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