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/mem_internal.h"
23 #include "libavutil/tx.h"
24
25 #include "avcodec.h"
26 #include "get_bits.h"
27 #include "internal.h"
28 #include "hca_data.h"
29
30 typedef struct ChannelContext {
31 float base[128];
32 DECLARE_ALIGNED(32, float, imdct_in)[128];
33 DECLARE_ALIGNED(32, float, imdct_out)[128];
34 DECLARE_ALIGNED(32, float, imdct_prev)[128];
35 int8_t scale_factors[128];
36 uint8_t scale[128];
37 int8_t intensity[8];
38 int8_t *hfr_scale;
39 unsigned count;
40 int chan_type;
41 } ChannelContext;
42
43 typedef struct HCAContext {
44 GetBitContext gb;
45
46 const AVCRC *crc_table;
47
48 ChannelContext ch[16];
49
50 uint8_t ath[128];
51
52 int ath_type;
53 unsigned hfr_group_count;
54 uint8_t track_count;
55 uint8_t channel_config;
56 uint8_t total_band_count;
57 uint8_t base_band_count;
58 uint8_t stereo_band_count;
59 uint8_t bands_per_hfr_group;
60
61 av_tx_fn tx_fn;
62 AVTXContext *tx_ctx;
63 AVFloatDSPContext *fdsp;
64 } HCAContext;
65
ath_init1(uint8_t * ath,int sample_rate)66 static void ath_init1(uint8_t *ath, int sample_rate)
67 {
68 unsigned int index;
69 unsigned int acc = 0;
70
71 for (int i = 0; i < 128; i++) {
72 acc += sample_rate;
73 index = acc >> 13;
74
75 if (index >= 654) {
76 memset(ath+i, 0xFF, (128 - i));
77 break;
78 }
79
80 ath[i] = ath_base_curve[index];
81 }
82 }
83
ath_init(uint8_t * ath,int type,int sample_rate)84 static int ath_init(uint8_t *ath, int type, int sample_rate)
85 {
86 switch (type) {
87 case 0:
88 /* nothing to do */
89 break;
90 case 1:
91 ath_init1(ath, sample_rate);
92 break;
93 default:
94 return AVERROR_INVALIDDATA;
95 }
96
97 return 0;
98 }
99
ceil2(unsigned a,unsigned b)100 static inline unsigned ceil2(unsigned a, unsigned b)
101 {
102 return (b > 0) ? (a / b + ((a % b) ? 1 : 0)) : 0;
103 }
104
decode_init(AVCodecContext * avctx)105 static av_cold int decode_init(AVCodecContext *avctx)
106 {
107 HCAContext *c = avctx->priv_data;
108 GetBitContext *gb = &c->gb;
109 int8_t r[16] = { 0 };
110 float scale = 1.f / 8.f;
111 unsigned b, chunk;
112 int version, ret;
113
114 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
115 c->crc_table = av_crc_get_table(AV_CRC_16_ANSI);
116
117 if (avctx->channels <= 0 || avctx->channels > 16)
118 return AVERROR(EINVAL);
119
120 ret = init_get_bits8(gb, avctx->extradata, avctx->extradata_size);
121 if (ret < 0)
122 return ret;
123 skip_bits_long(gb, 32);
124 version = get_bits(gb, 16);
125 skip_bits_long(gb, 16);
126
127 c->ath_type = version >= 0x200 ? 0 : 1;
128
129 if (get_bits_long(gb, 32) != MKBETAG('f', 'm', 't', 0))
130 return AVERROR_INVALIDDATA;
131 skip_bits_long(gb, 32);
132 skip_bits_long(gb, 32);
133 skip_bits_long(gb, 32);
134
135 chunk = get_bits_long(gb, 32);
136 if (chunk == MKBETAG('c', 'o', 'm', 'p')) {
137 skip_bits_long(gb, 16);
138 skip_bits_long(gb, 8);
139 skip_bits_long(gb, 8);
140 c->track_count = get_bits(gb, 8);
141 c->channel_config = get_bits(gb, 8);
142 c->total_band_count = get_bits(gb, 8);
143 c->base_band_count = get_bits(gb, 8);
144 c->stereo_band_count = get_bits(gb, 8);
145 c->bands_per_hfr_group = get_bits(gb, 8);
146 } else if (chunk == MKBETAG('d', 'e', 'c', 0)) {
147 skip_bits_long(gb, 16);
148 skip_bits_long(gb, 8);
149 skip_bits_long(gb, 8);
150 c->total_band_count = get_bits(gb, 8) + 1;
151 c->base_band_count = get_bits(gb, 8) + 1;
152 c->track_count = get_bits(gb, 4);
153 c->channel_config = get_bits(gb, 4);
154 if (!get_bits(gb, 8))
155 c->base_band_count = c->total_band_count;
156 c->stereo_band_count = c->total_band_count - c->base_band_count;
157 c->bands_per_hfr_group = 0;
158 } else
159 return AVERROR_INVALIDDATA;
160
161 if (c->total_band_count > FF_ARRAY_ELEMS(c->ch->imdct_in))
162 return AVERROR_INVALIDDATA;
163
164
165 while (get_bits_left(gb) >= 32) {
166 chunk = get_bits_long(gb, 32);
167 if (chunk == MKBETAG('v', 'b', 'r', 0)) {
168 skip_bits_long(gb, 16);
169 skip_bits_long(gb, 16);
170 } else if (chunk == MKBETAG('a', 't', 'h', 0)) {
171 c->ath_type = get_bits(gb, 16);
172 } else if (chunk == MKBETAG('r', 'v', 'a', 0)) {
173 skip_bits_long(gb, 32);
174 } else if (chunk == MKBETAG('c', 'o', 'm', 'm')) {
175 skip_bits_long(gb, get_bits(gb, 8) * 8);
176 } else if (chunk == MKBETAG('c', 'i', 'p', 'h')) {
177 skip_bits_long(gb, 16);
178 } else if (chunk == MKBETAG('l', 'o', 'o', 'p')) {
179 skip_bits_long(gb, 32);
180 skip_bits_long(gb, 32);
181 skip_bits_long(gb, 16);
182 skip_bits_long(gb, 16);
183 } else if (chunk == MKBETAG('p', 'a', 'd', 0)) {
184 break;
185 } else {
186 break;
187 }
188 }
189
190 ret = ath_init(c->ath, c->ath_type, avctx->sample_rate);
191 if (ret < 0)
192 return ret;
193
194 if (!c->track_count)
195 c->track_count = 1;
196
197 b = avctx->channels / c->track_count;
198 if (c->stereo_band_count && b > 1) {
199 int8_t *x = r;
200
201 for (int i = 0; i < c->track_count; i++, x+=b) {
202 switch (b) {
203 case 2:
204 case 3:
205 x[0] = 1;
206 x[1] = 2;
207 break;
208 case 4:
209 x[0]=1; x[1] = 2;
210 if (c->channel_config == 0) {
211 x[2]=1;
212 x[3]=2;
213 }
214 break;
215 case 5:
216 x[0]=1; x[1] = 2;
217 if (c->channel_config <= 2) {
218 x[3]=1;
219 x[4]=2;
220 }
221 break;
222 case 6:
223 case 7:
224 x[0] = 1; x[1] = 2; x[4] = 1; x[5] = 2;
225 break;
226 case 8:
227 x[0] = 1; x[1] = 2; x[4] = 1; x[5] = 2; x[6] = 1; x[7] = 2;
228 break;
229 }
230 }
231 }
232
233 if (c->total_band_count < c->base_band_count)
234 return AVERROR_INVALIDDATA;
235
236 c->hfr_group_count = ceil2(c->total_band_count - (c->base_band_count + c->stereo_band_count),
237 c->bands_per_hfr_group);
238
239 if (c->base_band_count + c->stereo_band_count + (unsigned long)c->hfr_group_count > 128ULL)
240 return AVERROR_INVALIDDATA;
241
242 for (int i = 0; i < avctx->channels; i++) {
243 c->ch[i].chan_type = r[i];
244 c->ch[i].count = c->base_band_count + ((r[i] != 2) ? c->stereo_band_count : 0);
245 c->ch[i].hfr_scale = &c->ch[i].scale_factors[c->base_band_count + c->stereo_band_count];
246 if (c->ch[i].count > 128)
247 return AVERROR_INVALIDDATA;
248 }
249
250 c->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
251 if (!c->fdsp)
252 return AVERROR(ENOMEM);
253
254 return av_tx_init(&c->tx_ctx, &c->tx_fn, AV_TX_FLOAT_MDCT, 1, 128, &scale, 0);
255 }
256
run_imdct(HCAContext * c,ChannelContext * ch,int index,float * out)257 static void run_imdct(HCAContext *c, ChannelContext *ch, int index, float *out)
258 {
259 c->tx_fn(c->tx_ctx, ch->imdct_out, ch->imdct_in, sizeof(float));
260
261 c->fdsp->vector_fmul_window(out, ch->imdct_prev + (128 >> 1),
262 ch->imdct_out, window, 128 >> 1);
263
264 memcpy(ch->imdct_prev, ch->imdct_out, 128 * sizeof(float));
265 }
266
apply_intensity_stereo(HCAContext * s,ChannelContext * ch1,ChannelContext * ch2,int index,unsigned band_count,unsigned base_band_count,unsigned stereo_band_count)267 static void apply_intensity_stereo(HCAContext *s, ChannelContext *ch1, ChannelContext *ch2,
268 int index, unsigned band_count, unsigned base_band_count,
269 unsigned stereo_band_count)
270 {
271 float ratio_l = intensity_ratio_table[ch2->intensity[index]];
272 float ratio_r = ratio_l - 2.0f;
273 float *c1 = &ch1->imdct_in[base_band_count];
274 float *c2 = &ch2->imdct_in[base_band_count];
275
276 if (ch1->chan_type != 1 || !stereo_band_count)
277 return;
278
279 for (int i = 0; i < band_count; i++) {
280 *(c2++) = *c1 * ratio_r;
281 *(c1++) *= ratio_l;
282 }
283 }
284
reconstruct_hfr(HCAContext * s,ChannelContext * ch,unsigned hfr_group_count,unsigned bands_per_hfr_group,unsigned start_band,unsigned total_band_count)285 static void reconstruct_hfr(HCAContext *s, ChannelContext *ch,
286 unsigned hfr_group_count,
287 unsigned bands_per_hfr_group,
288 unsigned start_band, unsigned total_band_count)
289 {
290 if (ch->chan_type == 2 || !bands_per_hfr_group)
291 return;
292
293 for (int i = 0, k = start_band, l = start_band - 1; i < hfr_group_count; i++){
294 for (int j = 0; j < bands_per_hfr_group && k < total_band_count && l >= 0; j++, k++, l--){
295 ch->imdct_in[k] = scale_conversion_table[ scale_conv_bias +
296 av_clip_intp2(ch->hfr_scale[i] - ch->scale_factors[l], 6) ] * ch->imdct_in[l];
297 }
298 }
299
300 ch->imdct_in[127] = 0;
301 }
302
dequantize_coefficients(HCAContext * c,ChannelContext * ch)303 static void dequantize_coefficients(HCAContext *c, ChannelContext *ch)
304 {
305 GetBitContext *gb = &c->gb;
306
307 for (int i = 0; i < ch->count; i++) {
308 unsigned scale = ch->scale[i];
309 int nb_bits = max_bits_table[scale];
310 int value = get_bitsz(gb, nb_bits);
311 float factor;
312
313 if (scale > 7) {
314 value = (1 - ((value & 1) << 1)) * (value >> 1);
315 if (!value)
316 skip_bits_long(gb, -1);
317 factor = value;
318 } else {
319 value += scale << 4;
320 skip_bits_long(gb, quant_spectrum_bits[value] - nb_bits);
321 factor = quant_spectrum_value[value];
322 }
323 ch->imdct_in[i] = factor * ch->base[i];
324 }
325
326 memset(ch->imdct_in + ch->count, 0, sizeof(ch->imdct_in) - ch->count * sizeof(ch->imdct_in[0]));
327 }
328
unpack(HCAContext * c,ChannelContext * ch,unsigned hfr_group_count,int packed_noise_level,const uint8_t * ath)329 static void unpack(HCAContext *c, ChannelContext *ch,
330 unsigned hfr_group_count,
331 int packed_noise_level,
332 const uint8_t *ath)
333 {
334 GetBitContext *gb = &c->gb;
335 int delta_bits = get_bits(gb, 3);
336
337 if (delta_bits > 5) {
338 for (int i = 0; i < ch->count; i++)
339 ch->scale_factors[i] = get_bits(gb, 6);
340 } else if (delta_bits) {
341 int factor = get_bits(gb, 6);
342 int max_value = (1 << delta_bits) - 1;
343 int half_max = max_value >> 1;
344
345 ch->scale_factors[0] = factor;
346 for (int i = 1; i < ch->count; i++){
347 int delta = get_bits(gb, delta_bits);
348
349 if (delta == max_value) {
350 factor = get_bits(gb, 6);
351 } else {
352 factor += delta - half_max;
353 }
354 factor = av_clip_uintp2(factor, 6);
355
356 ch->scale_factors[i] = factor;
357 }
358 } else {
359 memset(ch->scale_factors, 0, 128);
360 }
361
362 if (ch->chan_type == 2){
363 ch->intensity[0] = get_bits(gb, 4);
364 if (ch->intensity[0] < 15) {
365 for (int i = 1; i < 8; i++)
366 ch->intensity[i] = get_bits(gb, 4);
367 }
368 } else {
369 for (int i = 0; i < hfr_group_count; i++)
370 ch->hfr_scale[i] = get_bits(gb, 6);
371 }
372
373 for (int i = 0; i < ch->count; i++) {
374 int scale = ch->scale_factors[i];
375
376 if (scale) {
377 scale = c->ath[i] + ((packed_noise_level + i) >> 8) - ((scale * 5) >> 1) + 2;
378 scale = scale_table[av_clip(scale, 0, 58)];
379 }
380 ch->scale[i] = scale;
381 }
382
383 memset(ch->scale + ch->count, 0, sizeof(ch->scale) - ch->count);
384
385 for (int i = 0; i < ch->count; i++)
386 ch->base[i] = dequantizer_scaling_table[ch->scale_factors[i]] * quant_step_size[ch->scale[i]];
387 }
388
decode_frame(AVCodecContext * avctx,void * data,int * got_frame_ptr,AVPacket * avpkt)389 static int decode_frame(AVCodecContext *avctx, void *data,
390 int *got_frame_ptr, AVPacket *avpkt)
391 {
392 AVFrame *frame = data;
393 HCAContext *c = avctx->priv_data;
394 int ch, ret, packed_noise_level;
395 GetBitContext *gb = &c->gb;
396 float **samples;
397
398 if (avctx->err_recognition & AV_EF_CRCCHECK) {
399 if (av_crc(c->crc_table, 0, avpkt->data, avpkt->size))
400 return AVERROR_INVALIDDATA;
401 }
402
403 if ((ret = init_get_bits8(gb, avpkt->data, avpkt->size)) < 0)
404 return ret;
405
406 if (get_bits(gb, 16) != 0xFFFF)
407 return AVERROR_INVALIDDATA;
408
409 frame->nb_samples = 1024;
410 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
411 return ret;
412 samples = (float **)frame->extended_data;
413
414 packed_noise_level = (get_bits(gb, 9) << 8) - get_bits(gb, 7);
415
416 for (ch = 0; ch < avctx->channels; ch++)
417 unpack(c, &c->ch[ch], c->hfr_group_count, packed_noise_level, c->ath);
418
419 for (int i = 0; i < 8; i++) {
420 for (ch = 0; ch < avctx->channels; ch++)
421 dequantize_coefficients(c, &c->ch[ch]);
422 for (ch = 0; ch < avctx->channels; ch++)
423 reconstruct_hfr(c, &c->ch[ch], c->hfr_group_count, c->bands_per_hfr_group,
424 c->stereo_band_count + c->base_band_count, c->total_band_count);
425 for (ch = 0; ch < avctx->channels - 1; ch++)
426 apply_intensity_stereo(c, &c->ch[ch], &c->ch[ch+1], i,
427 c->total_band_count - c->base_band_count,
428 c->base_band_count, c->stereo_band_count);
429 for (ch = 0; ch < avctx->channels; ch++)
430 run_imdct(c, &c->ch[ch], i, samples[ch] + i * 128);
431 }
432
433 *got_frame_ptr = 1;
434
435 return avpkt->size;
436 }
437
decode_close(AVCodecContext * avctx)438 static av_cold int decode_close(AVCodecContext *avctx)
439 {
440 HCAContext *c = avctx->priv_data;
441
442 av_freep(&c->fdsp);
443 av_tx_uninit(&c->tx_ctx);
444
445 return 0;
446 }
447
448 AVCodec ff_hca_decoder = {
449 .name = "hca",
450 .long_name = NULL_IF_CONFIG_SMALL("CRI HCA"),
451 .type = AVMEDIA_TYPE_AUDIO,
452 .id = AV_CODEC_ID_HCA,
453 .priv_data_size = sizeof(HCAContext),
454 .init = decode_init,
455 .decode = decode_frame,
456 .close = decode_close,
457 .capabilities = AV_CODEC_CAP_DR1,
458 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
459 AV_SAMPLE_FMT_NONE },
460 };
461