1 /*
2 * This file is part of FFmpeg.
3 *
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
8 *
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19 #include "libavutil/crc.h"
20 #include "libavutil/float_dsp.h"
21 #include "libavutil/intreadwrite.h"
22 #include "libavutil/tx.h"
23
24 #include "avcodec.h"
25 #include "get_bits.h"
26 #include "internal.h"
27 #include "hca_data.h"
28
29 typedef struct ChannelContext {
30 float base[128];
31 DECLARE_ALIGNED(32, float, imdct_in)[128];
32 DECLARE_ALIGNED(32, float, imdct_out)[128];
33 DECLARE_ALIGNED(32, float, imdct_prev)[128];
34 int8_t scale_factors[128];
35 uint8_t scale[128];
36 int8_t intensity[8];
37 int8_t *hfr_scale;
38 unsigned count;
39 int chan_type;
40 } ChannelContext;
41
42 typedef struct HCAContext {
43 GetBitContext gb;
44
45 const AVCRC *crc_table;
46
47 ChannelContext ch[16];
48
49 uint8_t ath[128];
50
51 int ath_type;
52 unsigned hfr_group_count;
53 uint8_t track_count;
54 uint8_t channel_config;
55 uint8_t total_band_count;
56 uint8_t base_band_count;
57 uint8_t stereo_band_count;
58 uint8_t bands_per_hfr_group;
59
60 av_tx_fn tx_fn;
61 AVTXContext *tx_ctx;
62 AVFloatDSPContext *fdsp;
63 } HCAContext;
64
ath_init1(uint8_t * ath,int sample_rate)65 static void ath_init1(uint8_t *ath, int sample_rate)
66 {
67 unsigned int index;
68 unsigned int acc = 0;
69
70 for (int i = 0; i < 128; i++) {
71 acc += sample_rate;
72 index = acc >> 13;
73
74 if (index >= 654) {
75 memset(ath+i, 0xFF, (128 - i));
76 break;
77 }
78
79 ath[i] = ath_base_curve[index];
80 }
81 }
82
ath_init(uint8_t * ath,int type,int sample_rate)83 static int ath_init(uint8_t *ath, int type, int sample_rate)
84 {
85 switch (type) {
86 case 0:
87 /* nothing to do */
88 break;
89 case 1:
90 ath_init1(ath, sample_rate);
91 break;
92 default:
93 return AVERROR_INVALIDDATA;
94 }
95
96 return 0;
97 }
98
ceil2(unsigned a,unsigned b)99 static inline unsigned ceil2(unsigned a, unsigned b)
100 {
101 return (b > 0) ? (a / b + ((a % b) ? 1 : 0)) : 0;
102 }
103
decode_init(AVCodecContext * avctx)104 static av_cold int decode_init(AVCodecContext *avctx)
105 {
106 HCAContext *c = avctx->priv_data;
107 GetBitContext *gb = &c->gb;
108 int8_t r[16] = { 0 };
109 float scale = 1.f / 8.f;
110 unsigned b, chunk;
111 int version, ret;
112
113 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
114 c->crc_table = av_crc_get_table(AV_CRC_16_ANSI);
115
116 if (avctx->channels <= 0 || avctx->channels > 16)
117 return AVERROR(EINVAL);
118
119 ret = init_get_bits8(gb, avctx->extradata, avctx->extradata_size);
120 if (ret < 0)
121 return ret;
122 skip_bits_long(gb, 32);
123 version = get_bits(gb, 16);
124 skip_bits_long(gb, 16);
125
126 c->ath_type = version >= 0x200 ? 0 : 1;
127
128 if (get_bits_long(gb, 32) != MKBETAG('f', 'm', 't', 0))
129 return AVERROR_INVALIDDATA;
130 skip_bits_long(gb, 32);
131 skip_bits_long(gb, 32);
132 skip_bits_long(gb, 32);
133
134 chunk = get_bits_long(gb, 32);
135 if (chunk == MKBETAG('c', 'o', 'm', 'p')) {
136 skip_bits_long(gb, 16);
137 skip_bits_long(gb, 8);
138 skip_bits_long(gb, 8);
139 c->track_count = get_bits(gb, 8);
140 c->channel_config = get_bits(gb, 8);
141 c->total_band_count = get_bits(gb, 8);
142 c->base_band_count = get_bits(gb, 8);
143 c->stereo_band_count = get_bits(gb, 8);
144 c->bands_per_hfr_group = get_bits(gb, 8);
145 } else if (chunk == MKBETAG('d', 'e', 'c', 0)) {
146 skip_bits_long(gb, 16);
147 skip_bits_long(gb, 8);
148 skip_bits_long(gb, 8);
149 c->total_band_count = get_bits(gb, 8) + 1;
150 c->base_band_count = get_bits(gb, 8) + 1;
151 c->track_count = get_bits(gb, 4);
152 c->channel_config = get_bits(gb, 4);
153 if (!get_bits(gb, 8))
154 c->base_band_count = c->total_band_count;
155 c->stereo_band_count = c->total_band_count - c->base_band_count;
156 c->bands_per_hfr_group = 0;
157 } else
158 return AVERROR_INVALIDDATA;
159
160 if (c->total_band_count > FF_ARRAY_ELEMS(c->ch->imdct_in))
161 return AVERROR_INVALIDDATA;
162
163
164 while (get_bits_left(gb) >= 32) {
165 chunk = get_bits_long(gb, 32);
166 if (chunk == MKBETAG('v', 'b', 'r', 0)) {
167 skip_bits_long(gb, 16);
168 skip_bits_long(gb, 16);
169 } else if (chunk == MKBETAG('a', 't', 'h', 0)) {
170 c->ath_type = get_bits(gb, 16);
171 } else if (chunk == MKBETAG('r', 'v', 'a', 0)) {
172 skip_bits_long(gb, 32);
173 } else if (chunk == MKBETAG('c', 'o', 'm', 'm')) {
174 skip_bits_long(gb, get_bits(gb, 8) * 8);
175 } else if (chunk == MKBETAG('c', 'i', 'p', 'h')) {
176 skip_bits_long(gb, 16);
177 } else if (chunk == MKBETAG('l', 'o', 'o', 'p')) {
178 skip_bits_long(gb, 32);
179 skip_bits_long(gb, 32);
180 skip_bits_long(gb, 16);
181 skip_bits_long(gb, 16);
182 } else if (chunk == MKBETAG('p', 'a', 'd', 0)) {
183 break;
184 } else {
185 break;
186 }
187 }
188
189 ret = ath_init(c->ath, c->ath_type, avctx->sample_rate);
190 if (ret < 0)
191 return ret;
192
193 if (!c->track_count)
194 c->track_count = 1;
195
196 b = avctx->channels / c->track_count;
197 if (c->stereo_band_count && b > 1) {
198 int8_t *x = r;
199
200 for (int i = 0; i < c->track_count; i++, x+=b) {
201 switch (b) {
202 case 2:
203 case 3:
204 x[0] = 1;
205 x[1] = 2;
206 break;
207 case 4:
208 x[0]=1; x[1] = 2;
209 if (c->channel_config == 0) {
210 x[2]=1;
211 x[3]=2;
212 }
213 break;
214 case 5:
215 x[0]=1; x[1] = 2;
216 if (c->channel_config <= 2) {
217 x[3]=1;
218 x[4]=2;
219 }
220 break;
221 case 6:
222 case 7:
223 x[0] = 1; x[1] = 2; x[4] = 1; x[5] = 2;
224 break;
225 case 8:
226 x[0] = 1; x[1] = 2; x[4] = 1; x[5] = 2; x[6] = 1; x[7] = 2;
227 break;
228 }
229 }
230 }
231
232 if (c->total_band_count < c->base_band_count)
233 return AVERROR_INVALIDDATA;
234
235 c->hfr_group_count = ceil2(c->total_band_count - (c->base_band_count + c->stereo_band_count),
236 c->bands_per_hfr_group);
237
238 if (c->base_band_count + c->stereo_band_count + (unsigned long)c->hfr_group_count > 128ULL)
239 return AVERROR_INVALIDDATA;
240
241 for (int i = 0; i < avctx->channels; i++) {
242 c->ch[i].chan_type = r[i];
243 c->ch[i].count = c->base_band_count + ((r[i] != 2) ? c->stereo_band_count : 0);
244 c->ch[i].hfr_scale = &c->ch[i].scale_factors[c->base_band_count + c->stereo_band_count];
245 if (c->ch[i].count > 128)
246 return AVERROR_INVALIDDATA;
247 }
248
249 c->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
250 if (!c->fdsp)
251 return AVERROR(ENOMEM);
252
253 return av_tx_init(&c->tx_ctx, &c->tx_fn, AV_TX_FLOAT_MDCT, 1, 128, &scale, 0);
254 }
255
run_imdct(HCAContext * c,ChannelContext * ch,int index,float * out)256 static void run_imdct(HCAContext *c, ChannelContext *ch, int index, float *out)
257 {
258 c->tx_fn(c->tx_ctx, ch->imdct_out, ch->imdct_in, sizeof(float));
259
260 c->fdsp->vector_fmul_window(out, ch->imdct_prev + (128 >> 1),
261 ch->imdct_out, window, 128 >> 1);
262
263 memcpy(ch->imdct_prev, ch->imdct_out, 128 * sizeof(float));
264 }
265
apply_intensity_stereo(HCAContext * s,ChannelContext * ch1,ChannelContext * ch2,int index,unsigned band_count,unsigned base_band_count,unsigned stereo_band_count)266 static void apply_intensity_stereo(HCAContext *s, ChannelContext *ch1, ChannelContext *ch2,
267 int index, unsigned band_count, unsigned base_band_count,
268 unsigned stereo_band_count)
269 {
270 float ratio_l = intensity_ratio_table[ch1->intensity[index]];
271 float ratio_r = ratio_l - 2.0f;
272 float *c1 = &ch1->imdct_in[base_band_count];
273 float *c2 = &ch2->imdct_in[base_band_count];
274
275 if (ch1->chan_type != 1 || !stereo_band_count)
276 return;
277
278 for (int i = 0; i < band_count; i++) {
279 *(c2++) = *c1 * ratio_r;
280 *(c1++) *= ratio_l;
281 }
282 }
283
reconstruct_hfr(HCAContext * s,ChannelContext * ch,unsigned hfr_group_count,unsigned bands_per_hfr_group,unsigned start_band,unsigned total_band_count)284 static void reconstruct_hfr(HCAContext *s, ChannelContext *ch,
285 unsigned hfr_group_count,
286 unsigned bands_per_hfr_group,
287 unsigned start_band, unsigned total_band_count)
288 {
289 if (ch->chan_type == 2 || !bands_per_hfr_group)
290 return;
291
292 for (int i = 0, k = start_band, l = start_band - 1; i < hfr_group_count; i++){
293 for (int j = 0; j < bands_per_hfr_group && k < total_band_count && l >= 0; j++, k++, l--){
294 ch->imdct_in[k] = scale_conversion_table[ (ch->hfr_scale[i] - ch->scale_factors[l]) & 63 ] * ch->imdct_in[l];
295 }
296 }
297
298 ch->imdct_in[127] = 0;
299 }
300
dequantize_coefficients(HCAContext * c,ChannelContext * ch)301 static void dequantize_coefficients(HCAContext *c, ChannelContext *ch)
302 {
303 GetBitContext *gb = &c->gb;
304
305 for (int i = 0; i < ch->count; i++) {
306 unsigned scale = ch->scale[i];
307 int nb_bits = max_bits_table[scale];
308 int value = get_bitsz(gb, nb_bits);
309 float factor;
310
311 if (scale > 7) {
312 value = (1 - ((value & 1) << 1)) * (value >> 1);
313 if (!value)
314 skip_bits_long(gb, -1);
315 factor = value;
316 } else {
317 value += scale << 4;
318 skip_bits_long(gb, quant_spectrum_bits[value] - nb_bits);
319 factor = quant_spectrum_value[value];
320 }
321 ch->imdct_in[i] = factor * ch->base[i];
322 }
323
324 memset(ch->imdct_in + ch->count, 0, sizeof(ch->imdct_in) - ch->count * sizeof(ch->imdct_in[0]));
325 }
326
unpack(HCAContext * c,ChannelContext * ch,unsigned hfr_group_count,int packed_noise_level,const uint8_t * ath)327 static void unpack(HCAContext *c, ChannelContext *ch,
328 unsigned hfr_group_count,
329 int packed_noise_level,
330 const uint8_t *ath)
331 {
332 GetBitContext *gb = &c->gb;
333 int delta_bits = get_bits(gb, 3);
334
335 if (delta_bits > 5) {
336 for (int i = 0; i < ch->count; i++)
337 ch->scale_factors[i] = get_bits(gb, 6);
338 } else if (delta_bits) {
339 int factor = get_bits(gb, 6);
340 int max_value = (1 << delta_bits) - 1;
341 int half_max = max_value >> 1;
342
343 ch->scale_factors[0] = factor;
344 for (int i = 1; i < ch->count; i++){
345 int delta = get_bits(gb, delta_bits);
346
347 if (delta == max_value) {
348 factor = get_bits(gb, 6);
349 } else {
350 factor += delta - half_max;
351 }
352 factor = av_clip_uintp2(factor, 6);
353
354 ch->scale_factors[i] = factor;
355 }
356 } else {
357 memset(ch->scale_factors, 0, 128);
358 }
359
360 if (ch->chan_type == 2){
361 ch->intensity[0] = get_bits(gb, 4);
362 if (ch->intensity[0] < 15) {
363 for (int i = 1; i < 8; i++)
364 ch->intensity[i] = get_bits(gb, 4);
365 }
366 } else {
367 for (int i = 0; i < hfr_group_count; i++)
368 ch->hfr_scale[i] = get_bits(gb, 6);
369 }
370
371 for (int i = 0; i < ch->count; i++) {
372 int scale = ch->scale_factors[i];
373
374 if (scale) {
375 scale = c->ath[i] + ((packed_noise_level + i) >> 8) - ((scale * 5) >> 1) + 2;
376 scale = scale_table[av_clip(scale, 0, 58)];
377 }
378 ch->scale[i] = scale;
379 }
380
381 memset(ch->scale + ch->count, 0, sizeof(ch->scale) - ch->count);
382
383 for (int i = 0; i < ch->count; i++)
384 ch->base[i] = dequantizer_scaling_table[ch->scale_factors[i]] * quant_step_size[ch->scale[i]];
385 }
386
decode_frame(AVCodecContext * avctx,void * data,int * got_frame_ptr,AVPacket * avpkt)387 static int decode_frame(AVCodecContext *avctx, void *data,
388 int *got_frame_ptr, AVPacket *avpkt)
389 {
390 AVFrame *frame = data;
391 HCAContext *c = avctx->priv_data;
392 int ch, ret, packed_noise_level;
393 GetBitContext *gb = &c->gb;
394 float **samples;
395
396 if (avctx->err_recognition & AV_EF_CRCCHECK) {
397 if (av_crc(c->crc_table, 0, avpkt->data, avpkt->size))
398 return AVERROR_INVALIDDATA;
399 }
400
401 if ((ret = init_get_bits8(gb, avpkt->data, avpkt->size)) < 0)
402 return ret;
403
404 if (get_bits(gb, 16) != 0xFFFF)
405 return AVERROR_INVALIDDATA;
406
407 frame->nb_samples = 1024;
408 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
409 return ret;
410 samples = (float **)frame->extended_data;
411
412 packed_noise_level = (get_bits(gb, 9) << 8) - get_bits(gb, 7);
413
414 for (ch = 0; ch < avctx->channels; ch++)
415 unpack(c, &c->ch[ch], c->hfr_group_count, packed_noise_level, c->ath);
416
417 for (int i = 0; i < 8; i++) {
418 for (ch = 0; ch < avctx->channels; ch++)
419 dequantize_coefficients(c, &c->ch[ch]);
420 for (ch = 0; ch < avctx->channels; ch++)
421 reconstruct_hfr(c, &c->ch[ch], c->hfr_group_count, c->bands_per_hfr_group,
422 c->stereo_band_count + c->base_band_count, c->total_band_count);
423 for (ch = 0; ch < avctx->channels - 1; ch++)
424 apply_intensity_stereo(c, &c->ch[ch], &c->ch[ch+1], i,
425 c->total_band_count - c->base_band_count,
426 c->base_band_count, c->stereo_band_count);
427 for (ch = 0; ch < avctx->channels; ch++)
428 run_imdct(c, &c->ch[ch], i, samples[ch] + i * 128);
429 }
430
431 *got_frame_ptr = 1;
432
433 return avpkt->size;
434 }
435
decode_close(AVCodecContext * avctx)436 static av_cold int decode_close(AVCodecContext *avctx)
437 {
438 HCAContext *c = avctx->priv_data;
439
440 av_freep(&c->fdsp);
441 av_tx_uninit(&c->tx_ctx);
442
443 return 0;
444 }
445
446 AVCodec ff_hca_decoder = {
447 .name = "hca",
448 .long_name = NULL_IF_CONFIG_SMALL("CRI HCA"),
449 .type = AVMEDIA_TYPE_AUDIO,
450 .id = AV_CODEC_ID_HCA,
451 .priv_data_size = sizeof(HCAContext),
452 .init = decode_init,
453 .decode = decode_frame,
454 .close = decode_close,
455 .capabilities = AV_CODEC_CAP_DR1,
456 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
457 AV_SAMPLE_FMT_NONE },
458 };
459