4xm: remove unneeded check for remaining unused data.
[libav.git] / libavcodec / 4xm.c
1 /*
2 * 4XM codec
3 * Copyright (c) 2003 Michael Niedermayer
4 *
5 * This file is part of Libav.
6 *
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file
24 * 4XM codec.
25 */
26
27 #include "libavutil/intreadwrite.h"
28 #include "avcodec.h"
29 #include "dsputil.h"
30 #include "get_bits.h"
31 #include "bytestream.h"
32
33 //#undef NDEBUG
34 //#include <assert.h>
35
36 #define BLOCK_TYPE_VLC_BITS 5
37 #define ACDC_VLC_BITS 9
38
39 #define CFRAME_BUFFER_COUNT 100
40
41 static const uint8_t block_type_tab[2][4][8][2]={
42 {
43 { //{8,4,2}x{8,4,2}
44 { 0,1}, { 2,2}, { 6,3}, {14,4}, {30,5}, {31,5}, { 0,0}
45 },{ //{8,4}x1
46 { 0,1}, { 0,0}, { 2,2}, { 6,3}, {14,4}, {15,4}, { 0,0}
47 },{ //1x{8,4}
48 { 0,1}, { 2,2}, { 0,0}, { 6,3}, {14,4}, {15,4}, { 0,0}
49 },{ //1x2, 2x1
50 { 0,1}, { 0,0}, { 0,0}, { 2,2}, { 6,3}, {14,4}, {15,4}
51 }
52 },{
53 { //{8,4,2}x{8,4,2}
54 { 1,2}, { 4,3}, { 5,3}, {0,2}, {6,3}, {7,3}, {0,0}
55 },{//{8,4}x1
56 { 1,2}, { 0,0}, { 2,2}, {0,2}, {6,3}, {7,3}, {0,0}
57 },{//1x{8,4}
58 { 1,2}, { 2,2}, { 0,0}, {0,2}, {6,3}, {7,3}, {0,0}
59 },{//1x2, 2x1
60 { 1,2}, { 0,0}, { 0,0}, {0,2}, {2,2}, {6,3}, {7,3}
61 }
62 }
63 };
64
65 static const uint8_t size2index[4][4]={
66 {-1, 3, 1, 1},
67 { 3, 0, 0, 0},
68 { 2, 0, 0, 0},
69 { 2, 0, 0, 0},
70 };
71
72 static const int8_t mv[256][2]={
73 { 0, 0},{ 0, -1},{ -1, 0},{ 1, 0},{ 0, 1},{ -1, -1},{ 1, -1},{ -1, 1},
74 { 1, 1},{ 0, -2},{ -2, 0},{ 2, 0},{ 0, 2},{ -1, -2},{ 1, -2},{ -2, -1},
75 { 2, -1},{ -2, 1},{ 2, 1},{ -1, 2},{ 1, 2},{ -2, -2},{ 2, -2},{ -2, 2},
76 { 2, 2},{ 0, -3},{ -3, 0},{ 3, 0},{ 0, 3},{ -1, -3},{ 1, -3},{ -3, -1},
77 { 3, -1},{ -3, 1},{ 3, 1},{ -1, 3},{ 1, 3},{ -2, -3},{ 2, -3},{ -3, -2},
78 { 3, -2},{ -3, 2},{ 3, 2},{ -2, 3},{ 2, 3},{ 0, -4},{ -4, 0},{ 4, 0},
79 { 0, 4},{ -1, -4},{ 1, -4},{ -4, -1},{ 4, -1},{ 4, 1},{ -1, 4},{ 1, 4},
80 { -3, -3},{ -3, 3},{ 3, 3},{ -2, -4},{ -4, -2},{ 4, -2},{ -4, 2},{ -2, 4},
81 { 2, 4},{ -3, -4},{ 3, -4},{ 4, -3},{ -5, 0},{ -4, 3},{ -3, 4},{ 3, 4},
82 { -1, -5},{ -5, -1},{ -5, 1},{ -1, 5},{ -2, -5},{ 2, -5},{ 5, -2},{ 5, 2},
83 { -4, -4},{ -4, 4},{ -3, -5},{ -5, -3},{ -5, 3},{ 3, 5},{ -6, 0},{ 0, 6},
84 { -6, -1},{ -6, 1},{ 1, 6},{ 2, -6},{ -6, 2},{ 2, 6},{ -5, -4},{ 5, 4},
85 { 4, 5},{ -6, -3},{ 6, 3},{ -7, 0},{ -1, -7},{ 5, -5},{ -7, 1},{ -1, 7},
86 { 4, -6},{ 6, 4},{ -2, -7},{ -7, 2},{ -3, -7},{ 7, -3},{ 3, 7},{ 6, -5},
87 { 0, -8},{ -1, -8},{ -7, -4},{ -8, 1},{ 4, 7},{ 2, -8},{ -2, 8},{ 6, 6},
88 { -8, 3},{ 5, -7},{ -5, 7},{ 8, -4},{ 0, -9},{ -9, -1},{ 1, 9},{ 7, -6},
89 { -7, 6},{ -5, -8},{ -5, 8},{ -9, 3},{ 9, -4},{ 7, -7},{ 8, -6},{ 6, 8},
90 { 10, 1},{-10, 2},{ 9, -5},{ 10, -3},{ -8, -7},{-10, -4},{ 6, -9},{-11, 0},
91 { 11, 1},{-11, -2},{ -2, 11},{ 7, -9},{ -7, 9},{ 10, 6},{ -4, 11},{ 8, -9},
92 { 8, 9},{ 5, 11},{ 7,-10},{ 12, -3},{ 11, 6},{ -9, -9},{ 8, 10},{ 5, 12},
93 {-11, 7},{ 13, 2},{ 6,-12},{ 10, 9},{-11, 8},{ -7, 12},{ 0, 14},{ 14, -2},
94 { -9, 11},{ -6, 13},{-14, -4},{ -5,-14},{ 5, 14},{-15, -1},{-14, -6},{ 3,-15},
95 { 11,-11},{ -7, 14},{ -5, 15},{ 8,-14},{ 15, 6},{ 3, 16},{ 7,-15},{-16, 5},
96 { 0, 17},{-16, -6},{-10, 14},{-16, 7},{ 12, 13},{-16, 8},{-17, 6},{-18, 3},
97 { -7, 17},{ 15, 11},{ 16, 10},{ 2,-19},{ 3,-19},{-11,-16},{-18, 8},{-19, -6},
98 { 2,-20},{-17,-11},{-10,-18},{ 8, 19},{-21, -1},{-20, 7},{ -4, 21},{ 21, 5},
99 { 15, 16},{ 2,-22},{-10,-20},{-22, 5},{ 20,-11},{ -7,-22},{-12, 20},{ 23, -5},
100 { 13,-20},{ 24, -2},{-15, 19},{-11, 22},{ 16, 19},{ 23,-10},{-18,-18},{ -9,-24},
101 { 24,-10},{ -3, 26},{-23, 13},{-18,-20},{ 17, 21},{ -4, 27},{ 27, 6},{ 1,-28},
102 {-11, 26},{-17,-23},{ 7, 28},{ 11,-27},{ 29, 5},{-23,-19},{-28,-11},{-21, 22},
103 {-30, 7},{-17, 26},{-27, 16},{ 13, 29},{ 19,-26},{ 10,-31},{-14,-30},{ 20,-27},
104 {-29, 18},{-16,-31},{-28,-22},{ 21,-30},{-25, 28},{ 26,-29},{ 25,-32},{-32,-32}
105 };
106
107 // this is simply the scaled down elementwise product of the standard jpeg quantizer table and the AAN premul table
108 static const uint8_t dequant_table[64]={
109 16, 15, 13, 19, 24, 31, 28, 17,
110 17, 23, 25, 31, 36, 63, 45, 21,
111 18, 24, 27, 37, 52, 59, 49, 20,
112 16, 28, 34, 40, 60, 80, 51, 20,
113 18, 31, 48, 66, 68, 86, 56, 21,
114 19, 38, 56, 59, 64, 64, 48, 20,
115 27, 48, 55, 55, 56, 51, 35, 15,
116 20, 35, 34, 32, 31, 22, 15, 8,
117 };
118
119 static VLC block_type_vlc[2][4];
120
121
122 typedef struct CFrameBuffer{
123 unsigned int allocated_size;
124 unsigned int size;
125 int id;
126 uint8_t *data;
127 }CFrameBuffer;
128
129 typedef struct FourXContext{
130 AVCodecContext *avctx;
131 DSPContext dsp;
132 AVFrame current_picture, last_picture;
133 GetBitContext pre_gb; ///< ac/dc prefix
134 GetBitContext gb;
135 const uint8_t *bytestream;
136 const uint16_t *wordstream;
137 int mv[256];
138 VLC pre_vlc;
139 int last_dc;
140 DECLARE_ALIGNED(16, DCTELEM, block)[6][64];
141 void *bitstream_buffer;
142 unsigned int bitstream_buffer_size;
143 int version;
144 CFrameBuffer cfrm[CFRAME_BUFFER_COUNT];
145 } FourXContext;
146
147
148 #define FIX_1_082392200 70936
149 #define FIX_1_414213562 92682
150 #define FIX_1_847759065 121095
151 #define FIX_2_613125930 171254
152
153 #define MULTIPLY(var,const) (((var)*(const)) >> 16)
154
155 static void idct(DCTELEM block[64]){
156 int tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
157 int tmp10, tmp11, tmp12, tmp13;
158 int z5, z10, z11, z12, z13;
159 int i;
160 int temp[64];
161
162 for(i=0; i<8; i++){
163 tmp10 = block[8*0 + i] + block[8*4 + i];
164 tmp11 = block[8*0 + i] - block[8*4 + i];
165
166 tmp13 = block[8*2 + i] + block[8*6 + i];
167 tmp12 = MULTIPLY(block[8*2 + i] - block[8*6 + i], FIX_1_414213562) - tmp13;
168
169 tmp0 = tmp10 + tmp13;
170 tmp3 = tmp10 - tmp13;
171 tmp1 = tmp11 + tmp12;
172 tmp2 = tmp11 - tmp12;
173
174 z13 = block[8*5 + i] + block[8*3 + i];
175 z10 = block[8*5 + i] - block[8*3 + i];
176 z11 = block[8*1 + i] + block[8*7 + i];
177 z12 = block[8*1 + i] - block[8*7 + i];
178
179 tmp7 = z11 + z13;
180 tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562);
181
182 z5 = MULTIPLY(z10 + z12, FIX_1_847759065);
183 tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5;
184 tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5;
185
186 tmp6 = tmp12 - tmp7;
187 tmp5 = tmp11 - tmp6;
188 tmp4 = tmp10 + tmp5;
189
190 temp[8*0 + i] = tmp0 + tmp7;
191 temp[8*7 + i] = tmp0 - tmp7;
192 temp[8*1 + i] = tmp1 + tmp6;
193 temp[8*6 + i] = tmp1 - tmp6;
194 temp[8*2 + i] = tmp2 + tmp5;
195 temp[8*5 + i] = tmp2 - tmp5;
196 temp[8*4 + i] = tmp3 + tmp4;
197 temp[8*3 + i] = tmp3 - tmp4;
198 }
199
200 for(i=0; i<8*8; i+=8){
201 tmp10 = temp[0 + i] + temp[4 + i];
202 tmp11 = temp[0 + i] - temp[4 + i];
203
204 tmp13 = temp[2 + i] + temp[6 + i];
205 tmp12 = MULTIPLY(temp[2 + i] - temp[6 + i], FIX_1_414213562) - tmp13;
206
207 tmp0 = tmp10 + tmp13;
208 tmp3 = tmp10 - tmp13;
209 tmp1 = tmp11 + tmp12;
210 tmp2 = tmp11 - tmp12;
211
212 z13 = temp[5 + i] + temp[3 + i];
213 z10 = temp[5 + i] - temp[3 + i];
214 z11 = temp[1 + i] + temp[7 + i];
215 z12 = temp[1 + i] - temp[7 + i];
216
217 tmp7 = z11 + z13;
218 tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562);
219
220 z5 = MULTIPLY(z10 + z12, FIX_1_847759065);
221 tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5;
222 tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5;
223
224 tmp6 = tmp12 - tmp7;
225 tmp5 = tmp11 - tmp6;
226 tmp4 = tmp10 + tmp5;
227
228 block[0 + i] = (tmp0 + tmp7)>>6;
229 block[7 + i] = (tmp0 - tmp7)>>6;
230 block[1 + i] = (tmp1 + tmp6)>>6;
231 block[6 + i] = (tmp1 - tmp6)>>6;
232 block[2 + i] = (tmp2 + tmp5)>>6;
233 block[5 + i] = (tmp2 - tmp5)>>6;
234 block[4 + i] = (tmp3 + tmp4)>>6;
235 block[3 + i] = (tmp3 - tmp4)>>6;
236 }
237 }
238
239 static av_cold void init_vlcs(FourXContext *f){
240 static VLC_TYPE table[8][32][2];
241 int i;
242
243 for(i=0; i<8; i++){
244 block_type_vlc[0][i].table= table[i];
245 block_type_vlc[0][i].table_allocated= 32;
246 init_vlc(&block_type_vlc[0][i], BLOCK_TYPE_VLC_BITS, 7,
247 &block_type_tab[0][i][0][1], 2, 1,
248 &block_type_tab[0][i][0][0], 2, 1, INIT_VLC_USE_NEW_STATIC);
249 }
250 }
251
252 static void init_mv(FourXContext *f){
253 int i;
254
255 for(i=0; i<256; i++){
256 if(f->version>1)
257 f->mv[i] = mv[i][0] + mv[i][1] *f->current_picture.linesize[0]/2;
258 else
259 f->mv[i] = (i&15) - 8 + ((i>>4)-8)*f->current_picture.linesize[0]/2;
260 }
261 }
262
263 #if HAVE_BIGENDIAN
264 #define LE_CENTRIC_MUL(dst, src, scale, dc) \
265 { \
266 unsigned tmpval = AV_RN32(src); \
267 tmpval = (tmpval << 16) | (tmpval >> 16); \
268 tmpval = tmpval * (scale) + (dc); \
269 tmpval = (tmpval << 16) | (tmpval >> 16); \
270 AV_WN32A(dst, tmpval); \
271 }
272 #else
273 #define LE_CENTRIC_MUL(dst, src, scale, dc) \
274 { \
275 unsigned tmpval = AV_RN32(src) * (scale) + (dc); \
276 AV_WN32A(dst, tmpval); \
277 }
278 #endif
279
280 static inline void mcdc(uint16_t *dst, uint16_t *src, int log2w, int h, int stride, int scale, unsigned dc){
281 int i;
282 dc*= 0x10001;
283
284 switch(log2w){
285 case 0:
286 for(i=0; i<h; i++){
287 dst[0] = scale*src[0] + dc;
288 if(scale) src += stride;
289 dst += stride;
290 }
291 break;
292 case 1:
293 for(i=0; i<h; i++){
294 LE_CENTRIC_MUL(dst, src, scale, dc);
295 if(scale) src += stride;
296 dst += stride;
297 }
298 break;
299 case 2:
300 for(i=0; i<h; i++){
301 LE_CENTRIC_MUL(dst, src, scale, dc);
302 LE_CENTRIC_MUL(dst + 2, src + 2, scale, dc);
303 if(scale) src += stride;
304 dst += stride;
305 }
306 break;
307 case 3:
308 for(i=0; i<h; i++){
309 LE_CENTRIC_MUL(dst, src, scale, dc);
310 LE_CENTRIC_MUL(dst + 2, src + 2, scale, dc);
311 LE_CENTRIC_MUL(dst + 4, src + 4, scale, dc);
312 LE_CENTRIC_MUL(dst + 6, src + 6, scale, dc);
313 if(scale) src += stride;
314 dst += stride;
315 }
316 break;
317 default: assert(0);
318 }
319 }
320
321 static void decode_p_block(FourXContext *f, uint16_t *dst, uint16_t *src, int log2w, int log2h, int stride){
322 const int index= size2index[log2h][log2w];
323 const int h= 1<<log2h;
324 int code= get_vlc2(&f->gb, block_type_vlc[1-(f->version>1)][index].table, BLOCK_TYPE_VLC_BITS, 1);
325 uint16_t *start= (uint16_t*)f->last_picture.data[0];
326 uint16_t *end= start + stride*(f->avctx->height-h+1) - (1<<log2w);
327
328 assert(code>=0 && code<=6);
329
330 if(code == 0){
331 src += f->mv[ *f->bytestream++ ];
332 if(start > src || src > end){
333 av_log(f->avctx, AV_LOG_ERROR, "mv out of pic\n");
334 return;
335 }
336 mcdc(dst, src, log2w, h, stride, 1, 0);
337 }else if(code == 1){
338 log2h--;
339 decode_p_block(f, dst , src , log2w, log2h, stride);
340 decode_p_block(f, dst + (stride<<log2h), src + (stride<<log2h), log2w, log2h, stride);
341 }else if(code == 2){
342 log2w--;
343 decode_p_block(f, dst , src , log2w, log2h, stride);
344 decode_p_block(f, dst + (1<<log2w), src + (1<<log2w), log2w, log2h, stride);
345 }else if(code == 3 && f->version<2){
346 mcdc(dst, src, log2w, h, stride, 1, 0);
347 }else if(code == 4){
348 src += f->mv[ *f->bytestream++ ];
349 if(start > src || src > end){
350 av_log(f->avctx, AV_LOG_ERROR, "mv out of pic\n");
351 return;
352 }
353 mcdc(dst, src, log2w, h, stride, 1, av_le2ne16(*f->wordstream++));
354 }else if(code == 5){
355 mcdc(dst, src, log2w, h, stride, 0, av_le2ne16(*f->wordstream++));
356 }else if(code == 6){
357 if(log2w){
358 dst[0] = av_le2ne16(*f->wordstream++);
359 dst[1] = av_le2ne16(*f->wordstream++);
360 }else{
361 dst[0 ] = av_le2ne16(*f->wordstream++);
362 dst[stride] = av_le2ne16(*f->wordstream++);
363 }
364 }
365 }
366
367 static int decode_p_frame(FourXContext *f, const uint8_t *buf, int length){
368 int x, y;
369 const int width= f->avctx->width;
370 const int height= f->avctx->height;
371 uint16_t *src= (uint16_t*)f->last_picture.data[0];
372 uint16_t *dst= (uint16_t*)f->current_picture.data[0];
373 const int stride= f->current_picture.linesize[0]>>1;
374 unsigned int bitstream_size, bytestream_size, wordstream_size, extra;
375
376 if(f->version>1){
377 extra=20;
378 bitstream_size= AV_RL32(buf+8);
379 wordstream_size= AV_RL32(buf+12);
380 bytestream_size= AV_RL32(buf+16);
381 }else{
382 extra=0;
383 bitstream_size = AV_RL16(buf-4);
384 wordstream_size= AV_RL16(buf-2);
385 bytestream_size= FFMAX(length - bitstream_size - wordstream_size, 0);
386 }
387
388 if(bitstream_size+ bytestream_size+ wordstream_size + extra != length
389 || bitstream_size > (1<<26)
390 || bytestream_size > (1<<26)
391 || wordstream_size > (1<<26)
392 ){
393 av_log(f->avctx, AV_LOG_ERROR, "lengths %d %d %d %d\n", bitstream_size, bytestream_size, wordstream_size,
394 bitstream_size+ bytestream_size+ wordstream_size - length);
395 return -1;
396 }
397
398 av_fast_malloc(&f->bitstream_buffer, &f->bitstream_buffer_size, bitstream_size + FF_INPUT_BUFFER_PADDING_SIZE);
399 if (!f->bitstream_buffer)
400 return AVERROR(ENOMEM);
401 f->dsp.bswap_buf(f->bitstream_buffer, (const uint32_t*)(buf + extra), bitstream_size/4);
402 memset((uint8_t*)f->bitstream_buffer + bitstream_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);
403 init_get_bits(&f->gb, f->bitstream_buffer, 8*bitstream_size);
404
405 f->wordstream= (const uint16_t*)(buf + extra + bitstream_size);
406 f->bytestream= buf + extra + bitstream_size + wordstream_size;
407
408 init_mv(f);
409
410 for(y=0; y<height; y+=8){
411 for(x=0; x<width; x+=8){
412 decode_p_block(f, dst + x, src + x, 3, 3, stride);
413 }
414 src += 8*stride;
415 dst += 8*stride;
416 }
417
418 return 0;
419 }
420
421 /**
422 * decode block and dequantize.
423 * Note this is almost identical to MJPEG.
424 */
425 static int decode_i_block(FourXContext *f, DCTELEM *block){
426 int code, i, j, level, val;
427
428 /* DC coef */
429 val = get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3);
430 if (val>>4){
431 av_log(f->avctx, AV_LOG_ERROR, "error dc run != 0\n");
432 }
433
434 if(val)
435 val = get_xbits(&f->gb, val);
436
437 val = val * dequant_table[0] + f->last_dc;
438 f->last_dc =
439 block[0] = val;
440 /* AC coefs */
441 i = 1;
442 for(;;) {
443 code = get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3);
444
445 /* EOB */
446 if (code == 0)
447 break;
448 if (code == 0xf0) {
449 i += 16;
450 } else {
451 level = get_xbits(&f->gb, code & 0xf);
452 i += code >> 4;
453 if (i >= 64) {
454 av_log(f->avctx, AV_LOG_ERROR, "run %d oveflow\n", i);
455 return 0;
456 }
457
458 j= ff_zigzag_direct[i];
459 block[j] = level * dequant_table[j];
460 i++;
461 if (i >= 64)
462 break;
463 }
464 }
465
466 return 0;
467 }
468
469 static inline void idct_put(FourXContext *f, int x, int y){
470 DCTELEM (*block)[64]= f->block;
471 int stride= f->current_picture.linesize[0]>>1;
472 int i;
473 uint16_t *dst = ((uint16_t*)f->current_picture.data[0]) + y * stride + x;
474
475 for(i=0; i<4; i++){
476 block[i][0] += 0x80*8*8;
477 idct(block[i]);
478 }
479
480 if(!(f->avctx->flags&CODEC_FLAG_GRAY)){
481 for(i=4; i<6; i++) idct(block[i]);
482 }
483
484 /* Note transform is:
485 y= ( 1b + 4g + 2r)/14
486 cb=( 3b - 2g - 1r)/14
487 cr=(-1b - 4g + 5r)/14
488 */
489 for(y=0; y<8; y++){
490 for(x=0; x<8; x++){
491 DCTELEM *temp= block[(x>>2) + 2*(y>>2)] + 2*(x&3) + 2*8*(y&3); //FIXME optimize
492 int cb= block[4][x + 8*y];
493 int cr= block[5][x + 8*y];
494 int cg= (cb + cr)>>1;
495 int y;
496
497 cb+=cb;
498
499 y = temp[0];
500 dst[0 ]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);
501 y = temp[1];
502 dst[1 ]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);
503 y = temp[8];
504 dst[ stride]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);
505 y = temp[9];
506 dst[1+stride]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);
507 dst += 2;
508 }
509 dst += 2*stride - 2*8;
510 }
511 }
512
513 static int decode_i_mb(FourXContext *f){
514 int i;
515
516 f->dsp.clear_blocks(f->block[0]);
517
518 for(i=0; i<6; i++){
519 if(decode_i_block(f, f->block[i]) < 0)
520 return -1;
521 }
522
523 return 0;
524 }
525
526 static const uint8_t *read_huffman_tables(FourXContext *f, const uint8_t * const buf){
527 int frequency[512];
528 uint8_t flag[512];
529 int up[512];
530 uint8_t len_tab[257];
531 int bits_tab[257];
532 int start, end;
533 const uint8_t *ptr= buf;
534 int j;
535
536 memset(frequency, 0, sizeof(frequency));
537 memset(up, -1, sizeof(up));
538
539 start= *ptr++;
540 end= *ptr++;
541 for(;;){
542 int i;
543
544 for(i=start; i<=end; i++){
545 frequency[i]= *ptr++;
546 }
547 start= *ptr++;
548 if(start==0) break;
549
550 end= *ptr++;
551 }
552 frequency[256]=1;
553
554 while((ptr - buf)&3) ptr++; // 4byte align
555
556 for(j=257; j<512; j++){
557 int min_freq[2]= {256*256, 256*256};
558 int smallest[2]= {0, 0};
559 int i;
560 for(i=0; i<j; i++){
561 if(frequency[i] == 0) continue;
562 if(frequency[i] < min_freq[1]){
563 if(frequency[i] < min_freq[0]){
564 min_freq[1]= min_freq[0]; smallest[1]= smallest[0];
565 min_freq[0]= frequency[i];smallest[0]= i;
566 }else{
567 min_freq[1]= frequency[i];smallest[1]= i;
568 }
569 }
570 }
571 if(min_freq[1] == 256*256) break;
572
573 frequency[j]= min_freq[0] + min_freq[1];
574 flag[ smallest[0] ]= 0;
575 flag[ smallest[1] ]= 1;
576 up[ smallest[0] ]=
577 up[ smallest[1] ]= j;
578 frequency[ smallest[0] ]= frequency[ smallest[1] ]= 0;
579 }
580
581 for(j=0; j<257; j++){
582 int node;
583 int len=0;
584 int bits=0;
585
586 for(node= j; up[node] != -1; node= up[node]){
587 bits += flag[node]<<len;
588 len++;
589 if(len > 31) av_log(f->avctx, AV_LOG_ERROR, "vlc length overflow\n"); //can this happen at all ?
590 }
591
592 bits_tab[j]= bits;
593 len_tab[j]= len;
594 }
595
596 if (init_vlc(&f->pre_vlc, ACDC_VLC_BITS, 257,
597 len_tab , 1, 1,
598 bits_tab, 4, 4, 0))
599 return NULL;
600
601 return ptr;
602 }
603
604 static int mix(int c0, int c1){
605 int blue = 2*(c0&0x001F) + (c1&0x001F);
606 int green= (2*(c0&0x03E0) + (c1&0x03E0))>>5;
607 int red = 2*(c0>>10) + (c1>>10);
608 return red/3*1024 + green/3*32 + blue/3;
609 }
610
611 static int decode_i2_frame(FourXContext *f, const uint8_t *buf, int length){
612 int x, y, x2, y2;
613 const int width= f->avctx->width;
614 const int height= f->avctx->height;
615 uint16_t *dst= (uint16_t*)f->current_picture.data[0];
616 const int stride= f->current_picture.linesize[0]>>1;
617
618 for(y=0; y<height; y+=16){
619 for(x=0; x<width; x+=16){
620 unsigned int color[4], bits;
621 memset(color, 0, sizeof(color));
622 //warning following is purely guessed ...
623 color[0]= bytestream_get_le16(&buf);
624 color[1]= bytestream_get_le16(&buf);
625
626 if(color[0]&0x8000) av_log(NULL, AV_LOG_ERROR, "unk bit 1\n");
627 if(color[1]&0x8000) av_log(NULL, AV_LOG_ERROR, "unk bit 2\n");
628
629 color[2]= mix(color[0], color[1]);
630 color[3]= mix(color[1], color[0]);
631
632 bits= bytestream_get_le32(&buf);
633 for(y2=0; y2<16; y2++){
634 for(x2=0; x2<16; x2++){
635 int index= 2*(x2>>2) + 8*(y2>>2);
636 dst[y2*stride+x2]= color[(bits>>index)&3];
637 }
638 }
639 dst+=16;
640 }
641 dst += 16*stride - width;
642 }
643
644 return 0;
645 }
646
647 static int decode_i_frame(FourXContext *f, const uint8_t *buf, int length){
648 int x, y;
649 const int width= f->avctx->width;
650 const int height= f->avctx->height;
651 uint16_t *dst= (uint16_t*)f->current_picture.data[0];
652 const int stride= f->current_picture.linesize[0]>>1;
653 const unsigned int bitstream_size= AV_RL32(buf);
654 const int token_count av_unused = AV_RL32(buf + bitstream_size + 8);
655 unsigned int prestream_size= 4*AV_RL32(buf + bitstream_size + 4);
656 const uint8_t *prestream= buf + bitstream_size + 12;
657
658 if(prestream_size + bitstream_size + 12 != length
659 || bitstream_size > (1<<26)
660 || prestream_size > (1<<26)){
661 av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %d %d\n", prestream_size, bitstream_size, length);
662 return -1;
663 }
664
665 prestream= read_huffman_tables(f, prestream);
666
667 init_get_bits(&f->gb, buf + 4, 8*bitstream_size);
668
669 prestream_size= length + buf - prestream;
670
671 av_fast_malloc(&f->bitstream_buffer, &f->bitstream_buffer_size, prestream_size + FF_INPUT_BUFFER_PADDING_SIZE);
672 if (!f->bitstream_buffer)
673 return AVERROR(ENOMEM);
674 f->dsp.bswap_buf(f->bitstream_buffer, (const uint32_t*)prestream, prestream_size/4);
675 memset((uint8_t*)f->bitstream_buffer + prestream_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);
676 init_get_bits(&f->pre_gb, f->bitstream_buffer, 8*prestream_size);
677
678 f->last_dc= 0*128*8*8;
679
680 for(y=0; y<height; y+=16){
681 for(x=0; x<width; x+=16){
682 if(decode_i_mb(f) < 0)
683 return -1;
684
685 idct_put(f, x, y);
686 }
687 dst += 16*stride;
688 }
689
690 if(get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3) != 256)
691 av_log(f->avctx, AV_LOG_ERROR, "end mismatch\n");
692
693 return 0;
694 }
695
696 static int decode_frame(AVCodecContext *avctx,
697 void *data, int *data_size,
698 AVPacket *avpkt)
699 {
700 const uint8_t *buf = avpkt->data;
701 int buf_size = avpkt->size;
702 FourXContext * const f = avctx->priv_data;
703 AVFrame *picture = data;
704 AVFrame *p, temp;
705 int i, frame_4cc, frame_size;
706
707 frame_4cc= AV_RL32(buf);
708 if(buf_size != AV_RL32(buf+4)+8 || buf_size < 20){
709 av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %d\n", buf_size, AV_RL32(buf+4));
710 }
711
712 if(frame_4cc == AV_RL32("cfrm")){
713 int free_index=-1;
714 const int data_size= buf_size - 20;
715 const int id= AV_RL32(buf+12);
716 const int whole_size= AV_RL32(buf+16);
717 CFrameBuffer *cfrm;
718
719 for(i=0; i<CFRAME_BUFFER_COUNT; i++){
720 if(f->cfrm[i].id && f->cfrm[i].id < avctx->frame_number)
721 av_log(f->avctx, AV_LOG_ERROR, "lost c frame %d\n", f->cfrm[i].id);
722 }
723
724 for(i=0; i<CFRAME_BUFFER_COUNT; i++){
725 if(f->cfrm[i].id == id) break;
726 if(f->cfrm[i].size == 0 ) free_index= i;
727 }
728
729 if(i>=CFRAME_BUFFER_COUNT){
730 i= free_index;
731 f->cfrm[i].id= id;
732 }
733 cfrm= &f->cfrm[i];
734
735 cfrm->data= av_fast_realloc(cfrm->data, &cfrm->allocated_size, cfrm->size + data_size + FF_INPUT_BUFFER_PADDING_SIZE);
736 if(!cfrm->data){ //explicit check needed as memcpy below might not catch a NULL
737 av_log(f->avctx, AV_LOG_ERROR, "realloc falure");
738 return -1;
739 }
740
741 memcpy(cfrm->data + cfrm->size, buf+20, data_size);
742 cfrm->size += data_size;
743
744 if(cfrm->size >= whole_size){
745 buf= cfrm->data;
746 frame_size= cfrm->size;
747
748 if(id != avctx->frame_number){
749 av_log(f->avctx, AV_LOG_ERROR, "cframe id mismatch %d %d\n", id, avctx->frame_number);
750 }
751
752 cfrm->size= cfrm->id= 0;
753 frame_4cc= AV_RL32("pfrm");
754 }else
755 return buf_size;
756 }else{
757 buf= buf + 12;
758 frame_size= buf_size - 12;
759 }
760
761 temp= f->current_picture;
762 f->current_picture= f->last_picture;
763 f->last_picture= temp;
764
765 p= &f->current_picture;
766 avctx->coded_frame= p;
767
768 avctx->flags |= CODEC_FLAG_EMU_EDGE; // alternatively we would have to use our own buffer management
769
770 if(p->data[0])
771 avctx->release_buffer(avctx, p);
772
773 p->reference= 1;
774 if(avctx->get_buffer(avctx, p) < 0){
775 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
776 return -1;
777 }
778
779 if(frame_4cc == AV_RL32("ifr2")){
780 p->pict_type= AV_PICTURE_TYPE_I;
781 if(decode_i2_frame(f, buf-4, frame_size) < 0)
782 return -1;
783 }else if(frame_4cc == AV_RL32("ifrm")){
784 p->pict_type= AV_PICTURE_TYPE_I;
785 if(decode_i_frame(f, buf, frame_size) < 0)
786 return -1;
787 }else if(frame_4cc == AV_RL32("pfrm") || frame_4cc == AV_RL32("pfr2")){
788 if(!f->last_picture.data[0]){
789 f->last_picture.reference= 1;
790 if(avctx->get_buffer(avctx, &f->last_picture) < 0){
791 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
792 return -1;
793 }
794 }
795
796 p->pict_type= AV_PICTURE_TYPE_P;
797 if(decode_p_frame(f, buf, frame_size) < 0)
798 return -1;
799 }else if(frame_4cc == AV_RL32("snd_")){
800 av_log(avctx, AV_LOG_ERROR, "ignoring snd_ chunk length:%d\n", buf_size);
801 }else{
802 av_log(avctx, AV_LOG_ERROR, "ignoring unknown chunk length:%d\n", buf_size);
803 }
804
805 p->key_frame= p->pict_type == AV_PICTURE_TYPE_I;
806
807 *picture= *p;
808 *data_size = sizeof(AVPicture);
809
810 emms_c();
811
812 return buf_size;
813 }
814
815
816 static av_cold void common_init(AVCodecContext *avctx){
817 FourXContext * const f = avctx->priv_data;
818
819 dsputil_init(&f->dsp, avctx);
820
821 f->avctx= avctx;
822 }
823
824 static av_cold int decode_init(AVCodecContext *avctx){
825 FourXContext * const f = avctx->priv_data;
826
827 if(avctx->extradata_size != 4 || !avctx->extradata) {
828 av_log(avctx, AV_LOG_ERROR, "extradata wrong or missing\n");
829 return 1;
830 }
831
832 f->version= AV_RL32(avctx->extradata)>>16;
833 common_init(avctx);
834 init_vlcs(f);
835
836 if(f->version>2) avctx->pix_fmt= PIX_FMT_RGB565;
837 else avctx->pix_fmt= PIX_FMT_BGR555;
838
839 return 0;
840 }
841
842
843 static av_cold int decode_end(AVCodecContext *avctx){
844 FourXContext * const f = avctx->priv_data;
845 int i;
846
847 av_freep(&f->bitstream_buffer);
848 f->bitstream_buffer_size=0;
849 for(i=0; i<CFRAME_BUFFER_COUNT; i++){
850 av_freep(&f->cfrm[i].data);
851 f->cfrm[i].allocated_size= 0;
852 }
853 free_vlc(&f->pre_vlc);
854 if(f->current_picture.data[0])
855 avctx->release_buffer(avctx, &f->current_picture);
856 if(f->last_picture.data[0])
857 avctx->release_buffer(avctx, &f->last_picture);
858
859 return 0;
860 }
861
862 AVCodec ff_fourxm_decoder = {
863 .name = "4xm",
864 .type = AVMEDIA_TYPE_VIDEO,
865 .id = CODEC_ID_4XM,
866 .priv_data_size = sizeof(FourXContext),
867 .init = decode_init,
868 .close = decode_end,
869 .decode = decode_frame,
870 .capabilities = CODEC_CAP_DR1,
871 .long_name = NULL_IF_CONFIG_SMALL("4X Movie"),
872 };
873