Move/add COSTABLE/SINTABLE macros to dsputil to add extern definitions
[FFMpeg-mirror/lagarith.git] / libavcodec / 4xm.c
blobd2172ba256f996477725b8eb0ec5085a2f713f6b
1 /*
2 * 4XM codec
3 * Copyright (c) 2003 Michael Niedermayer
5 * This file is part of FFmpeg.
7 * FFmpeg 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.
12 * FFmpeg 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.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 /**
23 * @file libavcodec/4xm.c
24 * 4XM codec.
27 #include "libavutil/intreadwrite.h"
28 #include "avcodec.h"
29 #include "dsputil.h"
30 #include "get_bits.h"
31 #include "bytestream.h"
33 //#undef NDEBUG
34 //#include <assert.h>
36 #define BLOCK_TYPE_VLC_BITS 5
37 #define ACDC_VLC_BITS 9
39 #define CFRAME_BUFFER_COUNT 100
41 static const uint8_t block_type_tab[2][4][8][2]={
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}
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}
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},
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}
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,
119 static VLC block_type_vlc[2][4];
122 typedef struct CFrameBuffer{
123 unsigned int allocated_size;
124 unsigned int size;
125 int id;
126 uint8_t *data;
127 }CFrameBuffer;
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;
148 #define FIX_1_082392200 70936
149 #define FIX_1_414213562 92682
150 #define FIX_1_847759065 121095
151 #define FIX_2_613125930 171254
153 #define MULTIPLY(var,const) (((var)*(const)) >> 16)
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];
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];
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;
169 tmp0 = tmp10 + tmp13;
170 tmp3 = tmp10 - tmp13;
171 tmp1 = tmp11 + tmp12;
172 tmp2 = tmp11 - tmp12;
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];
179 tmp7 = z11 + z13;
180 tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562);
182 z5 = MULTIPLY(z10 + z12, FIX_1_847759065);
183 tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5;
184 tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5;
186 tmp6 = tmp12 - tmp7;
187 tmp5 = tmp11 - tmp6;
188 tmp4 = tmp10 + tmp5;
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;
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];
204 tmp13 = temp[2 + i] + temp[6 + i];
205 tmp12 = MULTIPLY(temp[2 + i] - temp[6 + i], FIX_1_414213562) - tmp13;
207 tmp0 = tmp10 + tmp13;
208 tmp3 = tmp10 - tmp13;
209 tmp1 = tmp11 + tmp12;
210 tmp2 = tmp11 - tmp12;
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];
217 tmp7 = z11 + z13;
218 tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562);
220 z5 = MULTIPLY(z10 + z12, FIX_1_847759065);
221 tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5;
222 tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5;
224 tmp6 = tmp12 - tmp7;
225 tmp5 = tmp11 - tmp6;
226 tmp4 = tmp10 + tmp5;
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;
239 static av_cold void init_vlcs(FourXContext *f){
240 static VLC_TYPE table[8][32][2];
241 int i;
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);
252 static void init_mv(FourXContext *f){
253 int i;
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;
263 static inline void mcdc(uint16_t *dst, uint16_t *src, int log2w, int h, int stride, int scale, int dc){
264 int i;
265 dc*= 0x10001;
267 switch(log2w){
268 case 0:
269 for(i=0; i<h; i++){
270 dst[0] = scale*src[0] + dc;
271 if(scale) src += stride;
272 dst += stride;
274 break;
275 case 1:
276 for(i=0; i<h; i++){
277 ((uint32_t*)dst)[0] = scale*((uint32_t*)src)[0] + dc;
278 if(scale) src += stride;
279 dst += stride;
281 break;
282 case 2:
283 for(i=0; i<h; i++){
284 ((uint32_t*)dst)[0] = scale*((uint32_t*)src)[0] + dc;
285 ((uint32_t*)dst)[1] = scale*((uint32_t*)src)[1] + dc;
286 if(scale) src += stride;
287 dst += stride;
289 break;
290 case 3:
291 for(i=0; i<h; i++){
292 ((uint32_t*)dst)[0] = scale*((uint32_t*)src)[0] + dc;
293 ((uint32_t*)dst)[1] = scale*((uint32_t*)src)[1] + dc;
294 ((uint32_t*)dst)[2] = scale*((uint32_t*)src)[2] + dc;
295 ((uint32_t*)dst)[3] = scale*((uint32_t*)src)[3] + dc;
296 if(scale) src += stride;
297 dst += stride;
299 break;
300 default: assert(0);
304 static void decode_p_block(FourXContext *f, uint16_t *dst, uint16_t *src, int log2w, int log2h, int stride){
305 const int index= size2index[log2h][log2w];
306 const int h= 1<<log2h;
307 int code= get_vlc2(&f->gb, block_type_vlc[1-(f->version>1)][index].table, BLOCK_TYPE_VLC_BITS, 1);
308 uint16_t *start= (uint16_t*)f->last_picture.data[0];
309 uint16_t *end= start + stride*(f->avctx->height-h+1) - (1<<log2w);
311 assert(code>=0 && code<=6);
313 if(code == 0){
314 src += f->mv[ *f->bytestream++ ];
315 if(start > src || src > end){
316 av_log(f->avctx, AV_LOG_ERROR, "mv out of pic\n");
317 return;
319 mcdc(dst, src, log2w, h, stride, 1, 0);
320 }else if(code == 1){
321 log2h--;
322 decode_p_block(f, dst , src , log2w, log2h, stride);
323 decode_p_block(f, dst + (stride<<log2h), src + (stride<<log2h), log2w, log2h, stride);
324 }else if(code == 2){
325 log2w--;
326 decode_p_block(f, dst , src , log2w, log2h, stride);
327 decode_p_block(f, dst + (1<<log2w), src + (1<<log2w), log2w, log2h, stride);
328 }else if(code == 3 && f->version<2){
329 mcdc(dst, src, log2w, h, stride, 1, 0);
330 }else if(code == 4){
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;
336 mcdc(dst, src, log2w, h, stride, 1, le2me_16(*f->wordstream++));
337 }else if(code == 5){
338 mcdc(dst, src, log2w, h, stride, 0, le2me_16(*f->wordstream++));
339 }else if(code == 6){
340 if(log2w){
341 dst[0] = le2me_16(*f->wordstream++);
342 dst[1] = le2me_16(*f->wordstream++);
343 }else{
344 dst[0 ] = le2me_16(*f->wordstream++);
345 dst[stride] = le2me_16(*f->wordstream++);
350 static int decode_p_frame(FourXContext *f, const uint8_t *buf, int length){
351 int x, y;
352 const int width= f->avctx->width;
353 const int height= f->avctx->height;
354 uint16_t *src= (uint16_t*)f->last_picture.data[0];
355 uint16_t *dst= (uint16_t*)f->current_picture.data[0];
356 const int stride= f->current_picture.linesize[0]>>1;
357 unsigned int bitstream_size, bytestream_size, wordstream_size, extra;
359 if(f->version>1){
360 extra=20;
361 bitstream_size= AV_RL32(buf+8);
362 wordstream_size= AV_RL32(buf+12);
363 bytestream_size= AV_RL32(buf+16);
364 }else{
365 extra=0;
366 bitstream_size = AV_RL16(buf-4);
367 wordstream_size= AV_RL16(buf-2);
368 bytestream_size= FFMAX(length - bitstream_size - wordstream_size, 0);
371 if(bitstream_size+ bytestream_size+ wordstream_size + extra != length
372 || bitstream_size > (1<<26)
373 || bytestream_size > (1<<26)
374 || wordstream_size > (1<<26)
376 av_log(f->avctx, AV_LOG_ERROR, "lengths %d %d %d %d\n", bitstream_size, bytestream_size, wordstream_size,
377 bitstream_size+ bytestream_size+ wordstream_size - length);
378 return -1;
381 av_fast_malloc(&f->bitstream_buffer, &f->bitstream_buffer_size, bitstream_size + FF_INPUT_BUFFER_PADDING_SIZE);
382 if (!f->bitstream_buffer)
383 return AVERROR(ENOMEM);
384 f->dsp.bswap_buf(f->bitstream_buffer, (const uint32_t*)(buf + extra), bitstream_size/4);
385 init_get_bits(&f->gb, f->bitstream_buffer, 8*bitstream_size);
387 f->wordstream= (const uint16_t*)(buf + extra + bitstream_size);
388 f->bytestream= buf + extra + bitstream_size + wordstream_size;
390 init_mv(f);
392 for(y=0; y<height; y+=8){
393 for(x=0; x<width; x+=8){
394 decode_p_block(f, dst + x, src + x, 3, 3, stride);
396 src += 8*stride;
397 dst += 8*stride;
400 if( bitstream_size != (get_bits_count(&f->gb)+31)/32*4
401 || (((const char*)f->wordstream - (const char*)buf + 2)&~2) != extra + bitstream_size + wordstream_size
402 || (((const char*)f->bytestream - (const char*)buf + 3)&~3) != extra + bitstream_size + wordstream_size + bytestream_size)
403 av_log(f->avctx, AV_LOG_ERROR, " %d %td %td bytes left\n",
404 bitstream_size - (get_bits_count(&f->gb)+31)/32*4,
405 -(((const char*)f->bytestream - (const char*)buf + 3)&~3) + (extra + bitstream_size + wordstream_size + bytestream_size),
406 -(((const char*)f->wordstream - (const char*)buf + 2)&~2) + (extra + bitstream_size + wordstream_size)
409 return 0;
413 * decode block and dequantize.
414 * Note this is almost identical to MJPEG.
416 static int decode_i_block(FourXContext *f, DCTELEM *block){
417 int code, i, j, level, val;
419 /* DC coef */
420 val = get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3);
421 if (val>>4){
422 av_log(f->avctx, AV_LOG_ERROR, "error dc run != 0\n");
425 if(val)
426 val = get_xbits(&f->gb, val);
428 val = val * dequant_table[0] + f->last_dc;
429 f->last_dc =
430 block[0] = val;
431 /* AC coefs */
432 i = 1;
433 for(;;) {
434 code = get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3);
436 /* EOB */
437 if (code == 0)
438 break;
439 if (code == 0xf0) {
440 i += 16;
441 } else {
442 level = get_xbits(&f->gb, code & 0xf);
443 i += code >> 4;
444 if (i >= 64) {
445 av_log(f->avctx, AV_LOG_ERROR, "run %d oveflow\n", i);
446 return 0;
449 j= ff_zigzag_direct[i];
450 block[j] = level * dequant_table[j];
451 i++;
452 if (i >= 64)
453 break;
457 return 0;
460 static inline void idct_put(FourXContext *f, int x, int y){
461 DCTELEM (*block)[64]= f->block;
462 int stride= f->current_picture.linesize[0]>>1;
463 int i;
464 uint16_t *dst = ((uint16_t*)f->current_picture.data[0]) + y * stride + x;
466 for(i=0; i<4; i++){
467 block[i][0] += 0x80*8*8;
468 idct(block[i]);
471 if(!(f->avctx->flags&CODEC_FLAG_GRAY)){
472 for(i=4; i<6; i++) idct(block[i]);
475 /* Note transform is:
476 y= ( 1b + 4g + 2r)/14
477 cb=( 3b - 2g - 1r)/14
478 cr=(-1b - 4g + 5r)/14
480 for(y=0; y<8; y++){
481 for(x=0; x<8; x++){
482 DCTELEM *temp= block[(x>>2) + 2*(y>>2)] + 2*(x&3) + 2*8*(y&3); //FIXME optimize
483 int cb= block[4][x + 8*y];
484 int cr= block[5][x + 8*y];
485 int cg= (cb + cr)>>1;
486 int y;
488 cb+=cb;
490 y = temp[0];
491 dst[0 ]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);
492 y = temp[1];
493 dst[1 ]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);
494 y = temp[8];
495 dst[ stride]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);
496 y = temp[9];
497 dst[1+stride]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);
498 dst += 2;
500 dst += 2*stride - 2*8;
504 static int decode_i_mb(FourXContext *f){
505 int i;
507 f->dsp.clear_blocks(f->block[0]);
509 for(i=0; i<6; i++){
510 if(decode_i_block(f, f->block[i]) < 0)
511 return -1;
514 return 0;
517 static const uint8_t *read_huffman_tables(FourXContext *f, const uint8_t * const buf){
518 int frequency[512];
519 uint8_t flag[512];
520 int up[512];
521 uint8_t len_tab[257];
522 int bits_tab[257];
523 int start, end;
524 const uint8_t *ptr= buf;
525 int j;
527 memset(frequency, 0, sizeof(frequency));
528 memset(up, -1, sizeof(up));
530 start= *ptr++;
531 end= *ptr++;
532 for(;;){
533 int i;
535 for(i=start; i<=end; i++){
536 frequency[i]= *ptr++;
538 start= *ptr++;
539 if(start==0) break;
541 end= *ptr++;
543 frequency[256]=1;
545 while((ptr - buf)&3) ptr++; // 4byte align
547 for(j=257; j<512; j++){
548 int min_freq[2]= {256*256, 256*256};
549 int smallest[2]= {0, 0};
550 int i;
551 for(i=0; i<j; i++){
552 if(frequency[i] == 0) continue;
553 if(frequency[i] < min_freq[1]){
554 if(frequency[i] < min_freq[0]){
555 min_freq[1]= min_freq[0]; smallest[1]= smallest[0];
556 min_freq[0]= frequency[i];smallest[0]= i;
557 }else{
558 min_freq[1]= frequency[i];smallest[1]= i;
562 if(min_freq[1] == 256*256) break;
564 frequency[j]= min_freq[0] + min_freq[1];
565 flag[ smallest[0] ]= 0;
566 flag[ smallest[1] ]= 1;
567 up[ smallest[0] ]=
568 up[ smallest[1] ]= j;
569 frequency[ smallest[0] ]= frequency[ smallest[1] ]= 0;
572 for(j=0; j<257; j++){
573 int node;
574 int len=0;
575 int bits=0;
577 for(node= j; up[node] != -1; node= up[node]){
578 bits += flag[node]<<len;
579 len++;
580 if(len > 31) av_log(f->avctx, AV_LOG_ERROR, "vlc length overflow\n"); //can this happen at all ?
583 bits_tab[j]= bits;
584 len_tab[j]= len;
587 init_vlc(&f->pre_vlc, ACDC_VLC_BITS, 257,
588 len_tab , 1, 1,
589 bits_tab, 4, 4, 0);
591 return ptr;
594 static int mix(int c0, int c1){
595 int blue = 2*(c0&0x001F) + (c1&0x001F);
596 int green= (2*(c0&0x03E0) + (c1&0x03E0))>>5;
597 int red = 2*(c0>>10) + (c1>>10);
598 return red/3*1024 + green/3*32 + blue/3;
601 static int decode_i2_frame(FourXContext *f, const uint8_t *buf, int length){
602 int x, y, x2, y2;
603 const int width= f->avctx->width;
604 const int height= f->avctx->height;
605 uint16_t *dst= (uint16_t*)f->current_picture.data[0];
606 const int stride= f->current_picture.linesize[0]>>1;
608 for(y=0; y<height; y+=16){
609 for(x=0; x<width; x+=16){
610 unsigned int color[4], bits;
611 memset(color, 0, sizeof(color));
612 //warning following is purely guessed ...
613 color[0]= bytestream_get_le16(&buf);
614 color[1]= bytestream_get_le16(&buf);
616 if(color[0]&0x8000) av_log(NULL, AV_LOG_ERROR, "unk bit 1\n");
617 if(color[1]&0x8000) av_log(NULL, AV_LOG_ERROR, "unk bit 2\n");
619 color[2]= mix(color[0], color[1]);
620 color[3]= mix(color[1], color[0]);
622 bits= bytestream_get_le32(&buf);
623 for(y2=0; y2<16; y2++){
624 for(x2=0; x2<16; x2++){
625 int index= 2*(x2>>2) + 8*(y2>>2);
626 dst[y2*stride+x2]= color[(bits>>index)&3];
629 dst+=16;
631 dst += 16*stride - width;
634 return 0;
637 static int decode_i_frame(FourXContext *f, const uint8_t *buf, int length){
638 int x, y;
639 const int width= f->avctx->width;
640 const int height= f->avctx->height;
641 uint16_t *dst= (uint16_t*)f->current_picture.data[0];
642 const int stride= f->current_picture.linesize[0]>>1;
643 const unsigned int bitstream_size= AV_RL32(buf);
644 const int token_count av_unused = AV_RL32(buf + bitstream_size + 8);
645 unsigned int prestream_size= 4*AV_RL32(buf + bitstream_size + 4);
646 const uint8_t *prestream= buf + bitstream_size + 12;
648 if(prestream_size + bitstream_size + 12 != length
649 || bitstream_size > (1<<26)
650 || prestream_size > (1<<26)){
651 av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %d %d\n", prestream_size, bitstream_size, length);
652 return -1;
655 prestream= read_huffman_tables(f, prestream);
657 init_get_bits(&f->gb, buf + 4, 8*bitstream_size);
659 prestream_size= length + buf - prestream;
661 av_fast_malloc(&f->bitstream_buffer, &f->bitstream_buffer_size, prestream_size + FF_INPUT_BUFFER_PADDING_SIZE);
662 if (!f->bitstream_buffer)
663 return AVERROR(ENOMEM);
664 f->dsp.bswap_buf(f->bitstream_buffer, (const uint32_t*)prestream, prestream_size/4);
665 init_get_bits(&f->pre_gb, f->bitstream_buffer, 8*prestream_size);
667 f->last_dc= 0*128*8*8;
669 for(y=0; y<height; y+=16){
670 for(x=0; x<width; x+=16){
671 if(decode_i_mb(f) < 0)
672 return -1;
674 idct_put(f, x, y);
676 dst += 16*stride;
679 if(get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3) != 256)
680 av_log(f->avctx, AV_LOG_ERROR, "end mismatch\n");
682 return 0;
685 static int decode_frame(AVCodecContext *avctx,
686 void *data, int *data_size,
687 AVPacket *avpkt)
689 const uint8_t *buf = avpkt->data;
690 int buf_size = avpkt->size;
691 FourXContext * const f = avctx->priv_data;
692 AVFrame *picture = data;
693 AVFrame *p, temp;
694 int i, frame_4cc, frame_size;
696 frame_4cc= AV_RL32(buf);
697 if(buf_size != AV_RL32(buf+4)+8 || buf_size < 20){
698 av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %d\n", buf_size, AV_RL32(buf+4));
701 if(frame_4cc == AV_RL32("cfrm")){
702 int free_index=-1;
703 const int data_size= buf_size - 20;
704 const int id= AV_RL32(buf+12);
705 const int whole_size= AV_RL32(buf+16);
706 CFrameBuffer *cfrm;
708 for(i=0; i<CFRAME_BUFFER_COUNT; i++){
709 if(f->cfrm[i].id && f->cfrm[i].id < avctx->frame_number)
710 av_log(f->avctx, AV_LOG_ERROR, "lost c frame %d\n", f->cfrm[i].id);
713 for(i=0; i<CFRAME_BUFFER_COUNT; i++){
714 if(f->cfrm[i].id == id) break;
715 if(f->cfrm[i].size == 0 ) free_index= i;
718 if(i>=CFRAME_BUFFER_COUNT){
719 i= free_index;
720 f->cfrm[i].id= id;
722 cfrm= &f->cfrm[i];
724 cfrm->data= av_fast_realloc(cfrm->data, &cfrm->allocated_size, cfrm->size + data_size + FF_INPUT_BUFFER_PADDING_SIZE);
725 if(!cfrm->data){ //explicit check needed as memcpy below might not catch a NULL
726 av_log(f->avctx, AV_LOG_ERROR, "realloc falure");
727 return -1;
730 memcpy(cfrm->data + cfrm->size, buf+20, data_size);
731 cfrm->size += data_size;
733 if(cfrm->size >= whole_size){
734 buf= cfrm->data;
735 frame_size= cfrm->size;
737 if(id != avctx->frame_number){
738 av_log(f->avctx, AV_LOG_ERROR, "cframe id mismatch %d %d\n", id, avctx->frame_number);
741 cfrm->size= cfrm->id= 0;
742 frame_4cc= AV_RL32("pfrm");
743 }else
744 return buf_size;
745 }else{
746 buf= buf + 12;
747 frame_size= buf_size - 12;
750 temp= f->current_picture;
751 f->current_picture= f->last_picture;
752 f->last_picture= temp;
754 p= &f->current_picture;
755 avctx->coded_frame= p;
757 avctx->flags |= CODEC_FLAG_EMU_EDGE; // alternatively we would have to use our own buffer management
759 if(p->data[0])
760 avctx->release_buffer(avctx, p);
762 p->reference= 1;
763 if(avctx->get_buffer(avctx, p) < 0){
764 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
765 return -1;
768 if(frame_4cc == AV_RL32("ifr2")){
769 p->pict_type= FF_I_TYPE;
770 if(decode_i2_frame(f, buf-4, frame_size) < 0)
771 return -1;
772 }else if(frame_4cc == AV_RL32("ifrm")){
773 p->pict_type= FF_I_TYPE;
774 if(decode_i_frame(f, buf, frame_size) < 0)
775 return -1;
776 }else if(frame_4cc == AV_RL32("pfrm") || frame_4cc == AV_RL32("pfr2")){
777 p->pict_type= FF_P_TYPE;
778 if(decode_p_frame(f, buf, frame_size) < 0)
779 return -1;
780 }else if(frame_4cc == AV_RL32("snd_")){
781 av_log(avctx, AV_LOG_ERROR, "ignoring snd_ chunk length:%d\n", buf_size);
782 }else{
783 av_log(avctx, AV_LOG_ERROR, "ignoring unknown chunk length:%d\n", buf_size);
786 p->key_frame= p->pict_type == FF_I_TYPE;
788 *picture= *p;
789 *data_size = sizeof(AVPicture);
791 emms_c();
793 return buf_size;
797 static av_cold void common_init(AVCodecContext *avctx){
798 FourXContext * const f = avctx->priv_data;
800 dsputil_init(&f->dsp, avctx);
802 f->avctx= avctx;
805 static av_cold int decode_init(AVCodecContext *avctx){
806 FourXContext * const f = avctx->priv_data;
808 if(avctx->extradata_size != 4 || !avctx->extradata) {
809 av_log(avctx, AV_LOG_ERROR, "extradata wrong or missing\n");
810 return 1;
813 f->version= AV_RL32(avctx->extradata)>>16;
814 common_init(avctx);
815 init_vlcs(f);
817 if(f->version>2) avctx->pix_fmt= PIX_FMT_RGB565;
818 else avctx->pix_fmt= PIX_FMT_RGB555;
820 return 0;
824 static av_cold int decode_end(AVCodecContext *avctx){
825 FourXContext * const f = avctx->priv_data;
826 int i;
828 av_freep(&f->bitstream_buffer);
829 f->bitstream_buffer_size=0;
830 for(i=0; i<CFRAME_BUFFER_COUNT; i++){
831 av_freep(&f->cfrm[i].data);
832 f->cfrm[i].allocated_size= 0;
834 free_vlc(&f->pre_vlc);
835 if(f->current_picture.data[0])
836 avctx->release_buffer(avctx, &f->current_picture);
837 if(f->last_picture.data[0])
838 avctx->release_buffer(avctx, &f->last_picture);
840 return 0;
843 AVCodec fourxm_decoder = {
844 "4xm",
845 CODEC_TYPE_VIDEO,
846 CODEC_ID_4XM,
847 sizeof(FourXContext),
848 decode_init,
849 NULL,
850 decode_end,
851 decode_frame,
852 CODEC_CAP_DR1,
853 .long_name = NULL_IF_CONFIG_SMALL("4X Movie"),