27e8ff5a84995e311f49f2f72b2130a2b2b032cb
[libav.git] / libavcodec / binkaudio.c
1 /*
2 * Bink Audio decoder
3 * Copyright (c) 2007-2011 Peter Ross (pross@xvid.org)
4 * Copyright (c) 2009 Daniel Verkamp (daniel@drv.nu)
5 *
6 * This file is part of Libav.
7 *
8 * Libav is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * Libav is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with Libav; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 /**
24 * @file
25 * Bink Audio decoder
26 *
27 * Technical details here:
28 * http://wiki.multimedia.cx/index.php?title=Bink_Audio
29 */
30
31 #include "avcodec.h"
32 #define ALT_BITSTREAM_READER_LE
33 #include "get_bits.h"
34 #include "dsputil.h"
35 #include "dct.h"
36 #include "rdft.h"
37 #include "fmtconvert.h"
38 #include "libavutil/intfloat_readwrite.h"
39
40 extern const uint16_t ff_wma_critical_freqs[25];
41
42 #define MAX_CHANNELS 2
43 #define BINK_BLOCK_MAX_SIZE (MAX_CHANNELS << 11)
44
45 typedef struct {
46 GetBitContext gb;
47 DSPContext dsp;
48 FmtConvertContext fmt_conv;
49 int version_b; ///< Bink version 'b'
50 int first;
51 int channels;
52 int frame_len; ///< transform size (samples)
53 int overlap_len; ///< overlap size (samples)
54 int block_size;
55 int num_bands;
56 unsigned int *bands;
57 float root;
58 DECLARE_ALIGNED(32, FFTSample, coeffs)[BINK_BLOCK_MAX_SIZE];
59 DECLARE_ALIGNED(16, short, previous)[BINK_BLOCK_MAX_SIZE / 16]; ///< coeffs from previous audio block
60 float *coeffs_ptr[MAX_CHANNELS]; ///< pointers to the coeffs arrays for float_to_int16_interleave
61 union {
62 RDFTContext rdft;
63 DCTContext dct;
64 } trans;
65 } BinkAudioContext;
66
67
68 static av_cold int decode_init(AVCodecContext *avctx)
69 {
70 BinkAudioContext *s = avctx->priv_data;
71 int sample_rate = avctx->sample_rate;
72 int sample_rate_half;
73 int i;
74 int frame_len_bits;
75
76 dsputil_init(&s->dsp, avctx);
77 ff_fmt_convert_init(&s->fmt_conv, avctx);
78
79 /* determine frame length */
80 if (avctx->sample_rate < 22050) {
81 frame_len_bits = 9;
82 } else if (avctx->sample_rate < 44100) {
83 frame_len_bits = 10;
84 } else {
85 frame_len_bits = 11;
86 }
87
88 if (avctx->channels > MAX_CHANNELS) {
89 av_log(avctx, AV_LOG_ERROR, "too many channels: %d\n", avctx->channels);
90 return -1;
91 }
92
93 s->version_b = avctx->extradata && avctx->extradata[3] == 'b';
94
95 if (avctx->codec->id == CODEC_ID_BINKAUDIO_RDFT) {
96 // audio is already interleaved for the RDFT format variant
97 sample_rate *= avctx->channels;
98 s->channels = 1;
99 if (!s->version_b)
100 frame_len_bits += av_log2(avctx->channels);
101 } else {
102 s->channels = avctx->channels;
103 }
104
105 s->frame_len = 1 << frame_len_bits;
106 s->overlap_len = s->frame_len / 16;
107 s->block_size = (s->frame_len - s->overlap_len) * s->channels;
108 sample_rate_half = (sample_rate + 1) / 2;
109 s->root = 2.0 / sqrt(s->frame_len);
110
111 /* calculate number of bands */
112 for (s->num_bands = 1; s->num_bands < 25; s->num_bands++)
113 if (sample_rate_half <= ff_wma_critical_freqs[s->num_bands - 1])
114 break;
115
116 s->bands = av_malloc((s->num_bands + 1) * sizeof(*s->bands));
117 if (!s->bands)
118 return AVERROR(ENOMEM);
119
120 /* populate bands data */
121 s->bands[0] = 2;
122 for (i = 1; i < s->num_bands; i++)
123 s->bands[i] = (ff_wma_critical_freqs[i - 1] * s->frame_len / sample_rate_half) & ~1;
124 s->bands[s->num_bands] = s->frame_len;
125
126 s->first = 1;
127 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
128
129 for (i = 0; i < s->channels; i++)
130 s->coeffs_ptr[i] = s->coeffs + i * s->frame_len;
131
132 if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == CODEC_ID_BINKAUDIO_RDFT)
133 ff_rdft_init(&s->trans.rdft, frame_len_bits, DFT_C2R);
134 else if (CONFIG_BINKAUDIO_DCT_DECODER)
135 ff_dct_init(&s->trans.dct, frame_len_bits, DCT_III);
136 else
137 return -1;
138
139 return 0;
140 }
141
142 static float get_float(GetBitContext *gb)
143 {
144 int power = get_bits(gb, 5);
145 float f = ldexpf(get_bits_long(gb, 23), power - 23);
146 if (get_bits1(gb))
147 f = -f;
148 return f;
149 }
150
151 static const uint8_t rle_length_tab[16] = {
152 2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 32, 64
153 };
154
155 #define GET_BITS_SAFE(out, nbits) do { \
156 if (get_bits_left(gb) < nbits) \
157 return AVERROR_INVALIDDATA; \
158 out = get_bits(gb, nbits); \
159 } while (0)
160
161 /**
162 * Decode Bink Audio block
163 * @param[out] out Output buffer (must contain s->block_size elements)
164 * @return 0 on success, negative error code on failure
165 */
166 static int decode_block(BinkAudioContext *s, short *out, int use_dct)
167 {
168 int ch, i, j, k;
169 float q, quant[25];
170 int width, coeff;
171 GetBitContext *gb = &s->gb;
172
173 if (use_dct)
174 skip_bits(gb, 2);
175
176 for (ch = 0; ch < s->channels; ch++) {
177 FFTSample *coeffs = s->coeffs_ptr[ch];
178 if (s->version_b) {
179 if (get_bits_left(gb) < 64)
180 return AVERROR_INVALIDDATA;
181 coeffs[0] = av_int2flt(get_bits(gb, 32)) * s->root;
182 coeffs[1] = av_int2flt(get_bits(gb, 32)) * s->root;
183 } else {
184 if (get_bits_left(gb) < 58)
185 return AVERROR_INVALIDDATA;
186 coeffs[0] = get_float(gb) * s->root;
187 coeffs[1] = get_float(gb) * s->root;
188 }
189
190 if (get_bits_left(gb) < s->num_bands * 8)
191 return AVERROR_INVALIDDATA;
192 for (i = 0; i < s->num_bands; i++) {
193 /* constant is result of 0.066399999/log10(M_E) */
194 int value = get_bits(gb, 8);
195 quant[i] = expf(FFMIN(value, 95) * 0.15289164787221953823f) * s->root;
196 }
197
198 k = 0;
199 q = quant[0];
200
201 // parse coefficients
202 i = 2;
203 while (i < s->frame_len) {
204 if (s->version_b) {
205 j = i + 16;
206 } else {
207 int v;
208 GET_BITS_SAFE(v, 1);
209 if (v) {
210 GET_BITS_SAFE(v, 4);
211 j = i + rle_length_tab[v] * 8;
212 } else {
213 j = i + 8;
214 }
215 }
216
217 j = FFMIN(j, s->frame_len);
218
219 GET_BITS_SAFE(width, 4);
220 if (width == 0) {
221 memset(coeffs + i, 0, (j - i) * sizeof(*coeffs));
222 i = j;
223 while (s->bands[k] < i)
224 q = quant[k++];
225 } else {
226 while (i < j) {
227 if (s->bands[k] == i)
228 q = quant[k++];
229 GET_BITS_SAFE(coeff, width);
230 if (coeff) {
231 int v;
232 GET_BITS_SAFE(v, 1);
233 if (v)
234 coeffs[i] = -q * coeff;
235 else
236 coeffs[i] = q * coeff;
237 } else {
238 coeffs[i] = 0.0f;
239 }
240 i++;
241 }
242 }
243 }
244
245 if (CONFIG_BINKAUDIO_DCT_DECODER && use_dct) {
246 coeffs[0] /= 0.5;
247 s->trans.dct.dct_calc(&s->trans.dct, coeffs);
248 s->dsp.vector_fmul_scalar(coeffs, coeffs, s->frame_len / 2, s->frame_len);
249 }
250 else if (CONFIG_BINKAUDIO_RDFT_DECODER)
251 s->trans.rdft.rdft_calc(&s->trans.rdft, coeffs);
252 }
253
254 s->fmt_conv.float_to_int16_interleave(out, (const float **)s->coeffs_ptr,
255 s->frame_len, s->channels);
256
257 if (!s->first) {
258 int count = s->overlap_len * s->channels;
259 int shift = av_log2(count);
260 for (i = 0; i < count; i++) {
261 out[i] = (s->previous[i] * (count - i) + out[i] * i) >> shift;
262 }
263 }
264
265 memcpy(s->previous, out + s->block_size,
266 s->overlap_len * s->channels * sizeof(*out));
267
268 s->first = 0;
269
270 return 0;
271 }
272
273 static av_cold int decode_end(AVCodecContext *avctx)
274 {
275 BinkAudioContext * s = avctx->priv_data;
276 av_freep(&s->bands);
277 if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == CODEC_ID_BINKAUDIO_RDFT)
278 ff_rdft_end(&s->trans.rdft);
279 else if (CONFIG_BINKAUDIO_DCT_DECODER)
280 ff_dct_end(&s->trans.dct);
281 return 0;
282 }
283
284 static void get_bits_align32(GetBitContext *s)
285 {
286 int n = (-get_bits_count(s)) & 31;
287 if (n) skip_bits(s, n);
288 }
289
290 static int decode_frame(AVCodecContext *avctx,
291 void *data, int *data_size,
292 AVPacket *avpkt)
293 {
294 BinkAudioContext *s = avctx->priv_data;
295 const uint8_t *buf = avpkt->data;
296 int buf_size = avpkt->size;
297 short *samples = data;
298 short *samples_end = (short*)((uint8_t*)data + *data_size);
299 int reported_size;
300 GetBitContext *gb = &s->gb;
301
302 if (buf_size < 4) {
303 av_log(avctx, AV_LOG_ERROR, "Packet is too small\n");
304 return AVERROR_INVALIDDATA;
305 }
306
307 init_get_bits(gb, buf, buf_size * 8);
308
309 reported_size = get_bits_long(gb, 32);
310 while (samples + s->block_size <= samples_end) {
311 if (decode_block(s, samples, avctx->codec->id == CODEC_ID_BINKAUDIO_DCT))
312 break;
313 samples += s->block_size;
314 get_bits_align32(gb);
315 }
316
317 *data_size = FFMIN(reported_size, (uint8_t*)samples - (uint8_t*)data);
318 return buf_size;
319 }
320
321 AVCodec ff_binkaudio_rdft_decoder = {
322 .name = "binkaudio_rdft",
323 .type = AVMEDIA_TYPE_AUDIO,
324 .id = CODEC_ID_BINKAUDIO_RDFT,
325 .priv_data_size = sizeof(BinkAudioContext),
326 .init = decode_init,
327 .close = decode_end,
328 .decode = decode_frame,
329 .long_name = NULL_IF_CONFIG_SMALL("Bink Audio (RDFT)")
330 };
331
332 AVCodec ff_binkaudio_dct_decoder = {
333 .name = "binkaudio_dct",
334 .type = AVMEDIA_TYPE_AUDIO,
335 .id = CODEC_ID_BINKAUDIO_DCT,
336 .priv_data_size = sizeof(BinkAudioContext),
337 .init = decode_init,
338 .close = decode_end,
339 .decode = decode_frame,
340 .long_name = NULL_IF_CONFIG_SMALL("Bink Audio (DCT)")
341 };