_line = 1;
+    for (s->mb_y = 0; s->mb_y < s->end_mb_y; s->mb_y++) {
+        s->mb_x = 0;
+        init_block_index(v);
+        for (; s->mb_x < v->end_mb_x; s->mb_x++) {
+            uint8_t *dst[6];
+            ff_update_block_index(s);
+            dst[0] = s->dest[0];
+            dst[1] = dst[0] + 8;
+            dst[2] = s->dest[0] + s->linesize * 8;
+            dst[3] = dst[2] + 8;
+            dst[4] = s->dest[1];
+            dst[5] = s->dest[2];
+            s->bdsp.clear_blocks(s->block[0]);
+            mb_pos = s->mb_x + s->mb_y * s->mb_width;
+            s->current_picture.mb_type[mb_pos]                     = MB_TYPE_INTRA;
+            s->current_picture.qscale_table[mb_pos]                = v->pq;
+            s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
+            s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
+
+            // do actual MB decoding and displaying
+            cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
+            v->s.ac_pred = get_bits1(&v->s.gb);
+
+            for (k = 0; k < 6; k++) {
+                val = ((cbp >> (5 - k)) & 1);
+
+                if (k < 4) {
+                    int pred   = vc1_coded_block_pred(&v->s, k, &coded_val);
+                    val        = val ^ pred;
+                    *coded_val = val;
+                }
+                cbp |= val << (5 - k);
+
+                vc1_decode_i_block(v, s->block[k], k, val, (k < 4) ? v->codingset : v->codingset2);
+
+                if (k > 3 && (s->flags & CODEC_FLAG_GRAY))
+                    continue;
+                v->vc1dsp.vc1_inv_trans_8x8(s->block[k]);
+                if (v->pq >= 9 && v->overlap) {
+                    if (v->rangeredfrm)
+                        for (j = 0; j < 64; j++)
+                            s->block[k][j] <<= 1;
+                    s->idsp.put_signed_pixels_clamped(s->block[k], dst[k],
+                                                      k & 4 ? s->uvlinesize
+                                                            : s->linesize);
+                } else {
+                    if (v->rangeredfrm)
+                        for (j = 0; j < 64; j++)
+                            s->block[k][j] = (s->block[k][j] - 64) << 1;
+                    s->idsp.put_pixels_clamped(s->block[k], dst[k],
+                                               k & 4 ? s->uvlinesize
+                                                     : s->linesize);
+                }
+            }
+
+            if (v->pq >= 9 && v->overlap) {
+                if (s->mb_x) {
+                    v->vc1dsp.vc1_h_overlap(s->dest[0], s->linesize);
+                    v->vc1dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
+                    if (!(s->flags & CODEC_FLAG_GRAY)) {
+                        v->vc1dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
+                        v->vc1dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
+                    }
+                }
+                v->vc1dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
+                v->vc1dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
+                if (!s->first_slice_line) {
+                    v->vc1dsp.vc1_v_overlap(s->dest[0], s->linesize);
+                    v->vc1dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
+                    if (!(s->flags & CODEC_FLAG_GRAY)) {
+                        v->vc1dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
+                        v->vc1dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
+                    }
+                }
+                v->vc1dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
+                v->vc1dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
+            }
+            if (v->s.loop_filter)
+                ff_vc1_loop_filter_iblk(v, v->pq);
+
+            if (get_bits_count(&s->gb) > v->bits) {
+                ff_er_add_slice(&s->er, 0, 0, s->mb_x, s->mb_y, ER_MB_ERROR);
+                av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n",
+                       get_bits_count(&s->gb), v->bits);
+                return;
+            }
+        }
+        if (!v->s.loop_filter)
+            ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16);
+        else if (s->mb_y)
+            ff_mpeg_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);
+
+        s->first_slice_line = 0;
+    }
+    if (v->s.loop_filter)
+        ff_mpeg_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);
+
+    /* This is intentionally mb_height and not end_mb_y - unlike in advanced
+     * profile, these only differ are when decoding MSS2 rectangles. */
+    ff_er_add_slice(&s->er, 0, 0, s->mb_width - 1, s->mb_height - 1, ER_MB_END);
+}
+
+/** Decode blocks of I-frame for advanced profile
+ */
+static void vc1_decode_i_blocks_adv(VC1Context *v)
+{
+    int k;
+    MpegEncContext *s = &v->s;
+    int cbp, val;
+    uint8_t *coded_val;
+    int mb_pos;
+    int mquant = v->pq;
+    int mqdiff;
+    GetBitContext *gb = &s->gb;
+
+    /* select codingmode used for VLC tables selection */
+    switch (v->y_ac_table_index) {
+    case 0:
+        v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
+        break;
+    case 1:
+        v->codingset = CS_HIGH_MOT_INTRA;
+        break;
+    case 2:
+        v->codingset = CS_MID_RATE_INTRA;
+        break;
+    }
+
+    switch (v->c_ac_table_index) {
+    case 0:
+        v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
+        break;
+    case 1:
+        v->codingset2 = CS_HIGH_MOT_INTER;
+        break;
+    case 2:
+        v->codingset2 = CS_MID_RATE_INTER;
+        break;
+    }
+
+    // do frame decode
+    s->mb_x             = s->mb_y = 0;
+    s->mb_intra         = 1;
+    s->first_slice_line = 1;
+    s->mb_y             = s->start_mb_y;
+    if (s->start_mb_y) {
+        s->mb_x = 0;
+        init_block_index(v);
+        memset(&s->coded_block[s->block_index[0] - s->b8_stride], 0,
+               (1 + s->b8_stride) * sizeof(*s->coded_block));
+    }
+    for (; s->mb_y < s->end_mb_y; s->mb_y++) {
+        s->mb_x = 0;
+        init_block_index(v);
+        for (;s->mb_x < s->mb_width; s->mb_x++) {
+            int16_t (*block)[64] = v->block[v->cur_blk_idx];
+            ff_update_block_index(s);
+            s->bdsp.clear_blocks(block[0]);
+            mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+            s->current_picture.mb_type[mb_pos + v->mb_off]                         = MB_TYPE_INTRA;
+            s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;
+            s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;
+
+            // do actual MB decoding and displaying
+            if (v->fieldtx_is_raw)
+                v->fieldtx_plane[mb_pos] = get_bits1(&v->s.gb);
+            cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
+            if ( v->acpred_is_raw)
+                v->s.ac_pred = get_bits1(&v->s.gb);
+            else
+                v->s.ac_pred = v->acpred_plane[mb_pos];
+
+            if (v->condover == CONDOVER_SELECT && v->overflg_is_raw)
+                v->over_flags_plane[mb_pos] = get_bits1(&v->s.gb);
+
+            GET_MQUANT();
+
+            s->current_picture.qscale_table[mb_pos] = mquant;
+            /* Set DC scale - y and c use the same */
+            s->y_dc_scale = s->y_dc_scale_table[mquant];
+            s->c_dc_scale = s->c_dc_scale_table[mquant];
+
+            for (k = 0; k < 6; k++) {
+                val = ((cbp >> (5 - k)) & 1);
+
+                if (k < 4) {
+                    int pred   = vc1_coded_block_pred(&v->s, k, &coded_val);
+                    val        = val ^ pred;
+                    *coded_val = val;
+                }
+                cbp |= val << (5 - k);
+
+                v->a_avail = !s->first_slice_line || (k == 2 || k == 3);
+                v->c_avail = !!s->mb_x || (k == 1 || k == 3);
+
+                vc1_decode_i_block_adv(v, block[k], k, val,
+                                       (k < 4) ? v->codingset : v->codingset2, mquant);
+
+                if (k > 3 && (s->flags & CODEC_FLAG_GRAY))
+                    continue;
+                v->vc1dsp.vc1_inv_trans_8x8(block[k]);
+            }
+
+            ff_vc1_smooth_overlap_filter_iblk(v);
+            vc1_put_signed_blocks_clamped(v);
+            if (v->s.loop_filter)
+                ff_vc1_loop_filter_iblk_delayed(v, v->pq);
+
+            if (get_bits_count(&s->gb) > v->bits) {
+                // TODO: may need modification to handle slice coding
+                ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);
+                av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n",
+                       get_bits_count(&s->gb), v->bits);
+                return;
+            }
+        }
+        if (!v->s.loop_filter)
+            ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16);
+        else if (s->mb_y)
+            ff_mpeg_draw_horiz_band(s, (s->mb_y-1) * 16, 16);
+        s->first_slice_line = 0;
+    }
+
+    /* raw bottom MB row */
+    s->mb_x = 0;
+    init_block_index(v);
+
+    for (;s->mb_x < s->mb_width; s->mb_x++) {
+        ff_update_block_index(s);
+        vc1_put_signed_blocks_clamped(v);
+        if (v->s.loop_filter)
+            ff_vc1_loop_filter_iblk_delayed(v, v->pq);
+    }
+    if (v->s.loop_filter)
+        ff_mpeg_draw_horiz_band(s, (s->end_mb_y-1)*16, 16);
+    ff_er_add_slice(&s->er, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,
+                    (s->end_mb_y << v->field_mode) - 1, ER_MB_END);
+}
+
+static void vc1_decode_p_blocks(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    int apply_loop_filter;
+
+    /* select codingmode used for VLC tables selection */
+    switch (v->c_ac_table_index) {
+    case 0:
+        v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
+        break;
+    case 1:
+        v->codingset = CS_HIGH_MOT_INTRA;
+        break;
+    case 2:
+        v->codingset = CS_MID_RATE_INTRA;
+        break;
+    }
+
+    switch (v->c_ac_table_index) {
+    case 0:
+        v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
+        break;
+    case 1:
+        v->codingset2 = CS_HIGH_MOT_INTER;
+        break;
+    case 2:
+        v->codingset2 = CS_MID_RATE_INTER;
+        break;
+    }
+
+    apply_loop_filter   = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY) &&
+                          v->fcm == PROGRESSIVE;
+    s->first_slice_line = 1;
+    memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);
+    for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {
+        s->mb_x = 0;
+        init_block_index(v);
+        for (; s->mb_x < s->mb_width; s->mb_x++) {
+            ff_update_block_index(s);
+
+            if (v->fcm == ILACE_FIELD)
+                vc1_decode_p_mb_intfi(v);
+            else if (v->fcm == ILACE_FRAME)
+                vc1_decode_p_mb_intfr(v);
+            else vc1_decode_p_mb(v);
+            if (s->mb_y != s->start_mb_y && apply_loop_filter)
+                ff_vc1_apply_p_loop_filter(v);
+            if (get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
+                // TODO: may need modification to handle slice coding
+                ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);
+                av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n",
+                       get_bits_count(&s->gb), v->bits, s->mb_x, s->mb_y);
+                return;
+            }
+        }
+        memmove(v->cbp_base,      v->cbp,      sizeof(v->cbp_base[0])      * s->mb_stride);
+        memmove(v->ttblk_base,    v->ttblk,    sizeof(v->ttblk_base[0])    * s->mb_stride);
+        memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride);
+        memmove(v->luma_mv_base,  v->luma_mv,  sizeof(v->luma_mv_base[0])  * s->mb_stride);
+        if (s->mb_y != s->start_mb_y) ff_mpeg_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);
+        s->first_slice_line = 0;
+    }
+    if (apply_loop_filter) {
+        s->mb_x = 0;
+        init_block_index(v);
+        for (; s->mb_x < s->mb_width; s->mb_x++) {
+            ff_update_block_index(s);
+            ff_vc1_apply_p_loop_filter(v);
+        }
+    }
+    if (s->end_mb_y >= s->start_mb_y)
+        ff_mpeg_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);
+    ff_er_add_slice(&s->er, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,
+                    (s->end_mb_y << v->field_mode) - 1, ER_MB_END);
+}
+
+static void vc1_decode_b_blocks(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+
+    /* select codingmode used for VLC tables selection */
+    switch (v->c_ac_table_index) {
+    case 0:
+        v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
+        break;
+    case 1:
+        v->codingset = CS_HIGH_MOT_INTRA;
+        break;
+    case 2:
+        v->codingset = CS_MID_RATE_INTRA;
+        break;
+    }
+
+    switch (v->c_ac_table_index) {
+    case 0:
+        v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
+        break;
+    case 1:
+        v->codingset2 = CS_HIGH_MOT_INTER;
+        break;
+    case 2:
+        v->codingset2 = CS_MID_RATE_INTER;
+        break;
+    }
+
+    s->first_slice_line = 1;
+    for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {
+        s->mb_x = 0;
+        init_block_index(v);
+        for (; s->mb_x < s->mb_width; s->mb_x++) {
+            ff_update_block_index(s);
+
+            if (v->fcm == ILACE_FIELD)
+                vc1_decode_b_mb_intfi(v);
+            else if (v->fcm == ILACE_FRAME)
+                vc1_decode_b_mb_intfr(v);
+            else
+                vc1_decode_b_mb(v);
+            if (get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
+                // TODO: may need modification to handle slice coding
+                ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);
+                av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n",
+                       get_bits_count(&s->gb), v->bits, s->mb_x, s->mb_y);
+                return;
+            }
+            if (v->s.loop_filter)
+                ff_vc1_loop_filter_iblk(v, v->pq);
+        }
+        if (!v->s.loop_filter)
+            ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16);
+        else if (s->mb_y)
+            ff_mpeg_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);
+        s->first_slice_line = 0;
+    }
+    if (v->s.loop_filter)
+        ff_mpeg_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);
+    ff_er_add_slice(&s->er, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,
+                    (s->end_mb_y << v->field_mode) - 1, ER_MB_END);
+}
+
+static void vc1_decode_skip_blocks(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+
+    if (!v->s.last_picture.f->data[0])
+        return;
+
+    ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_width - 1, s->end_mb_y - 1, ER_MB_END);
+    s->first_slice_line = 1;
+    for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {
+        s->mb_x = 0;
+        init_block_index(v);
+        ff_update_block_index(s);
+        memcpy(s->dest[0], s->last_picture.f->data[0] + s->mb_y * 16 * s->linesize,   s->linesize   * 16);
+        memcpy(s->dest[1], s->last_picture.f->data[1] + s->mb_y *  8 * s->uvlinesize, s->uvlinesize *  8);
+        memcpy(s->dest[2], s->last_picture.f->data[2] + s->mb_y *  8 * s->uvlinesize, s->uvlinesize *  8);
+        ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16);
+        s->first_slice_line = 0;
+    }
+    s->pict_type = AV_PICTURE_TYPE_P;
+}
+
+void ff_vc1_decode_blocks(VC1Context *v)
+{
+
+    v->s.esc3_level_length = 0;
+    if (v->x8_type) {
+        ff_intrax8_decode_picture(&v->x8, 2*v->pq + v->halfpq, v->pq * !v->pquantizer);
+    } else {
+        v->cur_blk_idx     =  0;
+        v->left_blk_idx    = -1;
+        v->topleft_blk_idx =  1;
+        v->top_blk_idx     =  2;
+        switch (v->s.pict_type) {
+        case AV_PICTURE_TYPE_I:
+            if (v->profile == PROFILE_ADVANCED)
+                vc1_decode_i_blocks_adv(v);
+            else
+                vc1_decode_i_blocks(v);
+            break;
+        case AV_PICTURE_TYPE_P:
+            if (v->p_frame_skipped)
+                vc1_decode_skip_blocks(v);
+            else
+                vc1_decode_p_blocks(v);
+            break;
+        case AV_PICTURE_TYPE_B:
+            if (v->bi_type) {
+                if (v->profile == PROFILE_ADVANCED)
+                    vc1_decode_i_blocks_adv(v);
+                else
+                    vc1_decode_i_blocks(v);
+            } else
+                vc1_decode_b_blocks(v);
+            break;
+        }
+    }
+}
diff --git a/libavcodec/vc1_loopfilter.c b/libavcodec/vc1_loopfilter.c
new file mode 100644 (file)
index 0000000..1b76d83
--- /dev/null
@@ -0,0 +1,353 @@
+/*
+ * VC-1 and WMV3 decoder
+ * Copyright (c) 2011 Mashiat Sarker Shakkhar
+ * Copyright (c) 2006-2007 Konstantin Shishkov
+ * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * VC-1 and WMV3 loopfilter
+ */
+
+#include "avcodec.h"
+#include "mpegvideo.h"
+#include "vc1.h"
+#include "vc1dsp.h"
+
+void ff_vc1_loop_filter_iblk(VC1Context *v, int pq)
+{
+    MpegEncContext *s = &v->s;
+    int j;
+    if (!s->first_slice_line) {
+        v->vc1dsp.vc1_v_loop_filter16(s->dest[0], s->linesize, pq);
+        if (s->mb_x)
+            v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);
+        v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize + 8, s->linesize, pq);
+        for (j = 0; j < 2; j++) {
+            v->vc1dsp.vc1_v_loop_filter8(s->dest[j + 1], s->uvlinesize, pq);
+            if (s->mb_x)
+                v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize, s->uvlinesize, pq);
+        }
+    }
+    v->vc1dsp.vc1_v_loop_filter16(s->dest[0] + 8 * s->linesize, s->linesize, pq);
+
+    if (s->mb_y == s->end_mb_y - 1) {
+        if (s->mb_x) {
+            v->vc1dsp.vc1_h_loop_filter16(s->dest[0], s->linesize, pq);
+            v->vc1dsp.vc1_h_loop_filter8(s->dest[1], s->uvlinesize, pq);
+            v->vc1dsp.vc1_h_loop_filter8(s->dest[2], s->uvlinesize, pq);
+        }
+        v->vc1dsp.vc1_h_loop_filter16(s->dest[0] + 8, s->linesize, pq);
+    }
+}
+
+void ff_vc1_loop_filter_iblk_delayed(VC1Context *v, int pq)
+{
+    MpegEncContext *s = &v->s;
+    int j;
+
+    /* The loopfilter runs 1 row and 1 column behind the overlap filter, which
+     * means it runs two rows/cols behind the decoding loop. */
+    if (!s->first_slice_line) {
+        if (s->mb_x) {
+            if (s->mb_y >= s->start_mb_y + 2) {
+                v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 16 * s->linesize - 16, s->linesize, pq);
+
+                if (s->mb_x >= 2)
+                    v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize - 16, s->linesize, pq);
+                v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize - 8, s->linesize, pq);
+                for (j = 0; j < 2; j++) {
+                    v->vc1dsp.vc1_v_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize - 8, s->uvlinesize, pq);
+                    if (s->mb_x >= 2) {
+                        v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 16 * s->uvlinesize - 8, s->uvlinesize, pq);
+                    }
+                }
+            }
+            v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 8 * s->linesize - 16, s->linesize, pq);
+        }
+
+        if (s->mb_x == s->mb_width - 1) {
+            if (s->mb_y >= s->start_mb_y + 2) {
+                v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);
+
+                if (s->mb_x)
+                    v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize, s->linesize, pq);
+                v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize + 8, s->linesize, pq);
+                for (j = 0; j < 2; j++) {
+                    v->vc1dsp.vc1_v_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize, s->uvlinesize, pq);
+                    if (s->mb_x >= 2) {
+                        v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 16 * s->uvlinesize, s->uvlinesize, pq);
+                    }
+                }
+            }
+            v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 8 * s->linesize, s->linesize, pq);
+        }
+
+        if (s->mb_y == s->end_mb_y) {
+            if (s->mb_x) {
+                if (s->mb_x >= 2)
+                    v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize - 16, s->linesize, pq);
+                v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize - 8, s->linesize, pq);
+                if (s->mb_x >= 2) {
+                    for (j = 0; j < 2; j++) {
+                        v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize - 8, s->uvlinesize, pq);
+                    }
+                }
+            }
+
+            if (s->mb_x == s->mb_width - 1) {
+                if (s->mb_x)
+                    v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);
+                v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize + 8, s->linesize, pq);
+                if (s->mb_x) {
+                    for (j = 0; j < 2; j++) {
+                        v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize, s->uvlinesize, pq);
+                    }
+                }
+            }
+        }
+    }
+}
+
+void ff_vc1_smooth_overlap_filter_iblk(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    int mb_pos;
+
+    if (v->condover == CONDOVER_NONE)
+        return;
+
+    mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+
+    /* Within a MB, the horizontal overlap always runs before the vertical.
+     * To accomplish that, we run the H on left and internal borders of the
+     * currently decoded MB. Then, we wait for the next overlap iteration
+     * to do H overlap on the right edge of this MB, before moving over and
+     * running the V overlap. Therefore, the V overlap makes us trail by one
+     * MB col and the H overlap filter makes us trail by one MB row. This
+     * is reflected in the time at which we run the put_pixels loop. */
+    if (v->condover == CONDOVER_ALL || v->pq >= 9 || v->over_flags_plane[mb_pos]) {
+        if (s->mb_x && (v->condover == CONDOVER_ALL || v->pq >= 9 ||
+                        v->over_flags_plane[mb_pos - 1])) {
+            v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][1],
+                                      v->block[v->cur_blk_idx][0]);
+            v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][3],
+                                      v->block[v->cur_blk_idx][2]);
+            if (!(s->flags & CODEC_FLAG_GRAY)) {
+                v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][4],
+                                          v->block[v->cur_blk_idx][4]);
+                v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][5],
+                                          v->block[v->cur_blk_idx][5]);
+            }
+        }
+        v->vc1dsp.vc1_h_s_overlap(v->block[v->cur_blk_idx][0],
+                                  v->block[v->cur_blk_idx][1]);
+        v->vc1dsp.vc1_h_s_overlap(v->block[v->cur_blk_idx][2],
+                                  v->block[v->cur_blk_idx][3]);
+
+        if (s->mb_x == s->mb_width - 1) {
+            if (!s->first_slice_line && (v->condover == CONDOVER_ALL || v->pq >= 9 ||
+                                         v->over_flags_plane[mb_pos - s->mb_stride])) {
+                v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][2],
+                                          v->block[v->cur_blk_idx][0]);
+                v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][3],
+                                          v->block[v->cur_blk_idx][1]);
+                if (!(s->flags & CODEC_FLAG_GRAY)) {
+                    v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][4],
+                                              v->block[v->cur_blk_idx][4]);
+                    v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][5],
+                                              v->block[v->cur_blk_idx][5]);
+                }
+            }
+            v->vc1dsp.vc1_v_s_overlap(v->block[v->cur_blk_idx][0],
+                                      v->block[v->cur_blk_idx][2]);
+            v->vc1dsp.vc1_v_s_overlap(v->block[v->cur_blk_idx][1],
+                                      v->block[v->cur_blk_idx][3]);
+        }
+    }
+    if (s->mb_x && (v->condover == CONDOVER_ALL || v->over_flags_plane[mb_pos - 1])) {
+        if (!s->first_slice_line && (v->condover == CONDOVER_ALL || v->pq >= 9 ||
+                                     v->over_flags_plane[mb_pos - s->mb_stride - 1])) {
+            v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][2],
+                                      v->block[v->left_blk_idx][0]);
+            v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][3],
+                                      v->block[v->left_blk_idx][1]);
+            if (!(s->flags & CODEC_FLAG_GRAY)) {
+                v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][4],
+                                          v->block[v->left_blk_idx][4]);
+                v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][5],
+                                          v->block[v->left_blk_idx][5]);
+            }
+        }
+        v->vc1dsp.vc1_v_s_overlap(v->block[v->left_blk_idx][0],
+                                  v->block[v->left_blk_idx][2]);
+        v->vc1dsp.vc1_v_s_overlap(v->block[v->left_blk_idx][1],
+                                  v->block[v->left_blk_idx][3]);
+    }
+}
+
+static av_always_inline void vc1_apply_p_v_loop_filter(VC1Context *v, int block_num)
+{
+    MpegEncContext *s  = &v->s;
+    int mb_cbp         = v->cbp[s->mb_x - s->mb_stride],
+        block_cbp      = mb_cbp      >> (block_num * 4), bottom_cbp,
+        mb_is_intra    = v->is_intra[s->mb_x - s->mb_stride],
+        block_is_intra = mb_is_intra >> (block_num * 4), bottom_is_intra;
+    int idx, linesize  = block_num > 3 ? s->uvlinesize : s->linesize, ttblk;
+    uint8_t *dst;
+
+    if (block_num > 3) {
+        dst      = s->dest[block_num - 3];
+    } else {
+        dst      = s->dest[0] + (block_num & 1) * 8 + ((block_num & 2) * 4 - 8) * linesize;
+    }
+    if (s->mb_y != s->end_mb_y || block_num < 2) {
+        int16_t (*mv)[2];
+        int mv_stride;
+
+        if (block_num > 3) {
+            bottom_cbp      = v->cbp[s->mb_x]      >> (block_num * 4);
+            bottom_is_intra = v->is_intra[s->mb_x] >> (block_num * 4);
+            mv              = &v->luma_mv[s->mb_x - s->mb_stride];
+            mv_stride       = s->mb_stride;
+        } else {
+            bottom_cbp      = (block_num < 2) ? (mb_cbp               >> ((block_num + 2) * 4))
+                                              : (v->cbp[s->mb_x]      >> ((block_num - 2) * 4));
+            bottom_is_intra = (block_num < 2) ? (mb_is_intra          >> ((block_num + 2) * 4))
+                                              : (v->is_intra[s->mb_x] >> ((block_num - 2) * 4));
+            mv_stride       = s->b8_stride;
+            mv              = &s->current_picture.motion_val[0][s->block_index[block_num] - 2 * mv_stride];
+        }
+
+        if (bottom_is_intra & 1 || block_is_intra & 1 ||
+            mv[0][0] != mv[mv_stride][0] || mv[0][1] != mv[mv_stride][1]) {
+            v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);
+        } else {
+            idx = ((bottom_cbp >> 2) | block_cbp) & 3;
+            if (idx == 3) {
+                v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);
+            } else if (idx) {
+                if (idx == 1)
+                    v->vc1dsp.vc1_v_loop_filter4(dst + 4, linesize, v->pq);
+                else
+                    v->vc1dsp.vc1_v_loop_filter4(dst,     linesize, v->pq);
+            }
+        }
+    }
+
+    dst -= 4 * linesize;
+    ttblk = (v->ttblk[s->mb_x - s->mb_stride] >> (block_num * 4)) & 0xF;
+    if (ttblk == TT_4X4 || ttblk == TT_8X4) {
+        idx = (block_cbp | (block_cbp >> 2)) & 3;
+        if (idx == 3) {
+            v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);
+        } else if (idx) {
+            if (idx == 1)
+                v->vc1dsp.vc1_v_loop_filter4(dst + 4, linesize, v->pq);
+            else
+                v->vc1dsp.vc1_v_loop_filter4(dst,     linesize, v->pq);
+        }
+    }
+}
+
+static av_always_inline void vc1_apply_p_h_loop_filter(VC1Context *v, int block_num)
+{
+    MpegEncContext *s  = &v->s;
+    int mb_cbp         = v->cbp[s->mb_x - 1 - s->mb_stride],
+        block_cbp      = mb_cbp      >> (block_num * 4), right_cbp,
+        mb_is_intra    = v->is_intra[s->mb_x - 1 - s->mb_stride],
+        block_is_intra = mb_is_intra >> (block_num * 4), right_is_intra;
+    int idx, linesize  = block_num > 3 ? s->uvlinesize : s->linesize, ttblk;
+    uint8_t *dst;
+
+    if (block_num > 3) {
+        dst = s->dest[block_num - 3] - 8 * linesize;
+    } else {
+        dst = s->dest[0] + (block_num & 1) * 8 + ((block_num & 2) * 4 - 16) * linesize - 8;
+    }
+
+    if (s->mb_x != s->mb_width || !(block_num & 5)) {
+        int16_t (*mv)[2];
+
+        if (block_num > 3) {
+            right_cbp      = v->cbp[s->mb_x - s->mb_stride] >> (block_num * 4);
+            right_is_intra = v->is_intra[s->mb_x - s->mb_stride] >> (block_num * 4);
+            mv             = &v->luma_mv[s->mb_x - s->mb_stride - 1];
+        } else {
+            right_cbp      = (block_num & 1) ? (v->cbp[s->mb_x - s->mb_stride]      >> ((block_num - 1) * 4))
+                                             : (mb_cbp                              >> ((block_num + 1) * 4));
+            right_is_intra = (block_num & 1) ? (v->is_intra[s->mb_x - s->mb_stride] >> ((block_num - 1) * 4))
+                                             : (mb_is_intra                         >> ((block_num + 1) * 4));
+            mv             = &s->current_picture.motion_val[0][s->block_index[block_num] - s->b8_stride * 2 - 2];
+        }
+        if (block_is_intra & 1 || right_is_intra & 1 || mv[0][0] != mv[1][0] || mv[0][1] != mv[1][1]) {
+            v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);
+        } else {
+            idx = ((right_cbp >> 1) | block_cbp) & 5; // FIXME check
+            if (idx == 5) {
+                v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);
+            } else if (idx) {
+                if (idx == 1)
+                    v->vc1dsp.vc1_h_loop_filter4(dst + 4 * linesize, linesize, v->pq);
+                else
+                    v->vc1dsp.vc1_h_loop_filter4(dst,                linesize, v->pq);
+            }
+        }
+    }
+
+    dst -= 4;
+    ttblk = (v->ttblk[s->mb_x - s->mb_stride - 1] >> (block_num * 4)) & 0xf;
+    if (ttblk == TT_4X4 || ttblk == TT_4X8) {
+        idx = (block_cbp | (block_cbp >> 1)) & 5;
+        if (idx == 5) {
+            v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);
+        } else if (idx) {
+            if (idx == 1)
+                v->vc1dsp.vc1_h_loop_filter4(dst + linesize * 4, linesize, v->pq);
+            else
+                v->vc1dsp.vc1_h_loop_filter4(dst,                linesize, v->pq);
+        }
+    }
+}
+
+void ff_vc1_apply_p_loop_filter(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    int i;
+
+    for (i = 0; i < 6; i++) {
+        vc1_apply_p_v_loop_filter(v, i);
+    }
+
+    /* V always precedes H, therefore we run H one MB before V;
+     * at the end of a row, we catch up to complete the row */
+    if (s->mb_x) {
+        for (i = 0; i < 6; i++) {
+            vc1_apply_p_h_loop_filter(v, i);
+        }
+        if (s->mb_x == s->mb_width - 1) {
+            s->mb_x++;
+            ff_update_block_index(s);
+            for (i = 0; i < 6; i++) {
+                vc1_apply_p_h_loop_filter(v, i);
+            }
+        }
+    }
+}
diff --git a/libavcodec/vc1_mc.c b/libavcodec/vc1_mc.c
new file mode 100644 (file)
index 0000000..54c4e72
--- /dev/null
@@ -0,0 +1,948 @@
+/*
+ * VC-1 and WMV3 decoder
+ * Copyright (c) 2011 Mashiat Sarker Shakkhar
+ * Copyright (c) 2006-2007 Konstantin Shishkov
+ * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * VC-1 and WMV3 block decoding routines
+ */
+
+#include "avcodec.h"
+#include "h264chroma.h"
+#include "mathops.h"
+#include "mpegvideo.h"
+#include "vc1.h"
+
+/** Do motion compensation over 1 macroblock
+ * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
+ */
+void ff_vc1_mc_1mv(VC1Context *v, int dir)
+{
+    MpegEncContext *s = &v->s;
+    H264ChromaContext *h264chroma = &v->h264chroma;
+    uint8_t *srcY, *srcU, *srcV;
+    int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
+    int v_edge_pos = s->v_edge_pos >> v->field_mode;
+    int i;
+    uint8_t (*luty)[256], (*lutuv)[256];
+    int use_ic;
+
+    if ((!v->field_mode ||
+         (v->ref_field_type[dir] == 1 && v->cur_field_type == 1)) &&
+        !v->s.last_picture.f->data[0])
+        return;
+
+    mx = s->mv[dir][0][0];
+    my = s->mv[dir][0][1];
+
+    // store motion vectors for further use in B frames
+    if (s->pict_type == AV_PICTURE_TYPE_P) {
+        for (i = 0; i < 4; i++) {
+            s->current_picture.motion_val[1][s->block_index[i] + v->blocks_off][0] = mx;
+            s->current_picture.motion_val[1][s->block_index[i] + v->blocks_off][1] = my;
+        }
+    }
+
+    uvmx = (mx + ((mx & 3) == 3)) >> 1;
+    uvmy = (my + ((my & 3) == 3)) >> 1;
+    v->luma_mv[s->mb_x][0] = uvmx;
+    v->luma_mv[s->mb_x][1] = uvmy;
+
+    if (v->field_mode &&
+        v->cur_field_type != v->ref_field_type[dir]) {
+        my   = my   - 2 + 4 * v->cur_field_type;
+        uvmy = uvmy - 2 + 4 * v->cur_field_type;
+    }
+
+    // fastuvmc shall be ignored for interlaced frame picture
+    if (v->fastuvmc && (v->fcm != ILACE_FRAME)) {
+        uvmx = uvmx + ((uvmx < 0) ? (uvmx & 1) : -(uvmx & 1));
+        uvmy = uvmy + ((uvmy < 0) ? (uvmy & 1) : -(uvmy & 1));
+    }
+    if (!dir) {
+        if (v->field_mode && (v->cur_field_type != v->ref_field_type[dir]) && v->second_field) {
+            srcY = s->current_picture.f->data[0];
+            srcU = s->current_picture.f->data[1];
+            srcV = s->current_picture.f->data[2];
+            luty  = v->curr_luty;
+            lutuv = v->curr_lutuv;
+            use_ic = v->curr_use_ic;
+        } else {
+            srcY = s->last_picture.f->data[0];
+            srcU = s->last_picture.f->data[1];
+            srcV = s->last_picture.f->data[2];
+            luty  = v->last_luty;
+            lutuv = v->last_lutuv;
+            use_ic = v->last_use_ic;
+        }
+    } else {
+        srcY = s->next_picture.f->data[0];
+        srcU = s->next_picture.f->data[1];
+        srcV = s->next_picture.f->data[2];
+        luty  = v->next_luty;
+        lutuv = v->next_lutuv;
+        use_ic = v->next_use_ic;
+    }
+
+    if (!srcY || !srcU) {
+        av_log(v->s.avctx, AV_LOG_ERROR, "Referenced frame missing.\n");
+        return;
+    }
+
+    src_x   = s->mb_x * 16 + (mx   >> 2);
+    src_y   = s->mb_y * 16 + (my   >> 2);
+    uvsrc_x = s->mb_x *  8 + (uvmx >> 2);
+    uvsrc_y = s->mb_y *  8 + (uvmy >> 2);
+
+    if (v->profile != PROFILE_ADVANCED) {
+        src_x   = av_clip(  src_x, -16, s->mb_width  * 16);
+        src_y   = av_clip(  src_y, -16, s->mb_height * 16);
+        uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);
+        uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);
+    } else {
+        src_x   = av_clip(  src_x, -17, s->avctx->coded_width);
+        src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);
+        uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);
+        uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);
+    }
+
+    srcY += src_y   * s->linesize   + src_x;
+    srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
+    srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
+
+    if (v->field_mode && v->ref_field_type[dir]) {
+        srcY += s->current_picture_ptr->f->linesize[0];
+        srcU += s->current_picture_ptr->f->linesize[1];
+        srcV += s->current_picture_ptr->f->linesize[2];
+    }
+
+    /* for grayscale we should not try to read from unknown area */
+    if (s->flags & CODEC_FLAG_GRAY) {
+        srcU = s->edge_emu_buffer + 18 * s->linesize;
+        srcV = s->edge_emu_buffer + 18 * s->linesize;
+    }
+
+    if (v->rangeredfrm || use_ic
+        || s->h_edge_pos < 22 || v_edge_pos < 22
+        || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel * 3
+        || (unsigned)(src_y - 1)        > v_edge_pos    - (my&3) - 16 - 3) {
+        uint8_t *uvbuf = s->edge_emu_buffer + 19 * s->linesize;
+
+        srcY -= s->mspel * (1 + s->linesize);
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY,
+                                 s->linesize, s->linesize,
+                                 17 + s->mspel * 2, 17 + s->mspel * 2,
+                                 src_x - s->mspel, src_y - s->mspel,
+                                 s->h_edge_pos, v_edge_pos);
+        srcY = s->edge_emu_buffer;
+        s->vdsp.emulated_edge_mc(uvbuf, srcU,
+                                 s->uvlinesize, s->uvlinesize,
+                                 8 + 1, 8 + 1,
+                                 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);
+        s->vdsp.emulated_edge_mc(uvbuf + 16, srcV,
+                                 s->uvlinesize, s->uvlinesize,
+                                 8 + 1, 8 + 1,
+                                 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);
+        srcU = uvbuf;
+        srcV = uvbuf + 16;
+        /* if we deal with range reduction we need to scale source blocks */
+        if (v->rangeredfrm) {
+            int i, j;
+            uint8_t *src, *src2;
+
+            src = srcY;
+            for (j = 0; j < 17 + s->mspel * 2; j++) {
+                for (i = 0; i < 17 + s->mspel * 2; i++)
+                    src[i] = ((src[i] - 128) >> 1) + 128;
+                src += s->linesize;
+            }
+            src  = srcU;
+            src2 = srcV;
+            for (j = 0; j < 9; j++) {
+                for (i = 0; i < 9; i++) {
+                    src[i]  = ((src[i]  - 128) >> 1) + 128;
+                    src2[i] = ((src2[i] - 128) >> 1) + 128;
+                }
+                src  += s->uvlinesize;
+                src2 += s->uvlinesize;
+            }
+        }
+        /* if we deal with intensity compensation we need to scale source blocks */
+        if (use_ic) {
+            int i, j;
+            uint8_t *src, *src2;
+
+            src = srcY;
+            for (j = 0; j < 17 + s->mspel * 2; j++) {
+                int f = v->field_mode ? v->ref_field_type[dir] : ((j + src_y - s->mspel) & 1) ;
+                for (i = 0; i < 17 + s->mspel * 2; i++)
+                    src[i] = luty[f][src[i]];
+                src += s->linesize;
+            }
+            src  = srcU;
+            src2 = srcV;
+            for (j = 0; j < 9; j++) {
+                int f = v->field_mode ? v->ref_field_type[dir] : ((j + uvsrc_y) & 1);
+                for (i = 0; i < 9; i++) {
+                    src[i]  = lutuv[f][src[i]];
+                    src2[i] = lutuv[f][src2[i]];
+                }
+                src  += s->uvlinesize;
+                src2 += s->uvlinesize;
+            }
+        }
+        srcY += s->mspel * (1 + s->linesize);
+    }
+
+    if (s->mspel) {
+        dxy = ((my & 3) << 2) | (mx & 3);
+        v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0]    , srcY    , s->linesize, v->rnd);
+        v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
+        srcY += s->linesize * 8;
+        v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize    , srcY    , s->linesize, v->rnd);
+        v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
+    } else { // hpel mc - always used for luma
+        dxy = (my & 2) | ((mx & 2) >> 1);
+        if (!v->rnd)
+            s->hdsp.put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
+        else
+            s->hdsp.put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
+    }
+
+    if (s->flags & CODEC_FLAG_GRAY) return;
+    /* Chroma MC always uses qpel bilinear */
+    uvmx = (uvmx & 3) << 1;
+    uvmy = (uvmy & 3) << 1;
+    if (!v->rnd) {
+        h264chroma->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
+        h264chroma->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
+    } else {
+        v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
+        v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
+    }
+}
+
+static inline int median4(int a, int b, int c, int d)
+{
+    if (a < b) {
+        if (c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;
+        else       return (FFMIN(b, c) + FFMAX(a, d)) / 2;
+    } else {
+        if (c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;
+        else       return (FFMIN(a, c) + FFMAX(b, d)) / 2;
+    }
+}
+
+/** Do motion compensation for 4-MV macroblock - luminance block
+ */
+void ff_vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg)
+{
+    MpegEncContext *s = &v->s;
+    uint8_t *srcY;
+    int dxy, mx, my, src_x, src_y;
+    int off;
+    int fieldmv = (v->fcm == ILACE_FRAME) ? v->blk_mv_type[s->block_index[n]] : 0;
+    int v_edge_pos = s->v_edge_pos >> v->field_mode;
+    uint8_t (*luty)[256];
+    int use_ic;
+
+    if ((!v->field_mode ||
+         (v->ref_field_type[dir] == 1 && v->cur_field_type == 1)) &&
+        !v->s.last_picture.f->data[0])
+        return;
+
+    mx = s->mv[dir][n][0];
+    my = s->mv[dir][n][1];
+
+    if (!dir) {
+        if (v->field_mode && (v->cur_field_type != v->ref_field_type[dir]) && v->second_field) {
+            srcY = s->current_picture.f->data[0];
+            luty = v->curr_luty;
+            use_ic = v->curr_use_ic;
+        } else {
+            srcY = s->last_picture.f->data[0];
+            luty = v->last_luty;
+            use_ic = v->last_use_ic;
+        }
+    } else {
+        srcY = s->next_picture.f->data[0];
+        luty = v->next_luty;
+        use_ic = v->next_use_ic;
+    }
+
+    if (!srcY) {
+        av_log(v->s.avctx, AV_LOG_ERROR, "Referenced frame missing.\n");
+        return;
+    }
+
+    if (v->field_mode) {
+        if (v->cur_field_type != v->ref_field_type[dir])
+            my = my - 2 + 4 * v->cur_field_type;
+    }
+
+    if (s->pict_type == AV_PICTURE_TYPE_P && n == 3 && v->field_mode) {
+        int same_count = 0, opp_count = 0, k;
+        int chosen_mv[2][4][2], f;
+        int tx, ty;
+        for (k = 0; k < 4; k++) {
+            f = v->mv_f[0][s->block_index[k] + v->blocks_off];
+            chosen_mv[f][f ? opp_count : same_count][0] = s->mv[0][k][0];
+            chosen_mv[f][f ? opp_count : same_count][1] = s->mv[0][k][1];
+            opp_count  += f;
+            same_count += 1 - f;
+        }
+        f = opp_count > same_count;
+        switch (f ? opp_count : same_count) {
+        case 4:
+            tx = median4(chosen_mv[f][0][0], chosen_mv[f][1][0],
+                         chosen_mv[f][2][0], chosen_mv[f][3][0]);
+            ty = median4(chosen_mv[f][0][1], chosen_mv[f][1][1],
+                         chosen_mv[f][2][1], chosen_mv[f][3][1]);
+            break;
+        case 3:
+            tx = mid_pred(chosen_mv[f][0][0], chosen_mv[f][1][0], chosen_mv[f][2][0]);
+            ty = mid_pred(chosen_mv[f][0][1], chosen_mv[f][1][1], chosen_mv[f][2][1]);
+            break;
+        case 2:
+            tx = (chosen_mv[f][0][0] + chosen_mv[f][1][0]) / 2;
+            ty = (chosen_mv[f][0][1] + chosen_mv[f][1][1]) / 2;
+            break;
+        }
+        s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = tx;
+        s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = ty;
+        for (k = 0; k < 4; k++)
+            v->mv_f[1][s->block_index[k] + v->blocks_off] = f;
+    }
+
+    if (v->fcm == ILACE_FRAME) {  // not sure if needed for other types of picture
+        int qx, qy;
+        int width  = s->avctx->coded_width;
+        int height = s->avctx->coded_height >> 1;
+        if (s->pict_type == AV_PICTURE_TYPE_P) {
+            s->current_picture.motion_val[1][s->block_index[n] + v->blocks_off][0] = mx;
+            s->current_picture.motion_val[1][s->block_index[n] + v->blocks_off][1] = my;
+        }
+        qx = (s->mb_x * 16) + (mx >> 2);
+        qy = (s->mb_y *  8) + (my >> 3);
+
+        if (qx < -17)
+            mx -= 4 * (qx + 17);
+        else if (qx > width)
+            mx -= 4 * (qx - width);
+        if (qy < -18)
+            my -= 8 * (qy + 18);
+        else if (qy > height + 1)
+            my -= 8 * (qy - height - 1);
+    }
+
+    if ((v->fcm == ILACE_FRAME) && fieldmv)
+        off = ((n > 1) ? s->linesize : 0) + (n & 1) * 8;
+    else
+        off = s->linesize * 4 * (n & 2) + (n & 1) * 8;
+
+    src_x = s->mb_x * 16 + (n & 1) * 8 + (mx >> 2);
+    if (!fieldmv)
+        src_y = s->mb_y * 16 + (n & 2) * 4 + (my >> 2);
+    else
+        src_y = s->mb_y * 16 + ((n > 1) ? 1 : 0) + (my >> 2);
+
+    if (v->profile != PROFILE_ADVANCED) {
+        src_x = av_clip(src_x, -16, s->mb_width  * 16);
+        src_y = av_clip(src_y, -16, s->mb_height * 16);
+    } else {
+        src_x = av_clip(src_x, -17, s->avctx->coded_width);
+        if (v->fcm == ILACE_FRAME) {
+            if (src_y & 1)
+                src_y = av_clip(src_y, -17, s->avctx->coded_height + 1);
+            else
+                src_y = av_clip(src_y, -18, s->avctx->coded_height);
+        } else {
+            src_y = av_clip(src_y, -18, s->avctx->coded_height + 1);
+        }
+    }
+
+    srcY += src_y * s->linesize + src_x;
+    if (v->field_mode && v->ref_field_type[dir])
+        srcY += s->current_picture_ptr->f->linesize[0];
+
+    if (fieldmv && !(src_y & 1))
+        v_edge_pos--;
+    if (fieldmv && (src_y & 1) && src_y < 4)
+        src_y--;
+    if (v->rangeredfrm || use_ic
+        || s->h_edge_pos < 13 || v_edge_pos < 23
+        || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx & 3) - 8 - s->mspel * 2
+        || (unsigned)(src_y - (s->mspel << fieldmv)) > v_edge_pos - (my & 3) - ((8 + s->mspel * 2) << fieldmv)) {
+        srcY -= s->mspel * (1 + (s->linesize << fieldmv));
+        /* check emulate edge stride and offset */
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY,
+                                 s->linesize, s->linesize,
+                                 9 + s->mspel * 2, (9 + s->mspel * 2) << fieldmv,
+                                 src_x - s->mspel, src_y - (s->mspel << fieldmv),
+                                 s->h_edge_pos, v_edge_pos);
+        srcY = s->edge_emu_buffer;
+        /* if we deal with range reduction we need to scale source blocks */
+        if (v->rangeredfrm) {
+            int i, j;
+            uint8_t *src;
+
+            src = srcY;
+            for (j = 0; j < 9 + s->mspel * 2; j++) {
+                for (i = 0; i < 9 + s->mspel * 2; i++)
+                    src[i] = ((src[i] - 128) >> 1) + 128;
+                src += s->linesize << fieldmv;
+            }
+        }
+        /* if we deal with intensity compensation we need to scale source blocks */
+        if (use_ic) {
+            int i, j;
+            uint8_t *src;
+
+            src = srcY;
+            for (j = 0; j < 9 + s->mspel * 2; j++) {
+                int f = v->field_mode ? v->ref_field_type[dir] : (((j<<fieldmv)+src_y - (s->mspel << fieldmv)) & 1);
+                for (i = 0; i < 9 + s->mspel * 2; i++)
+                    src[i] = luty[f][src[i]];
+                src += s->linesize << fieldmv;
+            }
+        }
+        srcY += s->mspel * (1 + (s->linesize << fieldmv));
+    }
+
+    if (s->mspel) {
+        dxy = ((my & 3) << 2) | (mx & 3);
+        if (avg)
+            v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd);
+        else
+            v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd);
+    } else { // hpel mc - always used for luma
+        dxy = (my & 2) | ((mx & 2) >> 1);
+        if (!v->rnd)
+            s->hdsp.put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
+        else
+            s->hdsp.put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
+    }
+}
+
+static av_always_inline int get_chroma_mv(int *mvx, int *mvy, int *a, int flag, int *tx, int *ty)
+{
+    int idx, i;
+    static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
+
+    idx =  ((a[3] != flag) << 3)
+         | ((a[2] != flag) << 2)
+         | ((a[1] != flag) << 1)
+         |  (a[0] != flag);
+    if (!idx) {
+        *tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);
+        *ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);
+        return 4;
+    } else if (count[idx] == 1) {
+        switch (idx) {
+        case 0x1:
+            *tx = mid_pred(mvx[1], mvx[2], mvx[3]);
+            *ty = mid_pred(mvy[1], mvy[2], mvy[3]);
+            return 3;
+        case 0x2:
+            *tx = mid_pred(mvx[0], mvx[2], mvx[3]);
+            *ty = mid_pred(mvy[0], mvy[2], mvy[3]);
+            return 3;
+        case 0x4:
+            *tx = mid_pred(mvx[0], mvx[1], mvx[3]);
+            *ty = mid_pred(mvy[0], mvy[1], mvy[3]);
+            return 3;
+        case 0x8:
+            *tx = mid_pred(mvx[0], mvx[1], mvx[2]);
+            *ty = mid_pred(mvy[0], mvy[1], mvy[2]);
+            return 3;
+        }
+    } else if (count[idx] == 2) {
+        int t1 = 0, t2 = 0;
+        for (i = 0; i < 3; i++)
+            if (!a[i]) {
+                t1 = i;
+                break;
+            }
+        for (i = t1 + 1; i < 4; i++)
+            if (!a[i]) {
+                t2 = i;
+                break;
+            }
+        *tx = (mvx[t1] + mvx[t2]) / 2;
+        *ty = (mvy[t1] + mvy[t2]) / 2;
+        return 2;
+    } else {
+        return 0;
+    }
+    return -1;
+}
+
+/** Do motion compensation for 4-MV macroblock - both chroma blocks
+ */
+void ff_vc1_mc_4mv_chroma(VC1Context *v, int dir)
+{
+    MpegEncContext *s = &v->s;
+    H264ChromaContext *h264chroma = &v->h264chroma;
+    uint8_t *srcU, *srcV;
+    int uvmx, uvmy, uvsrc_x, uvsrc_y;
+    int k, tx = 0, ty = 0;
+    int mvx[4], mvy[4], intra[4], mv_f[4];
+    int valid_count;
+    int chroma_ref_type = v->cur_field_type;
+    int v_edge_pos = s->v_edge_pos >> v->field_mode;
+    uint8_t (*lutuv)[256];
+    int use_ic;
+
+    if (!v->field_mode && !v->s.last_picture.f->data[0])
+        return;
+    if (s->flags & CODEC_FLAG_GRAY)
+        return;
+
+    for (k = 0; k < 4; k++) {
+        mvx[k] = s->mv[dir][k][0];
+        mvy[k] = s->mv[dir][k][1];
+        intra[k] = v->mb_type[0][s->block_index[k]];
+        if (v->field_mode)
+            mv_f[k] = v->mv_f[dir][s->block_index[k] + v->blocks_off];
+    }
+
+    /* calculate chroma MV vector from four luma MVs */
+    if (!v->field_mode || (v->field_mode && !v->numref)) {
+        valid_count = get_chroma_mv(mvx, mvy, intra, 0, &tx, &ty);
+        chroma_ref_type = v->reffield;
+        if (!valid_count) {
+            s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;
+            s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;
+            v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;
+            return; //no need to do MC for intra blocks
+        }
+    } else {
+        int dominant = 0;
+        if (mv_f[0] + mv_f[1] + mv_f[2] + mv_f[3] > 2)
+            dominant = 1;
+        valid_count = get_chroma_mv(mvx, mvy, mv_f, dominant, &tx, &ty);
+        if (dominant)
+            chroma_ref_type = !v->cur_field_type;
+    }
+    if (v->field_mode && chroma_ref_type == 1 && v->cur_field_type == 1 && !v->s.last_picture.f->data[0])
+        return;
+    s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = tx;
+    s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = ty;
+    uvmx = (tx + ((tx & 3) == 3)) >> 1;
+    uvmy = (ty + ((ty & 3) == 3)) >> 1;
+
+    v->luma_mv[s->mb_x][0] = uvmx;
+    v->luma_mv[s->mb_x][1] = uvmy;
+
+    if (v->fastuvmc) {
+        uvmx = uvmx + ((uvmx < 0) ? (uvmx & 1) : -(uvmx & 1));
+        uvmy = uvmy + ((uvmy < 0) ? (uvmy & 1) : -(uvmy & 1));
+    }
+    // Field conversion bias
+    if (v->cur_field_type != chroma_ref_type)
+        uvmy += 2 - 4 * chroma_ref_type;
+
+    uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
+    uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
+
+    if (v->profile != PROFILE_ADVANCED) {
+        uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width  * 8);
+        uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
+    } else {
+        uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width  >> 1);
+        uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
+    }
+
+    if (!dir) {
+        if (v->field_mode && (v->cur_field_type != chroma_ref_type) && v->second_field) {
+            srcU = s->current_picture.f->data[1];
+            srcV = s->current_picture.f->data[2];
+            lutuv = v->curr_lutuv;
+            use_ic = v->curr_use_ic;
+        } else {
+            srcU = s->last_picture.f->data[1];
+            srcV = s->last_picture.f->data[2];
+            lutuv = v->last_lutuv;
+            use_ic = v->last_use_ic;
+        }
+    } else {
+        srcU = s->next_picture.f->data[1];
+        srcV = s->next_picture.f->data[2];
+        lutuv = v->next_lutuv;
+        use_ic = v->next_use_ic;
+    }
+
+    if (!srcU) {
+        av_log(v->s.avctx, AV_LOG_ERROR, "Referenced frame missing.\n");
+        return;
+    }
+
+    srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
+    srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
+
+    if (v->field_mode) {
+        if (chroma_ref_type) {
+            srcU += s->current_picture_ptr->f->linesize[1];
+            srcV += s->current_picture_ptr->f->linesize[2];
+        }
+    }
+
+    if (v->rangeredfrm || use_ic
+        || s->h_edge_pos < 18 || v_edge_pos < 18
+        || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
+        || (unsigned)uvsrc_y > (v_edge_pos    >> 1) - 9) {
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcU,
+                                 s->uvlinesize, s->uvlinesize,
+                                 8 + 1, 8 + 1, uvsrc_x, uvsrc_y,
+                                 s->h_edge_pos >> 1, v_edge_pos >> 1);
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV,
+                                 s->uvlinesize, s->uvlinesize,
+                                 8 + 1, 8 + 1, uvsrc_x, uvsrc_y,
+                                 s->h_edge_pos >> 1, v_edge_pos >> 1);
+        srcU = s->edge_emu_buffer;
+        srcV = s->edge_emu_buffer + 16;
+
+        /* if we deal with range reduction we need to scale source blocks */
+        if (v->rangeredfrm) {
+            int i, j;
+            uint8_t *src, *src2;
+
+            src  = srcU;
+            src2 = srcV;
+            for (j = 0; j < 9; j++) {
+                for (i = 0; i < 9; i++) {
+                    src[i]  = ((src[i]  - 128) >> 1) + 128;
+                    src2[i] = ((src2[i] - 128) >> 1) + 128;
+                }
+                src  += s->uvlinesize;
+                src2 += s->uvlinesize;
+            }
+        }
+        /* if we deal with intensity compensation we need to scale source blocks */
+        if (use_ic) {
+            int i, j;
+            uint8_t *src, *src2;
+
+            src  = srcU;
+            src2 = srcV;
+            for (j = 0; j < 9; j++) {
+                int f = v->field_mode ? chroma_ref_type : ((j + uvsrc_y) & 1);
+                for (i = 0; i < 9; i++) {
+                    src[i]  = lutuv[f][src[i]];
+                    src2[i] = lutuv[f][src2[i]];
+                }
+                src  += s->uvlinesize;
+                src2 += s->uvlinesize;
+            }
+        }
+    }
+
+    /* Chroma MC always uses qpel bilinear */
+    uvmx = (uvmx & 3) << 1;
+    uvmy = (uvmy & 3) << 1;
+    if (!v->rnd) {
+        h264chroma->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
+        h264chroma->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
+    } else {
+        v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
+        v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
+    }
+}
+
+/** Do motion compensation for 4-MV interlaced frame chroma macroblock (both U and V)
+ */
+void ff_vc1_mc_4mv_chroma4(VC1Context *v, int dir, int dir2, int avg)
+{
+    MpegEncContext *s = &v->s;
+    H264ChromaContext *h264chroma = &v->h264chroma;
+    uint8_t *srcU, *srcV;
+    int uvsrc_x, uvsrc_y;
+    int uvmx_field[4], uvmy_field[4];
+    int i, off, tx, ty;
+    int fieldmv = v->blk_mv_type[s->block_index[0]];
+    static const int s_rndtblfield[16] = { 0, 0, 1, 2, 4, 4, 5, 6, 2, 2, 3, 8, 6, 6, 7, 12 };
+    int v_dist = fieldmv ? 1 : 4; // vertical offset for lower sub-blocks
+    int v_edge_pos = s->v_edge_pos >> 1;
+    int use_ic;
+    uint8_t (*lutuv)[256];
+
+    if (s->flags & CODEC_FLAG_GRAY)
+        return;
+
+    for (i = 0; i < 4; i++) {
+        int d = i < 2 ? dir: dir2;
+        tx = s->mv[d][i][0];
+        uvmx_field[i] = (tx + ((tx & 3) == 3)) >> 1;
+        ty = s->mv[d][i][1];
+        if (fieldmv)
+            uvmy_field[i] = (ty >> 4) * 8 + s_rndtblfield[ty & 0xF];
+        else
+            uvmy_field[i] = (ty + ((ty & 3) == 3)) >> 1;
+    }
+
+    for (i = 0; i < 4; i++) {
+        off = (i & 1) * 4 + ((i & 2) ? v_dist * s->uvlinesize : 0);
+        uvsrc_x = s->mb_x * 8 +  (i & 1) * 4           + (uvmx_field[i] >> 2);
+        uvsrc_y = s->mb_y * 8 + ((i & 2) ? v_dist : 0) + (uvmy_field[i] >> 2);
+        // FIXME: implement proper pull-back (see vc1cropmv.c, vc1CROPMV_ChromaPullBack())
+        uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width  >> 1);
+        uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
+        if (i < 2 ? dir : dir2) {
+            srcU = s->next_picture.f->data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
+            srcV = s->next_picture.f->data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
+            lutuv  = v->next_lutuv;
+            use_ic = v->next_use_ic;
+        } else {
+            srcU = s->last_picture.f->data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
+            srcV = s->last_picture.f->data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
+            lutuv  = v->last_lutuv;
+            use_ic = v->last_use_ic;
+        }
+        uvmx_field[i] = (uvmx_field[i] & 3) << 1;
+        uvmy_field[i] = (uvmy_field[i] & 3) << 1;
+
+        if (fieldmv && !(uvsrc_y & 1))
+            v_edge_pos--;
+        if (fieldmv && (uvsrc_y & 1) && uvsrc_y < 2)
+            uvsrc_y--;
+        if (use_ic
+            || s->h_edge_pos < 10 || v_edge_pos < (5 << fieldmv)
+            || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 5
+            || (unsigned)uvsrc_y > v_edge_pos - (5 << fieldmv)) {
+            s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcU,
+                                     s->uvlinesize, s->uvlinesize,
+                                     5, (5 << fieldmv), uvsrc_x, uvsrc_y,
+                                     s->h_edge_pos >> 1, v_edge_pos);
+            s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV,
+                                     s->uvlinesize, s->uvlinesize,
+                                     5, (5 << fieldmv), uvsrc_x, uvsrc_y,
+                                     s->h_edge_pos >> 1, v_edge_pos);
+            srcU = s->edge_emu_buffer;
+            srcV = s->edge_emu_buffer + 16;
+
+            /* if we deal with intensity compensation we need to scale source blocks */
+            if (use_ic) {
+                int i, j;
+                uint8_t *src, *src2;
+
+                src  = srcU;
+                src2 = srcV;
+                for (j = 0; j < 5; j++) {
+                    int f = (uvsrc_y + (j << fieldmv)) & 1;
+                    for (i = 0; i < 5; i++) {
+                        src[i]  = lutuv[f][src[i]];
+                        src2[i] = lutuv[f][src2[i]];
+                    }
+                    src  += s->uvlinesize << fieldmv;
+                    src2 += s->uvlinesize << fieldmv;
+                }
+            }
+        }
+        if (avg) {
+            if (!v->rnd) {
+                h264chroma->avg_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
+                h264chroma->avg_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
+            } else {
+                v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
+                v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
+            }
+        } else {
+            if (!v->rnd) {
+                h264chroma->put_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
+                h264chroma->put_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
+            } else {
+                v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
+                v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
+            }
+        }
+    }
+}
+
+/** Motion compensation for direct or interpolated blocks in B-frames
+ */
+void ff_vc1_interp_mc(VC1Context *v)
+{
+    MpegEncContext *s = &v->s;
+    H264ChromaContext *h264chroma = &v->h264chroma;
+    uint8_t *srcY, *srcU, *srcV;
+    int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
+    int off, off_uv;
+    int v_edge_pos = s->v_edge_pos >> v->field_mode;
+    int use_ic = v->next_use_ic;
+
+    if (!v->field_mode && !v->s.next_picture.f->data[0])
+        return;
+
+    mx   = s->mv[1][0][0];
+    my   = s->mv[1][0][1];
+    uvmx = (mx + ((mx & 3) == 3)) >> 1;
+    uvmy = (my + ((my & 3) == 3)) >> 1;
+    if (v->field_mode) {
+        if (v->cur_field_type != v->ref_field_type[1])
+            my   = my   - 2 + 4 * v->cur_field_type;
+            uvmy = uvmy - 2 + 4 * v->cur_field_type;
+    }
+    if (v->fastuvmc) {
+        uvmx = uvmx + ((uvmx < 0) ? -(uvmx & 1) : (uvmx & 1));
+        uvmy = uvmy + ((uvmy < 0) ? -(uvmy & 1) : (uvmy & 1));
+    }
+    srcY = s->next_picture.f->data[0];
+    srcU = s->next_picture.f->data[1];
+    srcV = s->next_picture.f->data[2];
+
+    src_x   = s->mb_x * 16 + (mx   >> 2);
+    src_y   = s->mb_y * 16 + (my   >> 2);
+    uvsrc_x = s->mb_x *  8 + (uvmx >> 2);
+    uvsrc_y = s->mb_y *  8 + (uvmy >> 2);
+
+    if (v->profile != PROFILE_ADVANCED) {
+        src_x   = av_clip(  src_x, -16, s->mb_width  * 16);
+        src_y   = av_clip(  src_y, -16, s->mb_height * 16);
+        uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);
+        uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);
+    } else {
+        src_x   = av_clip(  src_x, -17, s->avctx->coded_width);
+        src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);
+        uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);
+        uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);
+    }
+
+    srcY += src_y   * s->linesize   + src_x;
+    srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
+    srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
+
+    if (v->field_mode && v->ref_field_type[1]) {
+        srcY += s->current_picture_ptr->f->linesize[0];
+        srcU += s->current_picture_ptr->f->linesize[1];
+        srcV += s->current_picture_ptr->f->linesize[2];
+    }
+
+    /* for grayscale we should not try to read from unknown area */
+    if (s->flags & CODEC_FLAG_GRAY) {
+        srcU = s->edge_emu_buffer + 18 * s->linesize;
+        srcV = s->edge_emu_buffer + 18 * s->linesize;
+    }
+
+    if (v->rangeredfrm || s->h_edge_pos < 22 || v_edge_pos < 22 || use_ic
+        || (unsigned)(src_x - 1) > s->h_edge_pos - (mx & 3) - 16 - 3
+        || (unsigned)(src_y - 1) > v_edge_pos    - (my & 3) - 16 - 3) {
+        uint8_t *uvbuf = s->edge_emu_buffer + 19 * s->linesize;
+
+        srcY -= s->mspel * (1 + s->linesize);
+        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY,
+                                 s->linesize, s->linesize,
+                                 17 + s->mspel * 2, 17 + s->mspel * 2,
+                                 src_x - s->mspel, src_y - s->mspel,
+                                 s->h_edge_pos, v_edge_pos);
+        srcY = s->edge_emu_buffer;
+        s->vdsp.emulated_edge_mc(uvbuf, srcU,
+                                 s->uvlinesize, s->uvlinesize,
+                                 8 + 1, 8 + 1,
+                                 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);
+        s->vdsp.emulated_edge_mc(uvbuf + 16, srcV,
+                                 s->uvlinesize, s->uvlinesize,
+                                 8 + 1, 8 + 1,
+                                 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);
+        srcU = uvbuf;
+        srcV = uvbuf + 16;
+        /* if we deal with range reduction we need to scale source blocks */
+        if (v->rangeredfrm) {
+            int i, j;
+            uint8_t *src, *src2;
+
+            src = srcY;
+            for (j = 0; j < 17 + s->mspel * 2; j++) {
+                for (i = 0; i < 17 + s->mspel * 2; i++)
+                    src[i] = ((src[i] - 128) >> 1) + 128;
+                src += s->linesize;
+            }
+            src = srcU;
+            src2 = srcV;
+            for (j = 0; j < 9; j++) {
+                for (i = 0; i < 9; i++) {
+                    src[i]  = ((src[i]  - 128) >> 1) + 128;
+                    src2[i] = ((src2[i] - 128) >> 1) + 128;
+                }
+                src  += s->uvlinesize;
+                src2 += s->uvlinesize;
+            }
+        }
+
+        if (use_ic) {
+            uint8_t (*luty )[256] = v->next_luty;
+            uint8_t (*lutuv)[256] = v->next_lutuv;
+            int i, j;
+            uint8_t *src, *src2;
+
+            src = srcY;
+            for (j = 0; j < 17 + s->mspel * 2; j++) {
+                int f = v->field_mode ? v->ref_field_type[1] : ((j+src_y - s->mspel) & 1);
+                for (i = 0; i < 17 + s->mspel * 2; i++)
+                    src[i] = luty[f][src[i]];
+                src += s->linesize;
+            }
+            src  = srcU;
+            src2 = srcV;
+            for (j = 0; j < 9; j++) {
+                int f = v->field_mode ? v->ref_field_type[1] : ((j+uvsrc_y) & 1);
+                for (i = 0; i < 9; i++) {
+                    src[i]  = lutuv[f][src[i]];
+                    src2[i] = lutuv[f][src2[i]];
+                }
+                src  += s->uvlinesize;
+                src2 += s->uvlinesize;
+            }
+        }
+        srcY += s->mspel * (1 + s->linesize);
+    }
+
+    off    = 0;
+    off_uv = 0;
+
+    if (s->mspel) {
+        dxy = ((my & 3) << 2) | (mx & 3);
+        v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off    , srcY    , s->linesize, v->rnd);
+        v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8, srcY + 8, s->linesize, v->rnd);
+        srcY += s->linesize * 8;
+        v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize    , srcY    , s->linesize, v->rnd);
+        v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
+    } else { // hpel mc
+        dxy = (my & 2) | ((mx & 2) >> 1);
+
+        if (!v->rnd)
+            s->hdsp.avg_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16);
+        else
+            s->hdsp.avg_no_rnd_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, 16);
+    }
+
+    if (s->flags & CODEC_FLAG_GRAY) return;
+    /* Chroma MC always uses qpel blilinear */
+    uvmx = (uvmx & 3) << 1;
+    uvmy = (uvmy & 3) << 1;
+    if (!v->rnd) {
+        h264chroma->avg_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy);
+        h264chroma->avg_h264_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy);
+    } else {
+        v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy);
+        v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy);
+    }
+}
diff --git a/libavcodec/vc1_pred.c b/libavcodec/vc1_pred.c
new file mode 100644 (file)
index 0000000..6a54fe4
--- /dev/null
@@ -0,0 +1,959 @@
+/*
+ * VC-1 and WMV3 decoder
+ * Copyright (c) 2011 Mashiat Sarker Shakkhar
+ * Copyright (c) 2006-2007 Konstantin Shishkov
+ * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * VC-1 and WMV3 block decoding routines
+ */
+
+#include "mathops.h"
+#include "mpegutils.h"
+#include "mpegvideo.h"
+#include "vc1.h"
+#include "vc1_pred.h"
+#include "vc1data.h"
+
+static av_always_inline int scaleforsame_x(VC1Context *v, int n /* MV */, int dir)
+{
+    int scaledvalue, refdist;
+    int scalesame1, scalesame2;
+    int scalezone1_x, zone1offset_x;
+    int table_index = dir ^ v->second_field;
+
+    if (v->s.pict_type != AV_PICTURE_TYPE_B)
+        refdist = v->refdist;
+    else
+        refdist = dir ? v->brfd : v->frfd;
+    if (refdist > 3)
+        refdist = 3;
+    scalesame1    = ff_vc1_field_mvpred_scales[table_index][1][refdist];
+    scalesame2    = ff_vc1_field_mvpred_scales[table_index][2][refdist];
+    scalezone1_x  = ff_vc1_field_mvpred_scales[table_index][3][refdist];
+    zone1offset_x = ff_vc1_field_mvpred_scales[table_index][5][refdist];
+
+    if (FFABS(n) > 255)
+        scaledvalue = n;
+    else {
+        if (FFABS(n) < scalezone1_x)
+            scaledvalue = (n * scalesame1) >> 8;
+        else {
+            if (n < 0)
+                scaledvalue = ((n * scalesame2) >> 8) - zone1offset_x;
+            else
+                scaledvalue = ((n * scalesame2) >> 8) + zone1offset_x;
+        }
+    }
+    return av_clip(scaledvalue, -v->range_x, v->range_x - 1);
+}
+
+static av_always_inline int scaleforsame_y(VC1Context *v, int i, int n /* MV */, int dir)
+{
+    int scaledvalue, refdist;
+    int scalesame1, scalesame2;
+    int scalezone1_y, zone1offset_y;
+    int table_index = dir ^ v->second_field;
+
+    if (v->s.pict_type != AV_PICTURE_TYPE_B)
+        refdist = v->refdist;
+    else
+        refdist = dir ? v->brfd : v->frfd;
+    if (refdist > 3)
+        refdist = 3;
+    scalesame1    = ff_vc1_field_mvpred_scales[table_index][1][refdist];
+    scalesame2    = ff_vc1_field_mvpred_scales[table_index][2][refdist];
+    scalezone1_y  = ff_vc1_field_mvpred_scales[table_index][4][refdist];
+    zone1offset_y = ff_vc1_field_mvpred_scales[table_index][6][refdist];
+
+    if (FFABS(n) > 63)
+        scaledvalue = n;
+    else {
+        if (FFABS(n) < scalezone1_y)
+            scaledvalue = (n * scalesame1) >> 8;
+        else {
+            if (n < 0)
+                scaledvalue = ((n * scalesame2) >> 8) - zone1offset_y;
+            else
+                scaledvalue = ((n * scalesame2) >> 8) + zone1offset_y;
+        }
+    }
+
+    if (v->cur_field_type && !v->ref_field_type[dir])
+        return av_clip(scaledvalue, -v->range_y / 2 + 1, v->range_y / 2);
+    else
+        return av_clip(scaledvalue, -v->range_y / 2, v->range_y / 2 - 1);
+}
+
+static av_always_inline int scaleforopp_x(VC1Context *v, int n /* MV */)
+{
+    int scalezone1_x, zone1offset_x;
+    int scaleopp1, scaleopp2, brfd;
+    int scaledvalue;
+
+    brfd = FFMIN(v->brfd, 3);
+    scalezone1_x  = ff_vc1_b_field_mvpred_scales[3][brfd];
+    zone1offset_x = ff_vc1_b_field_mvpred_scales[5][brfd];
+    scaleopp1     = ff_vc1_b_field_mvpred_scales[1][brfd];
+    scaleopp2     = ff_vc1_b_field_mvpred_scales[2][brfd];
+
+    if (FFABS(n) > 255)
+        scaledvalue = n;
+    else {
+        if (FFABS(n) < scalezone1_x)
+            scaledvalue = (n * scaleopp1) >> 8;
+        else {
+            if (n < 0)
+                scaledvalue = ((n * scaleopp2) >> 8) - zone1offset_x;
+            else
+                scaledvalue = ((n * scaleopp2) >> 8) + zone1offset_x;
+        }
+    }
+    return av_clip(scaledvalue, -v->range_x, v->range_x - 1);
+}
+
+static av_always_inline int scaleforopp_y(VC1Context *v, int n /* MV */, int dir)
+{
+    int scalezone1_y, zone1offset_y;
+    int scaleopp1, scaleopp2, brfd;
+    int scaledvalue;
+
+    brfd = FFMIN(v->brfd, 3);
+    scalezone1_y  = ff_vc1_b_field_mvpred_scales[4][brfd];
+    zone1offset_y = ff_vc1_b_field_mvpred_scales[6][brfd];
+    scaleopp1     = ff_vc1_b_field_mvpred_scales[1][brfd];
+    scaleopp2     = ff_vc1_b_field_mvpred_scales[2][brfd];
+
+    if (FFABS(n) > 63)
+        scaledvalue = n;
+    else {
+        if (FFABS(n) < scalezone1_y)
+            scaledvalue = (n * scaleopp1) >> 8;
+        else {
+            if (n < 0)
+                scaledvalue = ((n * scaleopp2) >> 8) - zone1offset_y;
+            else
+                scaledvalue = ((n * scaleopp2) >> 8) + zone1offset_y;
+        }
+    }
+    if (v->cur_field_type && !v->ref_field_type[dir]) {
+        return av_clip(scaledvalue, -v->range_y / 2 + 1, v->range_y / 2);
+    } else {
+        return av_clip(scaledvalue, -v->range_y / 2, v->range_y / 2 - 1);
+    }
+}
+
+static av_always_inline int scaleforsame(VC1Context *v, int i, int n /* MV */,
+                                         int dim, int dir)
+{
+    int brfd, scalesame;
+    int hpel = 1 - v->s.quarter_sample;
+
+    n >>= hpel;
+    if (v->s.pict_type != AV_PICTURE_TYPE_B || v->second_field || !dir) {
+        if (dim)
+            n = scaleforsame_y(v, i, n, dir) << hpel;
+        else
+            n = scaleforsame_x(v, n, dir) << hpel;
+        return n;
+    }
+    brfd      = FFMIN(v->brfd, 3);
+    scalesame = ff_vc1_b_field_mvpred_scales[0][brfd];
+
+    n = (n * scalesame >> 8) << hpel;
+    return n;
+}
+
+static av_always_inline int scaleforopp(VC1Context *v, int n /* MV */,
+                                        int dim, int dir)
+{
+    int refdist, scaleopp;
+    int hpel = 1 - v->s.quarter_sample;
+
+    n >>= hpel;
+    if (v->s.pict_type == AV_PICTURE_TYPE_B && !v->second_field && dir == 1) {
+        if (dim)
+            n = scaleforopp_y(v, n, dir) << hpel;
+        else
+            n = scaleforopp_x(v, n) << hpel;
+        return n;
+    }
+    if (v->s.pict_type != AV_PICTURE_TYPE_B)
+        refdist = FFMIN(v->refdist, 3);
+    else
+        refdist = dir ? v->brfd : v->frfd;
+    scaleopp = ff_vc1_field_mvpred_scales[dir ^ v->second_field][0][refdist];
+
+    n = (n * scaleopp >> 8) << hpel;
+    return n;
+}
+
+/** Predict and set motion vector
+ */
+void ff_vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y,
+                    int mv1, int r_x, int r_y, uint8_t* is_intra,
+                    int pred_flag, int dir)
+{
+    MpegEncContext *s = &v->s;
+    int xy, wrap, off = 0;
+    int16_t *A, *B, *C;
+    int px, py;
+    int sum;
+    int mixedmv_pic, num_samefield = 0, num_oppfield = 0;
+    int opposite, a_f, b_f, c_f;
+    int16_t field_predA[2];
+    int16_t field_predB[2];
+    int16_t field_predC[2];
+    int a_valid, b_valid, c_valid;
+    int hybridmv_thresh, y_bias = 0;
+
+    if (v->mv_mode == MV_PMODE_MIXED_MV ||
+        ((v->mv_mode == MV_PMODE_INTENSITY_COMP) && (v->mv_mode2 == MV_PMODE_MIXED_MV)))
+        mixedmv_pic = 1;
+    else
+        mixedmv_pic = 0;
+    /* scale MV difference to be quad-pel */
+    dmv_x <<= 1 - s->quarter_sample;
+    dmv_y <<= 1 - s->quarter_sample;
+
+    wrap = s->b8_stride;
+    xy   = s->block_index[n];
+
+    if (s->mb_intra) {
+        s->mv[0][n][0] = s->current_picture.motion_val[0][xy + v->blocks_off][0] = 0;
+        s->mv[0][n][1] = s->current_picture.motion_val[0][xy + v->blocks_off][1] = 0;
+        s->current_picture.motion_val[1][xy + v->blocks_off][0] = 0;
+        s->current_picture.motion_val[1][xy + v->blocks_off][1] = 0;
+        if (mv1) { /* duplicate motion data for 1-MV block */
+            s->current_picture.motion_val[0][xy + 1 + v->blocks_off][0]        = 0;
+            s->current_picture.motion_val[0][xy + 1 + v->blocks_off][1]        = 0;
+            s->current_picture.motion_val[0][xy + wrap + v->blocks_off][0]     = 0;
+            s->current_picture.motion_val[0][xy + wrap + v->blocks_off][1]     = 0;
+            s->current_picture.motion_val[0][xy + wrap + 1 + v->blocks_off][0] = 0;
+            s->current_picture.motion_val[0][xy + wrap + 1 + v->blocks_off][1] = 0;
+            v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;
+            s->current_picture.motion_val[1][xy + 1 + v->blocks_off][0]        = 0;
+            s->current_picture.motion_val[1][xy + 1 + v->blocks_off][1]        = 0;
+            s->current_picture.motion_val[1][xy + wrap][0]                     = 0;
+            s->current_picture.motion_val[1][xy + wrap + v->blocks_off][1]     = 0;
+            s->current_picture.motion_val[1][xy + wrap + 1 + v->blocks_off][0] = 0;
+            s->current_picture.motion_val[1][xy + wrap + 1 + v->blocks_off][1] = 0;
+        }
+        return;
+    }
+
+    C = s->current_picture.motion_val[dir][xy -    1 + v->blocks_off];
+    A = s->current_picture.motion_val[dir][xy - wrap + v->blocks_off];
+    if (mv1) {
+        if (v->field_mode && mixedmv_pic)
+            off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
+        else
+            off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;
+    } else {
+        //in 4-MV mode different blocks have different B predictor position
+        switch (n) {
+        case 0:
+            off = (s->mb_x > 0) ? -1 : 1;
+            break;
+        case 1:
+            off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;
+            break;
+        case 2:
+            off = 1;
+            break;
+        case 3:
+            off = -1;
+        }
+    }
+    B = s->current_picture.motion_val[dir][xy - wrap + off + v->blocks_off];
+
+    a_valid = !s->first_slice_line || (n == 2 || n == 3);
+    b_valid = a_valid && (s->mb_width > 1);
+    c_valid = s->mb_x || (n == 1 || n == 3);
+    if (v->field_mode) {
+        a_valid = a_valid && !is_intra[xy - wrap];
+        b_valid = b_valid && !is_intra[xy - wrap + off];
+        c_valid = c_valid && !is_intra[xy - 1];
+    }
+
+    if (a_valid) {
+        a_f = v->mv_f[dir][xy - wrap + v->blocks_off];
+        num_oppfield  += a_f;
+        num_samefield += 1 - a_f;
+        field_predA[0] = A[0];
+        field_predA[1] = A[1];
+    } else {
+        field_predA[0] = field_predA[1] = 0;
+        a_f = 0;
+    }
+    if (b_valid) {
+        b_f = v->mv_f[dir][xy - wrap + off + v->blocks_off];
+        num_oppfield  += b_f;
+        num_samefield += 1 - b_f;
+        field_predB[0] = B[0];
+        field_predB[1] = B[1];
+    } else {
+        field_predB[0] = field_predB[1] = 0;
+        b_f = 0;
+    }
+    if (c_valid) {
+        c_f = v->mv_f[dir][xy - 1 + v->blocks_off];
+        num_oppfield  += c_f;
+        num_samefield += 1 - c_f;
+        field_predC[0] = C[0];
+        field_predC[1] = C[1];
+    } else {
+        field_predC[0] = field_predC[1] = 0;
+        c_f = 0;
+    }
+
+    if (v->field_mode) {
+        if (!v->numref)
+            // REFFIELD determines if the last field or the second-last field is
+            // to be used as reference
+            opposite = 1 - v->reffield;
+        else {
+            if (num_samefield <= num_oppfield)
+                opposite = 1 - pred_flag;
+            else
+                opposite = pred_flag;
+        }
+    } else
+        opposite = 0;
+    if (opposite) {
+        if (a_valid && !a_f) {
+            field_predA[0] = scaleforopp(v, field_predA[0], 0, dir);
+            field_predA[1] = scaleforopp(v, field_predA[1], 1, dir);
+        }
+        if (b_valid && !b_f) {
+            field_predB[0] = scaleforopp(v, field_predB[0], 0, dir);
+            field_predB[1] = scaleforopp(v, field_predB[1], 1, dir);
+        }
+        if (c_valid && !c_f) {
+            field_predC[0] = scaleforopp(v, field_predC[0], 0, dir);
+            field_predC[1] = scaleforopp(v, field_predC[1], 1, dir);
+        }
+        v->mv_f[dir][xy + v->blocks_off] = 1;
+        v->ref_field_type[dir] = !v->cur_field_type;
+    } else {
+        if (a_valid && a_f) {
+            field_predA[0] = scaleforsame(v, n, field_predA[0], 0, dir);
+            field_predA[1] = scaleforsame(v, n, field_predA[1], 1, dir);
+        }
+        if (b_valid && b_f) {
+            field_predB[0] = scaleforsame(v, n, field_predB[0], 0, dir);
+            field_predB[1] = scaleforsame(v, n, field_predB[1], 1, dir);
+        }
+        if (c_valid && c_f) {
+            field_predC[0] = scaleforsame(v, n, field_predC[0], 0, dir);
+            field_predC[1] = scaleforsame(v, n, field_predC[1], 1, dir);
+        }
+        v->mv_f[dir][xy + v->blocks_off] = 0;
+        v->ref_field_type[dir] = v->cur_field_type;
+    }
+
+    if (a_valid) {
+        px = field_predA[0];
+        py = field_predA[1];
+    } else if (c_valid) {
+        px = field_predC[0];
+        py = field_predC[1];
+    } else if (b_valid) {
+        px = field_predB[0];
+        py = field_predB[1];
+    } else {
+        px = 0;
+        py = 0;
+    }
+
+    if (num_samefield + num_oppfield > 1) {
+        px = mid_pred(field_predA[0], field_predB[0], field_predC[0]);
+        py = mid_pred(field_predA[1], field_predB[1], field_predC[1]);
+    }
+
+    /* Pullback MV as specified in 8.3.5.3.4 */
+    if (!v->field_mode) {
+        int qx, qy, X, Y;
+        qx = (s->mb_x << 6) + ((n == 1 || n == 3) ? 32 : 0);
+        qy = (s->mb_y << 6) + ((n == 2 || n == 3) ? 32 : 0);
+        X  = (s->mb_width  << 6) - 4;
+        Y  = (s->mb_height << 6) - 4;
+        if (mv1) {
+            if (qx + px < -60) px = -60 - qx;
+            if (qy + py < -60) py = -60 - qy;
+        } else {
+            if (qx + px < -28) px = -28 - qx;
+            if (qy + py < -28) py = -28 - qy;
+        }
+        if (qx + px > X) px = X - qx;
+        if (qy + py > Y) py = Y - qy;
+    }
+
+    if (!v->field_mode || s->pict_type != AV_PICTURE_TYPE_B) {
+        /* Calculate hybrid prediction as specified in 8.3.5.3.5 (also 10.3.5.4.3.5) */
+        hybridmv_thresh = 32;
+        if (a_valid && c_valid) {
+            if (is_intra[xy - wrap])
+                sum = FFABS(px) + FFABS(py);
+            else
+                sum = FFABS(px - field_predA[0]) + FFABS(py - field_predA[1]);
+            if (sum > hybridmv_thresh) {
+                if (get_bits1(&s->gb)) {     // read HYBRIDPRED bit
+                    px = field_predA[0];
+                    py = field_predA[1];
+                } else {
+                    px = field_predC[0];
+                    py = field_predC[1];
+                }
+            } else {
+                if (is_intra[xy - 1])
+                    sum = FFABS(px) + FFABS(py);
+                else
+                    sum = FFABS(px - field_predC[0]) + FFABS(py - field_predC[1]);
+                if (sum > hybridmv_thresh) {
+                    if (get_bits1(&s->gb)) {
+                        px = field_predA[0];
+                        py = field_predA[1];
+                    } else {
+                        px = field_predC[0];
+                        py = field_predC[1];
+                    }
+                }
+            }
+        }
+    }
+
+    if (v->field_mode && v->numref)
+        r_y >>= 1;
+    if (v->field_mode && v->cur_field_type && v->ref_field_type[dir] == 0)
+        y_bias = 1;
+    /* store MV using signed modulus of MV range defined in 4.11 */
+    s->mv[dir][n][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;
+    s->mv[dir][n][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1] = ((py + dmv_y + r_y - y_bias) & ((r_y << 1) - 1)) - r_y + y_bias;
+    if (mv1) { /* duplicate motion data for 1-MV block */
+        s->current_picture.motion_val[dir][xy +    1 +     v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0];
+        s->current_picture.motion_val[dir][xy +    1 +     v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1];
+        s->current_picture.motion_val[dir][xy + wrap +     v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0];
+        s->current_picture.motion_val[dir][xy + wrap +     v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1];
+        s->current_picture.motion_val[dir][xy + wrap + 1 + v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0];
+        s->current_picture.motion_val[dir][xy + wrap + 1 + v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1];
+        v->mv_f[dir][xy +    1 + v->blocks_off] = v->mv_f[dir][xy +            v->blocks_off];
+        v->mv_f[dir][xy + wrap + v->blocks_off] = v->mv_f[dir][xy + wrap + 1 + v->blocks_off] = v->mv_f[dir][xy + v->blocks_off];
+    }
+}
+
+/** Predict and set motion vector for interlaced frame picture MBs
+ */
+void ff_vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y,
+                          int mvn, int r_x, int r_y, uint8_t* is_intra, int dir)
+{
+    MpegEncContext *s = &v->s;
+    int xy, wrap, off = 0;
+    int A[2], B[2], C[2];
+    int px, py;
+    int a_valid = 0, b_valid = 0, c_valid = 0;
+    int field_a, field_b, field_c; // 0: same, 1: opposit
+    int total_valid, num_samefield, num_oppfield;
+    int pos_c, pos_b, n_adj;
+
+    wrap = s->b8_stride;
+    xy = s->block_index[n];
+
+    if (s->mb_intra) {
+        s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;
+        s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;
+        s->current_picture.motion_val[1][xy][0] = 0;
+        s->current_picture.motion_val[1][xy][1] = 0;
+        if (mvn == 1) { /* duplicate motion data for 1-MV block */
+            s->current_picture.motion_val[0][xy + 1][0]        = 0;
+            s->current_picture.motion_val[0][xy + 1][1]        = 0;
+            s->current_picture.motion_val[0][xy + wrap][0]     = 0;
+            s->current_picture.motion_val[0][xy + wrap][1]     = 0;
+            s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;
+            s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;
+            v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;
+            s->current_picture.motion_val[1][xy + 1][0]        = 0;
+            s->current_picture.motion_val[1][xy + 1][1]        = 0;
+            s->current_picture.motion_val[1][xy + wrap][0]     = 0;
+            s->current_picture.motion_val[1][xy + wrap][1]     = 0;
+            s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;
+            s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;
+        }
+        return;
+    }
+
+    off = ((n == 0) || (n == 1)) ? 1 : -1;
+    /* predict A */
+    if (s->mb_x || (n == 1) || (n == 3)) {
+        if ((v->blk_mv_type[xy]) // current block (MB) has a field MV
+            || (!v->blk_mv_type[xy] && !v->blk_mv_type[xy - 1])) { // or both have frame MV
+            A[0] = s->current_picture.motion_val[dir][xy - 1][0];
+            A[1] = s->current_picture.motion_val[dir][xy - 1][1];
+            a_valid = 1;
+        } else { // current block has frame mv and cand. has field MV (so average)
+            A[0] = (s->current_picture.motion_val[dir][xy - 1][0]
+                    + s->current_picture.motion_val[dir][xy - 1 + off * wrap][0] + 1) >> 1;
+            A[1] = (s->current_picture.motion_val[dir][xy - 1][1]
+                    + s->current_picture.motion_val[dir][xy - 1 + off * wrap][1] + 1) >> 1;
+            a_valid = 1;
+        }
+        if (!(n & 1) && v->is_intra[s->mb_x - 1]) {
+            a_valid = 0;
+            A[0] = A[1] = 0;
+        }
+    } else
+        A[0] = A[1] = 0;
+    /* Predict B and C */
+    B[0] = B[1] = C[0] = C[1] = 0;
+    if (n == 0 || n == 1 || v->blk_mv_type[xy]) {
+        if (!s->first_slice_line) {
+            if (!v->is_intra[s->mb_x - s->mb_stride]) {
+                b_valid = 1;
+                n_adj   = n | 2;
+                pos_b   = s->block_index[n_adj] - 2 * wrap;
+                if (v->blk_mv_type[pos_b] && v->blk_mv_type[xy]) {
+                    n_adj = (n & 2) | (n & 1);
+                }
+                B[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap][0];
+                B[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap][1];
+                if (v->blk_mv_type[pos_b] && !v->blk_mv_type[xy]) {
+                    B[0] = (B[0] + s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap][0] + 1) >> 1;
+                    B[1] = (B[1] + s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap][1] + 1) >> 1;
+                }
+            }
+            if (s->mb_width > 1) {
+                if (!v->is_intra[s->mb_x - s->mb_stride + 1]) {
+                    c_valid = 1;
+                    n_adj   = 2;
+                    pos_c   = s->block_index[2] - 2 * wrap + 2;
+                    if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) {
+                        n_adj = n & 2;
+                    }
+                    C[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap + 2][0];
+                    C[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap + 2][1];
+                    if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) {
+                        C[0] = (1 + C[0] + (s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap + 2][0])) >> 1;
+                        C[1] = (1 + C[1] + (s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap + 2][1])) >> 1;
+                    }
+                    if (s->mb_x == s->mb_width - 1) {
+                        if (!v->is_intra[s->mb_x - s->mb_stride - 1]) {
+                            c_valid = 1;
+                            n_adj   = 3;
+                            pos_c   = s->block_index[3] - 2 * wrap - 2;
+                            if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) {
+                                n_adj = n | 1;
+                            }
+                            C[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap - 2][0];
+                            C[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap - 2][1];
+                            if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) {
+                                C[0] = (1 + C[0] + s->current_picture.motion_val[dir][s->block_index[1] - 2 * wrap - 2][0]) >> 1;
+                                C[1] = (1 + C[1] + s->current_picture.motion_val[dir][s->block_index[1] - 2 * wrap - 2][1]) >> 1;
+                            }
+                        } else
+                            c_valid = 0;
+                    }
+                }
+            }
+        }
+    } else {
+        pos_b   = s->block_index[1];
+        b_valid = 1;
+        B[0]    = s->current_picture.motion_val[dir][pos_b][0];
+        B[1]    = s->current_picture.motion_val[dir][pos_b][1];
+        pos_c   = s->block_index[0];
+        c_valid = 1;
+        C[0]    = s->current_picture.motion_val[dir][pos_c][0];
+        C[1]    = s->current_picture.motion_val[dir][pos_c][1];
+    }
+
+    total_valid = a_valid + b_valid + c_valid;
+    // check if predictor A is out of bounds
+    if (!s->mb_x && !(n == 1 || n == 3)) {
+        A[0] = A[1] = 0;
+    }
+    // check if predictor B is out of bounds
+    if ((s->first_slice_line && v->blk_mv_type[xy]) || (s->first_slice_line && !(n & 2))) {
+        B[0] = B[1] = C[0] = C[1] = 0;
+    }
+    if (!v->blk_mv_type[xy]) {
+        if (s->mb_width == 1) {
+            px = B[0];
+            py = B[1];
+        } else {
+            if (total_valid >= 2) {
+                px = mid_pred(A[0], B[0], C[0]);
+                py = mid_pred(A[1], B[1], C[1]);
+            } else if (total_valid) {
+                if (a_valid) { px = A[0]; py = A[1]; }
+                if (b_valid) { px = B[0]; py = B[1]; }
+                if (c_valid) { px = C[0]; py = C[1]; }
+            } else
+                px = py = 0;
+        }
+    } else {
+        if (a_valid)
+            field_a = (A[1] & 4) ? 1 : 0;
+        else
+            field_a = 0;
+        if (b_valid)
+            field_b = (B[1] & 4) ? 1 : 0;
+        else
+            field_b = 0;
+        if (c_valid)
+            field_c = (C[1] & 4) ? 1 : 0;
+        else
+            field_c = 0;
+
+        num_oppfield  = field_a + field_b + field_c;
+        num_samefield = total_valid - num_oppfield;
+        if (total_valid == 3) {
+            if ((num_samefield == 3) || (num_oppfield == 3)) {
+                px = mid_pred(A[0], B[0], C[0]);
+                py = mid_pred(A[1], B[1], C[1]);
+            } else if (num_samefield >= num_oppfield) {
+                /* take one MV from same field set depending on priority
+                the check for B may not be necessary */
+                px = !field_a ? A[0] : B[0];
+                py = !field_a ? A[1] : B[1];
+            } else {
+                px =  field_a ? A[0] : B[0];
+                py =  field_a ? A[1] : B[1];
+            }
+        } else if (total_valid == 2) {
+            if (num_samefield >= num_oppfield) {
+                if (!field_a && a_valid) {
+                    px = A[0];
+                    py = A[1];
+                } else if (!field_b && b_valid) {
+                    px = B[0];
+                    py = B[1];
+                } else if (c_valid) {
+                    px = C[0];
+                    py = C[1];
+                } else px = py = 0;
+            } else {
+                if (field_a && a_valid) {
+                    px = A[0];
+                    py = A[1];
+                } else if (field_b && b_valid) {
+                    px = B[0];
+                    py = B[1];
+                } else if (c_valid) {
+                    px = C[0];
+                    py = C[1];
+                } else
+                    px = py = 0;
+            }
+        } else if (total_valid == 1) {
+            px = (a_valid) ? A[0] : ((b_valid) ? B[0] : C[0]);
+            py = (a_valid) ? A[1] : ((b_valid) ? B[1] : C[1]);
+        } else
+            px = py = 0;
+    }
+
+    /* store MV using signed modulus of MV range defined in 4.11 */
+    s->mv[dir][n][0] = s->current_picture.motion_val[dir][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;
+    s->mv[dir][n][1] = s->current_picture.motion_val[dir][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;
+    if (mvn == 1) { /* duplicate motion data for 1-MV block */
+        s->current_picture.motion_val[dir][xy +    1    ][0] = s->current_picture.motion_val[dir][xy][0];
+        s->current_picture.motion_val[dir][xy +    1    ][1] = s->current_picture.motion_val[dir][xy][1];
+        s->current_picture.motion_val[dir][xy + wrap    ][0] = s->current_picture.motion_val[dir][xy][0];
+        s->current_picture.motion_val[dir][xy + wrap    ][1] = s->current_picture.motion_val[dir][xy][1];
+        s->current_picture.motion_val[dir][xy + wrap + 1][0] = s->current_picture.motion_val[dir][xy][0];
+        s->current_picture.motion_val[dir][xy + wrap + 1][1] = s->current_picture.motion_val[dir][xy][1];
+    } else if (mvn == 2) { /* duplicate motion data for 2-Field MV block */
+        s->current_picture.motion_val[dir][xy + 1][0] = s->current_picture.motion_val[dir][xy][0];
+        s->current_picture.motion_val[dir][xy + 1][1] = s->current_picture.motion_val[dir][xy][1];
+        s->mv[dir][n + 1][0] = s->mv[dir][n][0];
+        s->mv[dir][n + 1][1] = s->mv[dir][n][1];
+    }
+}
+
+void ff_vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2],
+                      int direct, int mvtype)
+{
+    MpegEncContext *s = &v->s;
+    int xy, wrap, off = 0;
+    int16_t *A, *B, *C;
+    int px, py;
+    int sum;
+    int r_x, r_y;
+    const uint8_t *is_intra = v->mb_type[0];
+
+    r_x = v->range_x;
+    r_y = v->range_y;
+    /* scale MV difference to be quad-pel */
+    dmv_x[0] <<= 1 - s->quarter_sample;
+    dmv_y[0] <<= 1 - s->quarter_sample;
+    dmv_x[1] <<= 1 - s->quarter_sample;
+    dmv_y[1] <<= 1 - s->quarter_sample;
+
+    wrap = s->b8_stride;
+    xy = s->block_index[0];
+
+    if (s->mb_intra) {
+        s->current_picture.motion_val[0][xy + v->blocks_off][0] =
+        s->current_picture.motion_val[0][xy + v->blocks_off][1] =
+        s->current_picture.motion_val[1][xy + v->blocks_off][0] =
+        s->current_picture.motion_val[1][xy + v->blocks_off][1] = 0;
+        return;
+    }
+    if (!v->field_mode) {
+        s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);
+        s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);
+        s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);
+        s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);
+
+        /* Pullback predicted motion vectors as specified in 8.4.5.4 */
+        s->mv[0][0][0] = av_clip(s->mv[0][0][0], -60 - (s->mb_x << 6), (s->mb_width  << 6) - 4 - (s->mb_x << 6));
+        s->mv[0][0][1] = av_clip(s->mv[0][0][1], -60 - (s->mb_y << 6), (s->mb_height << 6) - 4 - (s->mb_y << 6));
+        s->mv[1][0][0] = av_clip(s->mv[1][0][0], -60 - (s->mb_x << 6), (s->mb_width  << 6) - 4 - (s->mb_x << 6));
+        s->mv[1][0][1] = av_clip(s->mv[1][0][1], -60 - (s->mb_y << 6), (s->mb_height << 6) - 4 - (s->mb_y << 6));
+    }
+    if (direct) {
+        s->current_picture.motion_val[0][xy + v->blocks_off][0] = s->mv[0][0][0];
+        s->current_picture.motion_val[0][xy + v->blocks_off][1] = s->mv[0][0][1];
+        s->current_picture.motion_val[1][xy + v->blocks_off][0] = s->mv[1][0][0];
+        s->current_picture.motion_val[1][xy + v->blocks_off][1] = s->mv[1][0][1];
+        return;
+    }
+
+    if ((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
+        C   = s->current_picture.motion_val[0][xy - 2];
+        A   = s->current_picture.motion_val[0][xy - wrap * 2];
+        off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
+        B   = s->current_picture.motion_val[0][xy - wrap * 2 + off];
+
+        if (!s->mb_x) C[0] = C[1] = 0;
+        if (!s->first_slice_line) { // predictor A is not out of bounds
+            if (s->mb_width == 1) {
+                px = A[0];
+                py = A[1];
+            } else {
+                px = mid_pred(A[0], B[0], C[0]);
+                py = mid_pred(A[1], B[1], C[1]);
+            }
+        } else if (s->mb_x) { // predictor C is not out of bounds
+            px = C[0];
+            py = C[1];
+        } else {
+            px = py = 0;
+        }
+        /* Pullback MV as specified in 8.3.5.3.4 */
+        {
+            int qx, qy, X, Y;
+            if (v->profile < PROFILE_ADVANCED) {
+                qx = (s->mb_x << 5);
+                qy = (s->mb_y << 5);
+                X  = (s->mb_width  << 5) - 4;
+                Y  = (s->mb_height << 5) - 4;
+                if (qx + px < -28) px = -28 - qx;
+                if (qy + py < -28) py = -28 - qy;
+                if (qx + px > X) px = X - qx;
+                if (qy + py > Y) py = Y - qy;
+            } else {
+                qx = (s->mb_x << 6);
+                qy = (s->mb_y << 6);
+                X  = (s->mb_width  << 6) - 4;
+                Y  = (s->mb_height << 6) - 4;
+                if (qx + px < -60) px = -60 - qx;
+                if (qy + py < -60) py = -60 - qy;
+                if (qx + px > X) px = X - qx;
+                if (qy + py > Y) py = Y - qy;
+            }
+        }
+        /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
+        if (0 && !s->first_slice_line && s->mb_x) {
+            if (is_intra[xy - wrap])
+                sum = FFABS(px) + FFABS(py);
+            else
+                sum = FFABS(px - A[0]) + FFABS(py - A[1]);
+            if (sum > 32) {
+                if (get_bits1(&s->gb)) {
+                    px = A[0];
+                    py = A[1];
+                } else {
+                    px = C[0];
+                    py = C[1];
+                }
+            } else {
+                if (is_intra[xy - 2])
+                    sum = FFABS(px) + FFABS(py);
+                else
+                    sum = FFABS(px - C[0]) + FFABS(py - C[1]);
+                if (sum > 32) {
+                    if (get_bits1(&s->gb)) {
+                        px = A[0];
+                        py = A[1];
+                    } else {
+                        px = C[0];
+                        py = C[1];
+                    }
+                }
+            }
+        }
+        /* store MV using signed modulus of MV range defined in 4.11 */
+        s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;
+        s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;
+    }
+    if ((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
+        C   = s->current_picture.motion_val[1][xy - 2];
+        A   = s->current_picture.motion_val[1][xy - wrap * 2];
+        off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
+        B   = s->current_picture.motion_val[1][xy - wrap * 2 + off];
+
+        if (!s->mb_x)
+            C[0] = C[1] = 0;
+        if (!s->first_slice_line) { // predictor A is not out of bounds
+            if (s->mb_width == 1) {
+                px = A[0];
+                py = A[1];
+            } else {
+                px = mid_pred(A[0], B[0], C[0]);
+                py = mid_pred(A[1], B[1], C[1]);
+            }
+        } else if (s->mb_x) { // predictor C is not out of bounds
+            px = C[0];
+            py = C[1];
+        } else {
+            px = py = 0;
+        }
+        /* Pullback MV as specified in 8.3.5.3.4 */
+        {
+            int qx, qy, X, Y;
+            if (v->profile < PROFILE_ADVANCED) {
+                qx = (s->mb_x << 5);
+                qy = (s->mb_y << 5);
+                X  = (s->mb_width  << 5) - 4;
+                Y  = (s->mb_height << 5) - 4;
+                if (qx + px < -28) px = -28 - qx;
+                if (qy + py < -28) py = -28 - qy;
+                if (qx + px > X) px = X - qx;
+                if (qy + py > Y) py = Y - qy;
+            } else {
+                qx = (s->mb_x << 6);
+                qy = (s->mb_y << 6);
+                X  = (s->mb_width  << 6) - 4;
+                Y  = (s->mb_height << 6) - 4;
+                if (qx + px < -60) px = -60 - qx;
+                if (qy + py < -60) py = -60 - qy;
+                if (qx + px > X) px = X - qx;
+                if (qy + py > Y) py = Y - qy;
+            }
+        }
+        /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
+        if (0 && !s->first_slice_line && s->mb_x) {
+            if (is_intra[xy - wrap])
+                sum = FFABS(px) + FFABS(py);
+            else
+                sum = FFABS(px - A[0]) + FFABS(py - A[1]);
+            if (sum > 32) {
+                if (get_bits1(&s->gb)) {
+                    px = A[0];
+                    py = A[1];
+                } else {
+                    px = C[0];
+                    py = C[1];
+                }
+            } else {
+                if (is_intra[xy - 2])
+                    sum = FFABS(px) + FFABS(py);
+                else
+                    sum = FFABS(px - C[0]) + FFABS(py - C[1]);
+                if (sum > 32) {
+                    if (get_bits1(&s->gb)) {
+                        px = A[0];
+                        py = A[1];
+                    } else {
+                        px = C[0];
+                        py = C[1];
+                    }
+                }
+            }
+        }
+        /* store MV using signed modulus of MV range defined in 4.11 */
+
+        s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;
+        s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;
+    }
+    s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
+    s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
+    s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
+    s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
+}
+
+void ff_vc1_pred_b_mv_intfi(VC1Context *v, int n, int *dmv_x, int *dmv_y,
+                            int mv1, int *pred_flag)
+{
+    int dir = (v->bmvtype == BMV_TYPE_BACKWARD) ? 1 : 0;
+    MpegEncContext *s = &v->s;
+    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+
+    if (v->bmvtype == BMV_TYPE_DIRECT) {
+        int total_opp, k, f;
+        if (s->next_picture.mb_type[mb_pos + v->mb_off] != MB_TYPE_INTRA) {
+            s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][0],
+                                      v->bfraction, 0, s->quarter_sample);
+            s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][1],
+                                      v->bfraction, 0, s->quarter_sample);
+            s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][0],
+                                      v->bfraction, 1, s->quarter_sample);
+            s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][1],
+                                      v->bfraction, 1, s->quarter_sample);
+
+            total_opp = v->mv_f_next[0][s->block_index[0] + v->blocks_off]
+                      + v->mv_f_next[0][s->block_index[1] + v->blocks_off]
+                      + v->mv_f_next[0][s->block_index[2] + v->blocks_off]
+                      + v->mv_f_next[0][s->block_index[3] + v->blocks_off];
+            f = (total_opp > 2) ? 1 : 0;
+        } else {
+            s->mv[0][0][0] = s->mv[0][0][1] = 0;
+            s->mv[1][0][0] = s->mv[1][0][1] = 0;
+            f = 0;
+        }
+        v->ref_field_type[0] = v->ref_field_type[1] = v->cur_field_type ^ f;
+        for (k = 0; k < 4; k++) {
+            s->current_picture.motion_val[0][s->block_index[k] + v->blocks_off][0] = s->mv[0][0][0];
+            s->current_picture.motion_val[0][s->block_index[k] + v->blocks_off][1] = s->mv[0][0][1];
+            s->current_picture.motion_val[1][s->block_index[k] + v->blocks_off][0] = s->mv[1][0][0];
+            s->current_picture.motion_val[1][s->block_index[k] + v->blocks_off][1] = s->mv[1][0][1];
+            v->mv_f[0][s->block_index[k] + v->blocks_off] = f;
+            v->mv_f[1][s->block_index[k] + v->blocks_off] = f;
+        }
+        return;
+    }
+    if (v->bmvtype == BMV_TYPE_INTERPOLATED) {
+        ff_vc1_pred_mv(v, 0, dmv_x[0], dmv_y[0],   1, v->range_x, v->range_y, v->mb_type[0], pred_flag[0], 0);
+        ff_vc1_pred_mv(v, 0, dmv_x[1], dmv_y[1],   1, v->range_x, v->range_y, v->mb_type[0], pred_flag[1], 1);
+        return;
+    }
+    if (dir) { // backward
+        ff_vc1_pred_mv(v, n, dmv_x[1], dmv_y[1], mv1, v->range_x, v->range_y, v->mb_type[0], pred_flag[1], 1);
+        if (n == 3 || mv1) {
+            ff_vc1_pred_mv(v, 0, dmv_x[0], dmv_y[0],   1, v->range_x, v->range_y, v->mb_type[0], 0, 0);
+        }
+    } else { // forward
+        ff_vc1_pred_mv(v, n, dmv_x[0], dmv_y[0], mv1, v->range_x, v->range_y, v->mb_type[0], pred_flag[0], 0);
+        if (n == 3 || mv1) {
+            ff_vc1_pred_mv(v, 0, dmv_x[1], dmv_y[1],   1, v->range_x, v->range_y, v->mb_type[0], 0, 1);
+        }
+    }
+}
diff --git a/libavcodec/vc1_pred.h b/libavcodec/vc1_pred.h
new file mode 100644 (file)
index 0000000..34c9c1a
--- /dev/null
@@ -0,0 +1,59 @@
+/*
+ * VC-1 and WMV3 decoder
+ * Copyright (c) 2006-2007 Konstantin Shishkov
+ * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_VC1_PRED_H
+#define AVCODEC_VC1_PRED_H
+
+#include "vc1.h"
+#include "vc1data.h"
+
+void ff_vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y,
+                    int mv1, int r_x, int r_y, uint8_t* is_intra,
+                    int pred_flag, int dir);
+void ff_vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y,
+                          int mvn, int r_x, int r_y, uint8_t* is_intra,
+                          int dir);
+void ff_vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2],
+                      int direct, int mvtype);
+void ff_vc1_pred_b_mv_intfi(VC1Context *v, int n, int *dmv_x, int *dmv_y,
+                            int mv1, int *pred_flag);
+
+static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)
+{
+    int n = bfrac;
+
+#if B_FRACTION_DEN==256
+    if (inv)
+        n -= 256;
+    if (!qs)
+        return 2 * ((value * n + 255) >> 9);
+    return (value * n + 128) >> 8;
+#else
+    if (inv)
+        n -= B_FRACTION_DEN;
+    if (!qs)
+        return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));
+    return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;
+#endif
+}
+
+#endif /* AVCODEC_VC1_PRED_H */
index 84e8188..66c569b 100644 (file)
@@ -94,8 +94,6 @@ extern VLC ff_vc1_ac_coeff_table[8];
 #define VC1_IF_MBMODE_VLC_BITS 5
 //@}
 
-
-/* Denominator used for ff_vc1_bfraction_lut */
 #define B_FRACTION_DEN  256
 
 /* pre-computed scales for all bfractions and base=256 */
index 7eb8549..610b13e 100644 (file)
  * VC-1 and WMV3 decoder
  */
 
-#include "internal.h"
 #include "avcodec.h"
-#include "error_resilience.h"
+#include "blockdsp.h"
+#include "get_bits.h"
+#include "internal.h"
 #include "mpeg_er.h"
-#include "mpegutils.h"
 #include "mpegvideo.h"
-#include "h263.h"
-#include "h264chroma.h"
-#include "qpeldsp.h"
+#include "msmpeg4data.h"
 #include "vc1.h"
 #include "vc1data.h"
-#include "vc1acdata.h"
-#include "msmpeg4data.h"
-#include "unary.h"
-#include "mathops.h"
-
-#undef NDEBUG
-#include <assert.h>
-
-#define MB_INTRA_VLC_BITS 9
-#define DC_VLC_BITS 9
-
-
-// offset tables for interlaced picture MVDATA decoding
-static const int offset_table1[9] = {  0,  1,  2,  4,  8, 16, 32,  64, 128 };
-static const int offset_table2[9] = {  0,  1,  3,  7, 15, 31, 63, 127, 255 };
-
-/***********************************************************************/
-/**
- * @name VC-1 Bitplane decoding
- * @see 8.7, p56
- * @{
- */
-
-/**
- * Imode types
- * @{
- */
-enum Imode {
-    IMODE_RAW,
-    IMODE_NORM2,
-    IMODE_DIFF2,
-    IMODE_NORM6,
-    IMODE_DIFF6,
-    IMODE_ROWSKIP,
-    IMODE_COLSKIP
-};
-/** @} */ //imode defines
-
-static void init_block_index(VC1Context *v)
-{
-    MpegEncContext *s = &v->s;
-    ff_init_block_index(s);
-    if (v->field_mode && !(v->second_field ^ v->tff)) {
-        s->dest[0] += s->current_picture_ptr->f->linesize[0];
-        s->dest[1] += s->current_picture_ptr->f->linesize[1];
-        s->dest[2] += s->current_picture_ptr->f->linesize[2];
-    }
-}
-
-/** @} */ //Bitplane group
-
-static void vc1_put_signed_blocks_clamped(VC1Context *v)
-{
-    MpegEncContext *s = &v->s;
-    int topleft_mb_pos, top_mb_pos;
-    int stride_y, fieldtx = 0;
-    int v_dist;
-
-    /* The put pixels loop is always one MB row behind the decoding loop,
-     * because we can only put pixels when overlap filtering is done, and
-     * for filtering of the bottom edge of a MB, we need the next MB row
-     * present as well.
-     * Within the row, the put pixels loop is also one MB col behind the
-     * decoding loop. The reason for this is again, because for filtering
-     * of the right MB edge, we need the next MB present. */
-    if (!s->first_slice_line) {
-        if (s->mb_x) {
-            topleft_mb_pos = (s->mb_y - 1) * s->mb_stride + s->mb_x - 1;
-            if (v->fcm == ILACE_FRAME)
-                fieldtx = v->fieldtx_plane[topleft_mb_pos];
-            stride_y       = s->linesize << fieldtx;
-            v_dist         = (16 - fieldtx) >> (fieldtx == 0);
-            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][0],
-                                              s->dest[0] - 16 * s->linesize - 16,
-                                              stride_y);
-            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][1],
-                                              s->dest[0] - 16 * s->linesize - 8,
-                                              stride_y);
-            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][2],
-                                              s->dest[0] - v_dist * s->linesize - 16,
-                                              stride_y);
-            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][3],
-                                              s->dest[0] - v_dist * s->linesize - 8,
-                                              stride_y);
-            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][4],
-                                              s->dest[1] - 8 * s->uvlinesize - 8,
-                                              s->uvlinesize);
-            s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][5],
-                                              s->dest[2] - 8 * s->uvlinesize - 8,
-                                              s->uvlinesize);
-        }
-        if (s->mb_x == s->mb_width - 1) {
-            top_mb_pos = (s->mb_y - 1) * s->mb_stride + s->mb_x;
-            if (v->fcm == ILACE_FRAME)
-                fieldtx = v->fieldtx_plane[top_mb_pos];
-            stride_y   = s->linesize << fieldtx;
-            v_dist     = fieldtx ? 15 : 8;
-            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][0],
-                                              s->dest[0] - 16 * s->linesize,
-                                              stride_y);
-            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][1],
-                                              s->dest[0] - 16 * s->linesize + 8,
-                                              stride_y);
-            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][2],
-                                              s->dest[0] - v_dist * s->linesize,
-                                              stride_y);
-            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][3],
-                                              s->dest[0] - v_dist * s->linesize + 8,
-                                              stride_y);
-            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][4],
-                                              s->dest[1] - 8 * s->uvlinesize,
-                                              s->uvlinesize);
-            s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][5],
-                                              s->dest[2] - 8 * s->uvlinesize,
-                                              s->uvlinesize);
-        }
-    }
-
-#define inc_blk_idx(idx) do { \
-        idx++; \
-        if (idx >= v->n_allocated_blks) \
-            idx = 0; \
-    } while (0)
-
-    inc_blk_idx(v->topleft_blk_idx);
-    inc_blk_idx(v->top_blk_idx);
-    inc_blk_idx(v->left_blk_idx);
-    inc_blk_idx(v->cur_blk_idx);
-}
-
-static void vc1_loop_filter_iblk(VC1Context *v, int pq)
-{
-    MpegEncContext *s = &v->s;
-    int j;
-    if (!s->first_slice_line) {
-        v->vc1dsp.vc1_v_loop_filter16(s->dest[0], s->linesize, pq);
-        if (s->mb_x)
-            v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);
-        v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize + 8, s->linesize, pq);
-        for (j = 0; j < 2; j++) {
-            v->vc1dsp.vc1_v_loop_filter8(s->dest[j + 1], s->uvlinesize, pq);
-            if (s->mb_x)
-                v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize, s->uvlinesize, pq);
-        }
-    }
-    v->vc1dsp.vc1_v_loop_filter16(s->dest[0] + 8 * s->linesize, s->linesize, pq);
-
-    if (s->mb_y == s->end_mb_y - 1) {
-        if (s->mb_x) {
-            v->vc1dsp.vc1_h_loop_filter16(s->dest[0], s->linesize, pq);
-            v->vc1dsp.vc1_h_loop_filter8(s->dest[1], s->uvlinesize, pq);
-            v->vc1dsp.vc1_h_loop_filter8(s->dest[2], s->uvlinesize, pq);
-        }
-        v->vc1dsp.vc1_h_loop_filter16(s->dest[0] + 8, s->linesize, pq);
-    }
-}
-
-static void vc1_loop_filter_iblk_delayed(VC1Context *v, int pq)
-{
-    MpegEncContext *s = &v->s;
-    int j;
-
-    /* The loopfilter runs 1 row and 1 column behind the overlap filter, which
-     * means it runs two rows/cols behind the decoding loop. */
-    if (!s->first_slice_line) {
-        if (s->mb_x) {
-            if (s->mb_y >= s->start_mb_y + 2) {
-                v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 16 * s->linesize - 16, s->linesize, pq);
-
-                if (s->mb_x >= 2)
-                    v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize - 16, s->linesize, pq);
-                v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize - 8, s->linesize, pq);
-                for (j = 0; j < 2; j++) {
-                    v->vc1dsp.vc1_v_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize - 8, s->uvlinesize, pq);
-                    if (s->mb_x >= 2) {
-                        v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 16 * s->uvlinesize - 8, s->uvlinesize, pq);
-                    }
-                }
-            }
-            v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 8 * s->linesize - 16, s->linesize, pq);
-        }
-
-        if (s->mb_x == s->mb_width - 1) {
-            if (s->mb_y >= s->start_mb_y + 2) {
-                v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);
-
-                if (s->mb_x)
-                    v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize, s->linesize, pq);
-                v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize + 8, s->linesize, pq);
-                for (j = 0; j < 2; j++) {
-                    v->vc1dsp.vc1_v_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize, s->uvlinesize, pq);
-                    if (s->mb_x >= 2) {
-                        v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 16 * s->uvlinesize, s->uvlinesize, pq);
-                    }
-                }
-            }
-            v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 8 * s->linesize, s->linesize, pq);
-        }
-
-        if (s->mb_y == s->end_mb_y) {
-            if (s->mb_x) {
-                if (s->mb_x >= 2)
-                    v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize - 16, s->linesize, pq);
-                v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize - 8, s->linesize, pq);
-                if (s->mb_x >= 2) {
-                    for (j = 0; j < 2; j++) {
-                        v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize - 8, s->uvlinesize, pq);
-                    }
-                }
-            }
-
-            if (s->mb_x == s->mb_width - 1) {
-                if (s->mb_x)
-                    v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);
-                v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize + 8, s->linesize, pq);
-                if (s->mb_x) {
-                    for (j = 0; j < 2; j++) {
-                        v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize, s->uvlinesize, pq);
-                    }
-                }
-            }
-        }
-    }
-}
-
-static void vc1_smooth_overlap_filter_iblk(VC1Context *v)
-{
-    MpegEncContext *s = &v->s;
-    int mb_pos;
-
-    if (v->condover == CONDOVER_NONE)
-        return;
-
-    mb_pos = s->mb_x + s->mb_y * s->mb_stride;
-
-    /* Within a MB, the horizontal overlap always runs before the vertical.
-     * To accomplish that, we run the H on left and internal borders of the
-     * currently decoded MB. Then, we wait for the next overlap iteration
-     * to do H overlap on the right edge of this MB, before moving over and
-     * running the V overlap. Therefore, the V overlap makes us trail by one
-     * MB col and the H overlap filter makes us trail by one MB row. This
-     * is reflected in the time at which we run the put_pixels loop. */
-    if (v->condover == CONDOVER_ALL || v->pq >= 9 || v->over_flags_plane[mb_pos]) {
-        if (s->mb_x && (v->condover == CONDOVER_ALL || v->pq >= 9 ||
-                        v->over_flags_plane[mb_pos - 1])) {
-            v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][1],
-                                      v->block[v->cur_blk_idx][0]);
-            v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][3],
-                                      v->block[v->cur_blk_idx][2]);
-            if (!(s->flags & CODEC_FLAG_GRAY)) {
-                v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][4],
-                                          v->block[v->cur_blk_idx][4]);
-                v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][5],
-                                          v->block[v->cur_blk_idx][5]);
-            }
-        }
-        v->vc1dsp.vc1_h_s_overlap(v->block[v->cur_blk_idx][0],
-                                  v->block[v->cur_blk_idx][1]);
-        v->vc1dsp.vc1_h_s_overlap(v->block[v->cur_blk_idx][2],
-                                  v->block[v->cur_blk_idx][3]);
-
-        if (s->mb_x == s->mb_width - 1) {
-            if (!s->first_slice_line && (v->condover == CONDOVER_ALL || v->pq >= 9 ||
-                                         v->over_flags_plane[mb_pos - s->mb_stride])) {
-                v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][2],
-                                          v->block[v->cur_blk_idx][0]);
-                v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][3],
-                                          v->block[v->cur_blk_idx][1]);
-                if (!(s->flags & CODEC_FLAG_GRAY)) {
-                    v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][4],
-                                              v->block[v->cur_blk_idx][4]);
-                    v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][5],
-                                              v->block[v->cur_blk_idx][5]);
-                }
-            }
-            v->vc1dsp.vc1_v_s_overlap(v->block[v->cur_blk_idx][0],
-                                      v->block[v->cur_blk_idx][2]);
-            v->vc1dsp.vc1_v_s_overlap(v->block[v->cur_blk_idx][1],
-                                      v->block[v->cur_blk_idx][3]);
-        }
-    }
-    if (s->mb_x && (v->condover == CONDOVER_ALL || v->over_flags_plane[mb_pos - 1])) {
-        if (!s->first_slice_line && (v->condover == CONDOVER_ALL || v->pq >= 9 ||
-                                     v->over_flags_plane[mb_pos - s->mb_stride - 1])) {
-            v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][2],
-                                      v->block[v->left_blk_idx][0]);
-            v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][3],
-                                      v->block[v->left_blk_idx][1]);
-            if (!(s->flags & CODEC_FLAG_GRAY)) {
-                v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][4],
-                                          v->block[v->left_blk_idx][4]);
-                v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][5],
-                                          v->block[v->left_blk_idx][5]);
-            }
-        }
-        v->vc1dsp.vc1_v_s_overlap(v->block[v->left_blk_idx][0],
-                                  v->block[v->left_blk_idx][2]);
-        v->vc1dsp.vc1_v_s_overlap(v->block[v->left_blk_idx][1],
-                                  v->block[v->left_blk_idx][3]);
-    }
-}
-
-/** Do motion compensation over 1 macroblock
- * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
- */
-static void vc1_mc_1mv(VC1Context *v, int dir)
-{
-    MpegEncContext *s = &v->s;
-    H264ChromaContext *h264chroma = &v->h264chroma;
-    uint8_t *srcY, *srcU, *srcV;
-    int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
-    int v_edge_pos = s->v_edge_pos >> v->field_mode;
-    int i;
-    uint8_t (*luty)[256], (*lutuv)[256];
-    int use_ic;
-
-    if ((!v->field_mode ||
-         (v->ref_field_type[dir] == 1 && v->cur_field_type == 1)) &&
-        !v->s.last_picture.f->data[0])
-        return;
-
-    mx = s->mv[dir][0][0];
-    my = s->mv[dir][0][1];
-
-    // store motion vectors for further use in B frames
-    if (s->pict_type == AV_PICTURE_TYPE_P) {
-        for (i = 0; i < 4; i++) {
-            s->current_picture.motion_val[1][s->block_index[i] + v->blocks_off][0] = mx;
-            s->current_picture.motion_val[1][s->block_index[i] + v->blocks_off][1] = my;
-        }
-    }
-
-    uvmx = (mx + ((mx & 3) == 3)) >> 1;
-    uvmy = (my + ((my & 3) == 3)) >> 1;
-    v->luma_mv[s->mb_x][0] = uvmx;
-    v->luma_mv[s->mb_x][1] = uvmy;
-
-    if (v->field_mode &&
-        v->cur_field_type != v->ref_field_type[dir]) {
-        my   = my   - 2 + 4 * v->cur_field_type;
-        uvmy = uvmy - 2 + 4 * v->cur_field_type;
-    }
-
-    // fastuvmc shall be ignored for interlaced frame picture
-    if (v->fastuvmc && (v->fcm != ILACE_FRAME)) {
-        uvmx = uvmx + ((uvmx < 0) ? (uvmx & 1) : -(uvmx & 1));
-        uvmy = uvmy + ((uvmy < 0) ? (uvmy & 1) : -(uvmy & 1));
-    }
-    if (!dir) {
-        if (v->field_mode && (v->cur_field_type != v->ref_field_type[dir]) && v->second_field) {
-            srcY = s->current_picture.f->data[0];
-            srcU = s->current_picture.f->data[1];
-            srcV = s->current_picture.f->data[2];
-            luty  = v->curr_luty;
-            lutuv = v->curr_lutuv;
-            use_ic = v->curr_use_ic;
-        } else {
-            srcY = s->last_picture.f->data[0];
-            srcU = s->last_picture.f->data[1];
-            srcV = s->last_picture.f->data[2];
-            luty  = v->last_luty;
-            lutuv = v->last_lutuv;
-            use_ic = v->last_use_ic;
-        }
-    } else {
-        srcY = s->next_picture.f->data[0];
-        srcU = s->next_picture.f->data[1];
-        srcV = s->next_picture.f->data[2];
-        luty  = v->next_luty;
-        lutuv = v->next_lutuv;
-        use_ic = v->next_use_ic;
-    }
-
-    if (!srcY || !srcU) {
-        av_log(v->s.avctx, AV_LOG_ERROR, "Referenced frame missing.\n");
-        return;
-    }
-
-    src_x   = s->mb_x * 16 + (mx   >> 2);
-    src_y   = s->mb_y * 16 + (my   >> 2);
-    uvsrc_x = s->mb_x *  8 + (uvmx >> 2);
-    uvsrc_y = s->mb_y *  8 + (uvmy >> 2);
-
-    if (v->profile != PROFILE_ADVANCED) {
-        src_x   = av_clip(  src_x, -16, s->mb_width  * 16);
-        src_y   = av_clip(  src_y, -16, s->mb_height * 16);
-        uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);
-        uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);
-    } else {
-        src_x   = av_clip(  src_x, -17, s->avctx->coded_width);
-        src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);
-        uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);
-        uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);
-    }
-
-    srcY += src_y   * s->linesize   + src_x;
-    srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
-    srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
-
-    if (v->field_mode && v->ref_field_type[dir]) {
-        srcY += s->current_picture_ptr->f->linesize[0];
-        srcU += s->current_picture_ptr->f->linesize[1];
-        srcV += s->current_picture_ptr->f->linesize[2];
-    }
-
-    /* for grayscale we should not try to read from unknown area */
-    if (s->flags & CODEC_FLAG_GRAY) {
-        srcU = s->edge_emu_buffer + 18 * s->linesize;
-        srcV = s->edge_emu_buffer + 18 * s->linesize;
-    }
-
-    if (v->rangeredfrm || use_ic
-        || s->h_edge_pos < 22 || v_edge_pos < 22
-        || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel * 3
-        || (unsigned)(src_y - 1)        > v_edge_pos    - (my&3) - 16 - 3) {
-        uint8_t *uvbuf = s->edge_emu_buffer + 19 * s->linesize;
-
-        srcY -= s->mspel * (1 + s->linesize);
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY,
-                                 s->linesize, s->linesize,
-                                 17 + s->mspel * 2, 17 + s->mspel * 2,
-                                 src_x - s->mspel, src_y - s->mspel,
-                                 s->h_edge_pos, v_edge_pos);
-        srcY = s->edge_emu_buffer;
-        s->vdsp.emulated_edge_mc(uvbuf, srcU,
-                                 s->uvlinesize, s->uvlinesize,
-                                 8 + 1, 8 + 1,
-                                 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);
-        s->vdsp.emulated_edge_mc(uvbuf + 16, srcV,
-                                 s->uvlinesize, s->uvlinesize,
-                                 8 + 1, 8 + 1,
-                                 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);
-        srcU = uvbuf;
-        srcV = uvbuf + 16;
-        /* if we deal with range reduction we need to scale source blocks */
-        if (v->rangeredfrm) {
-            int i, j;
-            uint8_t *src, *src2;
-
-            src = srcY;
-            for (j = 0; j < 17 + s->mspel * 2; j++) {
-                for (i = 0; i < 17 + s->mspel * 2; i++)
-                    src[i] = ((src[i] - 128) >> 1) + 128;
-                src += s->linesize;
-            }
-            src  = srcU;
-            src2 = srcV;
-            for (j = 0; j < 9; j++) {
-                for (i = 0; i < 9; i++) {
-                    src[i]  = ((src[i]  - 128) >> 1) + 128;
-                    src2[i] = ((src2[i] - 128) >> 1) + 128;
-                }
-                src  += s->uvlinesize;
-                src2 += s->uvlinesize;
-            }
-        }
-        /* if we deal with intensity compensation we need to scale source blocks */
-        if (use_ic) {
-            int i, j;
-            uint8_t *src, *src2;
-
-            src = srcY;
-            for (j = 0; j < 17 + s->mspel * 2; j++) {
-                int f = v->field_mode ? v->ref_field_type[dir] : ((j + src_y - s->mspel) & 1) ;
-                for (i = 0; i < 17 + s->mspel * 2; i++)
-                    src[i] = luty[f][src[i]];
-                src += s->linesize;
-            }
-            src  = srcU;
-            src2 = srcV;
-            for (j = 0; j < 9; j++) {
-                int f = v->field_mode ? v->ref_field_type[dir] : ((j + uvsrc_y) & 1);
-                for (i = 0; i < 9; i++) {
-                    src[i]  = lutuv[f][src[i]];
-                    src2[i] = lutuv[f][src2[i]];
-                }
-                src  += s->uvlinesize;
-                src2 += s->uvlinesize;
-            }
-        }
-        srcY += s->mspel * (1 + s->linesize);
-    }
-
-    if (s->mspel) {
-        dxy = ((my & 3) << 2) | (mx & 3);
-        v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0]    , srcY    , s->linesize, v->rnd);
-        v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
-        srcY += s->linesize * 8;
-        v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize    , srcY    , s->linesize, v->rnd);
-        v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
-    } else { // hpel mc - always used for luma
-        dxy = (my & 2) | ((mx & 2) >> 1);
-        if (!v->rnd)
-            s->hdsp.put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
-        else
-            s->hdsp.put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
-    }
-
-    if (s->flags & CODEC_FLAG_GRAY) return;
-    /* Chroma MC always uses qpel bilinear */
-    uvmx = (uvmx & 3) << 1;
-    uvmy = (uvmy & 3) << 1;
-    if (!v->rnd) {
-        h264chroma->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
-        h264chroma->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
-    } else {
-        v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
-        v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
-    }
-}
-
-static inline int median4(int a, int b, int c, int d)
-{
-    if (a < b) {
-        if (c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;
-        else       return (FFMIN(b, c) + FFMAX(a, d)) / 2;
-    } else {
-        if (c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;
-        else       return (FFMIN(a, c) + FFMAX(b, d)) / 2;
-    }
-}
-
-/** Do motion compensation for 4-MV macroblock - luminance block
- */
-static void vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg)
-{
-    MpegEncContext *s = &v->s;
-    uint8_t *srcY;
-    int dxy, mx, my, src_x, src_y;
-    int off;
-    int fieldmv = (v->fcm == ILACE_FRAME) ? v->blk_mv_type[s->block_index[n]] : 0;
-    int v_edge_pos = s->v_edge_pos >> v->field_mode;
-    uint8_t (*luty)[256];
-    int use_ic;
-
-    if ((!v->field_mode ||
-         (v->ref_field_type[dir] == 1 && v->cur_field_type == 1)) &&
-        !v->s.last_picture.f->data[0])
-        return;
-
-    mx = s->mv[dir][n][0];
-    my = s->mv[dir][n][1];
-
-    if (!dir) {
-        if (v->field_mode && (v->cur_field_type != v->ref_field_type[dir]) && v->second_field) {
-            srcY = s->current_picture.f->data[0];
-            luty = v->curr_luty;
-            use_ic = v->curr_use_ic;
-        } else {
-            srcY = s->last_picture.f->data[0];
-            luty = v->last_luty;
-            use_ic = v->last_use_ic;
-        }
-    } else {
-        srcY = s->next_picture.f->data[0];
-        luty = v->next_luty;
-        use_ic = v->next_use_ic;
-    }
-
-    if (!srcY) {
-        av_log(v->s.avctx, AV_LOG_ERROR, "Referenced frame missing.\n");
-        return;
-    }
-
-    if (v->field_mode) {
-        if (v->cur_field_type != v->ref_field_type[dir])
-            my = my - 2 + 4 * v->cur_field_type;
-    }
-
-    if (s->pict_type == AV_PICTURE_TYPE_P && n == 3 && v->field_mode) {
-        int same_count = 0, opp_count = 0, k;
-        int chosen_mv[2][4][2], f;
-        int tx, ty;
-        for (k = 0; k < 4; k++) {
-            f = v->mv_f[0][s->block_index[k] + v->blocks_off];
-            chosen_mv[f][f ? opp_count : same_count][0] = s->mv[0][k][0];
-            chosen_mv[f][f ? opp_count : same_count][1] = s->mv[0][k][1];
-            opp_count  += f;
-            same_count += 1 - f;
-        }
-        f = opp_count > same_count;
-        switch (f ? opp_count : same_count) {
-        case 4:
-            tx = median4(chosen_mv[f][0][0], chosen_mv[f][1][0],
-                         chosen_mv[f][2][0], chosen_mv[f][3][0]);
-            ty = median4(chosen_mv[f][0][1], chosen_mv[f][1][1],
-                         chosen_mv[f][2][1], chosen_mv[f][3][1]);
-            break;
-        case 3:
-            tx = mid_pred(chosen_mv[f][0][0], chosen_mv[f][1][0], chosen_mv[f][2][0]);
-            ty = mid_pred(chosen_mv[f][0][1], chosen_mv[f][1][1], chosen_mv[f][2][1]);
-            break;
-        case 2:
-            tx = (chosen_mv[f][0][0] + chosen_mv[f][1][0]) / 2;
-            ty = (chosen_mv[f][0][1] + chosen_mv[f][1][1]) / 2;
-            break;
-        }
-        s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = tx;
-        s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = ty;
-        for (k = 0; k < 4; k++)
-            v->mv_f[1][s->block_index[k] + v->blocks_off] = f;
-    }
-
-    if (v->fcm == ILACE_FRAME) {  // not sure if needed for other types of picture
-        int qx, qy;
-        int width  = s->avctx->coded_width;
-        int height = s->avctx->coded_height >> 1;
-        if (s->pict_type == AV_PICTURE_TYPE_P) {
-            s->current_picture.motion_val[1][s->block_index[n] + v->blocks_off][0] = mx;
-            s->current_picture.motion_val[1][s->block_index[n] + v->blocks_off][1] = my;
-        }
-        qx = (s->mb_x * 16) + (mx >> 2);
-        qy = (s->mb_y *  8) + (my >> 3);
-
-        if (qx < -17)
-            mx -= 4 * (qx + 17);
-        else if (qx > width)
-            mx -= 4 * (qx - width);
-        if (qy < -18)
-            my -= 8 * (qy + 18);
-        else if (qy > height + 1)
-            my -= 8 * (qy - height - 1);
-    }
-
-    if ((v->fcm == ILACE_FRAME) && fieldmv)
-        off = ((n > 1) ? s->linesize : 0) + (n & 1) * 8;
-    else
-        off = s->linesize * 4 * (n & 2) + (n & 1) * 8;
-
-    src_x = s->mb_x * 16 + (n & 1) * 8 + (mx >> 2);
-    if (!fieldmv)
-        src_y = s->mb_y * 16 + (n & 2) * 4 + (my >> 2);
-    else
-        src_y = s->mb_y * 16 + ((n > 1) ? 1 : 0) + (my >> 2);
-
-    if (v->profile != PROFILE_ADVANCED) {
-        src_x = av_clip(src_x, -16, s->mb_width  * 16);
-        src_y = av_clip(src_y, -16, s->mb_height * 16);
-    } else {
-        src_x = av_clip(src_x, -17, s->avctx->coded_width);
-        if (v->fcm == ILACE_FRAME) {
-            if (src_y & 1)
-                src_y = av_clip(src_y, -17, s->avctx->coded_height + 1);
-            else
-                src_y = av_clip(src_y, -18, s->avctx->coded_height);
-        } else {
-            src_y = av_clip(src_y, -18, s->avctx->coded_height + 1);
-        }
-    }
-
-    srcY += src_y * s->linesize + src_x;
-    if (v->field_mode && v->ref_field_type[dir])
-        srcY += s->current_picture_ptr->f->linesize[0];
-
-    if (fieldmv && !(src_y & 1))
-        v_edge_pos--;
-    if (fieldmv && (src_y & 1) && src_y < 4)
-        src_y--;
-    if (v->rangeredfrm || use_ic
-        || s->h_edge_pos < 13 || v_edge_pos < 23
-        || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx & 3) - 8 - s->mspel * 2
-        || (unsigned)(src_y - (s->mspel << fieldmv)) > v_edge_pos - (my & 3) - ((8 + s->mspel * 2) << fieldmv)) {
-        srcY -= s->mspel * (1 + (s->linesize << fieldmv));
-        /* check emulate edge stride and offset */
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY,
-                                 s->linesize, s->linesize,
-                                 9 + s->mspel * 2, (9 + s->mspel * 2) << fieldmv,
-                                 src_x - s->mspel, src_y - (s->mspel << fieldmv),
-                                 s->h_edge_pos, v_edge_pos);
-        srcY = s->edge_emu_buffer;
-        /* if we deal with range reduction we need to scale source blocks */
-        if (v->rangeredfrm) {
-            int i, j;
-            uint8_t *src;
-
-            src = srcY;
-            for (j = 0; j < 9 + s->mspel * 2; j++) {
-                for (i = 0; i < 9 + s->mspel * 2; i++)
-                    src[i] = ((src[i] - 128) >> 1) + 128;
-                src += s->linesize << fieldmv;
-            }
-        }
-        /* if we deal with intensity compensation we need to scale source blocks */
-        if (use_ic) {
-            int i, j;
-            uint8_t *src;
-
-            src = srcY;
-            for (j = 0; j < 9 + s->mspel * 2; j++) {
-                int f = v->field_mode ? v->ref_field_type[dir] : (((j<<fieldmv)+src_y - (s->mspel << fieldmv)) & 1);
-                for (i = 0; i < 9 + s->mspel * 2; i++)
-                    src[i] = luty[f][src[i]];
-                src += s->linesize << fieldmv;
-            }
-        }
-        srcY += s->mspel * (1 + (s->linesize << fieldmv));
-    }
-
-    if (s->mspel) {
-        dxy = ((my & 3) << 2) | (mx & 3);
-        if (avg)
-            v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd);
-        else
-            v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd);
-    } else { // hpel mc - always used for luma
-        dxy = (my & 2) | ((mx & 2) >> 1);
-        if (!v->rnd)
-            s->hdsp.put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
-        else
-            s->hdsp.put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
-    }
-}
-
-static av_always_inline int get_chroma_mv(int *mvx, int *mvy, int *a, int flag, int *tx, int *ty)
-{
-    int idx, i;
-    static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
-
-    idx =  ((a[3] != flag) << 3)
-         | ((a[2] != flag) << 2)
-         | ((a[1] != flag) << 1)
-         |  (a[0] != flag);
-    if (!idx) {
-        *tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);
-        *ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);
-        return 4;
-    } else if (count[idx] == 1) {
-        switch (idx) {
-        case 0x1:
-            *tx = mid_pred(mvx[1], mvx[2], mvx[3]);
-            *ty = mid_pred(mvy[1], mvy[2], mvy[3]);
-            return 3;
-        case 0x2:
-            *tx = mid_pred(mvx[0], mvx[2], mvx[3]);
-            *ty = mid_pred(mvy[0], mvy[2], mvy[3]);
-            return 3;
-        case 0x4:
-            *tx = mid_pred(mvx[0], mvx[1], mvx[3]);
-            *ty = mid_pred(mvy[0], mvy[1], mvy[3]);
-            return 3;
-        case 0x8:
-            *tx = mid_pred(mvx[0], mvx[1], mvx[2]);
-            *ty = mid_pred(mvy[0], mvy[1], mvy[2]);
-            return 3;
-        }
-    } else if (count[idx] == 2) {
-        int t1 = 0, t2 = 0;
-        for (i = 0; i < 3; i++)
-            if (!a[i]) {
-                t1 = i;
-                break;
-            }
-        for (i = t1 + 1; i < 4; i++)
-            if (!a[i]) {
-                t2 = i;
-                break;
-            }
-        *tx = (mvx[t1] + mvx[t2]) / 2;
-        *ty = (mvy[t1] + mvy[t2]) / 2;
-        return 2;
-    } else {
-        return 0;
-    }
-    return -1;
-}
-
-/** Do motion compensation for 4-MV macroblock - both chroma blocks
- */
-static void vc1_mc_4mv_chroma(VC1Context *v, int dir)
-{
-    MpegEncContext *s = &v->s;
-    H264ChromaContext *h264chroma = &v->h264chroma;
-    uint8_t *srcU, *srcV;
-    int uvmx, uvmy, uvsrc_x, uvsrc_y;
-    int k, tx = 0, ty = 0;
-    int mvx[4], mvy[4], intra[4], mv_f[4];
-    int valid_count;
-    int chroma_ref_type = v->cur_field_type;
-    int v_edge_pos = s->v_edge_pos >> v->field_mode;
-    uint8_t (*lutuv)[256];
-    int use_ic;
-
-    if (!v->field_mode && !v->s.last_picture.f->data[0])
-        return;
-    if (s->flags & CODEC_FLAG_GRAY)
-        return;
-
-    for (k = 0; k < 4; k++) {
-        mvx[k] = s->mv[dir][k][0];
-        mvy[k] = s->mv[dir][k][1];
-        intra[k] = v->mb_type[0][s->block_index[k]];
-        if (v->field_mode)
-            mv_f[k] = v->mv_f[dir][s->block_index[k] + v->blocks_off];
-    }
-
-    /* calculate chroma MV vector from four luma MVs */
-    if (!v->field_mode || (v->field_mode && !v->numref)) {
-        valid_count = get_chroma_mv(mvx, mvy, intra, 0, &tx, &ty);
-        chroma_ref_type = v->reffield;
-        if (!valid_count) {
-            s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;
-            s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;
-            v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;
-            return; //no need to do MC for intra blocks
-        }
-    } else {
-        int dominant = 0;
-        if (mv_f[0] + mv_f[1] + mv_f[2] + mv_f[3] > 2)
-            dominant = 1;
-        valid_count = get_chroma_mv(mvx, mvy, mv_f, dominant, &tx, &ty);
-        if (dominant)
-            chroma_ref_type = !v->cur_field_type;
-    }
-    if (v->field_mode && chroma_ref_type == 1 && v->cur_field_type == 1 && !v->s.last_picture.f->data[0])
-        return;
-    s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = tx;
-    s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = ty;
-    uvmx = (tx + ((tx & 3) == 3)) >> 1;
-    uvmy = (ty + ((ty & 3) == 3)) >> 1;
-
-    v->luma_mv[s->mb_x][0] = uvmx;
-    v->luma_mv[s->mb_x][1] = uvmy;
-
-    if (v->fastuvmc) {
-        uvmx = uvmx + ((uvmx < 0) ? (uvmx & 1) : -(uvmx & 1));
-        uvmy = uvmy + ((uvmy < 0) ? (uvmy & 1) : -(uvmy & 1));
-    }
-    // Field conversion bias
-    if (v->cur_field_type != chroma_ref_type)
-        uvmy += 2 - 4 * chroma_ref_type;
-
-    uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
-    uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
-
-    if (v->profile != PROFILE_ADVANCED) {
-        uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width  * 8);
-        uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
-    } else {
-        uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width  >> 1);
-        uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
-    }
-
-    if (!dir) {
-        if (v->field_mode && (v->cur_field_type != chroma_ref_type) && v->second_field) {
-            srcU = s->current_picture.f->data[1];
-            srcV = s->current_picture.f->data[2];
-            lutuv = v->curr_lutuv;
-            use_ic = v->curr_use_ic;
-        } else {
-            srcU = s->last_picture.f->data[1];
-            srcV = s->last_picture.f->data[2];
-            lutuv = v->last_lutuv;
-            use_ic = v->last_use_ic;
-        }
-    } else {
-        srcU = s->next_picture.f->data[1];
-        srcV = s->next_picture.f->data[2];
-        lutuv = v->next_lutuv;
-        use_ic = v->next_use_ic;
-    }
-
-    if (!srcU) {
-        av_log(v->s.avctx, AV_LOG_ERROR, "Referenced frame missing.\n");
-        return;
-    }
-
-    srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
-    srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
-
-    if (v->field_mode) {
-        if (chroma_ref_type) {
-            srcU += s->current_picture_ptr->f->linesize[1];
-            srcV += s->current_picture_ptr->f->linesize[2];
-        }
-    }
-
-    if (v->rangeredfrm || use_ic
-        || s->h_edge_pos < 18 || v_edge_pos < 18
-        || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
-        || (unsigned)uvsrc_y > (v_edge_pos    >> 1) - 9) {
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcU,
-                                 s->uvlinesize, s->uvlinesize,
-                                 8 + 1, 8 + 1, uvsrc_x, uvsrc_y,
-                                 s->h_edge_pos >> 1, v_edge_pos >> 1);
-        s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV,
-                                 s->uvlinesize, s->uvlinesize,
-                                 8 + 1, 8 + 1, uvsrc_x, uvsrc_y,
-                                 s->h_edge_pos >> 1, v_edge_pos >> 1);
-        srcU = s->edge_emu_buffer;
-        srcV = s->edge_emu_buffer + 16;
-
-        /* if we deal with range reduction we need to scale source blocks */
-        if (v->rangeredfrm) {
-            int i, j;
-            uint8_t *src, *src2;
-
-            src  = srcU;
-            src2 = srcV;
-            for (j = 0; j < 9; j++) {
-                for (i = 0; i < 9; i++) {
-                    src[i]  = ((src[i]  - 128) >> 1) + 128;
-                    src2[i] = ((src2[i] - 128) >> 1) + 128;
-                }
-                src  += s->uvlinesize;
-                src2 += s->uvlinesize;
-            }
-        }
-        /* if we deal with intensity compensation we need to scale source blocks */
-        if (use_ic) {
-            int i, j;
-            uint8_t *src, *src2;
-
-            src  = srcU;
-            src2 = srcV;
-            for (j = 0; j < 9; j++) {
-                int f = v->field_mode ? chroma_ref_type : ((j + uvsrc_y) & 1);
-                for (i = 0; i < 9; i++) {
-                    src[i]  = lutuv[f][src[i]];
-                    src2[i] = lutuv[f][src2[i]];
-                }
-                src  += s->uvlinesize;
-                src2 += s->uvlinesize;
-            }
-        }
-    }
-
-    /* Chroma MC always uses qpel bilinear */
-    uvmx = (uvmx & 3) << 1;
-    uvmy = (uvmy & 3) << 1;
-    if (!v->rnd) {
-        h264chroma->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
-        h264chroma->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
-    } else {
-        v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
-        v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
-    }
-}
-
-/** Do motion compensation for 4-MV interlaced frame chroma macroblock (both U and V)
- */
-static void vc1_mc_4mv_chroma4(VC1Context *v, int dir, int dir2, int avg)
-{
-    MpegEncContext *s = &v->s;
-    H264ChromaContext *h264chroma = &v->h264chroma;
-    uint8_t *srcU, *srcV;
-    int uvsrc_x, uvsrc_y;
-    int uvmx_field[4], uvmy_field[4];
-    int i, off, tx, ty;
-    int fieldmv = v->blk_mv_type[s->block_index[0]];
-    static const int s_rndtblfield[16] = { 0, 0, 1, 2, 4, 4, 5, 6, 2, 2, 3, 8, 6, 6, 7, 12 };
-    int v_dist = fieldmv ? 1 : 4; // vertical offset for lower sub-blocks
-    int v_edge_pos = s->v_edge_pos >> 1;
-    int use_ic;
-    uint8_t (*lutuv)[256];
-
-    if (s->flags & CODEC_FLAG_GRAY)
-        return;
-
-    for (i = 0; i < 4; i++) {
-        int d = i < 2 ? dir: dir2;
-        tx = s->mv[d][i][0];
-        uvmx_field[i] = (tx + ((tx & 3) == 3)) >> 1;
-        ty = s->mv[d][i][1];
-        if (fieldmv)
-            uvmy_field[i] = (ty >> 4) * 8 + s_rndtblfield[ty & 0xF];
-        else
-            uvmy_field[i] = (ty + ((ty & 3) == 3)) >> 1;
-    }
-
-    for (i = 0; i < 4; i++) {
-        off = (i & 1) * 4 + ((i & 2) ? v_dist * s->uvlinesize : 0);
-        uvsrc_x = s->mb_x * 8 +  (i & 1) * 4           + (uvmx_field[i] >> 2);
-        uvsrc_y = s->mb_y * 8 + ((i & 2) ? v_dist : 0) + (uvmy_field[i] >> 2);
-        // FIXME: implement proper pull-back (see vc1cropmv.c, vc1CROPMV_ChromaPullBack())
-        uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width  >> 1);
-        uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
-        if (i < 2 ? dir : dir2) {
-            srcU = s->next_picture.f->data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
-            srcV = s->next_picture.f->data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
-            lutuv  = v->next_lutuv;
-            use_ic = v->next_use_ic;
-        } else {
-            srcU = s->last_picture.f->data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
-            srcV = s->last_picture.f->data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
-            lutuv  = v->last_lutuv;
-            use_ic = v->last_use_ic;
-        }
-        uvmx_field[i] = (uvmx_field[i] & 3) << 1;
-        uvmy_field[i] = (uvmy_field[i] & 3) << 1;
-
-        if (fieldmv && !(uvsrc_y & 1))
-            v_edge_pos--;
-        if (fieldmv && (uvsrc_y & 1) && uvsrc_y < 2)
-            uvsrc_y--;
-        if (use_ic
-            || s->h_edge_pos < 10 || v_edge_pos < (5 << fieldmv)
-            || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 5
-            || (unsigned)uvsrc_y > v_edge_pos - (5 << fieldmv)) {
-            s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcU,
-                                     s->uvlinesize, s->uvlinesize,
-                                     5, (5 << fieldmv), uvsrc_x, uvsrc_y,
-                                     s->h_edge_pos >> 1, v_edge_pos);
-            s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV,
-                                     s->uvlinesize, s->uvlinesize,
-                                     5, (5 << fieldmv), uvsrc_x, uvsrc_y,
-                                     s->h_edge_pos >> 1, v_edge_pos);
-            srcU = s->edge_emu_buffer;
-            srcV = s->edge_emu_buffer + 16;
-
-            /* if we deal with intensity compensation we need to scale source blocks */
-            if (use_ic) {
-                int i, j;
-                uint8_t *src, *src2;
-
-                src  = srcU;
-                src2 = srcV;
-                for (j = 0; j < 5; j++) {
-                    int f = (uvsrc_y + (j << fieldmv)) & 1;
-                    for (i = 0; i < 5; i++) {
-                        src[i]  = lutuv[f][src[i]];
-                        src2[i] = lutuv[f][src2[i]];
-                    }
-                    src  += s->uvlinesize << fieldmv;
-                    src2 += s->uvlinesize << fieldmv;
-                }
-            }
-        }
-        if (avg) {
-            if (!v->rnd) {
-                h264chroma->avg_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
-                h264chroma->avg_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
-            } else {
-                v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
-                v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
-            }
-        } else {
-            if (!v->rnd) {
-                h264chroma->put_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
-                h264chroma->put_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
-            } else {
-                v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
-                v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);
-            }
-        }
-    }
-}
-
-/***********************************************************************/
-/**
- * @name VC-1 Block-level functions
- * @see 7.1.4, p91 and 8.1.1.7, p(1)04
- * @{
- */
-
-/**
- * @def GET_MQUANT
- * @brief Get macroblock-level quantizer scale
- */
-#define GET_MQUANT()                                           \
-    if (v->dquantfrm) {                                        \
-        int edges = 0;                                         \
-        if (v->dqprofile == DQPROFILE_ALL_MBS) {               \
-            if (v->dqbilevel) {                                \
-                mquant = (get_bits1(gb)) ? v->altpq : v->pq;   \
-            } else {                                           \
-                mqdiff = get_bits(gb, 3);                      \
-                if (mqdiff != 7)                               \
-                    mquant = v->pq + mqdiff;                   \
-                else                                           \
-                    mquant = get_bits(gb, 5);                  \
-            }                                                  \
-        }                                                      \
-        if (v->dqprofile == DQPROFILE_SINGLE_EDGE)             \
-            edges = 1 << v->dqsbedge;                          \
-        else if (v->dqprofile == DQPROFILE_DOUBLE_EDGES)       \
-            edges = (3 << v->dqsbedge) % 15;                   \
-        else if (v->dqprofile == DQPROFILE_FOUR_EDGES)         \
-            edges = 15;                                        \
-        if ((edges&1) && !s->mb_x)                             \
-            mquant = v->altpq;                                 \
-        if ((edges&2) && s->first_slice_line)                  \
-            mquant = v->altpq;                                 \
-        if ((edges&4) && s->mb_x == (s->mb_width - 1))         \
-            mquant = v->altpq;                                 \
-        if ((edges&8) && s->mb_y == (s->mb_height - 1))        \
-            mquant = v->altpq;                                 \
-        if (!mquant || mquant > 31) {                          \
-            av_log(v->s.avctx, AV_LOG_ERROR,                   \
-                   "Overriding invalid mquant %d\n", mquant);  \
-            mquant = 1;                                        \
-        }                                                      \
-    }
-
-/**
- * @def GET_MVDATA(_dmv_x, _dmv_y)
- * @brief Get MV differentials
- * @see MVDATA decoding from 8.3.5.2, p(1)20
- * @param _dmv_x Horizontal differential for decoded MV
- * @param _dmv_y Vertical differential for decoded MV
- */
-#define GET_MVDATA(_dmv_x, _dmv_y)                                      \
-    index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table, \
-                         VC1_MV_DIFF_VLC_BITS, 2);                      \
-    if (index > 36) {                                                   \
-        mb_has_coeffs = 1;                                              \
-        index -= 37;                                                    \
-    } else                                                              \
-        mb_has_coeffs = 0;                                              \
-    s->mb_intra = 0;                                                    \
-    if (!index) {                                                       \
-        _dmv_x = _dmv_y = 0;                                            \
-    } else if (index == 35) {                                           \
-        _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample);          \
-        _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample);          \
-    } else if (index == 36) {                                           \
-        _dmv_x = 0;                                                     \
-        _dmv_y = 0;                                                     \
-        s->mb_intra = 1;                                                \
-    } else {                                                            \
-        index1 = index % 6;                                             \
-        if (!s->quarter_sample && index1 == 5) val = 1;                 \
-        else                                   val = 0;                 \
-        if (size_table[index1] - val > 0)                               \
-            val = get_bits(gb, size_table[index1] - val);               \
-        else                                   val = 0;                 \
-        sign = 0 - (val&1);                                             \
-        _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \
-                                                                        \
-        index1 = index / 6;                                             \
-        if (!s->quarter_sample && index1 == 5) val = 1;                 \
-        else                                   val = 0;                 \
-        if (size_table[index1] - val > 0)                               \
-            val = get_bits(gb, size_table[index1] - val);               \
-        else                                   val = 0;                 \
-        sign = 0 - (val & 1);                                           \
-        _dmv_y = (sign ^ ((val >> 1) + offset_table[index1])) - sign;   \
-    }
-
-static av_always_inline void get_mvdata_interlaced(VC1Context *v, int *dmv_x,
-                                                   int *dmv_y, int *pred_flag)
-{
-    int index, index1;
-    int extend_x = 0, extend_y = 0;
-    GetBitContext *gb = &v->s.gb;
-    int bits, esc;
-    int val, sign;
-    const int* offs_tab;
-
-    if (v->numref) {
-        bits = VC1_2REF_MVDATA_VLC_BITS;
-        esc  = 125;
-    } else {
-        bits = VC1_1REF_MVDATA_VLC_BITS;
-        esc  = 71;
-    }
-    switch (v->dmvrange) {
-    case 1:
-        extend_x = 1;
-        break;
-    case 2:
-        extend_y = 1;
-        break;
-    case 3:
-        extend_x = extend_y = 1;
-        break;
-    }
-    index = get_vlc2(gb, v->imv_vlc->table, bits, 3);
-    if (index == esc) {
-        *dmv_x = get_bits(gb, v->k_x);
-        *dmv_y = get_bits(gb, v->k_y);
-        if (v->numref) {
-            if (pred_flag) {
-                *pred_flag = *dmv_y & 1;
-                *dmv_y     = (*dmv_y + *pred_flag) >> 1;
-            } else {
-                *dmv_y     = (*dmv_y + (*dmv_y & 1)) >> 1;
-            }
-        }
-    }
-    else {
-        if (extend_x)
-            offs_tab = offset_table2;
-        else
-            offs_tab = offset_table1;
-        index1 = (index + 1) % 9;
-        if (index1 != 0) {
-            val    = get_bits(gb, index1 + extend_x);
-            sign   = 0 -(val & 1);
-            *dmv_x = (sign ^ ((val >> 1) + offs_tab[index1])) - sign;
-        } else
-            *dmv_x = 0;
-        if (extend_y)
-            offs_tab = offset_table2;
-        else
-            offs_tab = offset_table1;
-        index1 = (index + 1) / 9;
-        if (index1 > v->numref) {
-            val    = get_bits(gb, (index1 + (extend_y << v->numref)) >> v->numref);
-            sign   = 0 - (val & 1);
-            *dmv_y = (sign ^ ((val >> 1) + offs_tab[index1 >> v->numref])) - sign;
-        } else
-            *dmv_y = 0;
-        if (v->numref && pred_flag)
-            *pred_flag = index1 & 1;
-    }
-}
-
-static av_always_inline int scaleforsame_x(VC1Context *v, int n /* MV */, int dir)
-{
-    int scaledvalue, refdist;
-    int scalesame1, scalesame2;
-    int scalezone1_x, zone1offset_x;
-    int table_index = dir ^ v->second_field;
-
-    if (v->s.pict_type != AV_PICTURE_TYPE_B)
-        refdist = v->refdist;
-    else
-        refdist = dir ? v->brfd : v->frfd;
-    if (refdist > 3)
-        refdist = 3;
-    scalesame1    = ff_vc1_field_mvpred_scales[table_index][1][refdist];
-    scalesame2    = ff_vc1_field_mvpred_scales[table_index][2][refdist];
-    scalezone1_x  = ff_vc1_field_mvpred_scales[table_index][3][refdist];
-    zone1offset_x = ff_vc1_field_mvpred_scales[table_index][5][refdist];
-
-    if (FFABS(n) > 255)
-        scaledvalue = n;
-    else {
-        if (FFABS(n) < scalezone1_x)
-            scaledvalue = (n * scalesame1) >> 8;
-        else {
-            if (n < 0)
-                scaledvalue = ((n * scalesame2) >> 8) - zone1offset_x;
-            else
-                scaledvalue = ((n * scalesame2) >> 8) + zone1offset_x;
-        }
-    }
-    return av_clip(scaledvalue, -v->range_x, v->range_x - 1);
-}
-
-static av_always_inline int scaleforsame_y(VC1Context *v, int i, int n /* MV */, int dir)
-{
-    int scaledvalue, refdist;
-    int scalesame1, scalesame2;
-    int scalezone1_y, zone1offset_y;
-    int table_index = dir ^ v->second_field;
-
-    if (v->s.pict_type != AV_PICTURE_TYPE_B)
-        refdist = v->refdist;
-    else
-        refdist = dir ? v->brfd : v->frfd;
-    if (refdist > 3)
-        refdist = 3;
-    scalesame1    = ff_vc1_field_mvpred_scales[table_index][1][refdist];
-    scalesame2    = ff_vc1_field_mvpred_scales[table_index][2][refdist];
-    scalezone1_y  = ff_vc1_field_mvpred_scales[table_index][4][refdist];
-    zone1offset_y = ff_vc1_field_mvpred_scales[table_index][6][refdist];
-
-    if (FFABS(n) > 63)
-        scaledvalue = n;
-    else {
-        if (FFABS(n) < scalezone1_y)
-            scaledvalue = (n * scalesame1) >> 8;
-        else {
-            if (n < 0)
-                scaledvalue = ((n * scalesame2) >> 8) - zone1offset_y;
-            else
-                scaledvalue = ((n * scalesame2) >> 8) + zone1offset_y;
-        }
-    }
-
-    if (v->cur_field_type && !v->ref_field_type[dir])
-        return av_clip(scaledvalue, -v->range_y / 2 + 1, v->range_y / 2);
-    else
-        return av_clip(scaledvalue, -v->range_y / 2, v->range_y / 2 - 1);
-}
-
-static av_always_inline int scaleforopp_x(VC1Context *v, int n /* MV */)
-{
-    int scalezone1_x, zone1offset_x;
-    int scaleopp1, scaleopp2, brfd;
-    int scaledvalue;
-
-    brfd = FFMIN(v->brfd, 3);
-    scalezone1_x  = ff_vc1_b_field_mvpred_scales[3][brfd];
-    zone1offset_x = ff_vc1_b_field_mvpred_scales[5][brfd];
-    scaleopp1     = ff_vc1_b_field_mvpred_scales[1][brfd];
-    scaleopp2     = ff_vc1_b_field_mvpred_scales[2][brfd];
-
-    if (FFABS(n) > 255)
-        scaledvalue = n;
-    else {
-        if (FFABS(n) < scalezone1_x)
-            scaledvalue = (n * scaleopp1) >> 8;
-        else {
-            if (n < 0)
-                scaledvalue = ((n * scaleopp2) >> 8) - zone1offset_x;
-            else
-                scaledvalue = ((n * scaleopp2) >> 8) + zone1offset_x;
-        }
-    }
-    return av_clip(scaledvalue, -v->range_x, v->range_x - 1);
-}
-
-static av_always_inline int scaleforopp_y(VC1Context *v, int n /* MV */, int dir)
-{
-    int scalezone1_y, zone1offset_y;
-    int scaleopp1, scaleopp2, brfd;
-    int scaledvalue;
-
-    brfd = FFMIN(v->brfd, 3);
-    scalezone1_y  = ff_vc1_b_field_mvpred_scales[4][brfd];
-    zone1offset_y = ff_vc1_b_field_mvpred_scales[6][brfd];
-    scaleopp1     = ff_vc1_b_field_mvpred_scales[1][brfd];
-    scaleopp2     = ff_vc1_b_field_mvpred_scales[2][brfd];
-
-    if (FFABS(n) > 63)
-        scaledvalue = n;
-    else {
-        if (FFABS(n) < scalezone1_y)
-            scaledvalue = (n * scaleopp1) >> 8;
-        else {
-            if (n < 0)
-                scaledvalue = ((n * scaleopp2) >> 8) - zone1offset_y;
-            else
-                scaledvalue = ((n * scaleopp2) >> 8) + zone1offset_y;
-        }
-    }
-    if (v->cur_field_type && !v->ref_field_type[dir]) {
-        return av_clip(scaledvalue, -v->range_y / 2 + 1, v->range_y / 2);
-    } else {
-        return av_clip(scaledvalue, -v->range_y / 2, v->range_y / 2 - 1);
-    }
-}
-
-static av_always_inline int scaleforsame(VC1Context *v, int i, int n /* MV */,
-                                         int dim, int dir)
-{
-    int brfd, scalesame;
-    int hpel = 1 - v->s.quarter_sample;
-
-    n >>= hpel;
-    if (v->s.pict_type != AV_PICTURE_TYPE_B || v->second_field || !dir) {
-        if (dim)
-            n = scaleforsame_y(v, i, n, dir) << hpel;
-        else
-            n = scaleforsame_x(v, n, dir) << hpel;
-        return n;
-    }
-    brfd      = FFMIN(v->brfd, 3);
-    scalesame = ff_vc1_b_field_mvpred_scales[0][brfd];
-
-    n = (n * scalesame >> 8) << hpel;
-    return n;
-}
-
-static av_always_inline int scaleforopp(VC1Context *v, int n /* MV */,
-                                        int dim, int dir)
-{
-    int refdist, scaleopp;
-    int hpel = 1 - v->s.quarter_sample;
-
-    n >>= hpel;
-    if (v->s.pict_type == AV_PICTURE_TYPE_B && !v->second_field && dir == 1) {
-        if (dim)
-            n = scaleforopp_y(v, n, dir) << hpel;
-        else
-            n = scaleforopp_x(v, n) << hpel;
-        return n;
-    }
-    if (v->s.pict_type != AV_PICTURE_TYPE_B)
-        refdist = FFMIN(v->refdist, 3);
-    else
-        refdist = dir ? v->brfd : v->frfd;
-    scaleopp = ff_vc1_field_mvpred_scales[dir ^ v->second_field][0][refdist];
-
-    n = (n * scaleopp >> 8) << hpel;
-    return n;
-}
-
-/** Predict and set motion vector
- */
-static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y,
-                               int mv1, int r_x, int r_y, uint8_t* is_intra,
-                               int pred_flag, int dir)
-{
-    MpegEncContext *s = &v->s;
-    int xy, wrap, off = 0;
-    int16_t *A, *B, *C;
-    int px, py;
-    int sum;
-    int mixedmv_pic, num_samefield = 0, num_oppfield = 0;
-    int opposite, a_f, b_f, c_f;
-    int16_t field_predA[2];
-    int16_t field_predB[2];
-    int16_t field_predC[2];
-    int a_valid, b_valid, c_valid;
-    int hybridmv_thresh, y_bias = 0;
-
-    if (v->mv_mode == MV_PMODE_MIXED_MV ||
-        ((v->mv_mode == MV_PMODE_INTENSITY_COMP) && (v->mv_mode2 == MV_PMODE_MIXED_MV)))
-        mixedmv_pic = 1;
-    else
-        mixedmv_pic = 0;
-    /* scale MV difference to be quad-pel */
-    dmv_x <<= 1 - s->quarter_sample;
-    dmv_y <<= 1 - s->quarter_sample;
-
-    wrap = s->b8_stride;
-    xy   = s->block_index[n];
-
-    if (s->mb_intra) {
-        s->mv[0][n][0] = s->current_picture.motion_val[0][xy + v->blocks_off][0] = 0;
-        s->mv[0][n][1] = s->current_picture.motion_val[0][xy + v->blocks_off][1] = 0;
-        s->current_picture.motion_val[1][xy + v->blocks_off][0] = 0;
-        s->current_picture.motion_val[1][xy + v->blocks_off][1] = 0;
-        if (mv1) { /* duplicate motion data for 1-MV block */
-            s->current_picture.motion_val[0][xy + 1 + v->blocks_off][0]        = 0;
-            s->current_picture.motion_val[0][xy + 1 + v->blocks_off][1]        = 0;
-            s->current_picture.motion_val[0][xy + wrap + v->blocks_off][0]     = 0;
-            s->current_picture.motion_val[0][xy + wrap + v->blocks_off][1]     = 0;
-            s->current_picture.motion_val[0][xy + wrap + 1 + v->blocks_off][0] = 0;
-            s->current_picture.motion_val[0][xy + wrap + 1 + v->blocks_off][1] = 0;
-            v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;
-            s->current_picture.motion_val[1][xy + 1 + v->blocks_off][0]        = 0;
-            s->current_picture.motion_val[1][xy + 1 + v->blocks_off][1]        = 0;
-            s->current_picture.motion_val[1][xy + wrap][0]                     = 0;
-            s->current_picture.motion_val[1][xy + wrap + v->blocks_off][1]     = 0;
-            s->current_picture.motion_val[1][xy + wrap + 1 + v->blocks_off][0] = 0;
-            s->current_picture.motion_val[1][xy + wrap + 1 + v->blocks_off][1] = 0;
-        }
-        return;
-    }
-
-    C = s->current_picture.motion_val[dir][xy -    1 + v->blocks_off];
-    A = s->current_picture.motion_val[dir][xy - wrap + v->blocks_off];
-    if (mv1) {
-        if (v->field_mode && mixedmv_pic)
-            off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
-        else
-            off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;
-    } else {
-        //in 4-MV mode different blocks have different B predictor position
-        switch (n) {
-        case 0:
-            off = (s->mb_x > 0) ? -1 : 1;
-            break;
-        case 1:
-            off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;
-            break;
-        case 2:
-            off = 1;
-            break;
-        case 3:
-            off = -1;
-        }
-    }
-    B = s->current_picture.motion_val[dir][xy - wrap + off + v->blocks_off];
-
-    a_valid = !s->first_slice_line || (n == 2 || n == 3);
-    b_valid = a_valid && (s->mb_width > 1);
-    c_valid = s->mb_x || (n == 1 || n == 3);
-    if (v->field_mode) {
-        a_valid = a_valid && !is_intra[xy - wrap];
-        b_valid = b_valid && !is_intra[xy - wrap + off];
-        c_valid = c_valid && !is_intra[xy - 1];
-    }
-
-    if (a_valid) {
-        a_f = v->mv_f[dir][xy - wrap + v->blocks_off];
-        num_oppfield  += a_f;
-        num_samefield += 1 - a_f;
-        field_predA[0] = A[0];
-        field_predA[1] = A[1];
-    } else {
-        field_predA[0] = field_predA[1] = 0;
-        a_f = 0;
-    }
-    if (b_valid) {
-        b_f = v->mv_f[dir][xy - wrap + off + v->blocks_off];
-        num_oppfield  += b_f;
-        num_samefield += 1 - b_f;
-        field_predB[0] = B[0];
-        field_predB[1] = B[1];
-    } else {
-        field_predB[0] = field_predB[1] = 0;
-        b_f = 0;
-    }
-    if (c_valid) {
-        c_f = v->mv_f[dir][xy - 1 + v->blocks_off];
-        num_oppfield  += c_f;
-        num_samefield += 1 - c_f;
-        field_predC[0] = C[0];
-        field_predC[1] = C[1];
-    } else {
-        field_predC[0] = field_predC[1] = 0;
-        c_f = 0;
-    }
-
-    if (v->field_mode) {
-        if (!v->numref)
-            // REFFIELD determines if the last field or the second-last field is
-            // to be used as reference
-            opposite = 1 - v->reffield;
-        else {
-            if (num_samefield <= num_oppfield)
-                opposite = 1 - pred_flag;
-            else
-                opposite = pred_flag;
-        }
-    } else
-        opposite = 0;
-    if (opposite) {
-        if (a_valid && !a_f) {
-            field_predA[0] = scaleforopp(v, field_predA[0], 0, dir);
-            field_predA[1] = scaleforopp(v, field_predA[1], 1, dir);
-        }
-        if (b_valid && !b_f) {
-            field_predB[0] = scaleforopp(v, field_predB[0], 0, dir);
-            field_predB[1] = scaleforopp(v, field_predB[1], 1, dir);
-        }
-        if (c_valid && !c_f) {
-            field_predC[0] = scaleforopp(v, field_predC[0], 0, dir);
-            field_predC[1] = scaleforopp(v, field_predC[1], 1, dir);
-        }
-        v->mv_f[dir][xy + v->blocks_off] = 1;
-        v->ref_field_type[dir] = !v->cur_field_type;
-    } else {
-        if (a_valid && a_f) {
-            field_predA[0] = scaleforsame(v, n, field_predA[0], 0, dir);
-            field_predA[1] = scaleforsame(v, n, field_predA[1], 1, dir);
-        }
-        if (b_valid && b_f) {
-            field_predB[0] = scaleforsame(v, n, field_predB[0], 0, dir);
-            field_predB[1] = scaleforsame(v, n, field_predB[1], 1, dir);
-        }
-        if (c_valid && c_f) {
-            field_predC[0] = scaleforsame(v, n, field_predC[0], 0, dir);
-            field_predC[1] = scaleforsame(v, n, field_predC[1], 1, dir);
-        }
-        v->mv_f[dir][xy + v->blocks_off] = 0;
-        v->ref_field_type[dir] = v->cur_field_type;
-    }
-
-    if (a_valid) {
-        px = field_predA[0];
-        py = field_predA[1];
-    } else if (c_valid) {
-        px = field_predC[0];
-        py = field_predC[1];
-    } else if (b_valid) {
-        px = field_predB[0];
-        py = field_predB[1];
-    } else {
-        px = 0;
-        py = 0;
-    }
-
-    if (num_samefield + num_oppfield > 1) {
-        px = mid_pred(field_predA[0], field_predB[0], field_predC[0]);
-        py = mid_pred(field_predA[1], field_predB[1], field_predC[1]);
-    }
-
-    /* Pullback MV as specified in 8.3.5.3.4 */
-    if (!v->field_mode) {
-        int qx, qy, X, Y;
-        qx = (s->mb_x << 6) + ((n == 1 || n == 3) ? 32 : 0);
-        qy = (s->mb_y << 6) + ((n == 2 || n == 3) ? 32 : 0);
-        X  = (s->mb_width  << 6) - 4;
-        Y  = (s->mb_height << 6) - 4;
-        if (mv1) {
-            if (qx + px < -60) px = -60 - qx;
-            if (qy + py < -60) py = -60 - qy;
-        } else {
-            if (qx + px < -28) px = -28 - qx;
-            if (qy + py < -28) py = -28 - qy;
-        }
-        if (qx + px > X) px = X - qx;
-        if (qy + py > Y) py = Y - qy;
-    }
-
-    if (!v->field_mode || s->pict_type != AV_PICTURE_TYPE_B) {
-        /* Calculate hybrid prediction as specified in 8.3.5.3.5 (also 10.3.5.4.3.5) */
-        hybridmv_thresh = 32;
-        if (a_valid && c_valid) {
-            if (is_intra[xy - wrap])
-                sum = FFABS(px) + FFABS(py);
-            else
-                sum = FFABS(px - field_predA[0]) + FFABS(py - field_predA[1]);
-            if (sum > hybridmv_thresh) {
-                if (get_bits1(&s->gb)) {     // read HYBRIDPRED bit
-                    px = field_predA[0];
-                    py = field_predA[1];
-                } else {
-                    px = field_predC[0];
-                    py = field_predC[1];
-                }
-            } else {
-                if (is_intra[xy - 1])
-                    sum = FFABS(px) + FFABS(py);
-                else
-                    sum = FFABS(px - field_predC[0]) + FFABS(py - field_predC[1]);
-                if (sum > hybridmv_thresh) {
-                    if (get_bits1(&s->gb)) {
-                        px = field_predA[0];
-                        py = field_predA[1];
-                    } else {
-                        px = field_predC[0];
-                        py = field_predC[1];
-                    }
-                }
-            }
-        }
-    }
-
-    if (v->field_mode && v->numref)
-        r_y >>= 1;
-    if (v->field_mode && v->cur_field_type && v->ref_field_type[dir] == 0)
-        y_bias = 1;
-    /* store MV using signed modulus of MV range defined in 4.11 */
-    s->mv[dir][n][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;
-    s->mv[dir][n][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1] = ((py + dmv_y + r_y - y_bias) & ((r_y << 1) - 1)) - r_y + y_bias;
-    if (mv1) { /* duplicate motion data for 1-MV block */
-        s->current_picture.motion_val[dir][xy +    1 +     v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0];
-        s->current_picture.motion_val[dir][xy +    1 +     v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1];
-        s->current_picture.motion_val[dir][xy + wrap +     v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0];
-        s->current_picture.motion_val[dir][xy + wrap +     v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1];
-        s->current_picture.motion_val[dir][xy + wrap + 1 + v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0];
-        s->current_picture.motion_val[dir][xy + wrap + 1 + v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1];
-        v->mv_f[dir][xy +    1 + v->blocks_off] = v->mv_f[dir][xy +            v->blocks_off];
-        v->mv_f[dir][xy + wrap + v->blocks_off] = v->mv_f[dir][xy + wrap + 1 + v->blocks_off] = v->mv_f[dir][xy + v->blocks_off];
-    }
-}
-
-/** Predict and set motion vector for interlaced frame picture MBs
- */
-static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y,
-                                     int mvn, int r_x, int r_y, uint8_t* is_intra, int dir)
-{
-    MpegEncContext *s = &v->s;
-    int xy, wrap, off = 0;
-    int A[2], B[2], C[2];
-    int px, py;
-    int a_valid = 0, b_valid = 0, c_valid = 0;
-    int field_a, field_b, field_c; // 0: same, 1: opposit
-    int total_valid, num_samefield, num_oppfield;
-    int pos_c, pos_b, n_adj;
-
-    wrap = s->b8_stride;
-    xy = s->block_index[n];
-
-    if (s->mb_intra) {
-        s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;
-        s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;
-        s->current_picture.motion_val[1][xy][0] = 0;
-        s->current_picture.motion_val[1][xy][1] = 0;
-        if (mvn == 1) { /* duplicate motion data for 1-MV block */
-            s->current_picture.motion_val[0][xy + 1][0]        = 0;
-            s->current_picture.motion_val[0][xy + 1][1]        = 0;
-            s->current_picture.motion_val[0][xy + wrap][0]     = 0;
-            s->current_picture.motion_val[0][xy + wrap][1]     = 0;
-            s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;
-            s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;
-            v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;
-            s->current_picture.motion_val[1][xy + 1][0]        = 0;
-            s->current_picture.motion_val[1][xy + 1][1]        = 0;
-            s->current_picture.motion_val[1][xy + wrap][0]     = 0;
-            s->current_picture.motion_val[1][xy + wrap][1]     = 0;
-            s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;
-            s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;
-        }
-        return;
-    }
-
-    off = ((n == 0) || (n == 1)) ? 1 : -1;
-    /* predict A */
-    if (s->mb_x || (n == 1) || (n == 3)) {
-        if ((v->blk_mv_type[xy]) // current block (MB) has a field MV
-            || (!v->blk_mv_type[xy] && !v->blk_mv_type[xy - 1])) { // or both have frame MV
-            A[0] = s->current_picture.motion_val[dir][xy - 1][0];
-            A[1] = s->current_picture.motion_val[dir][xy - 1][1];
-            a_valid = 1;
-        } else { // current block has frame mv and cand. has field MV (so average)
-            A[0] = (s->current_picture.motion_val[dir][xy - 1][0]
-                    + s->current_picture.motion_val[dir][xy - 1 + off * wrap][0] + 1) >> 1;
-            A[1] = (s->current_picture.motion_val[dir][xy - 1][1]
-                    + s->current_picture.motion_val[dir][xy - 1 + off * wrap][1] + 1) >> 1;
-            a_valid = 1;
-        }
-        if (!(n & 1) && v->is_intra[s->mb_x - 1]) {
-            a_valid = 0;
-            A[0] = A[1] = 0;
-        }
-    } else
-        A[0] = A[1] = 0;
-    /* Predict B and C */
-    B[0] = B[1] = C[0] = C[1] = 0;
-    if (n == 0 || n == 1 || v->blk_mv_type[xy]) {
-        if (!s->first_slice_line) {
-            if (!v->is_intra[s->mb_x - s->mb_stride]) {
-                b_valid = 1;
-                n_adj   = n | 2;
-                pos_b   = s->block_index[n_adj] - 2 * wrap;
-                if (v->blk_mv_type[pos_b] && v->blk_mv_type[xy]) {
-                    n_adj = (n & 2) | (n & 1);
-                }
-                B[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap][0];
-                B[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap][1];
-                if (v->blk_mv_type[pos_b] && !v->blk_mv_type[xy]) {
-                    B[0] = (B[0] + s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap][0] + 1) >> 1;
-                    B[1] = (B[1] + s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap][1] + 1) >> 1;
-                }
-            }
-            if (s->mb_width > 1) {
-                if (!v->is_intra[s->mb_x - s->mb_stride + 1]) {
-                    c_valid = 1;
-                    n_adj   = 2;
-                    pos_c   = s->block_index[2] - 2 * wrap + 2;
-                    if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) {
-                        n_adj = n & 2;
-                    }
-                    C[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap + 2][0];
-                    C[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap + 2][1];
-                    if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) {
-                        C[0] = (1 + C[0] + (s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap + 2][0])) >> 1;
-                        C[1] = (1 + C[1] + (s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap + 2][1])) >> 1;
-                    }
-                    if (s->mb_x == s->mb_width - 1) {
-                        if (!v->is_intra[s->mb_x - s->mb_stride - 1]) {
-                            c_valid = 1;
-                            n_adj   = 3;
-                            pos_c   = s->block_index[3] - 2 * wrap - 2;
-                            if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) {
-                                n_adj = n | 1;
-                            }
-                            C[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap - 2][0];
-                            C[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap - 2][1];
-                            if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) {
-                                C[0] = (1 + C[0] + s->current_picture.motion_val[dir][s->block_index[1] - 2 * wrap - 2][0]) >> 1;
-                                C[1] = (1 + C[1] + s->current_picture.motion_val[dir][s->block_index[1] - 2 * wrap - 2][1]) >> 1;
-                            }
-                        } else
-                            c_valid = 0;
-                    }
-                }
-            }
-        }
-    } else {
-        pos_b   = s->block_index[1];
-        b_valid = 1;
-        B[0]    = s->current_picture.motion_val[dir][pos_b][0];
-        B[1]    = s->current_picture.motion_val[dir][pos_b][1];
-        pos_c   = s->block_index[0];
-        c_valid = 1;
-        C[0]    = s->current_picture.motion_val[dir][pos_c][0];
-        C[1]    = s->current_picture.motion_val[dir][pos_c][1];
-    }
-
-    total_valid = a_valid + b_valid + c_valid;
-    // check if predictor A is out of bounds
-    if (!s->mb_x && !(n == 1 || n == 3)) {
-        A[0] = A[1] = 0;
-    }
-    // check if predictor B is out of bounds
-    if ((s->first_slice_line && v->blk_mv_type[xy]) || (s->first_slice_line && !(n & 2))) {
-        B[0] = B[1] = C[0] = C[1] = 0;
-    }
-    if (!v->blk_mv_type[xy]) {
-        if (s->mb_width == 1) {
-            px = B[0];
-            py = B[1];
-        } else {
-            if (total_valid >= 2) {
-                px = mid_pred(A[0], B[0], C[0]);
-                py = mid_pred(A[1], B[1], C[1]);
-            } else if (total_valid) {