/src/ffmpeg/libavcodec/bonk.c
Line | Count | Source |
1 | | /* |
2 | | * Bonk audio decoder |
3 | | * |
4 | | * This file is part of FFmpeg. |
5 | | * |
6 | | * FFmpeg is free software; you can redistribute it and/or |
7 | | * modify it under the terms of the GNU Lesser General Public |
8 | | * License as published by the Free Software Foundation; either |
9 | | * version 2.1 of the License, or (at your option) any later version. |
10 | | * |
11 | | * FFmpeg is distributed in the hope that it will be useful, |
12 | | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
13 | | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
14 | | * Lesser General Public License for more details. |
15 | | * |
16 | | * You should have received a copy of the GNU Lesser General Public |
17 | | * License along with FFmpeg; if not, write to the Free Software |
18 | | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
19 | | */ |
20 | | |
21 | | #include "libavutil/intreadwrite.h" |
22 | | #include "libavutil/mem.h" |
23 | | #include "avcodec.h" |
24 | | #include "codec_internal.h" |
25 | | #include "decode.h" |
26 | | #define BITSTREAM_READER_LE |
27 | | #include "get_bits.h" |
28 | | |
29 | | typedef struct BitCount { |
30 | | uint8_t bit; |
31 | | unsigned count; |
32 | | } BitCount; |
33 | | |
34 | | typedef struct BonkContext { |
35 | | GetBitContext gb; |
36 | | int skip; |
37 | | |
38 | | uint8_t *bitstream; |
39 | | int64_t max_framesize; |
40 | | int bitstream_size; |
41 | | int bitstream_index; |
42 | | |
43 | | uint64_t nb_samples; |
44 | | int lossless; |
45 | | int mid_side; |
46 | | int n_taps; |
47 | | int down_sampling; |
48 | | int samples_per_packet; |
49 | | |
50 | | int state[2][2048], k[2048]; |
51 | | int *samples[2]; |
52 | | int *input_samples; |
53 | | uint8_t quant[2048]; |
54 | | BitCount *bits; |
55 | | } BonkContext; |
56 | | |
57 | | static av_cold int bonk_close(AVCodecContext *avctx) |
58 | 237 | { |
59 | 237 | BonkContext *s = avctx->priv_data; |
60 | | |
61 | 237 | av_freep(&s->bitstream); |
62 | 237 | av_freep(&s->input_samples); |
63 | 237 | av_freep(&s->samples[0]); |
64 | 237 | av_freep(&s->samples[1]); |
65 | 237 | av_freep(&s->bits); |
66 | 237 | s->bitstream_size = 0; |
67 | | |
68 | 237 | return 0; |
69 | 237 | } |
70 | | |
71 | | static av_cold int bonk_init(AVCodecContext *avctx) |
72 | 237 | { |
73 | 237 | BonkContext *s = avctx->priv_data; |
74 | | |
75 | 237 | avctx->sample_fmt = AV_SAMPLE_FMT_S16P; |
76 | 237 | if (avctx->extradata_size < 17) |
77 | 0 | return AVERROR(EINVAL); |
78 | | |
79 | 237 | if (avctx->extradata[0]) { |
80 | 8 | av_log(avctx, AV_LOG_ERROR, "Unsupported version.\n"); |
81 | 8 | return AVERROR_INVALIDDATA; |
82 | 8 | } |
83 | | |
84 | 229 | if (avctx->ch_layout.nb_channels < 1 || avctx->ch_layout.nb_channels > 2) |
85 | 3 | return AVERROR_INVALIDDATA; |
86 | | |
87 | 226 | s->nb_samples = AV_RL32(avctx->extradata + 1) / avctx->ch_layout.nb_channels; |
88 | 226 | if (!s->nb_samples) |
89 | 152 | s->nb_samples = UINT64_MAX; |
90 | 226 | s->lossless = avctx->extradata[10] != 0; |
91 | 226 | s->mid_side = avctx->extradata[11] != 0; |
92 | 226 | s->n_taps = AV_RL16(avctx->extradata + 12); |
93 | 226 | if (!s->n_taps || s->n_taps > 2048) |
94 | 1 | return AVERROR(EINVAL); |
95 | | |
96 | 225 | s->down_sampling = avctx->extradata[14]; |
97 | 225 | if (!s->down_sampling) |
98 | 0 | return AVERROR(EINVAL); |
99 | | |
100 | 225 | s->samples_per_packet = AV_RL16(avctx->extradata + 15); |
101 | 225 | if (!s->samples_per_packet) |
102 | 0 | return AVERROR(EINVAL); |
103 | | |
104 | 225 | if (s->down_sampling * s->samples_per_packet < s->n_taps) |
105 | 0 | return AVERROR_INVALIDDATA; |
106 | | |
107 | 225 | s->max_framesize = s->samples_per_packet * avctx->ch_layout.nb_channels * s->down_sampling * 16LL; |
108 | 225 | if (s->max_framesize > (INT32_MAX - AV_INPUT_BUFFER_PADDING_SIZE) / 8) |
109 | 0 | return AVERROR_INVALIDDATA; |
110 | | |
111 | 225 | s->bitstream = av_calloc(s->max_framesize + AV_INPUT_BUFFER_PADDING_SIZE, sizeof(*s->bitstream)); |
112 | 225 | if (!s->bitstream) |
113 | 0 | return AVERROR(ENOMEM); |
114 | | |
115 | 225 | s->input_samples = av_calloc(s->samples_per_packet, sizeof(*s->input_samples)); |
116 | 225 | if (!s->input_samples) |
117 | 0 | return AVERROR(ENOMEM); |
118 | | |
119 | 225 | s->samples[0] = av_calloc(s->samples_per_packet * s->down_sampling, sizeof(*s->samples[0])); |
120 | 225 | s->samples[1] = av_calloc(s->samples_per_packet * s->down_sampling, sizeof(*s->samples[0])); |
121 | 225 | if (!s->samples[0] || !s->samples[1]) |
122 | 0 | return AVERROR(ENOMEM); |
123 | | |
124 | 225 | s->bits = av_calloc(s->max_framesize * 8, sizeof(*s->bits)); |
125 | 225 | if (!s->bits) |
126 | 0 | return AVERROR(ENOMEM); |
127 | | |
128 | 115k | for (int i = 0; i < 512; i++) { |
129 | 115k | s->quant[i] = sqrt(i + 1); |
130 | 115k | } |
131 | | |
132 | 225 | return 0; |
133 | 225 | } |
134 | | |
135 | | static unsigned read_uint_max(BonkContext *s, uint32_t max) |
136 | 475k | { |
137 | 475k | unsigned value = 0; |
138 | | |
139 | 475k | if (max == 0) |
140 | 341k | return 0; |
141 | | |
142 | 133k | av_assert0(max >> 31 == 0); |
143 | | |
144 | 1.10M | for (unsigned i = 1; i <= max - value; i+=i) |
145 | 974k | if (get_bits1(&s->gb)) |
146 | 479k | value += i; |
147 | | |
148 | 133k | return value; |
149 | 133k | } |
150 | | |
151 | | static int intlist_read(BonkContext *s, int *buf, int entries, int base_2_part) |
152 | 148k | { |
153 | 148k | int i, low_bits = 0, x = 0, max_x; |
154 | 148k | int n_zeros = 0, step = 256, dominant = 0; |
155 | 148k | int pos = 0, level = 0; |
156 | 148k | BitCount *bits = s->bits; |
157 | 148k | int passes = 1; |
158 | | |
159 | 148k | memset(buf, 0, entries * sizeof(*buf)); |
160 | 148k | if (base_2_part) { |
161 | 75.2k | low_bits = get_bits(&s->gb, 4); |
162 | | |
163 | 75.2k | if (low_bits) |
164 | 1.25M | for (i = 0; i < entries; i++) |
165 | 1.23M | buf[i] = get_bits(&s->gb, low_bits); |
166 | 75.2k | } |
167 | | |
168 | 1.43M | while (n_zeros < entries) { |
169 | 1.28M | int steplet = step >> 8; |
170 | | |
171 | 1.28M | if (get_bits_left(&s->gb) <= 0) |
172 | 252 | return AVERROR_INVALIDDATA; |
173 | | |
174 | 1.28M | if (!get_bits1(&s->gb)) { |
175 | 808k | av_assert0(steplet >= 0); |
176 | | |
177 | 808k | if (steplet > 0) { |
178 | 808k | bits[x ].bit = dominant; |
179 | 808k | bits[x++].count = steplet; |
180 | 808k | } |
181 | | |
182 | 808k | if (!dominant) |
183 | 346k | n_zeros += steplet; |
184 | | |
185 | 808k | if (step > INT32_MAX*8LL/9 + 1) |
186 | 638 | return AVERROR_INVALIDDATA; |
187 | 807k | step += step / 8; |
188 | 807k | } else if (steplet > 0) { |
189 | 475k | int actual_run = read_uint_max(s, steplet - 1); |
190 | | |
191 | 475k | av_assert0(actual_run >= 0); |
192 | | |
193 | 475k | if (actual_run > 0) { |
194 | 111k | bits[x ].bit = dominant; |
195 | 111k | bits[x++].count = actual_run; |
196 | 111k | } |
197 | | |
198 | 475k | bits[x ].bit = !dominant; |
199 | 475k | bits[x++].count = 1; |
200 | | |
201 | 475k | if (!dominant) |
202 | 198k | n_zeros += actual_run; |
203 | 276k | else |
204 | 276k | n_zeros++; |
205 | | |
206 | 475k | step -= step / 8; |
207 | 475k | } |
208 | | |
209 | 1.28M | if (step < 256) { |
210 | 130k | step = 65536 / step; |
211 | 130k | dominant = !dominant; |
212 | 130k | } |
213 | 1.28M | } |
214 | | |
215 | 147k | max_x = x; |
216 | 147k | x = 0; |
217 | 147k | n_zeros = 0; |
218 | 16.6M | for (i = 0; n_zeros < entries; i++) { |
219 | 16.4M | if (x >= max_x) |
220 | 0 | return AVERROR_INVALIDDATA; |
221 | | |
222 | 16.4M | if (pos >= entries) { |
223 | 168k | pos = 0; |
224 | 168k | level += passes << low_bits; |
225 | 168k | passes = 1; |
226 | 168k | if (bits[x].bit && bits[x].count > entries - n_zeros) |
227 | 63.2k | passes = bits[x].count / (entries - n_zeros); |
228 | 168k | } |
229 | | |
230 | 16.4M | if (level > 1 << 16) |
231 | 700 | return AVERROR_INVALIDDATA; |
232 | | |
233 | 16.4M | if (buf[pos] >= level) { |
234 | 5.65M | if (bits[x].bit) |
235 | 3.37M | buf[pos] += passes << low_bits; |
236 | 2.27M | else |
237 | 2.27M | n_zeros++; |
238 | | |
239 | 5.65M | av_assert1(bits[x].count >= passes); |
240 | 5.65M | bits[x].count -= passes; |
241 | 5.65M | x += bits[x].count == 0; |
242 | 5.65M | } |
243 | | |
244 | 16.4M | pos++; |
245 | 16.4M | } |
246 | | |
247 | 2.35M | for (i = 0; i < entries; i++) { |
248 | 2.21M | if (buf[i] && get_bits1(&s->gb)) { |
249 | 229k | buf[i] = -buf[i]; |
250 | 229k | } |
251 | 2.21M | } |
252 | | |
253 | 146k | return 0; |
254 | 147k | } |
255 | | |
256 | | static inline int shift_down(int a, int b) |
257 | 2.97G | { |
258 | 2.97G | return (a >> b) + (a < 0); |
259 | 2.97G | } |
260 | | |
261 | | static inline int shift(int a, int b) |
262 | 6.23M | { |
263 | 6.23M | return a + (1 << b - 1) >> b; |
264 | 6.23M | } |
265 | | |
266 | 2.97G | #define LATTICE_SHIFT 10 |
267 | 17.3M | #define SAMPLE_SHIFT 4 |
268 | 17.4M | #define SAMPLE_FACTOR (1 << SAMPLE_SHIFT) |
269 | | |
270 | | static int predictor_calc_error(int *k, int *state, int order, int error) |
271 | 8.67M | { |
272 | 8.67M | int i, x = error - (unsigned)shift_down(k[order-1] * (unsigned)state[order-1], LATTICE_SHIFT); |
273 | 8.67M | int *k_ptr = &(k[order-2]), |
274 | 8.67M | *state_ptr = &(state[order-2]); |
275 | | |
276 | 1.42G | for (i = order-2; i >= 0; i--, k_ptr--, state_ptr--) { |
277 | 1.41G | unsigned k_value = *k_ptr, state_value = *state_ptr; |
278 | | |
279 | 1.41G | x -= (unsigned) shift_down(k_value * (unsigned)state_value, LATTICE_SHIFT); |
280 | 1.41G | state_ptr[1] = state_value + shift_down(k_value * x, LATTICE_SHIFT); |
281 | 1.41G | } |
282 | | |
283 | | // don't drift too far, to avoid overflows |
284 | 8.67M | x = av_clip(x, -(SAMPLE_FACTOR << 16), SAMPLE_FACTOR << 16); |
285 | | |
286 | 8.67M | state[0] = x; |
287 | | |
288 | 8.67M | return x; |
289 | 8.67M | } |
290 | | |
291 | | static void predictor_init_state(int *k, unsigned *state, int order) |
292 | 75.2k | { |
293 | 323k | for (int i = order - 2; i >= 0; i--) { |
294 | 247k | unsigned x = state[i]; |
295 | | |
296 | 68.6M | for (int j = 0, p = i + 1; p < order; j++, p++) { |
297 | 68.3M | int tmp = x + shift_down(k[j] * state[p], LATTICE_SHIFT); |
298 | | |
299 | 68.3M | state[p] += shift_down(k[j] * x, LATTICE_SHIFT); |
300 | 68.3M | x = tmp; |
301 | 68.3M | } |
302 | 247k | } |
303 | 75.2k | } |
304 | | |
305 | | static int bonk_decode(AVCodecContext *avctx, AVFrame *frame, |
306 | | int *got_frame_ptr, AVPacket *pkt) |
307 | 279k | { |
308 | 279k | BonkContext *s = avctx->priv_data; |
309 | 279k | GetBitContext *gb = &s->gb; |
310 | 279k | const uint8_t *buf; |
311 | 279k | int quant, n, buf_size, input_buf_size; |
312 | 279k | int ret = AVERROR_INVALIDDATA; |
313 | | |
314 | 279k | if ((!pkt->size && !s->bitstream_size) || s->nb_samples == 0) { |
315 | 3.92k | *got_frame_ptr = 0; |
316 | 3.92k | return pkt->size; |
317 | 3.92k | } |
318 | | |
319 | 275k | buf_size = FFMIN(pkt->size, s->max_framesize - s->bitstream_size); |
320 | 275k | input_buf_size = buf_size; |
321 | 275k | if (s->bitstream_index + s->bitstream_size + buf_size + AV_INPUT_BUFFER_PADDING_SIZE > s->max_framesize) { |
322 | 136k | memmove(s->bitstream, &s->bitstream[s->bitstream_index], s->bitstream_size); |
323 | 136k | s->bitstream_index = 0; |
324 | 136k | } |
325 | 275k | if (pkt->data) |
326 | 271k | memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], pkt->data, buf_size); |
327 | 275k | buf = &s->bitstream[s->bitstream_index]; |
328 | 275k | buf_size += s->bitstream_size; |
329 | 275k | s->bitstream_size = buf_size; |
330 | 275k | if (buf_size < s->max_framesize && pkt->data) { |
331 | 202k | *got_frame_ptr = 0; |
332 | 202k | return input_buf_size; |
333 | 202k | } |
334 | | |
335 | 73.0k | frame->nb_samples = FFMIN(s->samples_per_packet * s->down_sampling, s->nb_samples); |
336 | 73.0k | if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) |
337 | 7 | goto fail; |
338 | | |
339 | 73.0k | if ((ret = init_get_bits8(gb, buf, buf_size)) < 0) |
340 | 0 | goto fail; |
341 | | |
342 | 73.0k | skip_bits(gb, s->skip); |
343 | 73.0k | if ((ret = intlist_read(s, s->k, s->n_taps, 0)) < 0) |
344 | 520 | goto fail; |
345 | | |
346 | 392k | for (int i = 0; i < s->n_taps; i++) |
347 | 320k | s->k[i] *= s->quant[i]; |
348 | 72.5k | quant = s->lossless ? 1 : get_bits(&s->gb, 16) * SAMPLE_FACTOR; |
349 | | |
350 | 146k | for (int ch = 0; ch < avctx->ch_layout.nb_channels; ch++) { |
351 | 75.2k | const int samples_per_packet = s->samples_per_packet; |
352 | 75.2k | const int down_sampling = s->down_sampling; |
353 | 75.2k | const int offset = samples_per_packet * down_sampling - 1; |
354 | 75.2k | int *state = s->state[ch]; |
355 | 75.2k | int *sample = s->samples[ch]; |
356 | | |
357 | 75.2k | predictor_init_state(s->k, state, s->n_taps); |
358 | 75.2k | if ((ret = intlist_read(s, s->input_samples, samples_per_packet, 1)) < 0) |
359 | 1.07k | goto fail; |
360 | | |
361 | 1.96M | for (int i = 0; i < samples_per_packet; i++) { |
362 | 8.67M | for (int j = 0; j < s->down_sampling - 1; j++) { |
363 | 6.78M | sample[0] = predictor_calc_error(s->k, state, s->n_taps, 0); |
364 | 6.78M | sample++; |
365 | 6.78M | } |
366 | | |
367 | 1.89M | sample[0] = predictor_calc_error(s->k, state, s->n_taps, s->input_samples[i] * (unsigned)quant); |
368 | 1.89M | sample++; |
369 | 1.89M | } |
370 | | |
371 | 74.2k | sample = s->samples[ch]; |
372 | 368k | for (int i = 0; i < s->n_taps; i++) |
373 | 294k | state[i] = sample[offset - i]; |
374 | 74.2k | } |
375 | | |
376 | 71.4k | if (s->mid_side && avctx->ch_layout.nb_channels == 2) { |
377 | 71.2k | for (int i = 0; i < frame->nb_samples; i++) { |
378 | 68.7k | s->samples[1][i] += shift(s->samples[0][i], 1); |
379 | 68.7k | s->samples[0][i] -= s->samples[1][i]; |
380 | 68.7k | } |
381 | 2.47k | } |
382 | | |
383 | 71.4k | if (!s->lossless) { |
384 | 22.9k | for (int ch = 0; ch < avctx->ch_layout.nb_channels; ch++) { |
385 | 12.5k | int *samples = s->samples[ch]; |
386 | 6.18M | for (int i = 0; i < frame->nb_samples; i++) |
387 | 6.17M | samples[i] = shift(samples[i], 4); |
388 | 12.5k | } |
389 | 10.3k | } |
390 | | |
391 | 145k | for (int ch = 0; ch < avctx->ch_layout.nb_channels; ch++) { |
392 | 74.0k | int16_t *osamples = (int16_t *)frame->extended_data[ch]; |
393 | 74.0k | int *samples = s->samples[ch]; |
394 | 8.64M | for (int i = 0; i < frame->nb_samples; i++) |
395 | 8.57M | osamples[i] = av_clip_int16(samples[i]); |
396 | 74.0k | } |
397 | | |
398 | 71.4k | s->nb_samples -= frame->nb_samples; |
399 | | |
400 | 71.4k | s->skip = get_bits_count(gb) - 8 * (get_bits_count(gb) / 8); |
401 | 71.4k | n = get_bits_count(gb) / 8; |
402 | | |
403 | 71.4k | if (n > buf_size) { |
404 | 1.61k | fail: |
405 | 1.61k | s->bitstream_size = 0; |
406 | 1.61k | s->bitstream_index = 0; |
407 | 1.61k | return AVERROR_INVALIDDATA; |
408 | 16 | } |
409 | | |
410 | 71.4k | *got_frame_ptr = 1; |
411 | | |
412 | 71.4k | if (s->bitstream_size) { |
413 | 71.4k | s->bitstream_index += n; |
414 | 71.4k | s->bitstream_size -= n; |
415 | 71.4k | return input_buf_size; |
416 | 71.4k | } |
417 | 0 | return n; |
418 | 71.4k | } |
419 | | |
420 | | const FFCodec ff_bonk_decoder = { |
421 | | .p.name = "bonk", |
422 | | CODEC_LONG_NAME("Bonk audio"), |
423 | | .p.type = AVMEDIA_TYPE_AUDIO, |
424 | | .p.id = AV_CODEC_ID_BONK, |
425 | | .priv_data_size = sizeof(BonkContext), |
426 | | .init = bonk_init, |
427 | | FF_CODEC_DECODE_CB(bonk_decode), |
428 | | .close = bonk_close, |
429 | | .p.capabilities = AV_CODEC_CAP_DELAY | |
430 | | AV_CODEC_CAP_DR1, |
431 | | .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, |
432 | | CODEC_SAMPLEFMTS(AV_SAMPLE_FMT_S16P), |
433 | | }; |