| 1 | /* |
| 2 | * samplerate conversion for both audio and video |
| 3 | * Copyright (c) 2000 Fabrice Bellard |
| 4 | * |
| 5 | * This file is part of FFmpeg. |
| 6 | * |
| 7 | * FFmpeg is free software; you can redistribute it and/or |
| 8 | * modify it under the terms of the GNU Lesser General Public |
| 9 | * License as published by the Free Software Foundation; either |
| 10 | * version 2.1 of the License, or (at your option) any later version. |
| 11 | * |
| 12 | * FFmpeg is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | * Lesser General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU Lesser General Public |
| 18 | * License along with FFmpeg; if not, write to the Free Software |
| 19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 20 | */ |
| 21 | |
| 22 | /** |
| 23 | * @file |
| 24 | * samplerate conversion for both audio and video |
| 25 | */ |
| 26 | |
| 27 | #include <string.h> |
| 28 | |
| 29 | #include "avcodec.h" |
| 30 | #include "audioconvert.h" |
| 31 | #include "libavutil/opt.h" |
| 32 | #include "libavutil/mem.h" |
| 33 | #include "libavutil/samplefmt.h" |
| 34 | |
| 35 | #if FF_API_AVCODEC_RESAMPLE |
| 36 | |
| 37 | #define MAX_CHANNELS 8 |
| 38 | |
| 39 | struct AVResampleContext; |
| 40 | |
| 41 | static const char *context_to_name(void *ptr) |
| 42 | { |
| 43 | return "audioresample"; |
| 44 | } |
| 45 | |
| 46 | static const AVOption options[] = {{NULL}}; |
| 47 | static const AVClass audioresample_context_class = { |
| 48 | "ReSampleContext", context_to_name, options, LIBAVUTIL_VERSION_INT |
| 49 | }; |
| 50 | |
| 51 | struct ReSampleContext { |
| 52 | struct AVResampleContext *resample_context; |
| 53 | short *temp[MAX_CHANNELS]; |
| 54 | int temp_len; |
| 55 | float ratio; |
| 56 | /* channel convert */ |
| 57 | int input_channels, output_channels, filter_channels; |
| 58 | AVAudioConvert *convert_ctx[2]; |
| 59 | enum AVSampleFormat sample_fmt[2]; ///< input and output sample format |
| 60 | unsigned sample_size[2]; ///< size of one sample in sample_fmt |
| 61 | short *buffer[2]; ///< buffers used for conversion to S16 |
| 62 | unsigned buffer_size[2]; ///< sizes of allocated buffers |
| 63 | }; |
| 64 | |
| 65 | /* n1: number of samples */ |
| 66 | static void stereo_to_mono(short *output, short *input, int n1) |
| 67 | { |
| 68 | short *p, *q; |
| 69 | int n = n1; |
| 70 | |
| 71 | p = input; |
| 72 | q = output; |
| 73 | while (n >= 4) { |
| 74 | q[0] = (p[0] + p[1]) >> 1; |
| 75 | q[1] = (p[2] + p[3]) >> 1; |
| 76 | q[2] = (p[4] + p[5]) >> 1; |
| 77 | q[3] = (p[6] + p[7]) >> 1; |
| 78 | q += 4; |
| 79 | p += 8; |
| 80 | n -= 4; |
| 81 | } |
| 82 | while (n > 0) { |
| 83 | q[0] = (p[0] + p[1]) >> 1; |
| 84 | q++; |
| 85 | p += 2; |
| 86 | n--; |
| 87 | } |
| 88 | } |
| 89 | |
| 90 | /* n1: number of samples */ |
| 91 | static void mono_to_stereo(short *output, short *input, int n1) |
| 92 | { |
| 93 | short *p, *q; |
| 94 | int n = n1; |
| 95 | int v; |
| 96 | |
| 97 | p = input; |
| 98 | q = output; |
| 99 | while (n >= 4) { |
| 100 | v = p[0]; q[0] = v; q[1] = v; |
| 101 | v = p[1]; q[2] = v; q[3] = v; |
| 102 | v = p[2]; q[4] = v; q[5] = v; |
| 103 | v = p[3]; q[6] = v; q[7] = v; |
| 104 | q += 8; |
| 105 | p += 4; |
| 106 | n -= 4; |
| 107 | } |
| 108 | while (n > 0) { |
| 109 | v = p[0]; q[0] = v; q[1] = v; |
| 110 | q += 2; |
| 111 | p += 1; |
| 112 | n--; |
| 113 | } |
| 114 | } |
| 115 | |
| 116 | /* |
| 117 | 5.1 to stereo input: [fl, fr, c, lfe, rl, rr] |
| 118 | - Left = front_left + rear_gain * rear_left + center_gain * center |
| 119 | - Right = front_right + rear_gain * rear_right + center_gain * center |
| 120 | Where rear_gain is usually around 0.5-1.0 and |
| 121 | center_gain is almost always 0.7 (-3 dB) |
| 122 | */ |
| 123 | static void surround_to_stereo(short **output, short *input, int channels, int samples) |
| 124 | { |
| 125 | int i; |
| 126 | short l, r; |
| 127 | |
| 128 | for (i = 0; i < samples; i++) { |
| 129 | int fl,fr,c,rl,rr; |
| 130 | fl = input[0]; |
| 131 | fr = input[1]; |
| 132 | c = input[2]; |
| 133 | // lfe = input[3]; |
| 134 | rl = input[4]; |
| 135 | rr = input[5]; |
| 136 | |
| 137 | l = av_clip_int16(fl + (0.5 * rl) + (0.7 * c)); |
| 138 | r = av_clip_int16(fr + (0.5 * rr) + (0.7 * c)); |
| 139 | |
| 140 | /* output l & r. */ |
| 141 | *output[0]++ = l; |
| 142 | *output[1]++ = r; |
| 143 | |
| 144 | /* increment input. */ |
| 145 | input += channels; |
| 146 | } |
| 147 | } |
| 148 | |
| 149 | static void deinterleave(short **output, short *input, int channels, int samples) |
| 150 | { |
| 151 | int i, j; |
| 152 | |
| 153 | for (i = 0; i < samples; i++) { |
| 154 | for (j = 0; j < channels; j++) { |
| 155 | *output[j]++ = *input++; |
| 156 | } |
| 157 | } |
| 158 | } |
| 159 | |
| 160 | static void interleave(short *output, short **input, int channels, int samples) |
| 161 | { |
| 162 | int i, j; |
| 163 | |
| 164 | for (i = 0; i < samples; i++) { |
| 165 | for (j = 0; j < channels; j++) { |
| 166 | *output++ = *input[j]++; |
| 167 | } |
| 168 | } |
| 169 | } |
| 170 | |
| 171 | static void ac3_5p1_mux(short *output, short *input1, short *input2, int n) |
| 172 | { |
| 173 | int i; |
| 174 | short l, r; |
| 175 | |
| 176 | for (i = 0; i < n; i++) { |
| 177 | l = *input1++; |
| 178 | r = *input2++; |
| 179 | *output++ = l; /* left */ |
| 180 | *output++ = (l / 2) + (r / 2); /* center */ |
| 181 | *output++ = r; /* right */ |
| 182 | *output++ = 0; /* left surround */ |
| 183 | *output++ = 0; /* right surroud */ |
| 184 | *output++ = 0; /* low freq */ |
| 185 | } |
| 186 | } |
| 187 | |
| 188 | #define SUPPORT_RESAMPLE(ch1, ch2, ch3, ch4, ch5, ch6, ch7, ch8) \ |
| 189 | ch8<<7 | ch7<<6 | ch6<<5 | ch5<<4 | ch4<<3 | ch3<<2 | ch2<<1 | ch1<<0 |
| 190 | |
| 191 | static const uint8_t supported_resampling[MAX_CHANNELS] = { |
| 192 | // output ch: 1 2 3 4 5 6 7 8 |
| 193 | SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 0, 0, 0), // 1 input channel |
| 194 | SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 1, 0, 0), // 2 input channels |
| 195 | SUPPORT_RESAMPLE(0, 0, 1, 0, 0, 0, 0, 0), // 3 input channels |
| 196 | SUPPORT_RESAMPLE(0, 0, 0, 1, 0, 0, 0, 0), // 4 input channels |
| 197 | SUPPORT_RESAMPLE(0, 0, 0, 0, 1, 0, 0, 0), // 5 input channels |
| 198 | SUPPORT_RESAMPLE(0, 1, 0, 0, 0, 1, 0, 0), // 6 input channels |
| 199 | SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 1, 0), // 7 input channels |
| 200 | SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 0, 1), // 8 input channels |
| 201 | }; |
| 202 | |
| 203 | ReSampleContext *av_audio_resample_init(int output_channels, int input_channels, |
| 204 | int output_rate, int input_rate, |
| 205 | enum AVSampleFormat sample_fmt_out, |
| 206 | enum AVSampleFormat sample_fmt_in, |
| 207 | int filter_length, int log2_phase_count, |
| 208 | int linear, double cutoff) |
| 209 | { |
| 210 | ReSampleContext *s; |
| 211 | |
| 212 | if (input_channels > MAX_CHANNELS) { |
| 213 | av_log(NULL, AV_LOG_ERROR, |
| 214 | "Resampling with input channels greater than %d is unsupported.\n", |
| 215 | MAX_CHANNELS); |
| 216 | return NULL; |
| 217 | } |
| 218 | if (!(supported_resampling[input_channels-1] & (1<<(output_channels-1)))) { |
| 219 | int i; |
| 220 | av_log(NULL, AV_LOG_ERROR, "Unsupported audio resampling. Allowed " |
| 221 | "output channels for %d input channel%s", input_channels, |
| 222 | input_channels > 1 ? "s:" : ":"); |
| 223 | for (i = 0; i < MAX_CHANNELS; i++) |
| 224 | if (supported_resampling[input_channels-1] & (1<<i)) |
| 225 | av_log(NULL, AV_LOG_ERROR, " %d", i + 1); |
| 226 | av_log(NULL, AV_LOG_ERROR, "\n"); |
| 227 | return NULL; |
| 228 | } |
| 229 | |
| 230 | s = av_mallocz(sizeof(ReSampleContext)); |
| 231 | if (!s) { |
| 232 | av_log(NULL, AV_LOG_ERROR, "Can't allocate memory for resample context.\n"); |
| 233 | return NULL; |
| 234 | } |
| 235 | |
| 236 | s->ratio = (float)output_rate / (float)input_rate; |
| 237 | |
| 238 | s->input_channels = input_channels; |
| 239 | s->output_channels = output_channels; |
| 240 | |
| 241 | s->filter_channels = s->input_channels; |
| 242 | if (s->output_channels < s->filter_channels) |
| 243 | s->filter_channels = s->output_channels; |
| 244 | |
| 245 | s->sample_fmt[0] = sample_fmt_in; |
| 246 | s->sample_fmt[1] = sample_fmt_out; |
| 247 | s->sample_size[0] = av_get_bytes_per_sample(s->sample_fmt[0]); |
| 248 | s->sample_size[1] = av_get_bytes_per_sample(s->sample_fmt[1]); |
| 249 | |
| 250 | if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) { |
| 251 | if (!(s->convert_ctx[0] = av_audio_convert_alloc(AV_SAMPLE_FMT_S16, 1, |
| 252 | s->sample_fmt[0], 1, NULL, 0))) { |
| 253 | av_log(s, AV_LOG_ERROR, |
| 254 | "Cannot convert %s sample format to s16 sample format\n", |
| 255 | av_get_sample_fmt_name(s->sample_fmt[0])); |
| 256 | av_free(s); |
| 257 | return NULL; |
| 258 | } |
| 259 | } |
| 260 | |
| 261 | if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) { |
| 262 | if (!(s->convert_ctx[1] = av_audio_convert_alloc(s->sample_fmt[1], 1, |
| 263 | AV_SAMPLE_FMT_S16, 1, NULL, 0))) { |
| 264 | av_log(s, AV_LOG_ERROR, |
| 265 | "Cannot convert s16 sample format to %s sample format\n", |
| 266 | av_get_sample_fmt_name(s->sample_fmt[1])); |
| 267 | av_audio_convert_free(s->convert_ctx[0]); |
| 268 | av_free(s); |
| 269 | return NULL; |
| 270 | } |
| 271 | } |
| 272 | |
| 273 | s->resample_context = av_resample_init(output_rate, input_rate, |
| 274 | filter_length, log2_phase_count, |
| 275 | linear, cutoff); |
| 276 | |
| 277 | *(const AVClass**)s->resample_context = &audioresample_context_class; |
| 278 | |
| 279 | return s; |
| 280 | } |
| 281 | |
| 282 | /* resample audio. 'nb_samples' is the number of input samples */ |
| 283 | /* XXX: optimize it ! */ |
| 284 | int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples) |
| 285 | { |
| 286 | int i, nb_samples1; |
| 287 | short *bufin[MAX_CHANNELS]; |
| 288 | short *bufout[MAX_CHANNELS]; |
| 289 | short *buftmp2[MAX_CHANNELS], *buftmp3[MAX_CHANNELS]; |
| 290 | short *output_bak = NULL; |
| 291 | int lenout; |
| 292 | |
| 293 | if (s->input_channels == s->output_channels && s->ratio == 1.0 && 0) { |
| 294 | /* nothing to do */ |
| 295 | memcpy(output, input, nb_samples * s->input_channels * sizeof(short)); |
| 296 | return nb_samples; |
| 297 | } |
| 298 | |
| 299 | if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) { |
| 300 | int istride[1] = { s->sample_size[0] }; |
| 301 | int ostride[1] = { 2 }; |
| 302 | const void *ibuf[1] = { input }; |
| 303 | void *obuf[1]; |
| 304 | unsigned input_size = nb_samples * s->input_channels * 2; |
| 305 | |
| 306 | if (!s->buffer_size[0] || s->buffer_size[0] < input_size) { |
| 307 | av_free(s->buffer[0]); |
| 308 | s->buffer_size[0] = input_size; |
| 309 | s->buffer[0] = av_malloc(s->buffer_size[0]); |
| 310 | if (!s->buffer[0]) { |
| 311 | av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n"); |
| 312 | return 0; |
| 313 | } |
| 314 | } |
| 315 | |
| 316 | obuf[0] = s->buffer[0]; |
| 317 | |
| 318 | if (av_audio_convert(s->convert_ctx[0], obuf, ostride, |
| 319 | ibuf, istride, nb_samples * s->input_channels) < 0) { |
| 320 | av_log(s->resample_context, AV_LOG_ERROR, |
| 321 | "Audio sample format conversion failed\n"); |
| 322 | return 0; |
| 323 | } |
| 324 | |
| 325 | input = s->buffer[0]; |
| 326 | } |
| 327 | |
| 328 | lenout= 2*s->output_channels*nb_samples * s->ratio + 16; |
| 329 | |
| 330 | if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) { |
| 331 | int out_size = lenout * av_get_bytes_per_sample(s->sample_fmt[1]) * |
| 332 | s->output_channels; |
| 333 | output_bak = output; |
| 334 | |
| 335 | if (!s->buffer_size[1] || s->buffer_size[1] < out_size) { |
| 336 | av_free(s->buffer[1]); |
| 337 | s->buffer_size[1] = out_size; |
| 338 | s->buffer[1] = av_malloc(s->buffer_size[1]); |
| 339 | if (!s->buffer[1]) { |
| 340 | av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n"); |
| 341 | return 0; |
| 342 | } |
| 343 | } |
| 344 | |
| 345 | output = s->buffer[1]; |
| 346 | } |
| 347 | |
| 348 | /* XXX: move those malloc to resample init code */ |
| 349 | for (i = 0; i < s->filter_channels; i++) { |
| 350 | bufin[i] = av_malloc_array((nb_samples + s->temp_len), sizeof(short)); |
| 351 | bufout[i] = av_malloc_array(lenout, sizeof(short)); |
| 352 | |
| 353 | if (!bufin[i] || !bufout[i]) { |
| 354 | av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n"); |
| 355 | nb_samples1 = 0; |
| 356 | goto fail; |
| 357 | } |
| 358 | |
| 359 | memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short)); |
| 360 | buftmp2[i] = bufin[i] + s->temp_len; |
| 361 | } |
| 362 | |
| 363 | if (s->input_channels == 2 && s->output_channels == 1) { |
| 364 | buftmp3[0] = output; |
| 365 | stereo_to_mono(buftmp2[0], input, nb_samples); |
| 366 | } else if (s->output_channels >= 2 && s->input_channels == 1) { |
| 367 | buftmp3[0] = bufout[0]; |
| 368 | memcpy(buftmp2[0], input, nb_samples * sizeof(short)); |
| 369 | } else if (s->input_channels == 6 && s->output_channels ==2) { |
| 370 | buftmp3[0] = bufout[0]; |
| 371 | buftmp3[1] = bufout[1]; |
| 372 | surround_to_stereo(buftmp2, input, s->input_channels, nb_samples); |
| 373 | } else if (s->output_channels >= s->input_channels && s->input_channels >= 2) { |
| 374 | for (i = 0; i < s->input_channels; i++) { |
| 375 | buftmp3[i] = bufout[i]; |
| 376 | } |
| 377 | deinterleave(buftmp2, input, s->input_channels, nb_samples); |
| 378 | } else { |
| 379 | buftmp3[0] = output; |
| 380 | memcpy(buftmp2[0], input, nb_samples * sizeof(short)); |
| 381 | } |
| 382 | |
| 383 | nb_samples += s->temp_len; |
| 384 | |
| 385 | /* resample each channel */ |
| 386 | nb_samples1 = 0; /* avoid warning */ |
| 387 | for (i = 0; i < s->filter_channels; i++) { |
| 388 | int consumed; |
| 389 | int is_last = i + 1 == s->filter_channels; |
| 390 | |
| 391 | nb_samples1 = av_resample(s->resample_context, buftmp3[i], bufin[i], |
| 392 | &consumed, nb_samples, lenout, is_last); |
| 393 | s->temp_len = nb_samples - consumed; |
| 394 | s->temp[i] = av_realloc_array(s->temp[i], s->temp_len, sizeof(short)); |
| 395 | memcpy(s->temp[i], bufin[i] + consumed, s->temp_len * sizeof(short)); |
| 396 | } |
| 397 | |
| 398 | if (s->output_channels == 2 && s->input_channels == 1) { |
| 399 | mono_to_stereo(output, buftmp3[0], nb_samples1); |
| 400 | } else if (s->output_channels == 6 && s->input_channels == 2) { |
| 401 | ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1); |
| 402 | } else if ((s->output_channels == s->input_channels && s->input_channels >= 2) || |
| 403 | (s->output_channels == 2 && s->input_channels == 6)) { |
| 404 | interleave(output, buftmp3, s->output_channels, nb_samples1); |
| 405 | } |
| 406 | |
| 407 | if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) { |
| 408 | int istride[1] = { 2 }; |
| 409 | int ostride[1] = { s->sample_size[1] }; |
| 410 | const void *ibuf[1] = { output }; |
| 411 | void *obuf[1] = { output_bak }; |
| 412 | |
| 413 | if (av_audio_convert(s->convert_ctx[1], obuf, ostride, |
| 414 | ibuf, istride, nb_samples1 * s->output_channels) < 0) { |
| 415 | av_log(s->resample_context, AV_LOG_ERROR, |
| 416 | "Audio sample format conversion failed\n"); |
| 417 | return 0; |
| 418 | } |
| 419 | } |
| 420 | |
| 421 | fail: |
| 422 | for (i = 0; i < s->filter_channels; i++) { |
| 423 | av_free(bufin[i]); |
| 424 | av_free(bufout[i]); |
| 425 | } |
| 426 | |
| 427 | return nb_samples1; |
| 428 | } |
| 429 | |
| 430 | void audio_resample_close(ReSampleContext *s) |
| 431 | { |
| 432 | int i; |
| 433 | av_resample_close(s->resample_context); |
| 434 | for (i = 0; i < s->filter_channels; i++) |
| 435 | av_freep(&s->temp[i]); |
| 436 | av_freep(&s->buffer[0]); |
| 437 | av_freep(&s->buffer[1]); |
| 438 | av_audio_convert_free(s->convert_ctx[0]); |
| 439 | av_audio_convert_free(s->convert_ctx[1]); |
| 440 | av_free(s); |
| 441 | } |
| 442 | |
| 443 | #endif |