ac3enc: select bandwidth based on bit rate, sample rate, and number of
[ffmpeg.git] / libavcodec / rv40.c
index 791ed17..ea6c3c1 100644 (file)
@@ -2,31 +2,34 @@
  * RV40 decoder
  * Copyright (c) 2007 Konstantin Shishkov
  *
- * This file is part of FFmpeg.
+ * This file is part of Libav.
  *
- * FFmpeg is free software; you can redistribute it and/or
+ * Libav is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
  * License as published by the Free Software Foundation; either
  * version 2.1 of the License, or (at your option) any later version.
  *
- * FFmpeg is distributed in the hope that it will be useful,
+ * Libav is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  * Lesser General Public License for more details.
  *
  * You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
+ * License along with Libav; if not, write to the Free Software
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
 /**
- * @file rv40.c
+ * @file
  * RV40 decoder
  */
 
+#include "libavutil/imgutils.h"
+
 #include "avcodec.h"
 #include "dsputil.h"
 #include "mpegvideo.h"
+#include "golomb.h"
 
 #include "rv34.h"
 #include "rv40vlc2.h"
@@ -36,38 +39,60 @@ static VLC aic_top_vlc;
 static VLC aic_mode1_vlc[AIC_MODE1_NUM], aic_mode2_vlc[AIC_MODE2_NUM];
 static VLC ptype_vlc[NUM_PTYPE_VLCS], btype_vlc[NUM_BTYPE_VLCS];
 
+static const int16_t mode2_offs[] = {
+       0,  614, 1222, 1794, 2410,  3014,  3586,  4202,  4792, 5382, 5966, 6542,
+    7138, 7716, 8292, 8864, 9444, 10030, 10642, 11212, 11814
+};
+
 /**
  * Initialize all tables.
  */
-static void rv40_init_tables()
+static av_cold void rv40_init_tables(void)
 {
     int i;
+    static VLC_TYPE aic_table[1 << AIC_TOP_BITS][2];
+    static VLC_TYPE aic_mode1_table[AIC_MODE1_NUM << AIC_MODE1_BITS][2];
+    static VLC_TYPE aic_mode2_table[11814][2];
+    static VLC_TYPE ptype_table[NUM_PTYPE_VLCS << PTYPE_VLC_BITS][2];
+    static VLC_TYPE btype_table[NUM_BTYPE_VLCS << BTYPE_VLC_BITS][2];
 
+    aic_top_vlc.table = aic_table;
+    aic_top_vlc.table_allocated = 1 << AIC_TOP_BITS;
     init_vlc(&aic_top_vlc, AIC_TOP_BITS, AIC_TOP_SIZE,
              rv40_aic_top_vlc_bits,  1, 1,
-             rv40_aic_top_vlc_codes, 1, 1, INIT_VLC_USE_STATIC);
+             rv40_aic_top_vlc_codes, 1, 1, INIT_VLC_USE_NEW_STATIC);
     for(i = 0; i < AIC_MODE1_NUM; i++){
         // Every tenth VLC table is empty
         if((i % 10) == 9) continue;
+        aic_mode1_vlc[i].table = &aic_mode1_table[i << AIC_MODE1_BITS];
+        aic_mode1_vlc[i].table_allocated = 1 << AIC_MODE1_BITS;
         init_vlc(&aic_mode1_vlc[i], AIC_MODE1_BITS, AIC_MODE1_SIZE,
                  aic_mode1_vlc_bits[i],  1, 1,
-                 aic_mode1_vlc_codes[i], 1, 1, INIT_VLC_USE_STATIC);
+                 aic_mode1_vlc_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);
     }
     for(i = 0; i < AIC_MODE2_NUM; i++){
+        aic_mode2_vlc[i].table = &aic_mode2_table[mode2_offs[i]];
+        aic_mode2_vlc[i].table_allocated = mode2_offs[i + 1] - mode2_offs[i];
         init_vlc(&aic_mode2_vlc[i], AIC_MODE2_BITS, AIC_MODE2_SIZE,
                  aic_mode2_vlc_bits[i],  1, 1,
-                 aic_mode2_vlc_codes[i], 2, 2, INIT_VLC_USE_STATIC);
+                 aic_mode2_vlc_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
     }
-    for(i = 0; i < NUM_PTYPE_VLCS; i++)
-         init_vlc_sparse(&ptype_vlc[i], PTYPE_VLC_BITS, PTYPE_VLC_SIZE,
+    for(i = 0; i < NUM_PTYPE_VLCS; i++){
+        ptype_vlc[i].table = &ptype_table[i << PTYPE_VLC_BITS];
+        ptype_vlc[i].table_allocated = 1 << PTYPE_VLC_BITS;
+        init_vlc_sparse(&ptype_vlc[i], PTYPE_VLC_BITS, PTYPE_VLC_SIZE,
                          ptype_vlc_bits[i],  1, 1,
                          ptype_vlc_codes[i], 1, 1,
-                         ptype_vlc_syms,     1, 1, INIT_VLC_USE_STATIC);
-    for(i = 0; i < NUM_BTYPE_VLCS; i++)
-         init_vlc_sparse(&btype_vlc[i], BTYPE_VLC_BITS, BTYPE_VLC_SIZE,
+                         ptype_vlc_syms,     1, 1, INIT_VLC_USE_NEW_STATIC);
+    }
+    for(i = 0; i < NUM_BTYPE_VLCS; i++){
+        btype_vlc[i].table = &btype_table[i << BTYPE_VLC_BITS];
+        btype_vlc[i].table_allocated = 1 << BTYPE_VLC_BITS;
+        init_vlc_sparse(&btype_vlc[i], BTYPE_VLC_BITS, BTYPE_VLC_SIZE,
                          btype_vlc_bits[i],  1, 1,
                          btype_vlc_codes[i], 1, 1,
-                         btype_vlc_syms,     1, 1, INIT_VLC_USE_STATIC);
+                         btype_vlc_syms,     1, 1, INIT_VLC_USE_NEW_STATIC);
+    }
 }
 
 /**
@@ -102,7 +127,7 @@ static void rv40_parse_picture_size(GetBitContext *gb, int *w, int *h)
 
 static int rv40_parse_slice_header(RV34DecContext *r, GetBitContext *gb, SliceInfo *si)
 {
-    int t, mb_bits;
+    int mb_bits;
     int w = r->s.width, h = r->s.height;
     int mb_size;
 
@@ -116,9 +141,11 @@ static int rv40_parse_slice_header(RV34DecContext *r, GetBitContext *gb, SliceIn
         return -1;
     si->vlc_set = get_bits(gb, 2);
     skip_bits1(gb);
-    t = get_bits(gb, 13); /// ???
+    si->pts = get_bits(gb, 13);
     if(!si->type || !get_bits1(gb))
         rv40_parse_picture_size(gb, &w, &h);
+    if(av_image_check_size(w, h, 0, r->s.avctx) < 0)
+        return -1;
     si->width  = w;
     si->height = h;
     mb_size = ((w + 15) >> 4) * ((h + 15) >> 4);
@@ -131,15 +158,15 @@ static int rv40_parse_slice_header(RV34DecContext *r, GetBitContext *gb, SliceIn
 /**
  * Decode 4x4 intra types array.
  */
-static int rv40_decode_intra_types(RV34DecContext *r, GetBitContext *gb, int *dst)
+static int rv40_decode_intra_types(RV34DecContext *r, GetBitContext *gb, int8_t *dst)
 {
     MpegEncContext *s = &r->s;
     int i, j, k, v;
     int A, B, C;
     int pattern;
-    int *ptr;
+    int8_t *ptr;
 
-    for(i = 0; i < 4; i++, dst += s->b4_stride){
+    for(i = 0; i < 4; i++, dst += r->intra_types_stride){
         if(!i && s->first_slice_line){
             pattern = get_vlc2(gb, aic_top_vlc.table, AIC_TOP_BITS, 1);
             dst[0] = (pattern >> 2) & 2;
@@ -156,8 +183,8 @@ static int rv40_decode_intra_types(RV34DecContext *r, GetBitContext *gb, int *ds
              * The second one (used for retrieving only one coefficient) is
              * top + 10 * left.
              */
-            A = ptr[-s->b4_stride + 1]; // it won't be used for the last coefficient in a row
-            B = ptr[-s->b4_stride];
+            A = ptr[-r->intra_types_stride + 1]; // it won't be used for the last coefficient in a row
+            B = ptr[-r->intra_types_stride];
             C = ptr[-1];
             pattern = A + (B << 4) + (C << 8);
             for(k = 0; k < MODE2_PATTERNS_NUM; k++)
@@ -205,18 +232,18 @@ static int rv40_decode_mb_info(RV34DecContext *r)
     int count = 0;
 
     if(!r->s.mb_skip_run)
-        r->s.mb_skip_run = ff_rv34_get_gamma(gb);
+        r->s.mb_skip_run = svq3_get_ue_golomb(gb) + 1;
 
     if(--r->s.mb_skip_run)
          return RV34_MB_SKIP;
 
-    if(r->avail[0])
+    if(r->avail_cache[6-1])
         blocks[r->mb_type[mb_pos - 1]]++;
-    if(r->avail[1]){
+    if(r->avail_cache[6-4]){
         blocks[r->mb_type[mb_pos - s->mb_stride]]++;
-        if(r->avail[2])
+        if(r->avail_cache[6-2])
             blocks[r->mb_type[mb_pos - s->mb_stride + 1]]++;
-        if(r->avail[3])
+        if(r->avail_cache[6-5])
             blocks[r->mb_type[mb_pos - s->mb_stride - 1]]++;
     }
 
@@ -226,7 +253,7 @@ static int rv40_decode_mb_info(RV34DecContext *r)
             prev_type = i;
         }
     }
-    if(s->pict_type == P_TYPE){
+    if(s->pict_type == FF_P_TYPE){
         prev_type = block_num_to_ptype_vlc_num[prev_type];
         q = get_vlc2(gb, ptype_vlc[prev_type].table, PTYPE_VLC_BITS, 1);
         if(q < PBTYPE_ESCAPE)
@@ -244,10 +271,387 @@ static int rv40_decode_mb_info(RV34DecContext *r)
     return 0;
 }
 
+#define CLIP_SYMM(a, b) av_clip(a, -(b), b)
+/**
+ * weaker deblocking very similar to the one described in 4.4.2 of JVT-A003r1
+ */
+static inline void rv40_weak_loop_filter(uint8_t *src, const int step,
+                                         const int filter_p1, const int filter_q1,
+                                         const int alpha, const int beta,
+                                         const int lim_p0q0,
+                                         const int lim_q1, const int lim_p1,
+                                         const int diff_p1p0, const int diff_q1q0,
+                                         const int diff_p1p2, const int diff_q1q2)
+{
+    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
+    int t, u, diff;
+
+    t = src[0*step] - src[-1*step];
+    if(!t)
+        return;
+    u = (alpha * FFABS(t)) >> 7;
+    if(u > 3 - (filter_p1 && filter_q1))
+        return;
+
+    t <<= 2;
+    if(filter_p1 && filter_q1)
+        t += src[-2*step] - src[1*step];
+    diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0);
+    src[-1*step] = cm[src[-1*step] + diff];
+    src[ 0*step] = cm[src[ 0*step] - diff];
+    if(FFABS(diff_p1p2) <= beta && filter_p1){
+        t = (diff_p1p0 + diff_p1p2 - diff) >> 1;
+        src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)];
+    }
+    if(FFABS(diff_q1q2) <= beta && filter_q1){
+        t = (diff_q1q0 + diff_q1q2 + diff) >> 1;
+        src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)];
+    }
+}
+
+static av_always_inline void rv40_adaptive_loop_filter(uint8_t *src, const int step,
+                                             const int stride, const int dmode,
+                                             const int lim_q1, const int lim_p1,
+                                             const int alpha,
+                                             const int beta, const int beta2,
+                                             const int chroma, const int edge)
+{
+    int diff_p1p0[4], diff_q1q0[4], diff_p1p2[4], diff_q1q2[4];
+    int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0;
+    uint8_t *ptr;
+    int flag_strong0 = 1, flag_strong1 = 1;
+    int filter_p1, filter_q1;
+    int i;
+    int lims;
+
+    for(i = 0, ptr = src; i < 4; i++, ptr += stride){
+        diff_p1p0[i] = ptr[-2*step] - ptr[-1*step];
+        diff_q1q0[i] = ptr[ 1*step] - ptr[ 0*step];
+        sum_p1p0 += diff_p1p0[i];
+        sum_q1q0 += diff_q1q0[i];
+    }
+    filter_p1 = FFABS(sum_p1p0) < (beta<<2);
+    filter_q1 = FFABS(sum_q1q0) < (beta<<2);
+    if(!filter_p1 && !filter_q1)
+        return;
+
+    for(i = 0, ptr = src; i < 4; i++, ptr += stride){
+        diff_p1p2[i] = ptr[-2*step] - ptr[-3*step];
+        diff_q1q2[i] = ptr[ 1*step] - ptr[ 2*step];
+        sum_p1p2 += diff_p1p2[i];
+        sum_q1q2 += diff_q1q2[i];
+    }
+
+    if(edge){
+        flag_strong0 = filter_p1 && (FFABS(sum_p1p2) < beta2);
+        flag_strong1 = filter_q1 && (FFABS(sum_q1q2) < beta2);
+    }else{
+        flag_strong0 = flag_strong1 = 0;
+    }
+
+    lims = filter_p1 + filter_q1 + ((lim_q1 + lim_p1) >> 1) + 1;
+    if(flag_strong0 && flag_strong1){ /* strong filtering */
+        for(i = 0; i < 4; i++, src += stride){
+            int sflag, p0, q0, p1, q1;
+            int t = src[0*step] - src[-1*step];
+
+            if(!t) continue;
+            sflag = (alpha * FFABS(t)) >> 7;
+            if(sflag > 1) continue;
+
+            p0 = (25*src[-3*step] + 26*src[-2*step]
+                + 26*src[-1*step]
+                + 26*src[ 0*step] + 25*src[ 1*step] + rv40_dither_l[dmode + i]) >> 7;
+            q0 = (25*src[-2*step] + 26*src[-1*step]
+                + 26*src[ 0*step]
+                + 26*src[ 1*step] + 25*src[ 2*step] + rv40_dither_r[dmode + i]) >> 7;
+            if(sflag){
+                p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims);
+                q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims);
+            }
+            p1 = (25*src[-4*step] + 26*src[-3*step]
+                + 26*src[-2*step]
+                + 26*p0           + 25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;
+            q1 = (25*src[-1*step] + 26*q0
+                + 26*src[ 1*step]
+                + 26*src[ 2*step] + 25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;
+            if(sflag){
+                p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims);
+                q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims);
+            }
+            src[-2*step] = p1;
+            src[-1*step] = p0;
+            src[ 0*step] = q0;
+            src[ 1*step] = q1;
+            if(!chroma){
+                src[-3*step] = (25*src[-1*step] + 26*src[-2*step] + 51*src[-3*step] + 26*src[-4*step] + 64) >> 7;
+                src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] + 51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;
+            }
+        }
+    }else if(filter_p1 && filter_q1){
+        for(i = 0; i < 4; i++, src += stride)
+            rv40_weak_loop_filter(src, step, 1, 1, alpha, beta, lims, lim_q1, lim_p1,
+                                  diff_p1p0[i], diff_q1q0[i], diff_p1p2[i], diff_q1q2[i]);
+    }else{
+        for(i = 0; i < 4; i++, src += stride)
+            rv40_weak_loop_filter(src, step, filter_p1, filter_q1,
+                                  alpha, beta, lims>>1, lim_q1>>1, lim_p1>>1,
+                                  diff_p1p0[i], diff_q1q0[i], diff_p1p2[i], diff_q1q2[i]);
+    }
+}
+
+static void rv40_v_loop_filter(uint8_t *src, int stride, int dmode,
+                               int lim_q1, int lim_p1,
+                               int alpha, int beta, int beta2, int chroma, int edge){
+    rv40_adaptive_loop_filter(src, 1, stride, dmode, lim_q1, lim_p1,
+                              alpha, beta, beta2, chroma, edge);
+}
+static void rv40_h_loop_filter(uint8_t *src, int stride, int dmode,
+                               int lim_q1, int lim_p1,
+                               int alpha, int beta, int beta2, int chroma, int edge){
+    rv40_adaptive_loop_filter(src, stride, 1, dmode, lim_q1, lim_p1,
+                              alpha, beta, beta2, chroma, edge);
+}
+
+enum RV40BlockPos{
+    POS_CUR,
+    POS_TOP,
+    POS_LEFT,
+    POS_BOTTOM,
+};
+
+#define MASK_CUR          0x0001
+#define MASK_RIGHT        0x0008
+#define MASK_BOTTOM       0x0010
+#define MASK_TOP          0x1000
+#define MASK_Y_TOP_ROW    0x000F
+#define MASK_Y_LAST_ROW   0xF000
+#define MASK_Y_LEFT_COL   0x1111
+#define MASK_Y_RIGHT_COL  0x8888
+#define MASK_C_TOP_ROW    0x0003
+#define MASK_C_LAST_ROW   0x000C
+#define MASK_C_LEFT_COL   0x0005
+#define MASK_C_RIGHT_COL  0x000A
+
+static const int neighbour_offs_x[4] = { 0,  0, -1, 0 };
+static const int neighbour_offs_y[4] = { 0, -1,  0, 1 };
+
+/**
+ * RV40 loop filtering function
+ */
+static void rv40_loop_filter(RV34DecContext *r, int row)
+{
+    MpegEncContext *s = &r->s;
+    int mb_pos, mb_x;
+    int i, j, k;
+    uint8_t *Y, *C;
+    int alpha, beta, betaY, betaC;
+    int q;
+    int mbtype[4];   ///< current macroblock and its neighbours types
+    /**
+     * flags indicating that macroblock can be filtered with strong filter
+     * it is set only for intra coded MB and MB with DCs coded separately
+     */
+    int mb_strong[4];
+    int clip[4];     ///< MB filter clipping value calculated from filtering strength
+    /**
+     * coded block patterns for luma part of current macroblock and its neighbours
+     * Format:
+     * LSB corresponds to the top left block,
+     * each nibble represents one row of subblocks.
+     */
+    int cbp[4];
+    /**
+     * coded block patterns for chroma part of current macroblock and its neighbours
+     * Format is the same as for luma with two subblocks in a row.
+     */
+    int uvcbp[4][2];
+    /**
+     * This mask represents the pattern of luma subblocks that should be filtered
+     * in addition to the coded ones because because they lie at the edge of
+     * 8x8 block with different enough motion vectors
+     */
+    int mvmasks[4];
+
+    mb_pos = row * s->mb_stride;
+    for(mb_x = 0; mb_x < s->mb_width; mb_x++, mb_pos++){
+        int mbtype = s->current_picture_ptr->mb_type[mb_pos];
+        if(IS_INTRA(mbtype) || IS_SEPARATE_DC(mbtype))
+            r->cbp_luma  [mb_pos] = r->deblock_coefs[mb_pos] = 0xFFFF;
+        if(IS_INTRA(mbtype))
+            r->cbp_chroma[mb_pos] = 0xFF;
+    }
+    mb_pos = row * s->mb_stride;
+    for(mb_x = 0; mb_x < s->mb_width; mb_x++, mb_pos++){
+        int y_h_deblock, y_v_deblock;
+        int c_v_deblock[2], c_h_deblock[2];
+        int clip_left;
+        int avail[4];
+        int y_to_deblock, c_to_deblock[2];
+
+        q = s->current_picture_ptr->qscale_table[mb_pos];
+        alpha = rv40_alpha_tab[q];
+        beta  = rv40_beta_tab [q];
+        betaY = betaC = beta * 3;
+        if(s->width * s->height <= 176*144)
+            betaY += beta;
+
+        avail[0] = 1;
+        avail[1] = row;
+        avail[2] = mb_x;
+        avail[3] = row < s->mb_height - 1;
+        for(i = 0; i < 4; i++){
+            if(avail[i]){
+                int pos = mb_pos + neighbour_offs_x[i] + neighbour_offs_y[i]*s->mb_stride;
+                mvmasks[i] = r->deblock_coefs[pos];
+                mbtype [i] = s->current_picture_ptr->mb_type[pos];
+                cbp    [i] = r->cbp_luma[pos];
+                uvcbp[i][0] = r->cbp_chroma[pos] & 0xF;
+                uvcbp[i][1] = r->cbp_chroma[pos] >> 4;
+            }else{
+                mvmasks[i] = 0;
+                mbtype [i] = mbtype[0];
+                cbp    [i] = 0;
+                uvcbp[i][0] = uvcbp[i][1] = 0;
+            }
+            mb_strong[i] = IS_INTRA(mbtype[i]) || IS_SEPARATE_DC(mbtype[i]);
+            clip[i] = rv40_filter_clip_tbl[mb_strong[i] + 1][q];
+        }
+        y_to_deblock =  mvmasks[POS_CUR]
+                     | (mvmasks[POS_BOTTOM] << 16);
+        /* This pattern contains bits signalling that horizontal edges of
+         * the current block can be filtered.
+         * That happens when either of adjacent subblocks is coded or lies on
+         * the edge of 8x8 blocks with motion vectors differing by more than
+         * 3/4 pel in any component (any edge orientation for some reason).
+         */
+        y_h_deblock =   y_to_deblock
+                    | ((cbp[POS_CUR]                           <<  4) & ~MASK_Y_TOP_ROW)
+                    | ((cbp[POS_TOP]        & MASK_Y_LAST_ROW) >> 12);
+        /* This pattern contains bits signalling that vertical edges of
+         * the current block can be filtered.
+         * That happens when either of adjacent subblocks is coded or lies on
+         * the edge of 8x8 blocks with motion vectors differing by more than
+         * 3/4 pel in any component (any edge orientation for some reason).
+         */
+        y_v_deblock =   y_to_deblock
+                    | ((cbp[POS_CUR]                      << 1) & ~MASK_Y_LEFT_COL)
+                    | ((cbp[POS_LEFT] & MASK_Y_RIGHT_COL) >> 3);
+        if(!mb_x)
+            y_v_deblock &= ~MASK_Y_LEFT_COL;
+        if(!row)
+            y_h_deblock &= ~MASK_Y_TOP_ROW;
+        if(row == s->mb_height - 1 || (mb_strong[POS_CUR] || mb_strong[POS_BOTTOM]))
+            y_h_deblock &= ~(MASK_Y_TOP_ROW << 16);
+        /* Calculating chroma patterns is similar and easier since there is
+         * no motion vector pattern for them.
+         */
+        for(i = 0; i < 2; i++){
+            c_to_deblock[i] = (uvcbp[POS_BOTTOM][i] << 4) | uvcbp[POS_CUR][i];
+            c_v_deblock[i] =   c_to_deblock[i]
+                           | ((uvcbp[POS_CUR] [i]                       << 1) & ~MASK_C_LEFT_COL)
+                           | ((uvcbp[POS_LEFT][i]   & MASK_C_RIGHT_COL) >> 1);
+            c_h_deblock[i] =   c_to_deblock[i]
+                           | ((uvcbp[POS_TOP][i]    & MASK_C_LAST_ROW)  >> 2)
+                           |  (uvcbp[POS_CUR][i]                        << 2);
+            if(!mb_x)
+                c_v_deblock[i] &= ~MASK_C_LEFT_COL;
+            if(!row)
+                c_h_deblock[i] &= ~MASK_C_TOP_ROW;
+            if(row == s->mb_height - 1 || mb_strong[POS_CUR] || mb_strong[POS_BOTTOM])
+                c_h_deblock[i] &= ~(MASK_C_TOP_ROW << 4);
+        }
+
+        for(j = 0; j < 16; j += 4){
+            Y = s->current_picture_ptr->data[0] + mb_x*16 + (row*16 + j) * s->linesize;
+            for(i = 0; i < 4; i++, Y += 4){
+                int ij = i + j;
+                int clip_cur = y_to_deblock & (MASK_CUR << ij) ? clip[POS_CUR] : 0;
+                int dither = j ? ij : i*4;
+
+                // if bottom block is coded then we can filter its top edge
+                // (or bottom edge of this block, which is the same)
+                if(y_h_deblock & (MASK_BOTTOM << ij)){
+                    rv40_h_loop_filter(Y+4*s->linesize, s->linesize, dither,
+                                       y_to_deblock & (MASK_BOTTOM << ij) ? clip[POS_CUR] : 0,
+                                       clip_cur,
+                                       alpha, beta, betaY, 0, 0);
+                }
+                // filter left block edge in ordinary mode (with low filtering strength)
+                if(y_v_deblock & (MASK_CUR << ij) && (i || !(mb_strong[POS_CUR] || mb_strong[POS_LEFT]))){
+                    if(!i)
+                        clip_left = mvmasks[POS_LEFT] & (MASK_RIGHT << j) ? clip[POS_LEFT] : 0;
+                    else
+                        clip_left = y_to_deblock & (MASK_CUR << (ij-1)) ? clip[POS_CUR] : 0;
+                    rv40_v_loop_filter(Y, s->linesize, dither,
+                                       clip_cur,
+                                       clip_left,
+                                       alpha, beta, betaY, 0, 0);
+                }
+                // filter top edge of the current macroblock when filtering strength is high
+                if(!j && y_h_deblock & (MASK_CUR << i) && (mb_strong[POS_CUR] || mb_strong[POS_TOP])){
+                    rv40_h_loop_filter(Y, s->linesize, dither,
+                                       clip_cur,
+                                       mvmasks[POS_TOP] & (MASK_TOP << i) ? clip[POS_TOP] : 0,
+                                       alpha, beta, betaY, 0, 1);
+                }
+                // filter left block edge in edge mode (with high filtering strength)
+                if(y_v_deblock & (MASK_CUR << ij) && !i && (mb_strong[POS_CUR] || mb_strong[POS_LEFT])){
+                    clip_left = mvmasks[POS_LEFT] & (MASK_RIGHT << j) ? clip[POS_LEFT] : 0;
+                    rv40_v_loop_filter(Y, s->linesize, dither,
+                                       clip_cur,
+                                       clip_left,
+                                       alpha, beta, betaY, 0, 1);
+                }
+            }
+        }
+        for(k = 0; k < 2; k++){
+            for(j = 0; j < 2; j++){
+                C = s->current_picture_ptr->data[k+1] + mb_x*8 + (row*8 + j*4) * s->uvlinesize;
+                for(i = 0; i < 2; i++, C += 4){
+                    int ij = i + j*2;
+                    int clip_cur = c_to_deblock[k] & (MASK_CUR << ij) ? clip[POS_CUR] : 0;
+                    if(c_h_deblock[k] & (MASK_CUR << (ij+2))){
+                        int clip_bot = c_to_deblock[k] & (MASK_CUR << (ij+2)) ? clip[POS_CUR] : 0;
+                        rv40_h_loop_filter(C+4*s->uvlinesize, s->uvlinesize, i*8,
+                                           clip_bot,
+                                           clip_cur,
+                                           alpha, beta, betaC, 1, 0);
+                    }
+                    if((c_v_deblock[k] & (MASK_CUR << ij)) && (i || !(mb_strong[POS_CUR] || mb_strong[POS_LEFT]))){
+                        if(!i)
+                            clip_left = uvcbp[POS_LEFT][k] & (MASK_CUR << (2*j+1)) ? clip[POS_LEFT] : 0;
+                        else
+                            clip_left = c_to_deblock[k]    & (MASK_CUR << (ij-1))  ? clip[POS_CUR]  : 0;
+                        rv40_v_loop_filter(C, s->uvlinesize, j*8,
+                                           clip_cur,
+                                           clip_left,
+                                           alpha, beta, betaC, 1, 0);
+                    }
+                    if(!j && c_h_deblock[k] & (MASK_CUR << ij) && (mb_strong[POS_CUR] || mb_strong[POS_TOP])){
+                        int clip_top = uvcbp[POS_TOP][k] & (MASK_CUR << (ij+2)) ? clip[POS_TOP] : 0;
+                        rv40_h_loop_filter(C, s->uvlinesize, i*8,
+                                           clip_cur,
+                                           clip_top,
+                                           alpha, beta, betaC, 1, 1);
+                    }
+                    if(c_v_deblock[k] & (MASK_CUR << ij) && !i && (mb_strong[POS_CUR] || mb_strong[POS_LEFT])){
+                        clip_left = uvcbp[POS_LEFT][k] & (MASK_CUR << (2*j+1)) ? clip[POS_LEFT] : 0;
+                        rv40_v_loop_filter(C, s->uvlinesize, j*8,
+                                           clip_cur,
+                                           clip_left,
+                                           alpha, beta, betaC, 1, 1);
+                    }
+                }
+            }
+        }
+    }
+}
+
 /**
  * Initialize decoder.
  */
-static int rv40_decode_init(AVCodecContext *avctx)
+static av_cold int rv40_decode_init(AVCodecContext *avctx)
 {
     RV34DecContext *r = avctx->priv_data;
 
@@ -258,18 +662,23 @@ static int rv40_decode_init(AVCodecContext *avctx)
     r->parse_slice_header = rv40_parse_slice_header;
     r->decode_intra_types = rv40_decode_intra_types;
     r->decode_mb_info     = rv40_decode_mb_info;
+    r->loop_filter        = rv40_loop_filter;
     r->luma_dc_quant_i = rv40_luma_dc_quant[0];
     r->luma_dc_quant_p = rv40_luma_dc_quant[1];
     return 0;
 }
 
-AVCodec rv40_decoder = {
+AVCodec ff_rv40_decoder = {
     "rv40",
-    CODEC_TYPE_VIDEO,
+    AVMEDIA_TYPE_VIDEO,
     CODEC_ID_RV40,
     sizeof(RV34DecContext),
     rv40_decode_init,
     NULL,
     ff_rv34_decode_end,
     ff_rv34_decode_frame,
+    CODEC_CAP_DR1 | CODEC_CAP_DELAY,
+    .flush = ff_mpeg_flush,
+    .long_name = NULL_IF_CONFIG_SMALL("RealVideo 4.0"),
+    .pix_fmts= ff_pixfmt_list_420,
 };