yading@10: /*
yading@10:  * Microsoft Screen 3 (aka Microsoft ATC Screen) decoder
yading@10:  * Copyright (c) 2012 Konstantin Shishkov
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: /**
yading@10:  * @file
yading@10:  * Microsoft Screen 3 (aka Microsoft ATC Screen) decoder
yading@10:  */
yading@10: 
yading@10: #include "avcodec.h"
yading@10: #include "bytestream.h"
yading@10: #include "dsputil.h"
yading@10: #include "internal.h"
yading@10: #include "mss34dsp.h"
yading@10: 
yading@10: #define HEADER_SIZE 27
yading@10: 
yading@10: #define MODEL2_SCALE       13
yading@10: #define MODEL_SCALE        15
yading@10: #define MODEL256_SEC_SCALE  9
yading@10: 
yading@10: typedef struct Model2 {
yading@10:     int      upd_val, till_rescale;
yading@10:     unsigned zero_freq,  zero_weight;
yading@10:     unsigned total_freq, total_weight;
yading@10: } Model2;
yading@10: 
yading@10: typedef struct Model {
yading@10:     int weights[16], freqs[16];
yading@10:     int num_syms;
yading@10:     int tot_weight;
yading@10:     int upd_val, max_upd_val, till_rescale;
yading@10: } Model;
yading@10: 
yading@10: typedef struct Model256 {
yading@10:     int weights[256], freqs[256];
yading@10:     int tot_weight;
yading@10:     int secondary[68];
yading@10:     int sec_size;
yading@10:     int upd_val, max_upd_val, till_rescale;
yading@10: } Model256;
yading@10: 
yading@10: #define RAC_BOTTOM 0x01000000
yading@10: typedef struct RangeCoder {
yading@10:     const uint8_t *src, *src_end;
yading@10: 
yading@10:     uint32_t range, low;
yading@10:     int got_error;
yading@10: } RangeCoder;
yading@10: 
yading@10: enum BlockType {
yading@10:     FILL_BLOCK = 0,
yading@10:     IMAGE_BLOCK,
yading@10:     DCT_BLOCK,
yading@10:     HAAR_BLOCK,
yading@10:     SKIP_BLOCK
yading@10: };
yading@10: 
yading@10: typedef struct BlockTypeContext {
yading@10:     int      last_type;
yading@10:     Model    bt_model[5];
yading@10: } BlockTypeContext;
yading@10: 
yading@10: typedef struct FillBlockCoder {
yading@10:     int      fill_val;
yading@10:     Model    coef_model;
yading@10: } FillBlockCoder;
yading@10: 
yading@10: typedef struct ImageBlockCoder {
yading@10:     Model256 esc_model, vec_entry_model;
yading@10:     Model    vec_size_model;
yading@10:     Model    vq_model[125];
yading@10: } ImageBlockCoder;
yading@10: 
yading@10: typedef struct DCTBlockCoder {
yading@10:     int      *prev_dc;
yading@10:     int      prev_dc_stride;
yading@10:     int      prev_dc_height;
yading@10:     int      quality;
yading@10:     uint16_t qmat[64];
yading@10:     Model    dc_model;
yading@10:     Model2   sign_model;
yading@10:     Model256 ac_model;
yading@10: } DCTBlockCoder;
yading@10: 
yading@10: typedef struct HaarBlockCoder {
yading@10:     int      quality, scale;
yading@10:     Model256 coef_model;
yading@10:     Model    coef_hi_model;
yading@10: } HaarBlockCoder;
yading@10: 
yading@10: typedef struct MSS3Context {
yading@10:     AVCodecContext   *avctx;
yading@10:     AVFrame          pic;
yading@10: 
yading@10:     int              got_error;
yading@10:     RangeCoder       coder;
yading@10:     BlockTypeContext btype[3];
yading@10:     FillBlockCoder   fill_coder[3];
yading@10:     ImageBlockCoder  image_coder[3];
yading@10:     DCTBlockCoder    dct_coder[3];
yading@10:     HaarBlockCoder   haar_coder[3];
yading@10: 
yading@10:     int              dctblock[64];
yading@10:     int              hblock[16 * 16];
yading@10: } MSS3Context;
yading@10: 
yading@10: 
yading@10: static void model2_reset(Model2 *m)
yading@10: {
yading@10:     m->zero_weight  = 1;
yading@10:     m->total_weight = 2;
yading@10:     m->zero_freq    = 0x1000;
yading@10:     m->total_freq   = 0x2000;
yading@10:     m->upd_val      = 4;
yading@10:     m->till_rescale = 4;
yading@10: }
yading@10: 
yading@10: static void model2_update(Model2 *m, int bit)
yading@10: {
yading@10:     unsigned scale;
yading@10: 
yading@10:     if (!bit)
yading@10:         m->zero_weight++;
yading@10:     m->till_rescale--;
yading@10:     if (m->till_rescale)
yading@10:         return;
yading@10: 
yading@10:     m->total_weight += m->upd_val;
yading@10:     if (m->total_weight > 0x2000) {
yading@10:         m->total_weight = (m->total_weight + 1) >> 1;
yading@10:         m->zero_weight  = (m->zero_weight  + 1) >> 1;
yading@10:         if (m->total_weight == m->zero_weight)
yading@10:             m->total_weight = m->zero_weight + 1;
yading@10:     }
yading@10:     m->upd_val = m->upd_val * 5 >> 2;
yading@10:     if (m->upd_val > 64)
yading@10:         m->upd_val = 64;
yading@10:     scale = 0x80000000u / m->total_weight;
yading@10:     m->zero_freq    = m->zero_weight  * scale >> 18;
yading@10:     m->total_freq   = m->total_weight * scale >> 18;
yading@10:     m->till_rescale = m->upd_val;
yading@10: }
yading@10: 
yading@10: static void model_update(Model *m, int val)
yading@10: {
yading@10:     int i, sum = 0;
yading@10:     unsigned scale;
yading@10: 
yading@10:     m->weights[val]++;
yading@10:     m->till_rescale--;
yading@10:     if (m->till_rescale)
yading@10:         return;
yading@10:     m->tot_weight += m->upd_val;
yading@10: 
yading@10:     if (m->tot_weight > 0x8000) {
yading@10:         m->tot_weight = 0;
yading@10:         for (i = 0; i < m->num_syms; i++) {
yading@10:             m->weights[i]  = (m->weights[i] + 1) >> 1;
yading@10:             m->tot_weight +=  m->weights[i];
yading@10:         }
yading@10:     }
yading@10:     scale = 0x80000000u / m->tot_weight;
yading@10:     for (i = 0; i < m->num_syms; i++) {
yading@10:         m->freqs[i] = sum * scale >> 16;
yading@10:         sum += m->weights[i];
yading@10:     }
yading@10: 
yading@10:     m->upd_val = m->upd_val * 5 >> 2;
yading@10:     if (m->upd_val > m->max_upd_val)
yading@10:         m->upd_val = m->max_upd_val;
yading@10:     m->till_rescale = m->upd_val;
yading@10: }
yading@10: 
yading@10: static void model_reset(Model *m)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     m->tot_weight   = 0;
yading@10:     for (i = 0; i < m->num_syms - 1; i++)
yading@10:         m->weights[i] = 1;
yading@10:     m->weights[m->num_syms - 1] = 0;
yading@10: 
yading@10:     m->upd_val      = m->num_syms;
yading@10:     m->till_rescale = 1;
yading@10:     model_update(m, m->num_syms - 1);
yading@10:     m->till_rescale =
yading@10:     m->upd_val      = (m->num_syms + 6) >> 1;
yading@10: }
yading@10: 
yading@10: static av_cold void model_init(Model *m, int num_syms)
yading@10: {
yading@10:     m->num_syms    = num_syms;
yading@10:     m->max_upd_val = 8 * num_syms + 48;
yading@10: 
yading@10:     model_reset(m);
yading@10: }
yading@10: 
yading@10: static void model256_update(Model256 *m, int val)
yading@10: {
yading@10:     int i, sum = 0;
yading@10:     unsigned scale;
yading@10:     int send, sidx = 1;
yading@10: 
yading@10:     m->weights[val]++;
yading@10:     m->till_rescale--;
yading@10:     if (m->till_rescale)
yading@10:         return;
yading@10:     m->tot_weight += m->upd_val;
yading@10: 
yading@10:     if (m->tot_weight > 0x8000) {
yading@10:         m->tot_weight = 0;
yading@10:         for (i = 0; i < 256; i++) {
yading@10:             m->weights[i]  = (m->weights[i] + 1) >> 1;
yading@10:             m->tot_weight +=  m->weights[i];
yading@10:         }
yading@10:     }
yading@10:     scale = 0x80000000u / m->tot_weight;
yading@10:     m->secondary[0] = 0;
yading@10:     for (i = 0; i < 256; i++) {
yading@10:         m->freqs[i] = sum * scale >> 16;
yading@10:         sum += m->weights[i];
yading@10:         send = m->freqs[i] >> MODEL256_SEC_SCALE;
yading@10:         while (sidx <= send)
yading@10:             m->secondary[sidx++] = i - 1;
yading@10:     }
yading@10:     while (sidx < m->sec_size)
yading@10:         m->secondary[sidx++] = 255;
yading@10: 
yading@10:     m->upd_val = m->upd_val * 5 >> 2;
yading@10:     if (m->upd_val > m->max_upd_val)
yading@10:         m->upd_val = m->max_upd_val;
yading@10:     m->till_rescale = m->upd_val;
yading@10: }
yading@10: 
yading@10: static void model256_reset(Model256 *m)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for (i = 0; i < 255; i++)
yading@10:         m->weights[i] = 1;
yading@10:     m->weights[255] = 0;
yading@10: 
yading@10:     m->tot_weight   = 0;
yading@10:     m->upd_val      = 256;
yading@10:     m->till_rescale = 1;
yading@10:     model256_update(m, 255);
yading@10:     m->till_rescale =
yading@10:     m->upd_val      = (256 + 6) >> 1;
yading@10: }
yading@10: 
yading@10: static av_cold void model256_init(Model256 *m)
yading@10: {
yading@10:     m->max_upd_val = 8 * 256 + 48;
yading@10:     m->sec_size    = (1 << 6) + 2;
yading@10: 
yading@10:     model256_reset(m);
yading@10: }
yading@10: 
yading@10: static void rac_init(RangeCoder *c, const uint8_t *src, int size)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     c->src       = src;
yading@10:     c->src_end   = src + size;
yading@10:     c->low       = 0;
yading@10:     for (i = 0; i < FFMIN(size, 4); i++)
yading@10:         c->low = (c->low << 8) | *c->src++;
yading@10:     c->range     = 0xFFFFFFFF;
yading@10:     c->got_error = 0;
yading@10: }
yading@10: 
yading@10: static void rac_normalise(RangeCoder *c)
yading@10: {
yading@10:     for (;;) {
yading@10:         c->range <<= 8;
yading@10:         c->low   <<= 8;
yading@10:         if (c->src < c->src_end) {
yading@10:             c->low |= *c->src++;
yading@10:         } else if (!c->low) {
yading@10:             c->got_error = 1;
yading@10:             c->low = 1;
yading@10:         }
yading@10:         if (c->range >= RAC_BOTTOM)
yading@10:             return;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static int rac_get_bit(RangeCoder *c)
yading@10: {
yading@10:     int bit;
yading@10: 
yading@10:     c->range >>= 1;
yading@10: 
yading@10:     bit = (c->range <= c->low);
yading@10:     if (bit)
yading@10:         c->low -= c->range;
yading@10: 
yading@10:     if (c->range < RAC_BOTTOM)
yading@10:         rac_normalise(c);
yading@10: 
yading@10:     return bit;
yading@10: }
yading@10: 
yading@10: static int rac_get_bits(RangeCoder *c, int nbits)
yading@10: {
yading@10:     int val;
yading@10: 
yading@10:     c->range >>= nbits;
yading@10:     val = c->low / c->range;
yading@10:     c->low -= c->range * val;
yading@10: 
yading@10:     if (c->range < RAC_BOTTOM)
yading@10:         rac_normalise(c);
yading@10: 
yading@10:     return val;
yading@10: }
yading@10: 
yading@10: static int rac_get_model2_sym(RangeCoder *c, Model2 *m)
yading@10: {
yading@10:     int bit, helper;
yading@10: 
yading@10:     helper = m->zero_freq * (c->range >> MODEL2_SCALE);
yading@10:     bit    = (c->low >= helper);
yading@10:     if (bit) {
yading@10:         c->low   -= helper;
yading@10:         c->range -= helper;
yading@10:     } else {
yading@10:         c->range  = helper;
yading@10:     }
yading@10: 
yading@10:     if (c->range < RAC_BOTTOM)
yading@10:         rac_normalise(c);
yading@10: 
yading@10:     model2_update(m, bit);
yading@10: 
yading@10:     return bit;
yading@10: }
yading@10: 
yading@10: static int rac_get_model_sym(RangeCoder *c, Model *m)
yading@10: {
yading@10:     int prob, prob2, helper, val;
yading@10:     int end, end2;
yading@10: 
yading@10:     prob       = 0;
yading@10:     prob2      = c->range;
yading@10:     c->range >>= MODEL_SCALE;
yading@10:     val        = 0;
yading@10:     end        = m->num_syms >> 1;
yading@10:     end2       = m->num_syms;
yading@10:     do {
yading@10:         helper = m->freqs[end] * c->range;
yading@10:         if (helper <= c->low) {
yading@10:             val   = end;
yading@10:             prob  = helper;
yading@10:         } else {
yading@10:             end2  = end;
yading@10:             prob2 = helper;
yading@10:         }
yading@10:         end = (end2 + val) >> 1;
yading@10:     } while (end != val);
yading@10:     c->low  -= prob;
yading@10:     c->range = prob2 - prob;
yading@10:     if (c->range < RAC_BOTTOM)
yading@10:         rac_normalise(c);
yading@10: 
yading@10:     model_update(m, val);
yading@10: 
yading@10:     return val;
yading@10: }
yading@10: 
yading@10: static int rac_get_model256_sym(RangeCoder *c, Model256 *m)
yading@10: {
yading@10:     int prob, prob2, helper, val;
yading@10:     int start, end;
yading@10:     int ssym;
yading@10: 
yading@10:     prob2      = c->range;
yading@10:     c->range >>= MODEL_SCALE;
yading@10: 
yading@10:     helper     = c->low / c->range;
yading@10:     ssym       = helper >> MODEL256_SEC_SCALE;
yading@10:     val        = m->secondary[ssym];
yading@10: 
yading@10:     end = start = m->secondary[ssym + 1] + 1;
yading@10:     while (end > val + 1) {
yading@10:         ssym = (end + val) >> 1;
yading@10:         if (m->freqs[ssym] <= helper) {
yading@10:             end = start;
yading@10:             val = ssym;
yading@10:         } else {
yading@10:             end   = (end + val) >> 1;
yading@10:             start = ssym;
yading@10:         }
yading@10:     }
yading@10:     prob = m->freqs[val] * c->range;
yading@10:     if (val != 255)
yading@10:         prob2 = m->freqs[val + 1] * c->range;
yading@10: 
yading@10:     c->low  -= prob;
yading@10:     c->range = prob2 - prob;
yading@10:     if (c->range < RAC_BOTTOM)
yading@10:         rac_normalise(c);
yading@10: 
yading@10:     model256_update(m, val);
yading@10: 
yading@10:     return val;
yading@10: }
yading@10: 
yading@10: static int decode_block_type(RangeCoder *c, BlockTypeContext *bt)
yading@10: {
yading@10:     bt->last_type = rac_get_model_sym(c, &bt->bt_model[bt->last_type]);
yading@10: 
yading@10:     return bt->last_type;
yading@10: }
yading@10: 
yading@10: static int decode_coeff(RangeCoder *c, Model *m)
yading@10: {
yading@10:     int val, sign;
yading@10: 
yading@10:     val = rac_get_model_sym(c, m);
yading@10:     if (val) {
yading@10:         sign = rac_get_bit(c);
yading@10:         if (val > 1) {
yading@10:             val--;
yading@10:             val = (1 << val) + rac_get_bits(c, val);
yading@10:         }
yading@10:         if (!sign)
yading@10:             val = -val;
yading@10:     }
yading@10: 
yading@10:     return val;
yading@10: }
yading@10: 
yading@10: static void decode_fill_block(RangeCoder *c, FillBlockCoder *fc,
yading@10:                               uint8_t *dst, int stride, int block_size)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     fc->fill_val += decode_coeff(c, &fc->coef_model);
yading@10: 
yading@10:     for (i = 0; i < block_size; i++, dst += stride)
yading@10:         memset(dst, fc->fill_val, block_size);
yading@10: }
yading@10: 
yading@10: static void decode_image_block(RangeCoder *c, ImageBlockCoder *ic,
yading@10:                                uint8_t *dst, int stride, int block_size)
yading@10: {
yading@10:     int i, j;
yading@10:     int vec_size;
yading@10:     int vec[4];
yading@10:     int prev_line[16];
yading@10:     int A, B, C;
yading@10: 
yading@10:     vec_size = rac_get_model_sym(c, &ic->vec_size_model) + 2;
yading@10:     for (i = 0; i < vec_size; i++)
yading@10:         vec[i] = rac_get_model256_sym(c, &ic->vec_entry_model);
yading@10:     for (; i < 4; i++)
yading@10:         vec[i] = 0;
yading@10:     memset(prev_line, 0, sizeof(prev_line));
yading@10: 
yading@10:     for (j = 0; j < block_size; j++) {
yading@10:         A = 0;
yading@10:         B = 0;
yading@10:         for (i = 0; i < block_size; i++) {
yading@10:             C = B;
yading@10:             B = prev_line[i];
yading@10:             A = rac_get_model_sym(c, &ic->vq_model[A + B * 5 + C * 25]);
yading@10: 
yading@10:             prev_line[i] = A;
yading@10:             if (A < 4)
yading@10:                dst[i] = vec[A];
yading@10:             else
yading@10:                dst[i] = rac_get_model256_sym(c, &ic->esc_model);
yading@10:         }
yading@10:         dst += stride;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static int decode_dct(RangeCoder *c, DCTBlockCoder *bc, int *block,
yading@10:                       int bx, int by)
yading@10: {
yading@10:     int skip, val, sign, pos = 1, zz_pos, dc;
yading@10:     int blk_pos = bx + by * bc->prev_dc_stride;
yading@10: 
yading@10:     memset(block, 0, sizeof(*block) * 64);
yading@10: 
yading@10:     dc = decode_coeff(c, &bc->dc_model);
yading@10:     if (by) {
yading@10:         if (bx) {
yading@10:             int l, tl, t;
yading@10: 
yading@10:             l  = bc->prev_dc[blk_pos - 1];
yading@10:             tl = bc->prev_dc[blk_pos - 1 - bc->prev_dc_stride];
yading@10:             t  = bc->prev_dc[blk_pos     - bc->prev_dc_stride];
yading@10: 
yading@10:             if (FFABS(t - tl) <= FFABS(l - tl))
yading@10:                 dc += l;
yading@10:             else
yading@10:                 dc += t;
yading@10:         } else {
yading@10:             dc += bc->prev_dc[blk_pos - bc->prev_dc_stride];
yading@10:         }
yading@10:     } else if (bx) {
yading@10:         dc += bc->prev_dc[bx - 1];
yading@10:     }
yading@10:     bc->prev_dc[blk_pos] = dc;
yading@10:     block[0]             = dc * bc->qmat[0];
yading@10: 
yading@10:     while (pos < 64) {
yading@10:         val = rac_get_model256_sym(c, &bc->ac_model);
yading@10:         if (!val)
yading@10:             return 0;
yading@10:         if (val == 0xF0) {
yading@10:             pos += 16;
yading@10:             continue;
yading@10:         }
yading@10:         skip = val >> 4;
yading@10:         val  = val & 0xF;
yading@10:         if (!val)
yading@10:             return -1;
yading@10:         pos += skip;
yading@10:         if (pos >= 64)
yading@10:             return -1;
yading@10: 
yading@10:         sign = rac_get_model2_sym(c, &bc->sign_model);
yading@10:         if (val > 1) {
yading@10:             val--;
yading@10:             val = (1 << val) + rac_get_bits(c, val);
yading@10:         }
yading@10:         if (!sign)
yading@10:             val = -val;
yading@10: 
yading@10:         zz_pos = ff_zigzag_direct[pos];
yading@10:         block[zz_pos] = val * bc->qmat[zz_pos];
yading@10:         pos++;
yading@10:     }
yading@10: 
yading@10:     return pos == 64 ? 0 : -1;
yading@10: }
yading@10: 
yading@10: static void decode_dct_block(RangeCoder *c, DCTBlockCoder *bc,
yading@10:                              uint8_t *dst, int stride, int block_size,
yading@10:                              int *block, int mb_x, int mb_y)
yading@10: {
yading@10:     int i, j;
yading@10:     int bx, by;
yading@10:     int nblocks = block_size >> 3;
yading@10: 
yading@10:     bx = mb_x * nblocks;
yading@10:     by = mb_y * nblocks;
yading@10: 
yading@10:     for (j = 0; j < nblocks; j++) {
yading@10:         for (i = 0; i < nblocks; i++) {
yading@10:             if (decode_dct(c, bc, block, bx + i, by + j)) {
yading@10:                 c->got_error = 1;
yading@10:                 return;
yading@10:             }
yading@10:             ff_mss34_dct_put(dst + i * 8, stride, block);
yading@10:         }
yading@10:         dst += 8 * stride;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void decode_haar_block(RangeCoder *c, HaarBlockCoder *hc,
yading@10:                               uint8_t *dst, int stride, int block_size,
yading@10:                               int *block)
yading@10: {
yading@10:     const int hsize = block_size >> 1;
yading@10:     int A, B, C, D, t1, t2, t3, t4;
yading@10:     int i, j;
yading@10: 
yading@10:     for (j = 0; j < block_size; j++) {
yading@10:         for (i = 0; i < block_size; i++) {
yading@10:             if (i < hsize && j < hsize)
yading@10:                 block[i] = rac_get_model256_sym(c, &hc->coef_model);
yading@10:             else
yading@10:                 block[i] = decode_coeff(c, &hc->coef_hi_model);
yading@10:             block[i] *= hc->scale;
yading@10:         }
yading@10:         block += block_size;
yading@10:     }
yading@10:     block -= block_size * block_size;
yading@10: 
yading@10:     for (j = 0; j < hsize; j++) {
yading@10:         for (i = 0; i < hsize; i++) {
yading@10:             A = block[i];
yading@10:             B = block[i + hsize];
yading@10:             C = block[i + hsize * block_size];
yading@10:             D = block[i + hsize * block_size + hsize];
yading@10: 
yading@10:             t1 = A - B;
yading@10:             t2 = C - D;
yading@10:             t3 = A + B;
yading@10:             t4 = C + D;
yading@10:             dst[i * 2]              = av_clip_uint8(t1 - t2);
yading@10:             dst[i * 2 + stride]     = av_clip_uint8(t1 + t2);
yading@10:             dst[i * 2 + 1]          = av_clip_uint8(t3 - t4);
yading@10:             dst[i * 2 + 1 + stride] = av_clip_uint8(t3 + t4);
yading@10:         }
yading@10:         block += block_size;
yading@10:         dst   += stride * 2;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void reset_coders(MSS3Context *ctx, int quality)
yading@10: {
yading@10:     int i, j;
yading@10: 
yading@10:     for (i = 0; i < 3; i++) {
yading@10:         ctx->btype[i].last_type = SKIP_BLOCK;
yading@10:         for (j = 0; j < 5; j++)
yading@10:             model_reset(&ctx->btype[i].bt_model[j]);
yading@10:         ctx->fill_coder[i].fill_val = 0;
yading@10:         model_reset(&ctx->fill_coder[i].coef_model);
yading@10:         model256_reset(&ctx->image_coder[i].esc_model);
yading@10:         model256_reset(&ctx->image_coder[i].vec_entry_model);
yading@10:         model_reset(&ctx->image_coder[i].vec_size_model);
yading@10:         for (j = 0; j < 125; j++)
yading@10:             model_reset(&ctx->image_coder[i].vq_model[j]);
yading@10:         if (ctx->dct_coder[i].quality != quality) {
yading@10:             ctx->dct_coder[i].quality = quality;
yading@10:             ff_mss34_gen_quant_mat(ctx->dct_coder[i].qmat, quality, !i);
yading@10:         }
yading@10:         memset(ctx->dct_coder[i].prev_dc, 0,
yading@10:                sizeof(*ctx->dct_coder[i].prev_dc) *
yading@10:                ctx->dct_coder[i].prev_dc_stride *
yading@10:                ctx->dct_coder[i].prev_dc_height);
yading@10:         model_reset(&ctx->dct_coder[i].dc_model);
yading@10:         model2_reset(&ctx->dct_coder[i].sign_model);
yading@10:         model256_reset(&ctx->dct_coder[i].ac_model);
yading@10:         if (ctx->haar_coder[i].quality != quality) {
yading@10:             ctx->haar_coder[i].quality = quality;
yading@10:             ctx->haar_coder[i].scale   = 17 - 7 * quality / 50;
yading@10:         }
yading@10:         model_reset(&ctx->haar_coder[i].coef_hi_model);
yading@10:         model256_reset(&ctx->haar_coder[i].coef_model);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static av_cold void init_coders(MSS3Context *ctx)
yading@10: {
yading@10:     int i, j;
yading@10: 
yading@10:     for (i = 0; i < 3; i++) {
yading@10:         for (j = 0; j < 5; j++)
yading@10:             model_init(&ctx->btype[i].bt_model[j], 5);
yading@10:         model_init(&ctx->fill_coder[i].coef_model, 12);
yading@10:         model256_init(&ctx->image_coder[i].esc_model);
yading@10:         model256_init(&ctx->image_coder[i].vec_entry_model);
yading@10:         model_init(&ctx->image_coder[i].vec_size_model, 3);
yading@10:         for (j = 0; j < 125; j++)
yading@10:             model_init(&ctx->image_coder[i].vq_model[j], 5);
yading@10:         model_init(&ctx->dct_coder[i].dc_model, 12);
yading@10:         model256_init(&ctx->dct_coder[i].ac_model);
yading@10:         model_init(&ctx->haar_coder[i].coef_hi_model, 12);
yading@10:         model256_init(&ctx->haar_coder[i].coef_model);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static int mss3_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
yading@10:                              AVPacket *avpkt)
yading@10: {
yading@10:     const uint8_t *buf = avpkt->data;
yading@10:     int buf_size = avpkt->size;
yading@10:     MSS3Context *c = avctx->priv_data;
yading@10:     RangeCoder *acoder = &c->coder;
yading@10:     GetByteContext gb;
yading@10:     uint8_t *dst[3];
yading@10:     int dec_width, dec_height, dec_x, dec_y, quality, keyframe;
yading@10:     int x, y, i, mb_width, mb_height, blk_size, btype;
yading@10:     int ret;
yading@10: 
yading@10:     if (buf_size < HEADER_SIZE) {
yading@10:         av_log(avctx, AV_LOG_ERROR,
yading@10:                "Frame should have at least %d bytes, got %d instead\n",
yading@10:                HEADER_SIZE, buf_size);
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     bytestream2_init(&gb, buf, buf_size);
yading@10:     keyframe   = bytestream2_get_be32(&gb);
yading@10:     if (keyframe & ~0x301) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Invalid frame type %X\n", keyframe);
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     keyframe   = !(keyframe & 1);
yading@10:     bytestream2_skip(&gb, 6);
yading@10:     dec_x      = bytestream2_get_be16(&gb);
yading@10:     dec_y      = bytestream2_get_be16(&gb);
yading@10:     dec_width  = bytestream2_get_be16(&gb);
yading@10:     dec_height = bytestream2_get_be16(&gb);
yading@10: 
yading@10:     if (dec_x + dec_width > avctx->width ||
yading@10:         dec_y + dec_height > avctx->height ||
yading@10:         (dec_width | dec_height) & 0xF) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Invalid frame dimensions %dx%d +%d,%d\n",
yading@10:                dec_width, dec_height, dec_x, dec_y);
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     bytestream2_skip(&gb, 4);
yading@10:     quality    = bytestream2_get_byte(&gb);
yading@10:     if (quality < 1 || quality > 100) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Invalid quality setting %d\n", quality);
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     bytestream2_skip(&gb, 4);
yading@10: 
yading@10:     if (keyframe && !bytestream2_get_bytes_left(&gb)) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Keyframe without data found\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     if (!keyframe && c->got_error)
yading@10:         return buf_size;
yading@10:     c->got_error = 0;
yading@10: 
yading@10:     if ((ret = ff_reget_buffer(avctx, &c->pic)) < 0)
yading@10:         return ret;
yading@10:     c->pic.key_frame = keyframe;
yading@10:     c->pic.pict_type = keyframe ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P;
yading@10:     if (!bytestream2_get_bytes_left(&gb)) {
yading@10:         if ((ret = av_frame_ref(data, &c->pic)) < 0)
yading@10:             return ret;
yading@10:         *got_frame      = 1;
yading@10: 
yading@10:         return buf_size;
yading@10:     }
yading@10: 
yading@10:     reset_coders(c, quality);
yading@10: 
yading@10:     rac_init(acoder, buf + HEADER_SIZE, buf_size - HEADER_SIZE);
yading@10: 
yading@10:     mb_width  = dec_width  >> 4;
yading@10:     mb_height = dec_height >> 4;
yading@10:     dst[0] = c->pic.data[0] + dec_x     +  dec_y      * c->pic.linesize[0];
yading@10:     dst[1] = c->pic.data[1] + dec_x / 2 + (dec_y / 2) * c->pic.linesize[1];
yading@10:     dst[2] = c->pic.data[2] + dec_x / 2 + (dec_y / 2) * c->pic.linesize[2];
yading@10:     for (y = 0; y < mb_height; y++) {
yading@10:         for (x = 0; x < mb_width; x++) {
yading@10:             for (i = 0; i < 3; i++) {
yading@10:                 blk_size = 8 << !i;
yading@10: 
yading@10:                 btype = decode_block_type(acoder, c->btype + i);
yading@10:                 switch (btype) {
yading@10:                 case FILL_BLOCK:
yading@10:                     decode_fill_block(acoder, c->fill_coder + i,
yading@10:                                       dst[i] + x * blk_size,
yading@10:                                       c->pic.linesize[i], blk_size);
yading@10:                     break;
yading@10:                 case IMAGE_BLOCK:
yading@10:                     decode_image_block(acoder, c->image_coder + i,
yading@10:                                        dst[i] + x * blk_size,
yading@10:                                        c->pic.linesize[i], blk_size);
yading@10:                     break;
yading@10:                 case DCT_BLOCK:
yading@10:                     decode_dct_block(acoder, c->dct_coder + i,
yading@10:                                      dst[i] + x * blk_size,
yading@10:                                      c->pic.linesize[i], blk_size,
yading@10:                                      c->dctblock, x, y);
yading@10:                     break;
yading@10:                 case HAAR_BLOCK:
yading@10:                     decode_haar_block(acoder, c->haar_coder + i,
yading@10:                                       dst[i] + x * blk_size,
yading@10:                                       c->pic.linesize[i], blk_size,
yading@10:                                       c->hblock);
yading@10:                     break;
yading@10:                 }
yading@10:                 if (c->got_error || acoder->got_error) {
yading@10:                     av_log(avctx, AV_LOG_ERROR, "Error decoding block %d,%d\n",
yading@10:                            x, y);
yading@10:                     c->got_error = 1;
yading@10:                     return AVERROR_INVALIDDATA;
yading@10:                 }
yading@10:             }
yading@10:         }
yading@10:         dst[0] += c->pic.linesize[0] * 16;
yading@10:         dst[1] += c->pic.linesize[1] * 8;
yading@10:         dst[2] += c->pic.linesize[2] * 8;
yading@10:     }
yading@10: 
yading@10:     if ((ret = av_frame_ref(data, &c->pic)) < 0)
yading@10:         return ret;
yading@10: 
yading@10:     *got_frame      = 1;
yading@10: 
yading@10:     return buf_size;
yading@10: }
yading@10: 
yading@10: static av_cold int mss3_decode_init(AVCodecContext *avctx)
yading@10: {
yading@10:     MSS3Context * const c = avctx->priv_data;
yading@10:     int i;
yading@10: 
yading@10:     c->avctx = avctx;
yading@10: 
yading@10:     if ((avctx->width & 0xF) || (avctx->height & 0xF)) {
yading@10:         av_log(avctx, AV_LOG_ERROR,
yading@10:                "Image dimensions should be a multiple of 16.\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     c->got_error = 0;
yading@10:     for (i = 0; i < 3; i++) {
yading@10:         int b_width  = avctx->width  >> (2 + !!i);
yading@10:         int b_height = avctx->height >> (2 + !!i);
yading@10:         c->dct_coder[i].prev_dc_stride = b_width;
yading@10:         c->dct_coder[i].prev_dc_height = b_height;
yading@10:         c->dct_coder[i].prev_dc = av_malloc(sizeof(*c->dct_coder[i].prev_dc) *
yading@10:                                             b_width * b_height);
yading@10:         if (!c->dct_coder[i].prev_dc) {
yading@10:             av_log(avctx, AV_LOG_ERROR, "Cannot allocate buffer\n");
yading@10:             while (i >= 0) {
yading@10:                 av_freep(&c->dct_coder[i].prev_dc);
yading@10:                 i--;
yading@10:             }
yading@10:             return AVERROR(ENOMEM);
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     avctx->pix_fmt     = AV_PIX_FMT_YUV420P;
yading@10: 
yading@10:     init_coders(c);
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static av_cold int mss3_decode_end(AVCodecContext *avctx)
yading@10: {
yading@10:     MSS3Context * const c = avctx->priv_data;
yading@10:     int i;
yading@10: 
yading@10:     av_frame_unref(&c->pic);
yading@10:     for (i = 0; i < 3; i++)
yading@10:         av_freep(&c->dct_coder[i].prev_dc);
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: AVCodec ff_msa1_decoder = {
yading@10:     .name           = "msa1",
yading@10:     .type           = AVMEDIA_TYPE_VIDEO,
yading@10:     .id             = AV_CODEC_ID_MSA1,
yading@10:     .priv_data_size = sizeof(MSS3Context),
yading@10:     .init           = mss3_decode_init,
yading@10:     .close          = mss3_decode_end,
yading@10:     .decode         = mss3_decode_frame,
yading@10:     .capabilities   = CODEC_CAP_DR1,
yading@10:     .long_name      = NULL_IF_CONFIG_SMALL("MS ATC Screen"),
yading@10: };