b9d3df89c890653b10ee4e4ac432cd18ac4b26b7
[libav.git] / libavcodec / binkaudio.c
1 /*
2 * Bink Audio decoder
3 * Copyright (c) 2007-2010 Peter Ross (pross@xvid.org)
4 * Copyright (c) 2009 Daniel Verkamp (daniel@drv.nu)
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg 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 * FFmpeg 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 FFmpeg; 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 libavcodec/binkaudio.c
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 "fft.h"
36
37 extern const uint16_t ff_wma_critical_freqs[25];
38
39 #define MAX_CHANNELS 2
40 #define BINK_BLOCK_MAX_SIZE (MAX_CHANNELS << 11)
41
42 typedef struct {
43 AVCodecContext *avctx;
44 GetBitContext gb;
45 DSPContext dsp;
46 int first;
47 int channels;
48 int frame_len; ///< transform size (samples)
49 int overlap_len; ///< overlap size (samples)
50 int block_size;
51 int num_bands;
52 unsigned int *bands;
53 float root;
54 DECLARE_ALIGNED(16, FFTSample, coeffs)[BINK_BLOCK_MAX_SIZE];
55 DECLARE_ALIGNED(16, short, previous)[BINK_BLOCK_MAX_SIZE / 16]; ///< coeffs from previous audio block
56 float *coeffs_ptr[MAX_CHANNELS]; ///< pointers to the coeffs arrays for float_to_int16_interleave
57 union {
58 RDFTContext rdft;
59 DCTContext dct;
60 } trans;
61 } BinkAudioContext;
62
63
64 static av_cold int decode_init(AVCodecContext *avctx)
65 {
66 BinkAudioContext *s = avctx->priv_data;
67 int sample_rate = avctx->sample_rate;
68 int sample_rate_half;
69 int i;
70 int frame_len_bits;
71
72 s->avctx = avctx;
73 dsputil_init(&s->dsp, avctx);
74
75 /* determine frame length */
76 if (avctx->sample_rate < 22050) {
77 frame_len_bits = 9;
78 } else if (avctx->sample_rate < 44100) {
79 frame_len_bits = 10;
80 } else {
81 frame_len_bits = 11;
82 }
83 s->frame_len = 1 << frame_len_bits;
84
85 if (s->channels > MAX_CHANNELS) {
86 av_log(s->avctx, AV_LOG_ERROR, "too many channels: %d\n", s->channels);
87 return -1;
88 }
89
90 if (avctx->codec->id == CODEC_ID_BINKAUDIO_RDFT) {
91 // audio is already interleaved for the RDFT format variant
92 sample_rate *= avctx->channels;
93 s->frame_len *= avctx->channels;
94 s->channels = 1;
95 if (avctx->channels == 2)
96 frame_len_bits++;
97 } else {
98 s->channels = avctx->channels;
99 }
100
101 s->overlap_len = s->frame_len / 16;
102 s->block_size = (s->frame_len - s->overlap_len) * s->channels;
103 sample_rate_half = (sample_rate + 1) / 2;
104 s->root = 2.0 / sqrt(s->frame_len);
105
106 /* calculate number of bands */
107 for (s->num_bands = 1; s->num_bands < 25; s->num_bands++)
108 if (sample_rate_half <= ff_wma_critical_freqs[s->num_bands - 1])
109 break;
110
111 s->bands = av_malloc((s->num_bands + 1) * sizeof(*s->bands));
112 if (!s->bands)
113 return AVERROR(ENOMEM);
114
115 /* populate bands data */
116 s->bands[0] = 1;
117 for (i = 1; i < s->num_bands; i++)
118 s->bands[i] = ff_wma_critical_freqs[i - 1] * (s->frame_len / 2) / sample_rate_half;
119 s->bands[s->num_bands] = s->frame_len / 2;
120
121 s->first = 1;
122 avctx->sample_fmt = SAMPLE_FMT_S16;
123
124 for (i = 0; i < s->channels; i++)
125 s->coeffs_ptr[i] = s->coeffs + i * s->frame_len;
126
127 if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == CODEC_ID_BINKAUDIO_RDFT)
128 ff_rdft_init(&s->trans.rdft, frame_len_bits, DFT_C2R);
129 else if (CONFIG_BINKAUDIO_DCT_DECODER)
130 ff_dct_init(&s->trans.dct, frame_len_bits, 1);
131 else
132 return -1;
133
134 return 0;
135 }
136
137 static float get_float(GetBitContext *gb)
138 {
139 int power = get_bits(gb, 5);
140 float f = ldexpf(get_bits_long(gb, 23), power - 23);
141 if (get_bits1(gb))
142 f = -f;
143 return f;
144 }
145
146 static const uint8_t rle_length_tab[16] = {
147 2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 32, 64
148 };
149
150 /**
151 * Decode Bink Audio block
152 * @param[out] out Output buffer (must contain s->block_size elements)
153 */
154 static void decode_block(BinkAudioContext *s, short *out, int use_dct)
155 {
156 int ch, i, j, k;
157 float q, quant[25];
158 int width, coeff;
159 GetBitContext *gb = &s->gb;
160
161 if (use_dct)
162 skip_bits(gb, 2);
163
164 for (ch = 0; ch < s->channels; ch++) {
165 FFTSample *coeffs = s->coeffs_ptr[ch];
166 q = 0.0f;
167 coeffs[0] = get_float(gb) * s->root;
168 coeffs[1] = get_float(gb) * s->root;
169
170 for (i = 0; i < s->num_bands; i++) {
171 /* constant is result of 0.066399999/log10(M_E) */
172 int value = get_bits(gb, 8);
173 quant[i] = expf(FFMIN(value, 95) * 0.15289164787221953823f) * s->root;
174 }
175
176 // find band (k)
177 for (k = 0; s->bands[k] < 1; k++) {
178 q = quant[k];
179 }
180
181 // parse coefficients
182 i = 2;
183 while (i < s->frame_len) {
184 if (get_bits1(gb)) {
185 j = i + rle_length_tab[get_bits(gb, 4)] * 8;
186 } else {
187 j = i + 8;
188 }
189
190 j = FFMIN(j, s->frame_len);
191
192 width = get_bits(gb, 4);
193 if (width == 0) {
194 memset(coeffs + i, 0, (j - i) * sizeof(*coeffs));
195 i = j;
196 while (s->bands[k] * 2 < i)
197 q = quant[k++];
198 } else {
199 while (i < j) {
200 if (s->bands[k] * 2 == i)
201 q = quant[k++];
202 coeff = get_bits(gb, width);
203 if (coeff) {
204 if (get_bits1(gb))
205 coeffs[i] = -q * coeff;
206 else
207 coeffs[i] = q * coeff;
208 } else {
209 coeffs[i] = 0.0f;
210 }
211 i++;
212 }
213 }
214 }
215
216 if (CONFIG_BINKAUDIO_DCT_DECODER && use_dct) {
217 coeffs[0] /= 0.5;
218 ff_dct_calc (&s->trans.dct, coeffs);
219 s->dsp.vector_fmul_scalar(coeffs, coeffs, s->frame_len / 2, s->frame_len);
220 }
221 else if (CONFIG_BINKAUDIO_RDFT_DECODER)
222 ff_rdft_calc(&s->trans.rdft, coeffs);
223 }
224
225 s->dsp.float_to_int16_interleave(out, (const float **)s->coeffs_ptr, s->frame_len, s->channels);
226
227 if (!s->first) {
228 int count = s->overlap_len * s->channels;
229 int shift = av_log2(count);
230 for (i = 0; i < count; i++) {
231 out[i] = (s->previous[i] * (count - i) + out[i] * i) >> shift;
232 }
233 }
234
235 memcpy(s->previous, out + s->block_size,
236 s->overlap_len * s->channels * sizeof(*out));
237
238 s->first = 0;
239 }
240
241 static av_cold int decode_end(AVCodecContext *avctx)
242 {
243 BinkAudioContext * s = avctx->priv_data;
244 av_freep(&s->bands);
245 if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == CODEC_ID_BINKAUDIO_RDFT)
246 ff_rdft_end(&s->trans.rdft);
247 else if (CONFIG_BINKAUDIO_DCT_DECODER)
248 ff_dct_end(&s->trans.dct);
249 return 0;
250 }
251
252 static void get_bits_align32(GetBitContext *s)
253 {
254 int n = (-get_bits_count(s)) & 31;
255 if (n) skip_bits(s, n);
256 }
257
258 static int decode_frame(AVCodecContext *avctx,
259 void *data, int *data_size,
260 AVPacket *avpkt)
261 {
262 BinkAudioContext *s = avctx->priv_data;
263 const uint8_t *buf = avpkt->data;
264 int buf_size = avpkt->size;
265 short *samples = data;
266 short *samples_end = (short*)((uint8_t*)data + *data_size);
267 int reported_size;
268 GetBitContext *gb = &s->gb;
269
270 init_get_bits(gb, buf, buf_size * 8);
271
272 reported_size = get_bits_long(gb, 32);
273 while (get_bits_count(gb) / 8 < buf_size &&
274 samples + s->block_size <= samples_end) {
275 decode_block(s, samples, avctx->codec->id == CODEC_ID_BINKAUDIO_DCT);
276 samples += s->block_size;
277 get_bits_align32(gb);
278 }
279
280 *data_size = FFMIN(reported_size, (uint8_t*)samples - (uint8_t*)data);
281 return buf_size;
282 }
283
284 AVCodec binkaudio_rdft_decoder = {
285 "binkaudio_rdft",
286 CODEC_TYPE_AUDIO,
287 CODEC_ID_BINKAUDIO_RDFT,
288 sizeof(BinkAudioContext),
289 decode_init,
290 NULL,
291 decode_end,
292 decode_frame,
293 .long_name = NULL_IF_CONFIG_SMALL("Bink Audio (RDFT)")
294 };
295
296 AVCodec binkaudio_dct_decoder = {
297 "binkaudio_dct",
298 CODEC_TYPE_AUDIO,
299 CODEC_ID_BINKAUDIO_DCT,
300 sizeof(BinkAudioContext),
301 decode_init,
302 NULL,
303 decode_end,
304 decode_frame,
305 .long_name = NULL_IF_CONFIG_SMALL("Bink Audio (DCT)")
306 };