yading@10: /*
yading@10:  * Bink video decoder
yading@10:  * Copyright (c) 2009 Konstantin Shishkov
yading@10:  * Copyright (C) 2011 Peter Ross <pross@xvid.org>
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: #include "libavutil/imgutils.h"
yading@10: #include "libavutil/internal.h"
yading@10: #include "avcodec.h"
yading@10: #include "dsputil.h"
yading@10: #include "binkdata.h"
yading@10: #include "binkdsp.h"
yading@10: #include "hpeldsp.h"
yading@10: #include "internal.h"
yading@10: #include "mathops.h"
yading@10: 
yading@10: #define BITSTREAM_READER_LE
yading@10: #include "get_bits.h"
yading@10: 
yading@10: #define BINK_FLAG_ALPHA 0x00100000
yading@10: #define BINK_FLAG_GRAY  0x00020000
yading@10: 
yading@10: static VLC bink_trees[16];
yading@10: 
yading@10: /**
yading@10:  * IDs for different data types used in old version of Bink video codec
yading@10:  */
yading@10: enum OldSources {
yading@10:     BINKB_SRC_BLOCK_TYPES = 0, ///< 8x8 block types
yading@10:     BINKB_SRC_COLORS,          ///< pixel values used for different block types
yading@10:     BINKB_SRC_PATTERN,         ///< 8-bit values for 2-colour pattern fill
yading@10:     BINKB_SRC_X_OFF,           ///< X components of motion value
yading@10:     BINKB_SRC_Y_OFF,           ///< Y components of motion value
yading@10:     BINKB_SRC_INTRA_DC,        ///< DC values for intrablocks with DCT
yading@10:     BINKB_SRC_INTER_DC,        ///< DC values for interblocks with DCT
yading@10:     BINKB_SRC_INTRA_Q,         ///< quantizer values for intrablocks with DCT
yading@10:     BINKB_SRC_INTER_Q,         ///< quantizer values for interblocks with DCT
yading@10:     BINKB_SRC_INTER_COEFS,     ///< number of coefficients for residue blocks
yading@10: 
yading@10:     BINKB_NB_SRC
yading@10: };
yading@10: 
yading@10: static const int binkb_bundle_sizes[BINKB_NB_SRC] = {
yading@10:     4, 8, 8, 5, 5, 11, 11, 4, 4, 7
yading@10: };
yading@10: 
yading@10: static const int binkb_bundle_signed[BINKB_NB_SRC] = {
yading@10:     0, 0, 0, 1, 1, 0, 1, 0, 0, 0
yading@10: };
yading@10: 
yading@10: static int32_t binkb_intra_quant[16][64];
yading@10: static int32_t binkb_inter_quant[16][64];
yading@10: 
yading@10: /**
yading@10:  * IDs for different data types used in Bink video codec
yading@10:  */
yading@10: enum Sources {
yading@10:     BINK_SRC_BLOCK_TYPES = 0, ///< 8x8 block types
yading@10:     BINK_SRC_SUB_BLOCK_TYPES, ///< 16x16 block types (a subset of 8x8 block types)
yading@10:     BINK_SRC_COLORS,          ///< pixel values used for different block types
yading@10:     BINK_SRC_PATTERN,         ///< 8-bit values for 2-colour pattern fill
yading@10:     BINK_SRC_X_OFF,           ///< X components of motion value
yading@10:     BINK_SRC_Y_OFF,           ///< Y components of motion value
yading@10:     BINK_SRC_INTRA_DC,        ///< DC values for intrablocks with DCT
yading@10:     BINK_SRC_INTER_DC,        ///< DC values for interblocks with DCT
yading@10:     BINK_SRC_RUN,             ///< run lengths for special fill block
yading@10: 
yading@10:     BINK_NB_SRC
yading@10: };
yading@10: 
yading@10: /**
yading@10:  * data needed to decode 4-bit Huffman-coded value
yading@10:  */
yading@10: typedef struct Tree {
yading@10:     int     vlc_num;  ///< tree number (in bink_trees[])
yading@10:     uint8_t syms[16]; ///< leaf value to symbol mapping
yading@10: } Tree;
yading@10: 
yading@10: #define GET_HUFF(gb, tree)  (tree).syms[get_vlc2(gb, bink_trees[(tree).vlc_num].table,\
yading@10:                                                  bink_trees[(tree).vlc_num].bits, 1)]
yading@10: 
yading@10: /**
yading@10:  * data structure used for decoding single Bink data type
yading@10:  */
yading@10: typedef struct Bundle {
yading@10:     int     len;       ///< length of number of entries to decode (in bits)
yading@10:     Tree    tree;      ///< Huffman tree-related data
yading@10:     uint8_t *data;     ///< buffer for decoded symbols
yading@10:     uint8_t *data_end; ///< buffer end
yading@10:     uint8_t *cur_dec;  ///< pointer to the not yet decoded part of the buffer
yading@10:     uint8_t *cur_ptr;  ///< pointer to the data that is not read from buffer yet
yading@10: } Bundle;
yading@10: 
yading@10: /*
yading@10:  * Decoder context
yading@10:  */
yading@10: typedef struct BinkContext {
yading@10:     AVCodecContext *avctx;
yading@10:     DSPContext     dsp;
yading@10:     HpelDSPContext hdsp;
yading@10:     BinkDSPContext bdsp;
yading@10:     AVFrame        *last;
yading@10:     int            version;              ///< internal Bink file version
yading@10:     int            has_alpha;
yading@10:     int            swap_planes;
yading@10: 
yading@10:     Bundle         bundle[BINKB_NB_SRC]; ///< bundles for decoding all data types
yading@10:     Tree           col_high[16];         ///< trees for decoding high nibble in "colours" data type
yading@10:     int            col_lastval;          ///< value of last decoded high nibble in "colours" data type
yading@10: } BinkContext;
yading@10: 
yading@10: /**
yading@10:  * Bink video block types
yading@10:  */
yading@10: enum BlockTypes {
yading@10:     SKIP_BLOCK = 0, ///< skipped block
yading@10:     SCALED_BLOCK,   ///< block has size 16x16
yading@10:     MOTION_BLOCK,   ///< block is copied from previous frame with some offset
yading@10:     RUN_BLOCK,      ///< block is composed from runs of colours with custom scan order
yading@10:     RESIDUE_BLOCK,  ///< motion block with some difference added
yading@10:     INTRA_BLOCK,    ///< intra DCT block
yading@10:     FILL_BLOCK,     ///< block is filled with single colour
yading@10:     INTER_BLOCK,    ///< motion block with DCT applied to the difference
yading@10:     PATTERN_BLOCK,  ///< block is filled with two colours following custom pattern
yading@10:     RAW_BLOCK,      ///< uncoded 8x8 block
yading@10: };
yading@10: 
yading@10: /**
yading@10:  * Initialize length length in all bundles.
yading@10:  *
yading@10:  * @param c     decoder context
yading@10:  * @param width plane width
yading@10:  * @param bw    plane width in 8x8 blocks
yading@10:  */
yading@10: static void init_lengths(BinkContext *c, int width, int bw)
yading@10: {
yading@10:     width = FFALIGN(width, 8);
yading@10: 
yading@10:     c->bundle[BINK_SRC_BLOCK_TYPES].len = av_log2((width >> 3) + 511) + 1;
yading@10: 
yading@10:     c->bundle[BINK_SRC_SUB_BLOCK_TYPES].len = av_log2((width >> 4) + 511) + 1;
yading@10: 
yading@10:     c->bundle[BINK_SRC_COLORS].len = av_log2(bw*64 + 511) + 1;
yading@10: 
yading@10:     c->bundle[BINK_SRC_INTRA_DC].len =
yading@10:     c->bundle[BINK_SRC_INTER_DC].len =
yading@10:     c->bundle[BINK_SRC_X_OFF].len =
yading@10:     c->bundle[BINK_SRC_Y_OFF].len = av_log2((width >> 3) + 511) + 1;
yading@10: 
yading@10:     c->bundle[BINK_SRC_PATTERN].len = av_log2((bw << 3) + 511) + 1;
yading@10: 
yading@10:     c->bundle[BINK_SRC_RUN].len = av_log2(bw*48 + 511) + 1;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Allocate memory for bundles.
yading@10:  *
yading@10:  * @param c decoder context
yading@10:  */
yading@10: static av_cold int init_bundles(BinkContext *c)
yading@10: {
yading@10:     int bw, bh, blocks;
yading@10:     int i;
yading@10: 
yading@10:     bw = (c->avctx->width  + 7) >> 3;
yading@10:     bh = (c->avctx->height + 7) >> 3;
yading@10:     blocks = bw * bh;
yading@10: 
yading@10:     for (i = 0; i < BINKB_NB_SRC; i++) {
yading@10:         c->bundle[i].data = av_malloc(blocks * 64);
yading@10:         if (!c->bundle[i].data)
yading@10:             return AVERROR(ENOMEM);
yading@10:         c->bundle[i].data_end = c->bundle[i].data + blocks * 64;
yading@10:     }
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Free memory used by bundles.
yading@10:  *
yading@10:  * @param c decoder context
yading@10:  */
yading@10: static av_cold void free_bundles(BinkContext *c)
yading@10: {
yading@10:     int i;
yading@10:     for (i = 0; i < BINKB_NB_SRC; i++)
yading@10:         av_freep(&c->bundle[i].data);
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Merge two consequent lists of equal size depending on bits read.
yading@10:  *
yading@10:  * @param gb   context for reading bits
yading@10:  * @param dst  buffer where merged list will be written to
yading@10:  * @param src  pointer to the head of the first list (the second lists starts at src+size)
yading@10:  * @param size input lists size
yading@10:  */
yading@10: static void merge(GetBitContext *gb, uint8_t *dst, uint8_t *src, int size)
yading@10: {
yading@10:     uint8_t *src2 = src + size;
yading@10:     int size2 = size;
yading@10: 
yading@10:     do {
yading@10:         if (!get_bits1(gb)) {
yading@10:             *dst++ = *src++;
yading@10:             size--;
yading@10:         } else {
yading@10:             *dst++ = *src2++;
yading@10:             size2--;
yading@10:         }
yading@10:     } while (size && size2);
yading@10: 
yading@10:     while (size--)
yading@10:         *dst++ = *src++;
yading@10:     while (size2--)
yading@10:         *dst++ = *src2++;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Read information about Huffman tree used to decode data.
yading@10:  *
yading@10:  * @param gb   context for reading bits
yading@10:  * @param tree pointer for storing tree data
yading@10:  */
yading@10: static void read_tree(GetBitContext *gb, Tree *tree)
yading@10: {
yading@10:     uint8_t tmp1[16] = { 0 }, tmp2[16], *in = tmp1, *out = tmp2;
yading@10:     int i, t, len;
yading@10: 
yading@10:     tree->vlc_num = get_bits(gb, 4);
yading@10:     if (!tree->vlc_num) {
yading@10:         for (i = 0; i < 16; i++)
yading@10:             tree->syms[i] = i;
yading@10:         return;
yading@10:     }
yading@10:     if (get_bits1(gb)) {
yading@10:         len = get_bits(gb, 3);
yading@10:         for (i = 0; i <= len; i++) {
yading@10:             tree->syms[i] = get_bits(gb, 4);
yading@10:             tmp1[tree->syms[i]] = 1;
yading@10:         }
yading@10:         for (i = 0; i < 16 && len < 16 - 1; i++)
yading@10:             if (!tmp1[i])
yading@10:                 tree->syms[++len] = i;
yading@10:     } else {
yading@10:         len = get_bits(gb, 2);
yading@10:         for (i = 0; i < 16; i++)
yading@10:             in[i] = i;
yading@10:         for (i = 0; i <= len; i++) {
yading@10:             int size = 1 << i;
yading@10:             for (t = 0; t < 16; t += size << 1)
yading@10:                 merge(gb, out + t, in + t, size);
yading@10:             FFSWAP(uint8_t*, in, out);
yading@10:         }
yading@10:         memcpy(tree->syms, in, 16);
yading@10:     }
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Prepare bundle for decoding data.
yading@10:  *
yading@10:  * @param gb          context for reading bits
yading@10:  * @param c           decoder context
yading@10:  * @param bundle_num  number of the bundle to initialize
yading@10:  */
yading@10: static void read_bundle(GetBitContext *gb, BinkContext *c, int bundle_num)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     if (bundle_num == BINK_SRC_COLORS) {
yading@10:         for (i = 0; i < 16; i++)
yading@10:             read_tree(gb, &c->col_high[i]);
yading@10:         c->col_lastval = 0;
yading@10:     }
yading@10:     if (bundle_num != BINK_SRC_INTRA_DC && bundle_num != BINK_SRC_INTER_DC)
yading@10:         read_tree(gb, &c->bundle[bundle_num].tree);
yading@10:     c->bundle[bundle_num].cur_dec =
yading@10:     c->bundle[bundle_num].cur_ptr = c->bundle[bundle_num].data;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * common check before starting decoding bundle data
yading@10:  *
yading@10:  * @param gb context for reading bits
yading@10:  * @param b  bundle
yading@10:  * @param t  variable where number of elements to decode will be stored
yading@10:  */
yading@10: #define CHECK_READ_VAL(gb, b, t) \
yading@10:     if (!b->cur_dec || (b->cur_dec > b->cur_ptr)) \
yading@10:         return 0; \
yading@10:     t = get_bits(gb, b->len); \
yading@10:     if (!t) { \
yading@10:         b->cur_dec = NULL; \
yading@10:         return 0; \
yading@10:     } \
yading@10: 
yading@10: static int read_runs(AVCodecContext *avctx, GetBitContext *gb, Bundle *b)
yading@10: {
yading@10:     int t, v;
yading@10:     const uint8_t *dec_end;
yading@10: 
yading@10:     CHECK_READ_VAL(gb, b, t);
yading@10:     dec_end = b->cur_dec + t;
yading@10:     if (dec_end > b->data_end) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Run value went out of bounds\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     if (get_bits1(gb)) {
yading@10:         v = get_bits(gb, 4);
yading@10:         memset(b->cur_dec, v, t);
yading@10:         b->cur_dec += t;
yading@10:     } else {
yading@10:         while (b->cur_dec < dec_end)
yading@10:             *b->cur_dec++ = GET_HUFF(gb, b->tree);
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int read_motion_values(AVCodecContext *avctx, GetBitContext *gb, Bundle *b)
yading@10: {
yading@10:     int t, sign, v;
yading@10:     const uint8_t *dec_end;
yading@10: 
yading@10:     CHECK_READ_VAL(gb, b, t);
yading@10:     dec_end = b->cur_dec + t;
yading@10:     if (dec_end > b->data_end) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Too many motion values\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     if (get_bits1(gb)) {
yading@10:         v = get_bits(gb, 4);
yading@10:         if (v) {
yading@10:             sign = -get_bits1(gb);
yading@10:             v = (v ^ sign) - sign;
yading@10:         }
yading@10:         memset(b->cur_dec, v, t);
yading@10:         b->cur_dec += t;
yading@10:     } else {
yading@10:         while (b->cur_dec < dec_end) {
yading@10:             v = GET_HUFF(gb, b->tree);
yading@10:             if (v) {
yading@10:                 sign = -get_bits1(gb);
yading@10:                 v = (v ^ sign) - sign;
yading@10:             }
yading@10:             *b->cur_dec++ = v;
yading@10:         }
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static const uint8_t bink_rlelens[4] = { 4, 8, 12, 32 };
yading@10: 
yading@10: static int read_block_types(AVCodecContext *avctx, GetBitContext *gb, Bundle *b)
yading@10: {
yading@10:     int t, v;
yading@10:     int last = 0;
yading@10:     const uint8_t *dec_end;
yading@10: 
yading@10:     CHECK_READ_VAL(gb, b, t);
yading@10:     dec_end = b->cur_dec + t;
yading@10:     if (dec_end > b->data_end) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Too many block type values\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     if (get_bits1(gb)) {
yading@10:         v = get_bits(gb, 4);
yading@10:         memset(b->cur_dec, v, t);
yading@10:         b->cur_dec += t;
yading@10:     } else {
yading@10:         while (b->cur_dec < dec_end) {
yading@10:             v = GET_HUFF(gb, b->tree);
yading@10:             if (v < 12) {
yading@10:                 last = v;
yading@10:                 *b->cur_dec++ = v;
yading@10:             } else {
yading@10:                 int run = bink_rlelens[v - 12];
yading@10: 
yading@10:                 if (dec_end - b->cur_dec < run)
yading@10:                     return AVERROR_INVALIDDATA;
yading@10:                 memset(b->cur_dec, last, run);
yading@10:                 b->cur_dec += run;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int read_patterns(AVCodecContext *avctx, GetBitContext *gb, Bundle *b)
yading@10: {
yading@10:     int t, v;
yading@10:     const uint8_t *dec_end;
yading@10: 
yading@10:     CHECK_READ_VAL(gb, b, t);
yading@10:     dec_end = b->cur_dec + t;
yading@10:     if (dec_end > b->data_end) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Too many pattern values\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     while (b->cur_dec < dec_end) {
yading@10:         v  = GET_HUFF(gb, b->tree);
yading@10:         v |= GET_HUFF(gb, b->tree) << 4;
yading@10:         *b->cur_dec++ = v;
yading@10:     }
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int read_colors(GetBitContext *gb, Bundle *b, BinkContext *c)
yading@10: {
yading@10:     int t, sign, v;
yading@10:     const uint8_t *dec_end;
yading@10: 
yading@10:     CHECK_READ_VAL(gb, b, t);
yading@10:     dec_end = b->cur_dec + t;
yading@10:     if (dec_end > b->data_end) {
yading@10:         av_log(c->avctx, AV_LOG_ERROR, "Too many color values\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     if (get_bits1(gb)) {
yading@10:         c->col_lastval = GET_HUFF(gb, c->col_high[c->col_lastval]);
yading@10:         v = GET_HUFF(gb, b->tree);
yading@10:         v = (c->col_lastval << 4) | v;
yading@10:         if (c->version < 'i') {
yading@10:             sign = ((int8_t) v) >> 7;
yading@10:             v = ((v & 0x7F) ^ sign) - sign;
yading@10:             v += 0x80;
yading@10:         }
yading@10:         memset(b->cur_dec, v, t);
yading@10:         b->cur_dec += t;
yading@10:     } else {
yading@10:         while (b->cur_dec < dec_end) {
yading@10:             c->col_lastval = GET_HUFF(gb, c->col_high[c->col_lastval]);
yading@10:             v = GET_HUFF(gb, b->tree);
yading@10:             v = (c->col_lastval << 4) | v;
yading@10:             if (c->version < 'i') {
yading@10:                 sign = ((int8_t) v) >> 7;
yading@10:                 v = ((v & 0x7F) ^ sign) - sign;
yading@10:                 v += 0x80;
yading@10:             }
yading@10:             *b->cur_dec++ = v;
yading@10:         }
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: /** number of bits used to store first DC value in bundle */
yading@10: #define DC_START_BITS 11
yading@10: 
yading@10: static int read_dcs(AVCodecContext *avctx, GetBitContext *gb, Bundle *b,
yading@10:                     int start_bits, int has_sign)
yading@10: {
yading@10:     int i, j, len, len2, bsize, sign, v, v2;
yading@10:     int16_t *dst     = (int16_t*)b->cur_dec;
yading@10:     int16_t *dst_end = (int16_t*)b->data_end;
yading@10: 
yading@10:     CHECK_READ_VAL(gb, b, len);
yading@10:     v = get_bits(gb, start_bits - has_sign);
yading@10:     if (v && has_sign) {
yading@10:         sign = -get_bits1(gb);
yading@10:         v = (v ^ sign) - sign;
yading@10:     }
yading@10:     if (dst_end - dst < 1)
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     *dst++ = v;
yading@10:     len--;
yading@10:     for (i = 0; i < len; i += 8) {
yading@10:         len2 = FFMIN(len - i, 8);
yading@10:         if (dst_end - dst < len2)
yading@10:             return AVERROR_INVALIDDATA;
yading@10:         bsize = get_bits(gb, 4);
yading@10:         if (bsize) {
yading@10:             for (j = 0; j < len2; j++) {
yading@10:                 v2 = get_bits(gb, bsize);
yading@10:                 if (v2) {
yading@10:                     sign = -get_bits1(gb);
yading@10:                     v2 = (v2 ^ sign) - sign;
yading@10:                 }
yading@10:                 v += v2;
yading@10:                 *dst++ = v;
yading@10:                 if (v < -32768 || v > 32767) {
yading@10:                     av_log(avctx, AV_LOG_ERROR, "DC value went out of bounds: %d\n", v);
yading@10:                     return AVERROR_INVALIDDATA;
yading@10:                 }
yading@10:             }
yading@10:         } else {
yading@10:             for (j = 0; j < len2; j++)
yading@10:                 *dst++ = v;
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     b->cur_dec = (uint8_t*)dst;
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Retrieve next value from bundle.
yading@10:  *
yading@10:  * @param c      decoder context
yading@10:  * @param bundle bundle number
yading@10:  */
yading@10: static inline int get_value(BinkContext *c, int bundle)
yading@10: {
yading@10:     int ret;
yading@10: 
yading@10:     if (bundle < BINK_SRC_X_OFF || bundle == BINK_SRC_RUN)
yading@10:         return *c->bundle[bundle].cur_ptr++;
yading@10:     if (bundle == BINK_SRC_X_OFF || bundle == BINK_SRC_Y_OFF)
yading@10:         return (int8_t)*c->bundle[bundle].cur_ptr++;
yading@10:     ret = *(int16_t*)c->bundle[bundle].cur_ptr;
yading@10:     c->bundle[bundle].cur_ptr += 2;
yading@10:     return ret;
yading@10: }
yading@10: 
yading@10: static void binkb_init_bundle(BinkContext *c, int bundle_num)
yading@10: {
yading@10:     c->bundle[bundle_num].cur_dec =
yading@10:     c->bundle[bundle_num].cur_ptr = c->bundle[bundle_num].data;
yading@10:     c->bundle[bundle_num].len = 13;
yading@10: }
yading@10: 
yading@10: static void binkb_init_bundles(BinkContext *c)
yading@10: {
yading@10:     int i;
yading@10:     for (i = 0; i < BINKB_NB_SRC; i++)
yading@10:         binkb_init_bundle(c, i);
yading@10: }
yading@10: 
yading@10: static int binkb_read_bundle(BinkContext *c, GetBitContext *gb, int bundle_num)
yading@10: {
yading@10:     const int bits = binkb_bundle_sizes[bundle_num];
yading@10:     const int mask = 1 << (bits - 1);
yading@10:     const int issigned = binkb_bundle_signed[bundle_num];
yading@10:     Bundle *b = &c->bundle[bundle_num];
yading@10:     int i, len;
yading@10: 
yading@10:     CHECK_READ_VAL(gb, b, len);
yading@10:     if (b->data_end - b->cur_dec < len * (1 + (bits > 8)))
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     if (bits <= 8) {
yading@10:         if (!issigned) {
yading@10:             for (i = 0; i < len; i++)
yading@10:                 *b->cur_dec++ = get_bits(gb, bits);
yading@10:         } else {
yading@10:             for (i = 0; i < len; i++)
yading@10:                 *b->cur_dec++ = get_bits(gb, bits) - mask;
yading@10:         }
yading@10:     } else {
yading@10:         int16_t *dst = (int16_t*)b->cur_dec;
yading@10: 
yading@10:         if (!issigned) {
yading@10:             for (i = 0; i < len; i++)
yading@10:                 *dst++ = get_bits(gb, bits);
yading@10:         } else {
yading@10:             for (i = 0; i < len; i++)
yading@10:                 *dst++ = get_bits(gb, bits) - mask;
yading@10:         }
yading@10:         b->cur_dec = (uint8_t*)dst;
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static inline int binkb_get_value(BinkContext *c, int bundle_num)
yading@10: {
yading@10:     int16_t ret;
yading@10:     const int bits = binkb_bundle_sizes[bundle_num];
yading@10: 
yading@10:     if (bits <= 8) {
yading@10:         int val = *c->bundle[bundle_num].cur_ptr++;
yading@10:         return binkb_bundle_signed[bundle_num] ? (int8_t)val : val;
yading@10:     }
yading@10:     ret = *(int16_t*)c->bundle[bundle_num].cur_ptr;
yading@10:     c->bundle[bundle_num].cur_ptr += 2;
yading@10:     return ret;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Read 8x8 block of DCT coefficients.
yading@10:  *
yading@10:  * @param gb       context for reading bits
yading@10:  * @param block    place for storing coefficients
yading@10:  * @param scan     scan order table
yading@10:  * @param quant_matrices quantization matrices
yading@10:  * @return 0 for success, negative value in other cases
yading@10:  */
yading@10: static int read_dct_coeffs(GetBitContext *gb, int32_t block[64], const uint8_t *scan,
yading@10:                            const int32_t quant_matrices[16][64], int q)
yading@10: {
yading@10:     int coef_list[128];
yading@10:     int mode_list[128];
yading@10:     int i, t, bits, ccoef, mode, sign;
yading@10:     int list_start = 64, list_end = 64, list_pos;
yading@10:     int coef_count = 0;
yading@10:     int coef_idx[64];
yading@10:     int quant_idx;
yading@10:     const int32_t *quant;
yading@10: 
yading@10:     coef_list[list_end] = 4;  mode_list[list_end++] = 0;
yading@10:     coef_list[list_end] = 24; mode_list[list_end++] = 0;
yading@10:     coef_list[list_end] = 44; mode_list[list_end++] = 0;
yading@10:     coef_list[list_end] = 1;  mode_list[list_end++] = 3;
yading@10:     coef_list[list_end] = 2;  mode_list[list_end++] = 3;
yading@10:     coef_list[list_end] = 3;  mode_list[list_end++] = 3;
yading@10: 
yading@10:     for (bits = get_bits(gb, 4) - 1; bits >= 0; bits--) {
yading@10:         list_pos = list_start;
yading@10:         while (list_pos < list_end) {
yading@10:             if (!(mode_list[list_pos] | coef_list[list_pos]) || !get_bits1(gb)) {
yading@10:                 list_pos++;
yading@10:                 continue;
yading@10:             }
yading@10:             ccoef = coef_list[list_pos];
yading@10:             mode  = mode_list[list_pos];
yading@10:             switch (mode) {
yading@10:             case 0:
yading@10:                 coef_list[list_pos] = ccoef + 4;
yading@10:                 mode_list[list_pos] = 1;
yading@10:             case 2:
yading@10:                 if (mode == 2) {
yading@10:                     coef_list[list_pos]   = 0;
yading@10:                     mode_list[list_pos++] = 0;
yading@10:                 }
yading@10:                 for (i = 0; i < 4; i++, ccoef++) {
yading@10:                     if (get_bits1(gb)) {
yading@10:                         coef_list[--list_start] = ccoef;
yading@10:                         mode_list[  list_start] = 3;
yading@10:                     } else {
yading@10:                         if (!bits) {
yading@10:                             t = 1 - (get_bits1(gb) << 1);
yading@10:                         } else {
yading@10:                             t = get_bits(gb, bits) | 1 << bits;
yading@10:                             sign = -get_bits1(gb);
yading@10:                             t = (t ^ sign) - sign;
yading@10:                         }
yading@10:                         block[scan[ccoef]] = t;
yading@10:                         coef_idx[coef_count++] = ccoef;
yading@10:                     }
yading@10:                 }
yading@10:                 break;
yading@10:             case 1:
yading@10:                 mode_list[list_pos] = 2;
yading@10:                 for (i = 0; i < 3; i++) {
yading@10:                     ccoef += 4;
yading@10:                     coef_list[list_end]   = ccoef;
yading@10:                     mode_list[list_end++] = 2;
yading@10:                 }
yading@10:                 break;
yading@10:             case 3:
yading@10:                 if (!bits) {
yading@10:                     t = 1 - (get_bits1(gb) << 1);
yading@10:                 } else {
yading@10:                     t = get_bits(gb, bits) | 1 << bits;
yading@10:                     sign = -get_bits1(gb);
yading@10:                     t = (t ^ sign) - sign;
yading@10:                 }
yading@10:                 block[scan[ccoef]] = t;
yading@10:                 coef_idx[coef_count++] = ccoef;
yading@10:                 coef_list[list_pos]   = 0;
yading@10:                 mode_list[list_pos++] = 0;
yading@10:                 break;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     if (q == -1) {
yading@10:         quant_idx = get_bits(gb, 4);
yading@10:     } else {
yading@10:         quant_idx = q;
yading@10:         if (quant_idx > 15U) {
yading@10:             av_log(NULL, AV_LOG_ERROR, "quant_index %d out of range\n", quant_idx);
yading@10:             return AVERROR_INVALIDDATA;
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     quant = quant_matrices[quant_idx];
yading@10: 
yading@10:     block[0] = (block[0] * quant[0]) >> 11;
yading@10:     for (i = 0; i < coef_count; i++) {
yading@10:         int idx = coef_idx[i];
yading@10:         block[scan[idx]] = (block[scan[idx]] * quant[idx]) >> 11;
yading@10:     }
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Read 8x8 block with residue after motion compensation.
yading@10:  *
yading@10:  * @param gb          context for reading bits
yading@10:  * @param block       place to store read data
yading@10:  * @param masks_count number of masks to decode
yading@10:  * @return 0 on success, negative value in other cases
yading@10:  */
yading@10: static int read_residue(GetBitContext *gb, int16_t block[64], int masks_count)
yading@10: {
yading@10:     int coef_list[128];
yading@10:     int mode_list[128];
yading@10:     int i, sign, mask, ccoef, mode;
yading@10:     int list_start = 64, list_end = 64, list_pos;
yading@10:     int nz_coeff[64];
yading@10:     int nz_coeff_count = 0;
yading@10: 
yading@10:     coef_list[list_end] =  4; mode_list[list_end++] = 0;
yading@10:     coef_list[list_end] = 24; mode_list[list_end++] = 0;
yading@10:     coef_list[list_end] = 44; mode_list[list_end++] = 0;
yading@10:     coef_list[list_end] =  0; mode_list[list_end++] = 2;
yading@10: 
yading@10:     for (mask = 1 << get_bits(gb, 3); mask; mask >>= 1) {
yading@10:         for (i = 0; i < nz_coeff_count; i++) {
yading@10:             if (!get_bits1(gb))
yading@10:                 continue;
yading@10:             if (block[nz_coeff[i]] < 0)
yading@10:                 block[nz_coeff[i]] -= mask;
yading@10:             else
yading@10:                 block[nz_coeff[i]] += mask;
yading@10:             masks_count--;
yading@10:             if (masks_count < 0)
yading@10:                 return 0;
yading@10:         }
yading@10:         list_pos = list_start;
yading@10:         while (list_pos < list_end) {
yading@10:             if (!(coef_list[list_pos] | mode_list[list_pos]) || !get_bits1(gb)) {
yading@10:                 list_pos++;
yading@10:                 continue;
yading@10:             }
yading@10:             ccoef = coef_list[list_pos];
yading@10:             mode  = mode_list[list_pos];
yading@10:             switch (mode) {
yading@10:             case 0:
yading@10:                 coef_list[list_pos] = ccoef + 4;
yading@10:                 mode_list[list_pos] = 1;
yading@10:             case 2:
yading@10:                 if (mode == 2) {
yading@10:                     coef_list[list_pos]   = 0;
yading@10:                     mode_list[list_pos++] = 0;
yading@10:                 }
yading@10:                 for (i = 0; i < 4; i++, ccoef++) {
yading@10:                     if (get_bits1(gb)) {
yading@10:                         coef_list[--list_start] = ccoef;
yading@10:                         mode_list[  list_start] = 3;
yading@10:                     } else {
yading@10:                         nz_coeff[nz_coeff_count++] = bink_scan[ccoef];
yading@10:                         sign = -get_bits1(gb);
yading@10:                         block[bink_scan[ccoef]] = (mask ^ sign) - sign;
yading@10:                         masks_count--;
yading@10:                         if (masks_count < 0)
yading@10:                             return 0;
yading@10:                     }
yading@10:                 }
yading@10:                 break;
yading@10:             case 1:
yading@10:                 mode_list[list_pos] = 2;
yading@10:                 for (i = 0; i < 3; i++) {
yading@10:                     ccoef += 4;
yading@10:                     coef_list[list_end]   = ccoef;
yading@10:                     mode_list[list_end++] = 2;
yading@10:                 }
yading@10:                 break;
yading@10:             case 3:
yading@10:                 nz_coeff[nz_coeff_count++] = bink_scan[ccoef];
yading@10:                 sign = -get_bits1(gb);
yading@10:                 block[bink_scan[ccoef]] = (mask ^ sign) - sign;
yading@10:                 coef_list[list_pos]   = 0;
yading@10:                 mode_list[list_pos++] = 0;
yading@10:                 masks_count--;
yading@10:                 if (masks_count < 0)
yading@10:                     return 0;
yading@10:                 break;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Copy 8x8 block from source to destination, where src and dst may be overlapped
yading@10:  */
yading@10: static inline void put_pixels8x8_overlapped(uint8_t *dst, uint8_t *src, int stride)
yading@10: {
yading@10:     uint8_t tmp[64];
yading@10:     int i;
yading@10:     for (i = 0; i < 8; i++)
yading@10:         memcpy(tmp + i*8, src + i*stride, 8);
yading@10:     for (i = 0; i < 8; i++)
yading@10:         memcpy(dst + i*stride, tmp + i*8, 8);
yading@10: }
yading@10: 
yading@10: static int binkb_decode_plane(BinkContext *c, AVFrame *frame, GetBitContext *gb,
yading@10:                               int plane_idx, int is_key, int is_chroma)
yading@10: {
yading@10:     int blk, ret;
yading@10:     int i, j, bx, by;
yading@10:     uint8_t *dst, *ref, *ref_start, *ref_end;
yading@10:     int v, col[2];
yading@10:     const uint8_t *scan;
yading@10:     int xoff, yoff;
yading@10:     LOCAL_ALIGNED_16(int16_t, block, [64]);
yading@10:     LOCAL_ALIGNED_16(int32_t, dctblock, [64]);
yading@10:     int coordmap[64];
yading@10:     int ybias = is_key ? -15 : 0;
yading@10:     int qp;
yading@10: 
yading@10:     const int stride = frame->linesize[plane_idx];
yading@10:     int bw = is_chroma ? (c->avctx->width  + 15) >> 4 : (c->avctx->width  + 7) >> 3;
yading@10:     int bh = is_chroma ? (c->avctx->height + 15) >> 4 : (c->avctx->height + 7) >> 3;
yading@10: 
yading@10:     binkb_init_bundles(c);
yading@10:     ref_start = frame->data[plane_idx];
yading@10:     ref_end   = frame->data[plane_idx] + (bh * frame->linesize[plane_idx] + bw) * 8;
yading@10: 
yading@10:     for (i = 0; i < 64; i++)
yading@10:         coordmap[i] = (i & 7) + (i >> 3) * stride;
yading@10: 
yading@10:     for (by = 0; by < bh; by++) {
yading@10:         for (i = 0; i < BINKB_NB_SRC; i++) {
yading@10:             if ((ret = binkb_read_bundle(c, gb, i)) < 0)
yading@10:                 return ret;
yading@10:         }
yading@10: 
yading@10:         dst  = frame->data[plane_idx]  + 8*by*stride;
yading@10:         for (bx = 0; bx < bw; bx++, dst += 8) {
yading@10:             blk = binkb_get_value(c, BINKB_SRC_BLOCK_TYPES);
yading@10:             switch (blk) {
yading@10:             case 0:
yading@10:                 break;
yading@10:             case 1:
yading@10:                 scan = bink_patterns[get_bits(gb, 4)];
yading@10:                 i = 0;
yading@10:                 do {
yading@10:                     int mode, run;
yading@10: 
yading@10:                     mode = get_bits1(gb);
yading@10:                     run = get_bits(gb, binkb_runbits[i]) + 1;
yading@10: 
yading@10:                     i += run;
yading@10:                     if (i > 64) {
yading@10:                         av_log(c->avctx, AV_LOG_ERROR, "Run went out of bounds\n");
yading@10:                         return AVERROR_INVALIDDATA;
yading@10:                     }
yading@10:                     if (mode) {
yading@10:                         v = binkb_get_value(c, BINKB_SRC_COLORS);
yading@10:                         for (j = 0; j < run; j++)
yading@10:                             dst[coordmap[*scan++]] = v;
yading@10:                     } else {
yading@10:                         for (j = 0; j < run; j++)
yading@10:                             dst[coordmap[*scan++]] = binkb_get_value(c, BINKB_SRC_COLORS);
yading@10:                     }
yading@10:                 } while (i < 63);
yading@10:                 if (i == 63)
yading@10:                     dst[coordmap[*scan++]] = binkb_get_value(c, BINKB_SRC_COLORS);
yading@10:                 break;
yading@10:             case 2:
yading@10:                 memset(dctblock, 0, sizeof(*dctblock) * 64);
yading@10:                 dctblock[0] = binkb_get_value(c, BINKB_SRC_INTRA_DC);
yading@10:                 qp = binkb_get_value(c, BINKB_SRC_INTRA_Q);
yading@10:                 read_dct_coeffs(gb, dctblock, bink_scan, (const int32_t (*)[64])binkb_intra_quant, qp);
yading@10:                 c->bdsp.idct_put(dst, stride, dctblock);
yading@10:                 break;
yading@10:             case 3:
yading@10:                 xoff = binkb_get_value(c, BINKB_SRC_X_OFF);
yading@10:                 yoff = binkb_get_value(c, BINKB_SRC_Y_OFF) + ybias;
yading@10:                 ref = dst + xoff + yoff * stride;
yading@10:                 if (ref < ref_start || ref + 8*stride > ref_end) {
yading@10:                     av_log(c->avctx, AV_LOG_WARNING, "Reference block is out of bounds\n");
yading@10:                 } else if (ref + 8*stride < dst || ref >= dst + 8*stride) {
yading@10:                     c->hdsp.put_pixels_tab[1][0](dst, ref, stride, 8);
yading@10:                 } else {
yading@10:                     put_pixels8x8_overlapped(dst, ref, stride);
yading@10:                 }
yading@10:                 c->dsp.clear_block(block);
yading@10:                 v = binkb_get_value(c, BINKB_SRC_INTER_COEFS);
yading@10:                 read_residue(gb, block, v);
yading@10:                 c->dsp.add_pixels8(dst, block, stride);
yading@10:                 break;
yading@10:             case 4:
yading@10:                 xoff = binkb_get_value(c, BINKB_SRC_X_OFF);
yading@10:                 yoff = binkb_get_value(c, BINKB_SRC_Y_OFF) + ybias;
yading@10:                 ref = dst + xoff + yoff * stride;
yading@10:                 if (ref < ref_start || ref + 8 * stride > ref_end) {
yading@10:                     av_log(c->avctx, AV_LOG_WARNING, "Reference block is out of bounds\n");
yading@10:                 } else if (ref + 8*stride < dst || ref >= dst + 8*stride) {
yading@10:                     c->hdsp.put_pixels_tab[1][0](dst, ref, stride, 8);
yading@10:                 } else {
yading@10:                     put_pixels8x8_overlapped(dst, ref, stride);
yading@10:                 }
yading@10:                 memset(dctblock, 0, sizeof(*dctblock) * 64);
yading@10:                 dctblock[0] = binkb_get_value(c, BINKB_SRC_INTER_DC);
yading@10:                 qp = binkb_get_value(c, BINKB_SRC_INTER_Q);
yading@10:                 read_dct_coeffs(gb, dctblock, bink_scan, (const int32_t (*)[64])binkb_inter_quant, qp);
yading@10:                 c->bdsp.idct_add(dst, stride, dctblock);
yading@10:                 break;
yading@10:             case 5:
yading@10:                 v = binkb_get_value(c, BINKB_SRC_COLORS);
yading@10:                 c->dsp.fill_block_tab[1](dst, v, stride, 8);
yading@10:                 break;
yading@10:             case 6:
yading@10:                 for (i = 0; i < 2; i++)
yading@10:                     col[i] = binkb_get_value(c, BINKB_SRC_COLORS);
yading@10:                 for (i = 0; i < 8; i++) {
yading@10:                     v = binkb_get_value(c, BINKB_SRC_PATTERN);
yading@10:                     for (j = 0; j < 8; j++, v >>= 1)
yading@10:                         dst[i*stride + j] = col[v & 1];
yading@10:                 }
yading@10:                 break;
yading@10:             case 7:
yading@10:                 xoff = binkb_get_value(c, BINKB_SRC_X_OFF);
yading@10:                 yoff = binkb_get_value(c, BINKB_SRC_Y_OFF) + ybias;
yading@10:                 ref = dst + xoff + yoff * stride;
yading@10:                 if (ref < ref_start || ref + 8 * stride > ref_end) {
yading@10:                     av_log(c->avctx, AV_LOG_WARNING, "Reference block is out of bounds\n");
yading@10:                 } else if (ref + 8*stride < dst || ref >= dst + 8*stride) {
yading@10:                     c->hdsp.put_pixels_tab[1][0](dst, ref, stride, 8);
yading@10:                 } else {
yading@10:                     put_pixels8x8_overlapped(dst, ref, stride);
yading@10:                 }
yading@10:                 break;
yading@10:             case 8:
yading@10:                 for (i = 0; i < 8; i++)
yading@10:                     memcpy(dst + i*stride, c->bundle[BINKB_SRC_COLORS].cur_ptr + i*8, 8);
yading@10:                 c->bundle[BINKB_SRC_COLORS].cur_ptr += 64;
yading@10:                 break;
yading@10:             default:
yading@10:                 av_log(c->avctx, AV_LOG_ERROR, "Unknown block type %d\n", blk);
yading@10:                 return AVERROR_INVALIDDATA;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10:     if (get_bits_count(gb) & 0x1F) //next plane data starts at 32-bit boundary
yading@10:         skip_bits_long(gb, 32 - (get_bits_count(gb) & 0x1F));
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int bink_decode_plane(BinkContext *c, AVFrame *frame, GetBitContext *gb,
yading@10:                              int plane_idx, int is_chroma)
yading@10: {
yading@10:     int blk, ret;
yading@10:     int i, j, bx, by;
yading@10:     uint8_t *dst, *prev, *ref, *ref_start, *ref_end;
yading@10:     int v, col[2];
yading@10:     const uint8_t *scan;
yading@10:     int xoff, yoff;
yading@10:     LOCAL_ALIGNED_16(int16_t, block, [64]);
yading@10:     LOCAL_ALIGNED_16(uint8_t, ublock, [64]);
yading@10:     LOCAL_ALIGNED_16(int32_t, dctblock, [64]);
yading@10:     int coordmap[64];
yading@10: 
yading@10:     const int stride = frame->linesize[plane_idx];
yading@10:     int bw = is_chroma ? (c->avctx->width  + 15) >> 4 : (c->avctx->width  + 7) >> 3;
yading@10:     int bh = is_chroma ? (c->avctx->height + 15) >> 4 : (c->avctx->height + 7) >> 3;
yading@10:     int width = c->avctx->width >> is_chroma;
yading@10: 
yading@10:     init_lengths(c, FFMAX(width, 8), bw);
yading@10:     for (i = 0; i < BINK_NB_SRC; i++)
yading@10:         read_bundle(gb, c, i);
yading@10: 
yading@10:     ref_start = c->last->data[plane_idx] ? c->last->data[plane_idx]
yading@10:                                          : frame->data[plane_idx];
yading@10:     ref_end   = ref_start
yading@10:                 + (bw - 1 + c->last->linesize[plane_idx] * (bh - 1)) * 8;
yading@10: 
yading@10:     for (i = 0; i < 64; i++)
yading@10:         coordmap[i] = (i & 7) + (i >> 3) * stride;
yading@10: 
yading@10:     for (by = 0; by < bh; by++) {
yading@10:         if ((ret = read_block_types(c->avctx, gb, &c->bundle[BINK_SRC_BLOCK_TYPES])) < 0)
yading@10:             return ret;
yading@10:         if ((ret = read_block_types(c->avctx, gb, &c->bundle[BINK_SRC_SUB_BLOCK_TYPES])) < 0)
yading@10:             return ret;
yading@10:         if ((ret = read_colors(gb, &c->bundle[BINK_SRC_COLORS], c)) < 0)
yading@10:             return ret;
yading@10:         if ((ret = read_patterns(c->avctx, gb, &c->bundle[BINK_SRC_PATTERN])) < 0)
yading@10:             return ret;
yading@10:         if ((ret = read_motion_values(c->avctx, gb, &c->bundle[BINK_SRC_X_OFF])) < 0)
yading@10:             return ret;
yading@10:         if ((ret = read_motion_values(c->avctx, gb, &c->bundle[BINK_SRC_Y_OFF])) < 0)
yading@10:             return ret;
yading@10:         if ((ret = read_dcs(c->avctx, gb, &c->bundle[BINK_SRC_INTRA_DC], DC_START_BITS, 0)) < 0)
yading@10:             return ret;
yading@10:         if ((ret = read_dcs(c->avctx, gb, &c->bundle[BINK_SRC_INTER_DC], DC_START_BITS, 1)) < 0)
yading@10:             return ret;
yading@10:         if ((ret = read_runs(c->avctx, gb, &c->bundle[BINK_SRC_RUN])) < 0)
yading@10:             return ret;
yading@10: 
yading@10:         if (by == bh)
yading@10:             break;
yading@10:         dst  = frame->data[plane_idx]  + 8*by*stride;
yading@10:         prev = (c->last->data[plane_idx] ? c->last->data[plane_idx]
yading@10:                                          : frame->data[plane_idx]) + 8*by*stride;
yading@10:         for (bx = 0; bx < bw; bx++, dst += 8, prev += 8) {
yading@10:             blk = get_value(c, BINK_SRC_BLOCK_TYPES);
yading@10:             // 16x16 block type on odd line means part of the already decoded block, so skip it
yading@10:             if ((by & 1) && blk == SCALED_BLOCK) {
yading@10:                 bx++;
yading@10:                 dst  += 8;
yading@10:                 prev += 8;
yading@10:                 continue;
yading@10:             }
yading@10:             switch (blk) {
yading@10:             case SKIP_BLOCK:
yading@10:                 c->hdsp.put_pixels_tab[1][0](dst, prev, stride, 8);
yading@10:                 break;
yading@10:             case SCALED_BLOCK:
yading@10:                 blk = get_value(c, BINK_SRC_SUB_BLOCK_TYPES);
yading@10:                 switch (blk) {
yading@10:                 case RUN_BLOCK:
yading@10:                     scan = bink_patterns[get_bits(gb, 4)];
yading@10:                     i = 0;
yading@10:                     do {
yading@10:                         int run = get_value(c, BINK_SRC_RUN) + 1;
yading@10: 
yading@10:                         i += run;
yading@10:                         if (i > 64) {
yading@10:                             av_log(c->avctx, AV_LOG_ERROR, "Run went out of bounds\n");
yading@10:                             return AVERROR_INVALIDDATA;
yading@10:                         }
yading@10:                         if (get_bits1(gb)) {
yading@10:                             v = get_value(c, BINK_SRC_COLORS);
yading@10:                             for (j = 0; j < run; j++)
yading@10:                                 ublock[*scan++] = v;
yading@10:                         } else {
yading@10:                             for (j = 0; j < run; j++)
yading@10:                                 ublock[*scan++] = get_value(c, BINK_SRC_COLORS);
yading@10:                         }
yading@10:                     } while (i < 63);
yading@10:                     if (i == 63)
yading@10:                         ublock[*scan++] = get_value(c, BINK_SRC_COLORS);
yading@10:                     break;
yading@10:                 case INTRA_BLOCK:
yading@10:                     memset(dctblock, 0, sizeof(*dctblock) * 64);
yading@10:                     dctblock[0] = get_value(c, BINK_SRC_INTRA_DC);
yading@10:                     read_dct_coeffs(gb, dctblock, bink_scan, bink_intra_quant, -1);
yading@10:                     c->bdsp.idct_put(ublock, 8, dctblock);
yading@10:                     break;
yading@10:                 case FILL_BLOCK:
yading@10:                     v = get_value(c, BINK_SRC_COLORS);
yading@10:                     c->dsp.fill_block_tab[0](dst, v, stride, 16);
yading@10:                     break;
yading@10:                 case PATTERN_BLOCK:
yading@10:                     for (i = 0; i < 2; i++)
yading@10:                         col[i] = get_value(c, BINK_SRC_COLORS);
yading@10:                     for (j = 0; j < 8; j++) {
yading@10:                         v = get_value(c, BINK_SRC_PATTERN);
yading@10:                         for (i = 0; i < 8; i++, v >>= 1)
yading@10:                             ublock[i + j*8] = col[v & 1];
yading@10:                     }
yading@10:                     break;
yading@10:                 case RAW_BLOCK:
yading@10:                     for (j = 0; j < 8; j++)
yading@10:                         for (i = 0; i < 8; i++)
yading@10:                             ublock[i + j*8] = get_value(c, BINK_SRC_COLORS);
yading@10:                     break;
yading@10:                 default:
yading@10:                     av_log(c->avctx, AV_LOG_ERROR, "Incorrect 16x16 block type %d\n", blk);
yading@10:                     return AVERROR_INVALIDDATA;
yading@10:                 }
yading@10:                 if (blk != FILL_BLOCK)
yading@10:                 c->bdsp.scale_block(ublock, dst, stride);
yading@10:                 bx++;
yading@10:                 dst  += 8;
yading@10:                 prev += 8;
yading@10:                 break;
yading@10:             case MOTION_BLOCK:
yading@10:                 xoff = get_value(c, BINK_SRC_X_OFF);
yading@10:                 yoff = get_value(c, BINK_SRC_Y_OFF);
yading@10:                 ref = prev + xoff + yoff * stride;
yading@10:                 if (ref < ref_start || ref > ref_end) {
yading@10:                     av_log(c->avctx, AV_LOG_ERROR, "Copy out of bounds @%d, %d\n",
yading@10:                            bx*8 + xoff, by*8 + yoff);
yading@10:                     return AVERROR_INVALIDDATA;
yading@10:                 }
yading@10:                 c->hdsp.put_pixels_tab[1][0](dst, ref, stride, 8);
yading@10:                 break;
yading@10:             case RUN_BLOCK:
yading@10:                 scan = bink_patterns[get_bits(gb, 4)];
yading@10:                 i = 0;
yading@10:                 do {
yading@10:                     int run = get_value(c, BINK_SRC_RUN) + 1;
yading@10: 
yading@10:                     i += run;
yading@10:                     if (i > 64) {
yading@10:                         av_log(c->avctx, AV_LOG_ERROR, "Run went out of bounds\n");
yading@10:                         return AVERROR_INVALIDDATA;
yading@10:                     }
yading@10:                     if (get_bits1(gb)) {
yading@10:                         v = get_value(c, BINK_SRC_COLORS);
yading@10:                         for (j = 0; j < run; j++)
yading@10:                             dst[coordmap[*scan++]] = v;
yading@10:                     } else {
yading@10:                         for (j = 0; j < run; j++)
yading@10:                             dst[coordmap[*scan++]] = get_value(c, BINK_SRC_COLORS);
yading@10:                     }
yading@10:                 } while (i < 63);
yading@10:                 if (i == 63)
yading@10:                     dst[coordmap[*scan++]] = get_value(c, BINK_SRC_COLORS);
yading@10:                 break;
yading@10:             case RESIDUE_BLOCK:
yading@10:                 xoff = get_value(c, BINK_SRC_X_OFF);
yading@10:                 yoff = get_value(c, BINK_SRC_Y_OFF);
yading@10:                 ref = prev + xoff + yoff * stride;
yading@10:                 if (ref < ref_start || ref > ref_end) {
yading@10:                     av_log(c->avctx, AV_LOG_ERROR, "Copy out of bounds @%d, %d\n",
yading@10:                            bx*8 + xoff, by*8 + yoff);
yading@10:                     return AVERROR_INVALIDDATA;
yading@10:                 }
yading@10:                 c->hdsp.put_pixels_tab[1][0](dst, ref, stride, 8);
yading@10:                 c->dsp.clear_block(block);
yading@10:                 v = get_bits(gb, 7);
yading@10:                 read_residue(gb, block, v);
yading@10:                 c->dsp.add_pixels8(dst, block, stride);
yading@10:                 break;
yading@10:             case INTRA_BLOCK:
yading@10:                 memset(dctblock, 0, sizeof(*dctblock) * 64);
yading@10:                 dctblock[0] = get_value(c, BINK_SRC_INTRA_DC);
yading@10:                 read_dct_coeffs(gb, dctblock, bink_scan, bink_intra_quant, -1);
yading@10:                 c->bdsp.idct_put(dst, stride, dctblock);
yading@10:                 break;
yading@10:             case FILL_BLOCK:
yading@10:                 v = get_value(c, BINK_SRC_COLORS);
yading@10:                 c->dsp.fill_block_tab[1](dst, v, stride, 8);
yading@10:                 break;
yading@10:             case INTER_BLOCK:
yading@10:                 xoff = get_value(c, BINK_SRC_X_OFF);
yading@10:                 yoff = get_value(c, BINK_SRC_Y_OFF);
yading@10:                 ref = prev + xoff + yoff * stride;
yading@10:                 if (ref < ref_start || ref > ref_end) {
yading@10:                     av_log(c->avctx, AV_LOG_ERROR, "Copy out of bounds @%d, %d\n",
yading@10:                            bx*8 + xoff, by*8 + yoff);
yading@10:                     return -1;
yading@10:                 }
yading@10:                 c->hdsp.put_pixels_tab[1][0](dst, ref, stride, 8);
yading@10:                 memset(dctblock, 0, sizeof(*dctblock) * 64);
yading@10:                 dctblock[0] = get_value(c, BINK_SRC_INTER_DC);
yading@10:                 read_dct_coeffs(gb, dctblock, bink_scan, bink_inter_quant, -1);
yading@10:                 c->bdsp.idct_add(dst, stride, dctblock);
yading@10:                 break;
yading@10:             case PATTERN_BLOCK:
yading@10:                 for (i = 0; i < 2; i++)
yading@10:                     col[i] = get_value(c, BINK_SRC_COLORS);
yading@10:                 for (i = 0; i < 8; i++) {
yading@10:                     v = get_value(c, BINK_SRC_PATTERN);
yading@10:                     for (j = 0; j < 8; j++, v >>= 1)
yading@10:                         dst[i*stride + j] = col[v & 1];
yading@10:                 }
yading@10:                 break;
yading@10:             case RAW_BLOCK:
yading@10:                 for (i = 0; i < 8; i++)
yading@10:                     memcpy(dst + i*stride, c->bundle[BINK_SRC_COLORS].cur_ptr + i*8, 8);
yading@10:                 c->bundle[BINK_SRC_COLORS].cur_ptr += 64;
yading@10:                 break;
yading@10:             default:
yading@10:                 av_log(c->avctx, AV_LOG_ERROR, "Unknown block type %d\n", blk);
yading@10:                 return AVERROR_INVALIDDATA;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10:     if (get_bits_count(gb) & 0x1F) //next plane data starts at 32-bit boundary
yading@10:         skip_bits_long(gb, 32 - (get_bits_count(gb) & 0x1F));
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, AVPacket *pkt)
yading@10: {
yading@10:     BinkContext * const c = avctx->priv_data;
yading@10:     AVFrame *frame = data;
yading@10:     GetBitContext gb;
yading@10:     int plane, plane_idx, ret;
yading@10:     int bits_count = pkt->size << 3;
yading@10: 
yading@10:     if (c->version > 'b') {
yading@10:         if ((ret = ff_get_buffer(avctx, frame, AV_GET_BUFFER_FLAG_REF)) < 0)
yading@10:             return ret;
yading@10:     } else {
yading@10:         if ((ret = ff_reget_buffer(avctx, c->last)) < 0)
yading@10:             return ret;
yading@10:         if ((ret = av_frame_ref(frame, c->last)) < 0)
yading@10:             return ret;
yading@10:     }
yading@10: 
yading@10:     init_get_bits(&gb, pkt->data, bits_count);
yading@10:     if (c->has_alpha) {
yading@10:         if (c->version >= 'i')
yading@10:             skip_bits_long(&gb, 32);
yading@10:         if ((ret = bink_decode_plane(c, frame, &gb, 3, 0)) < 0)
yading@10:             return ret;
yading@10:     }
yading@10:     if (c->version >= 'i')
yading@10:         skip_bits_long(&gb, 32);
yading@10: 
yading@10:     for (plane = 0; plane < 3; plane++) {
yading@10:         plane_idx = (!plane || !c->swap_planes) ? plane : (plane ^ 3);
yading@10: 
yading@10:         if (c->version > 'b') {
yading@10:             if ((ret = bink_decode_plane(c, frame, &gb, plane_idx, !!plane)) < 0)
yading@10:                 return ret;
yading@10:         } else {
yading@10:             if ((ret = binkb_decode_plane(c, frame, &gb, plane_idx,
yading@10:                                           !avctx->frame_number, !!plane)) < 0)
yading@10:                 return ret;
yading@10:         }
yading@10:         if (get_bits_count(&gb) >= bits_count)
yading@10:             break;
yading@10:     }
yading@10:     emms_c();
yading@10: 
yading@10:     if (c->version > 'b') {
yading@10:         av_frame_unref(c->last);
yading@10:         if ((ret = av_frame_ref(c->last, frame)) < 0)
yading@10:             return ret;
yading@10:     }
yading@10: 
yading@10:     *got_frame = 1;
yading@10: 
yading@10:     /* always report that the buffer was completely consumed */
yading@10:     return pkt->size;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Caclulate quantization tables for version b
yading@10:  */
yading@10: static av_cold void binkb_calc_quant(void)
yading@10: {
yading@10:     uint8_t inv_bink_scan[64];
yading@10:     static const int s[64]={
yading@10:         1073741824,1489322693,1402911301,1262586814,1073741824, 843633538, 581104888, 296244703,
yading@10:         1489322693,2065749918,1945893874,1751258219,1489322693,1170153332, 806015634, 410903207,
yading@10:         1402911301,1945893874,1832991949,1649649171,1402911301,1102260336, 759250125, 387062357,
yading@10:         1262586814,1751258219,1649649171,1484645031,1262586814, 992008094, 683307060, 348346918,
yading@10:         1073741824,1489322693,1402911301,1262586814,1073741824, 843633538, 581104888, 296244703,
yading@10:          843633538,1170153332,1102260336, 992008094, 843633538, 662838617, 456571181, 232757969,
yading@10:          581104888, 806015634, 759250125, 683307060, 581104888, 456571181, 314491699, 160326478,
yading@10:          296244703, 410903207, 387062357, 348346918, 296244703, 232757969, 160326478,  81733730,
yading@10:     };
yading@10:     int i, j;
yading@10: #define C (1LL<<30)
yading@10:     for (i = 0; i < 64; i++)
yading@10:         inv_bink_scan[bink_scan[i]] = i;
yading@10: 
yading@10:     for (j = 0; j < 16; j++) {
yading@10:         for (i = 0; i < 64; i++) {
yading@10:             int k = inv_bink_scan[i];
yading@10:             binkb_intra_quant[j][k] = binkb_intra_seed[i] * (int64_t)s[i] *
yading@10:                                         binkb_num[j]/(binkb_den[j] * (C>>12));
yading@10:             binkb_inter_quant[j][k] = binkb_inter_seed[i] * (int64_t)s[i] *
yading@10:                                         binkb_num[j]/(binkb_den[j] * (C>>12));
yading@10:         }
yading@10:     }
yading@10: }
yading@10: 
yading@10: static av_cold int decode_init(AVCodecContext *avctx)
yading@10: {
yading@10:     BinkContext * const c = avctx->priv_data;
yading@10:     static VLC_TYPE table[16 * 128][2];
yading@10:     static int binkb_initialised = 0;
yading@10:     int i, ret;
yading@10:     int flags;
yading@10: 
yading@10:     c->version = avctx->codec_tag >> 24;
yading@10:     if (avctx->extradata_size < 4) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Extradata missing or too short\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     flags = AV_RL32(avctx->extradata);
yading@10:     c->has_alpha = flags & BINK_FLAG_ALPHA;
yading@10:     c->swap_planes = c->version >= 'h';
yading@10:     if (!bink_trees[15].table) {
yading@10:         for (i = 0; i < 16; i++) {
yading@10:             const int maxbits = bink_tree_lens[i][15];
yading@10:             bink_trees[i].table = table + i*128;
yading@10:             bink_trees[i].table_allocated = 1 << maxbits;
yading@10:             init_vlc(&bink_trees[i], maxbits, 16,
yading@10:                      bink_tree_lens[i], 1, 1,
yading@10:                      bink_tree_bits[i], 1, 1, INIT_VLC_USE_NEW_STATIC | INIT_VLC_LE);
yading@10:         }
yading@10:     }
yading@10:     c->avctx = avctx;
yading@10: 
yading@10:     c->last = av_frame_alloc();
yading@10:     if (!c->last)
yading@10:         return AVERROR(ENOMEM);
yading@10: 
yading@10:     if ((ret = av_image_check_size(avctx->width, avctx->height, 0, avctx)) < 0)
yading@10:         return ret;
yading@10: 
yading@10:     avctx->pix_fmt = c->has_alpha ? AV_PIX_FMT_YUVA420P : AV_PIX_FMT_YUV420P;
yading@10: 
yading@10:     ff_dsputil_init(&c->dsp, avctx);
yading@10:     ff_hpeldsp_init(&c->hdsp, avctx->flags);
yading@10:     ff_binkdsp_init(&c->bdsp);
yading@10: 
yading@10:     if ((ret = init_bundles(c)) < 0) {
yading@10:         free_bundles(c);
yading@10:         return ret;
yading@10:     }
yading@10: 
yading@10:     if (c->version == 'b') {
yading@10:         if (!binkb_initialised) {
yading@10:             binkb_calc_quant();
yading@10:             binkb_initialised = 1;
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static av_cold int decode_end(AVCodecContext *avctx)
yading@10: {
yading@10:     BinkContext * const c = avctx->priv_data;
yading@10: 
yading@10:     av_frame_free(&c->last);
yading@10: 
yading@10:     free_bundles(c);
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: AVCodec ff_bink_decoder = {
yading@10:     .name           = "binkvideo",
yading@10:     .type           = AVMEDIA_TYPE_VIDEO,
yading@10:     .id             = AV_CODEC_ID_BINKVIDEO,
yading@10:     .priv_data_size = sizeof(BinkContext),
yading@10:     .init           = decode_init,
yading@10:     .close          = decode_end,
yading@10:     .decode         = decode_frame,
yading@10:     .long_name      = NULL_IF_CONFIG_SMALL("Bink video"),
yading@10:     .capabilities   = CODEC_CAP_DR1,
yading@10: };