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 "bytestream.h"
27 #include "codec_internal.h"
28 #include "get_bits.h"
29 #include "internal.h"
30 #include "hca_data.h"
31
32 typedef struct ChannelContext {
33 float base[128];
34 DECLARE_ALIGNED(32, float, imdct_in)[128];
35 DECLARE_ALIGNED(32, float, imdct_out)[128];
36 DECLARE_ALIGNED(32, float, imdct_prev)[128];
37 int8_t scale_factors[128];
38 uint8_t scale[128];
39 int8_t intensity[8];
40 int8_t *hfr_scale;
41 unsigned count;
42 int chan_type;
43 } ChannelContext;
44
45 typedef struct HCAContext {
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 GetByteContext gb0, *const gb = &gb0;
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->ch_layout.nb_channels <= 0 || avctx->ch_layout.nb_channels > 16)
118 return AVERROR(EINVAL);
119
120 if (avctx->extradata_size < 36)
121 return AVERROR_INVALIDDATA;
122 bytestream2_init(gb, avctx->extradata, avctx->extradata_size);
123
124 bytestream2_skipu(gb, 4);
125 version = bytestream2_get_be16(gb);
126 bytestream2_skipu(gb, 2);
127
128 c->ath_type = version >= 0x200 ? 0 : 1;
129
130 if (bytestream2_get_be32u(gb) != MKBETAG('f', 'm', 't', 0))
131 return AVERROR_INVALIDDATA;
132 bytestream2_skipu(gb, 4);
133 bytestream2_skipu(gb, 4);
134 bytestream2_skipu(gb, 4);
135
136 chunk = bytestream2_get_be32u(gb);
137 if (chunk == MKBETAG('c', 'o', 'm', 'p')) {
138 bytestream2_skipu(gb, 2);
139 bytestream2_skipu(gb, 1);
140 bytestream2_skipu(gb, 1);
141 c->track_count = bytestream2_get_byteu(gb);
142 c->channel_config = bytestream2_get_byteu(gb);
143 c->total_band_count = bytestream2_get_byteu(gb);
144 c->base_band_count = bytestream2_get_byteu(gb);
145 c->stereo_band_count = bytestream2_get_byte (gb);
146 c->bands_per_hfr_group = bytestream2_get_byte (gb);
147 } else if (chunk == MKBETAG('d', 'e', 'c', 0)) {
148 bytestream2_skipu(gb, 2);
149 bytestream2_skipu(gb, 1);
150 bytestream2_skipu(gb, 1);
151 c->total_band_count = bytestream2_get_byteu(gb) + 1;
152 c->base_band_count = bytestream2_get_byteu(gb) + 1;
153 c->track_count = bytestream2_peek_byteu(gb) >> 4;
154 c->channel_config = bytestream2_get_byteu(gb) & 0xF;
155 if (!bytestream2_get_byteu(gb))
156 c->base_band_count = c->total_band_count;
157 c->stereo_band_count = c->total_band_count - c->base_band_count;
158 c->bands_per_hfr_group = 0;
159 } else
160 return AVERROR_INVALIDDATA;
161
162 if (c->total_band_count > FF_ARRAY_ELEMS(c->ch->imdct_in))
163 return AVERROR_INVALIDDATA;
164
165
166 while (bytestream2_get_bytes_left(gb) >= 4) {
167 chunk = bytestream2_get_be32u(gb);
168 if (chunk == MKBETAG('v', 'b', 'r', 0)) {
169 bytestream2_skip(gb, 2 + 2);
170 } else if (chunk == MKBETAG('a', 't', 'h', 0)) {
171 c->ath_type = bytestream2_get_be16(gb);
172 } else if (chunk == MKBETAG('r', 'v', 'a', 0)) {
173 bytestream2_skip(gb, 4);
174 } else if (chunk == MKBETAG('c', 'o', 'm', 'm')) {
175 bytestream2_skip(gb, bytestream2_get_byte(gb) * 8);
176 } else if (chunk == MKBETAG('c', 'i', 'p', 'h')) {
177 bytestream2_skip(gb, 2);
178 } else if (chunk == MKBETAG('l', 'o', 'o', 'p')) {
179 bytestream2_skip(gb, 4 + 4 + 2 + 2);
180 } else if (chunk == MKBETAG('p', 'a', 'd', 0)) {
181 break;
182 } else {
183 break;
184 }
185 }
186
187 ret = ath_init(c->ath, c->ath_type, avctx->sample_rate);
188 if (ret < 0)
189 return ret;
190
191 if (!c->track_count)
192 c->track_count = 1;
193
194 b = avctx->ch_layout.nb_channels / c->track_count;
195 if (c->stereo_band_count && b > 1) {
196 int8_t *x = r;
197
198 for (int i = 0; i < c->track_count; i++, x+=b) {
199 switch (b) {
200 case 2:
201 case 3:
202 x[0] = 1;
203 x[1] = 2;
204 break;
205 case 4:
206 x[0]=1; x[1] = 2;
207 if (c->channel_config == 0) {
208 x[2]=1;
209 x[3]=2;
210 }
211 break;
212 case 5:
213 x[0]=1; x[1] = 2;
214 if (c->channel_config <= 2) {
215 x[3]=1;
216 x[4]=2;
217 }
218 break;
219 case 6:
220 case 7:
221 x[0] = 1; x[1] = 2; x[4] = 1; x[5] = 2;
222 break;
223 case 8:
224 x[0] = 1; x[1] = 2; x[4] = 1; x[5] = 2; x[6] = 1; x[7] = 2;
225 break;
226 }
227 }
228 }
229
230 if (c->total_band_count < c->base_band_count)
231 return AVERROR_INVALIDDATA;
232
233 c->hfr_group_count = ceil2(c->total_band_count - (c->base_band_count + c->stereo_band_count),
234 c->bands_per_hfr_group);
235
236 if (c->base_band_count + c->stereo_band_count + (unsigned long)c->hfr_group_count > 128ULL)
237 return AVERROR_INVALIDDATA;
238
239 for (int i = 0; i < avctx->ch_layout.nb_channels; i++) {
240 c->ch[i].chan_type = r[i];
241 c->ch[i].count = c->base_band_count + ((r[i] != 2) ? c->stereo_band_count : 0);
242 c->ch[i].hfr_scale = &c->ch[i].scale_factors[c->base_band_count + c->stereo_band_count];
243 if (c->ch[i].count > 128)
244 return AVERROR_INVALIDDATA;
245 }
246
247 c->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
248 if (!c->fdsp)
249 return AVERROR(ENOMEM);
250
251 return av_tx_init(&c->tx_ctx, &c->tx_fn, AV_TX_FLOAT_MDCT, 1, 128, &scale, 0);
252 }
253
run_imdct(HCAContext *c, ChannelContext *ch, int index, float *out)254 static void run_imdct(HCAContext *c, ChannelContext *ch, int index, float *out)
255 {
256 c->tx_fn(c->tx_ctx, ch->imdct_out, ch->imdct_in, sizeof(float));
257
258 c->fdsp->vector_fmul_window(out, ch->imdct_prev + (128 >> 1),
259 ch->imdct_out, window, 128 >> 1);
260
261 memcpy(ch->imdct_prev, ch->imdct_out, 128 * sizeof(float));
262 }
263
apply_intensity_stereo(HCAContext *s, ChannelContext *ch1, ChannelContext *ch2, int index, unsigned band_count, unsigned base_band_count, unsigned stereo_band_count)264 static void apply_intensity_stereo(HCAContext *s, ChannelContext *ch1, ChannelContext *ch2,
265 int index, unsigned band_count, unsigned base_band_count,
266 unsigned stereo_band_count)
267 {
268 float ratio_l = intensity_ratio_table[ch2->intensity[index]];
269 float ratio_r = ratio_l - 2.0f;
270 float *c1 = &ch1->imdct_in[base_band_count];
271 float *c2 = &ch2->imdct_in[base_band_count];
272
273 if (ch1->chan_type != 1 || !stereo_band_count)
274 return;
275
276 for (int i = 0; i < band_count; i++) {
277 *(c2++) = *c1 * ratio_r;
278 *(c1++) *= ratio_l;
279 }
280 }
281
reconstruct_hfr(HCAContext *s, ChannelContext *ch, unsigned hfr_group_count, unsigned bands_per_hfr_group, unsigned start_band, unsigned total_band_count)282 static void reconstruct_hfr(HCAContext *s, ChannelContext *ch,
283 unsigned hfr_group_count,
284 unsigned bands_per_hfr_group,
285 unsigned start_band, unsigned total_band_count)
286 {
287 if (ch->chan_type == 2 || !bands_per_hfr_group)
288 return;
289
290 for (int i = 0, k = start_band, l = start_band - 1; i < hfr_group_count; i++){
291 for (int j = 0; j < bands_per_hfr_group && k < total_band_count && l >= 0; j++, k++, l--){
292 ch->imdct_in[k] = scale_conversion_table[ scale_conv_bias +
293 av_clip_intp2(ch->hfr_scale[i] - ch->scale_factors[l], 6) ] * ch->imdct_in[l];
294 }
295 }
296
297 ch->imdct_in[127] = 0;
298 }
299
dequantize_coefficients(HCAContext *c, ChannelContext *ch, GetBitContext *gb)300 static void dequantize_coefficients(HCAContext *c, ChannelContext *ch,
301 GetBitContext *gb)
302 {
303 for (int i = 0; i < ch->count; i++) {
304 unsigned scale = ch->scale[i];
305 int nb_bits = max_bits_table[scale];
306 int value = get_bitsz(gb, nb_bits);
307 float factor;
308
309 if (scale > 7) {
310 value = (1 - ((value & 1) << 1)) * (value >> 1);
311 if (!value)
312 skip_bits_long(gb, -1);
313 factor = value;
314 } else {
315 value += scale << 4;
316 skip_bits_long(gb, quant_spectrum_bits[value] - nb_bits);
317 factor = quant_spectrum_value[value];
318 }
319 ch->imdct_in[i] = factor * ch->base[i];
320 }
321
322 memset(ch->imdct_in + ch->count, 0, sizeof(ch->imdct_in) - ch->count * sizeof(ch->imdct_in[0]));
323 }
324
unpack(HCAContext *c, ChannelContext *ch, GetBitContext *gb, unsigned hfr_group_count, int packed_noise_level, const uint8_t *ath)325 static void unpack(HCAContext *c, ChannelContext *ch,
326 GetBitContext *gb,
327 unsigned hfr_group_count,
328 int packed_noise_level,
329 const uint8_t *ath)
330 {
331 int delta_bits = get_bits(gb, 3);
332
333 if (delta_bits > 5) {
334 for (int i = 0; i < ch->count; i++)
335 ch->scale_factors[i] = get_bits(gb, 6);
336 } else if (delta_bits) {
337 int factor = get_bits(gb, 6);
338 int max_value = (1 << delta_bits) - 1;
339 int half_max = max_value >> 1;
340
341 ch->scale_factors[0] = factor;
342 for (int i = 1; i < ch->count; i++){
343 int delta = get_bits(gb, delta_bits);
344
345 if (delta == max_value) {
346 factor = get_bits(gb, 6);
347 } else {
348 factor += delta - half_max;
349 }
350 factor = av_clip_uintp2(factor, 6);
351
352 ch->scale_factors[i] = factor;
353 }
354 } else {
355 memset(ch->scale_factors, 0, 128);
356 }
357
358 if (ch->chan_type == 2){
359 ch->intensity[0] = get_bits(gb, 4);
360 if (ch->intensity[0] < 15) {
361 for (int i = 1; i < 8; i++)
362 ch->intensity[i] = get_bits(gb, 4);
363 }
364 } else {
365 for (int i = 0; i < hfr_group_count; i++)
366 ch->hfr_scale[i] = get_bits(gb, 6);
367 }
368
369 for (int i = 0; i < ch->count; i++) {
370 int scale = ch->scale_factors[i];
371
372 if (scale) {
373 scale = c->ath[i] + ((packed_noise_level + i) >> 8) - ((scale * 5) >> 1) + 2;
374 scale = scale_table[av_clip(scale, 0, 58)];
375 }
376 ch->scale[i] = scale;
377 }
378
379 memset(ch->scale + ch->count, 0, sizeof(ch->scale) - ch->count);
380
381 for (int i = 0; i < ch->count; i++)
382 ch->base[i] = dequantizer_scaling_table[ch->scale_factors[i]] * quant_step_size[ch->scale[i]];
383 }
384
decode_frame(AVCodecContext *avctx, AVFrame *frame, int *got_frame_ptr, AVPacket *avpkt)385 static int decode_frame(AVCodecContext *avctx, AVFrame *frame,
386 int *got_frame_ptr, AVPacket *avpkt)
387 {
388 HCAContext *c = avctx->priv_data;
389 int ch, ret, packed_noise_level;
390 GetBitContext gb0, *const gb = &gb0;
391 float **samples;
392
393 if (avctx->err_recognition & AV_EF_CRCCHECK) {
394 if (av_crc(c->crc_table, 0, avpkt->data, avpkt->size))
395 return AVERROR_INVALIDDATA;
396 }
397
398 if ((ret = init_get_bits8(gb, avpkt->data, avpkt->size)) < 0)
399 return ret;
400
401 if (get_bits(gb, 16) != 0xFFFF)
402 return AVERROR_INVALIDDATA;
403
404 frame->nb_samples = 1024;
405 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
406 return ret;
407 samples = (float **)frame->extended_data;
408
409 packed_noise_level = (get_bits(gb, 9) << 8) - get_bits(gb, 7);
410
411 for (ch = 0; ch < avctx->ch_layout.nb_channels; ch++)
412 unpack(c, &c->ch[ch], gb, c->hfr_group_count, packed_noise_level, c->ath);
413
414 for (int i = 0; i < 8; i++) {
415 for (ch = 0; ch < avctx->ch_layout.nb_channels; ch++)
416 dequantize_coefficients(c, &c->ch[ch], gb);
417 for (ch = 0; ch < avctx->ch_layout.nb_channels; ch++)
418 reconstruct_hfr(c, &c->ch[ch], c->hfr_group_count, c->bands_per_hfr_group,
419 c->stereo_band_count + c->base_band_count, c->total_band_count);
420 for (ch = 0; ch < avctx->ch_layout.nb_channels - 1; ch++)
421 apply_intensity_stereo(c, &c->ch[ch], &c->ch[ch+1], i,
422 c->total_band_count - c->base_band_count,
423 c->base_band_count, c->stereo_band_count);
424 for (ch = 0; ch < avctx->ch_layout.nb_channels; ch++)
425 run_imdct(c, &c->ch[ch], i, samples[ch] + i * 128);
426 }
427
428 *got_frame_ptr = 1;
429
430 return avpkt->size;
431 }
432
decode_close(AVCodecContext *avctx)433 static av_cold int decode_close(AVCodecContext *avctx)
434 {
435 HCAContext *c = avctx->priv_data;
436
437 av_freep(&c->fdsp);
438 av_tx_uninit(&c->tx_ctx);
439
440 return 0;
441 }
442
443 const FFCodec ff_hca_decoder = {
444 .p.name = "hca",
445 .p.long_name = NULL_IF_CONFIG_SMALL("CRI HCA"),
446 .p.type = AVMEDIA_TYPE_AUDIO,
447 .p.id = AV_CODEC_ID_HCA,
448 .priv_data_size = sizeof(HCAContext),
449 .init = decode_init,
450 FF_CODEC_DECODE_CB(decode_frame),
451 .close = decode_close,
452 .p.capabilities = AV_CODEC_CAP_DR1,
453 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_INIT_CLEANUP,
454 .p.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
455 AV_SAMPLE_FMT_NONE },
456 };
457