3 * Copyright (c) 2006 Justin Ruggles <justin.ruggles@gmail.com>
5 * This file is part of Libav.
7 * Libav 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.
12 * Libav 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.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 #include "libavutil/crc.h"
23 #include "libavutil/intmath.h"
24 #include "libavutil/md5.h"
25 #include "libavutil/opt.h"
35 #define FLAC_SUBFRAME_CONSTANT 0
36 #define FLAC_SUBFRAME_VERBATIM 1
37 #define FLAC_SUBFRAME_FIXED 8
38 #define FLAC_SUBFRAME_LPC 32
40 #define MAX_FIXED_ORDER 4
41 #define MAX_PARTITION_ORDER 8
42 #define MAX_PARTITIONS (1 << MAX_PARTITION_ORDER)
43 #define MAX_LPC_PRECISION 15
44 #define MAX_LPC_SHIFT 15
45 #define MAX_RICE_PARAM 14
47 typedef struct CompressionOptions {
48 int compression_level;
50 enum FFLPCType lpc_type;
52 int lpc_coeff_precision;
53 int min_prediction_order;
54 int max_prediction_order;
55 int prediction_order_method;
56 int min_partition_order;
57 int max_partition_order;
61 typedef struct RiceContext {
63 int params[MAX_PARTITIONS];
66 typedef struct FlacSubframe {
72 int32_t coefs[MAX_LPC_ORDER];
75 int32_t samples[FLAC_MAX_BLOCKSIZE];
76 int32_t residual[FLAC_MAX_BLOCKSIZE+1];
79 typedef struct FlacFrame {
80 FlacSubframe subframes[FLAC_MAX_CHANNELS];
88 typedef struct FlacEncodeContext {
97 int max_encoded_framesize;
99 uint64_t sample_count;
102 CompressionOptions options;
103 AVCodecContext *avctx;
105 struct AVMD5 *md5ctx;
107 unsigned int md5_buffer_size;
113 * Write streaminfo metadata block to byte array.
115 static void write_streaminfo(FlacEncodeContext *s, uint8_t *header)
119 memset(header, 0, FLAC_STREAMINFO_SIZE);
120 init_put_bits(&pb, header, FLAC_STREAMINFO_SIZE);
122 /* streaminfo metadata block */
123 put_bits(&pb, 16, s->max_blocksize);
124 put_bits(&pb, 16, s->max_blocksize);
125 put_bits(&pb, 24, s->min_framesize);
126 put_bits(&pb, 24, s->max_framesize);
127 put_bits(&pb, 20, s->samplerate);
128 put_bits(&pb, 3, s->channels-1);
129 put_bits(&pb, 5, 15); /* bits per sample - 1 */
130 /* write 36-bit sample count in 2 put_bits() calls */
131 put_bits(&pb, 24, (s->sample_count & 0xFFFFFF000LL) >> 12);
132 put_bits(&pb, 12, s->sample_count & 0x000000FFFLL);
134 memcpy(&header[18], s->md5sum, 16);
139 * Set blocksize based on samplerate.
140 * Choose the closest predefined blocksize >= BLOCK_TIME_MS milliseconds.
142 static int select_blocksize(int samplerate, int block_time_ms)
148 assert(samplerate > 0);
149 blocksize = ff_flac_blocksize_table[1];
150 target = (samplerate * block_time_ms) / 1000;
151 for (i = 0; i < 16; i++) {
152 if (target >= ff_flac_blocksize_table[i] &&
153 ff_flac_blocksize_table[i] > blocksize) {
154 blocksize = ff_flac_blocksize_table[i];
161 static av_cold void dprint_compression_options(FlacEncodeContext *s)
163 AVCodecContext *avctx = s->avctx;
164 CompressionOptions *opt = &s->options;
166 av_log(avctx, AV_LOG_DEBUG, " compression: %d\n", opt->compression_level);
168 switch (opt->lpc_type) {
169 case FF_LPC_TYPE_NONE:
170 av_log(avctx, AV_LOG_DEBUG, " lpc type: None\n");
172 case FF_LPC_TYPE_FIXED:
173 av_log(avctx, AV_LOG_DEBUG, " lpc type: Fixed pre-defined coefficients\n");
175 case FF_LPC_TYPE_LEVINSON:
176 av_log(avctx, AV_LOG_DEBUG, " lpc type: Levinson-Durbin recursion with Welch window\n");
178 case FF_LPC_TYPE_CHOLESKY:
179 av_log(avctx, AV_LOG_DEBUG, " lpc type: Cholesky factorization, %d pass%s\n",
180 opt->lpc_passes, opt->lpc_passes == 1 ? "" : "es");
184 av_log(avctx, AV_LOG_DEBUG, " prediction order: %d, %d\n",
185 opt->min_prediction_order, opt->max_prediction_order);
187 switch (opt->prediction_order_method) {
188 case ORDER_METHOD_EST:
189 av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "estimate");
191 case ORDER_METHOD_2LEVEL:
192 av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "2-level");
194 case ORDER_METHOD_4LEVEL:
195 av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "4-level");
197 case ORDER_METHOD_8LEVEL:
198 av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "8-level");
200 case ORDER_METHOD_SEARCH:
201 av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "full search");
203 case ORDER_METHOD_LOG:
204 av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "log search");
209 av_log(avctx, AV_LOG_DEBUG, " partition order: %d, %d\n",
210 opt->min_partition_order, opt->max_partition_order);
212 av_log(avctx, AV_LOG_DEBUG, " block size: %d\n", avctx->frame_size);
214 av_log(avctx, AV_LOG_DEBUG, " lpc precision: %d\n",
215 opt->lpc_coeff_precision);
219 static av_cold int flac_encode_init(AVCodecContext *avctx)
221 int freq = avctx->sample_rate;
222 int channels = avctx->channels;
223 FlacEncodeContext *s = avctx->priv_data;
229 if (avctx->sample_fmt != AV_SAMPLE_FMT_S16)
232 if (channels < 1 || channels > FLAC_MAX_CHANNELS)
234 s->channels = channels;
236 /* find samplerate in table */
239 for (i = 4; i < 12; i++) {
240 if (freq == ff_flac_sample_rate_table[i]) {
241 s->samplerate = ff_flac_sample_rate_table[i];
247 /* if not in table, samplerate is non-standard */
249 if (freq % 1000 == 0 && freq < 255000) {
251 s->sr_code[1] = freq / 1000;
252 } else if (freq % 10 == 0 && freq < 655350) {
254 s->sr_code[1] = freq / 10;
255 } else if (freq < 65535) {
257 s->sr_code[1] = freq;
261 s->samplerate = freq;
264 /* set compression option defaults based on avctx->compression_level */
265 if (avctx->compression_level < 0)
266 s->options.compression_level = 5;
268 s->options.compression_level = avctx->compression_level;
270 level = s->options.compression_level;
272 av_log(avctx, AV_LOG_ERROR, "invalid compression level: %d\n",
273 s->options.compression_level);
277 s->options.block_time_ms = ((int[]){ 27, 27, 27,105,105,105,105,105,105,105,105,105,105})[level];
279 if (s->options.lpc_type == FF_LPC_TYPE_DEFAULT)
280 s->options.lpc_type = ((int[]){ FF_LPC_TYPE_FIXED, FF_LPC_TYPE_FIXED, FF_LPC_TYPE_FIXED,
281 FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON,
282 FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON,
283 FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON,
284 FF_LPC_TYPE_LEVINSON})[level];
286 s->options.min_prediction_order = ((int[]){ 2, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1})[level];
287 s->options.max_prediction_order = ((int[]){ 3, 4, 4, 6, 8, 8, 8, 8, 12, 12, 12, 32, 32})[level];
289 if (s->options.prediction_order_method < 0)
290 s->options.prediction_order_method = ((int[]){ ORDER_METHOD_EST, ORDER_METHOD_EST, ORDER_METHOD_EST,
291 ORDER_METHOD_EST, ORDER_METHOD_EST, ORDER_METHOD_EST,
292 ORDER_METHOD_4LEVEL, ORDER_METHOD_LOG, ORDER_METHOD_4LEVEL,
293 ORDER_METHOD_LOG, ORDER_METHOD_SEARCH, ORDER_METHOD_LOG,
294 ORDER_METHOD_SEARCH})[level];
296 if (s->options.min_partition_order > s->options.max_partition_order) {
297 av_log(avctx, AV_LOG_ERROR, "invalid partition orders: min=%d max=%d\n",
298 s->options.min_partition_order, s->options.max_partition_order);
299 return AVERROR(EINVAL);
301 if (s->options.min_partition_order < 0)
302 s->options.min_partition_order = ((int[]){ 2, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0})[level];
303 if (s->options.max_partition_order < 0)
304 s->options.max_partition_order = ((int[]){ 2, 2, 3, 3, 3, 8, 8, 8, 8, 8, 8, 8, 8})[level];
306 if (s->options.lpc_type == FF_LPC_TYPE_NONE) {
307 s->options.min_prediction_order = 0;
308 } else if (avctx->min_prediction_order >= 0) {
309 if (s->options.lpc_type == FF_LPC_TYPE_FIXED) {
310 if (avctx->min_prediction_order > MAX_FIXED_ORDER) {
311 av_log(avctx, AV_LOG_ERROR, "invalid min prediction order: %d\n",
312 avctx->min_prediction_order);
315 } else if (avctx->min_prediction_order < MIN_LPC_ORDER ||
316 avctx->min_prediction_order > MAX_LPC_ORDER) {
317 av_log(avctx, AV_LOG_ERROR, "invalid min prediction order: %d\n",
318 avctx->min_prediction_order);
321 s->options.min_prediction_order = avctx->min_prediction_order;
323 if (s->options.lpc_type == FF_LPC_TYPE_NONE) {
324 s->options.max_prediction_order = 0;
325 } else if (avctx->max_prediction_order >= 0) {
326 if (s->options.lpc_type == FF_LPC_TYPE_FIXED) {
327 if (avctx->max_prediction_order > MAX_FIXED_ORDER) {
328 av_log(avctx, AV_LOG_ERROR, "invalid max prediction order: %d\n",
329 avctx->max_prediction_order);
332 } else if (avctx->max_prediction_order < MIN_LPC_ORDER ||
333 avctx->max_prediction_order > MAX_LPC_ORDER) {
334 av_log(avctx, AV_LOG_ERROR, "invalid max prediction order: %d\n",
335 avctx->max_prediction_order);
338 s->options.max_prediction_order = avctx->max_prediction_order;
340 if (s->options.max_prediction_order < s->options.min_prediction_order) {
341 av_log(avctx, AV_LOG_ERROR, "invalid prediction orders: min=%d max=%d\n",
342 s->options.min_prediction_order, s->options.max_prediction_order);
346 if (avctx->frame_size > 0) {
347 if (avctx->frame_size < FLAC_MIN_BLOCKSIZE ||
348 avctx->frame_size > FLAC_MAX_BLOCKSIZE) {
349 av_log(avctx, AV_LOG_ERROR, "invalid block size: %d\n",
354 s->avctx->frame_size = select_blocksize(s->samplerate, s->options.block_time_ms);
356 s->max_blocksize = s->avctx->frame_size;
358 /* set maximum encoded frame size in verbatim mode */
359 s->max_framesize = ff_flac_get_max_frame_size(s->avctx->frame_size,
362 /* initialize MD5 context */
363 s->md5ctx = av_md5_alloc();
365 return AVERROR(ENOMEM);
366 av_md5_init(s->md5ctx);
368 streaminfo = av_malloc(FLAC_STREAMINFO_SIZE);
370 return AVERROR(ENOMEM);
371 write_streaminfo(s, streaminfo);
372 avctx->extradata = streaminfo;
373 avctx->extradata_size = FLAC_STREAMINFO_SIZE;
376 s->min_framesize = s->max_framesize;
378 #if FF_API_OLD_ENCODE_AUDIO
379 avctx->coded_frame = avcodec_alloc_frame();
380 if (!avctx->coded_frame)
381 return AVERROR(ENOMEM);
384 ret = ff_lpc_init(&s->lpc_ctx, avctx->frame_size,
385 s->options.max_prediction_order, FF_LPC_TYPE_LEVINSON);
387 ff_dsputil_init(&s->dsp, avctx);
389 dprint_compression_options(s);
395 static void init_frame(FlacEncodeContext *s, int nb_samples)
402 for (i = 0; i < 16; i++) {
403 if (nb_samples == ff_flac_blocksize_table[i]) {
404 frame->blocksize = ff_flac_blocksize_table[i];
405 frame->bs_code[0] = i;
406 frame->bs_code[1] = 0;
411 frame->blocksize = nb_samples;
412 if (frame->blocksize <= 256) {
413 frame->bs_code[0] = 6;
414 frame->bs_code[1] = frame->blocksize-1;
416 frame->bs_code[0] = 7;
417 frame->bs_code[1] = frame->blocksize-1;
421 for (ch = 0; ch < s->channels; ch++) {
422 frame->subframes[ch].wasted = 0;
423 frame->subframes[ch].obits = 16;
426 frame->verbatim_only = 0;
431 * Copy channel-interleaved input samples into separate subframes.
433 static void copy_samples(FlacEncodeContext *s, const int16_t *samples)
439 for (i = 0, j = 0; i < frame->blocksize; i++)
440 for (ch = 0; ch < s->channels; ch++, j++)
441 frame->subframes[ch].samples[i] = samples[j];
445 static uint64_t rice_count_exact(int32_t *res, int n, int k)
450 for (i = 0; i < n; i++) {
451 int32_t v = -2 * res[i] - 1;
453 count += (v >> k) + 1 + k;
459 static uint64_t subframe_count_exact(FlacEncodeContext *s, FlacSubframe *sub,
462 int p, porder, psize;
466 /* subframe header */
470 if (sub->type == FLAC_SUBFRAME_CONSTANT) {
472 } else if (sub->type == FLAC_SUBFRAME_VERBATIM) {
473 count += s->frame.blocksize * sub->obits;
475 /* warm-up samples */
476 count += pred_order * sub->obits;
478 /* LPC coefficients */
479 if (sub->type == FLAC_SUBFRAME_LPC)
480 count += 4 + 5 + pred_order * s->options.lpc_coeff_precision;
482 /* rice-encoded block */
485 /* partition order */
486 porder = sub->rc.porder;
487 psize = s->frame.blocksize >> porder;
493 for (p = 0; p < 1 << porder; p++) {
494 int k = sub->rc.params[p];
496 count += rice_count_exact(&sub->residual[i], part_end - i, k);
498 part_end = FFMIN(s->frame.blocksize, part_end + psize);
506 #define rice_encode_count(sum, n, k) (((n)*((k)+1))+((sum-(n>>1))>>(k)))
509 * Solve for d/dk(rice_encode_count) = n-((sum-(n>>1))>>(k+1)) = 0.
511 static int find_optimal_param(uint64_t sum, int n)
518 sum2 = sum - (n >> 1);
519 k = av_log2(av_clipl_int32(sum2 / n));
520 return FFMIN(k, MAX_RICE_PARAM);
524 static uint64_t calc_optimal_rice_params(RiceContext *rc, int porder,
525 uint64_t *sums, int n, int pred_order)
531 part = (1 << porder);
534 cnt = (n >> porder) - pred_order;
535 for (i = 0; i < part; i++) {
536 k = find_optimal_param(sums[i], cnt);
538 all_bits += rice_encode_count(sums[i], cnt, k);
548 static void calc_sums(int pmin, int pmax, uint32_t *data, int n, int pred_order,
549 uint64_t sums[][MAX_PARTITIONS])
553 uint32_t *res, *res_end;
555 /* sums for highest level */
557 res = &data[pred_order];
558 res_end = &data[n >> pmax];
559 for (i = 0; i < parts; i++) {
561 while (res < res_end)
564 res_end += n >> pmax;
566 /* sums for lower levels */
567 for (i = pmax - 1; i >= pmin; i--) {
569 for (j = 0; j < parts; j++)
570 sums[i][j] = sums[i+1][2*j] + sums[i+1][2*j+1];
575 static uint64_t calc_rice_params(RiceContext *rc, int pmin, int pmax,
576 int32_t *data, int n, int pred_order)
579 uint64_t bits[MAX_PARTITION_ORDER+1];
583 uint64_t sums[MAX_PARTITION_ORDER+1][MAX_PARTITIONS];
585 assert(pmin >= 0 && pmin <= MAX_PARTITION_ORDER);
586 assert(pmax >= 0 && pmax <= MAX_PARTITION_ORDER);
587 assert(pmin <= pmax);
589 udata = av_malloc(n * sizeof(uint32_t));
590 for (i = 0; i < n; i++)
591 udata[i] = (2*data[i]) ^ (data[i]>>31);
593 calc_sums(pmin, pmax, udata, n, pred_order, sums);
596 bits[pmin] = UINT32_MAX;
597 for (i = pmin; i <= pmax; i++) {
598 bits[i] = calc_optimal_rice_params(&tmp_rc, i, sums[i], n, pred_order);
599 if (bits[i] <= bits[opt_porder]) {
606 return bits[opt_porder];
610 static int get_max_p_order(int max_porder, int n, int order)
612 int porder = FFMIN(max_porder, av_log2(n^(n-1)));
614 porder = FFMIN(porder, av_log2(n/order));
619 static uint64_t find_subframe_rice_params(FlacEncodeContext *s,
620 FlacSubframe *sub, int pred_order)
622 int pmin = get_max_p_order(s->options.min_partition_order,
623 s->frame.blocksize, pred_order);
624 int pmax = get_max_p_order(s->options.max_partition_order,
625 s->frame.blocksize, pred_order);
627 uint64_t bits = 8 + pred_order * sub->obits + 2 + 4;
628 if (sub->type == FLAC_SUBFRAME_LPC)
629 bits += 4 + 5 + pred_order * s->options.lpc_coeff_precision;
630 bits += calc_rice_params(&sub->rc, pmin, pmax, sub->residual,
631 s->frame.blocksize, pred_order);
636 static void encode_residual_fixed(int32_t *res, const int32_t *smp, int n,
641 for (i = 0; i < order; i++)
645 for (i = order; i < n; i++)
647 } else if (order == 1) {
648 for (i = order; i < n; i++)
649 res[i] = smp[i] - smp[i-1];
650 } else if (order == 2) {
651 int a = smp[order-1] - smp[order-2];
652 for (i = order; i < n; i += 2) {
653 int b = smp[i ] - smp[i-1];
655 a = smp[i+1] - smp[i ];
658 } else if (order == 3) {
659 int a = smp[order-1] - smp[order-2];
660 int c = smp[order-1] - 2*smp[order-2] + smp[order-3];
661 for (i = order; i < n; i += 2) {
662 int b = smp[i ] - smp[i-1];
665 a = smp[i+1] - smp[i ];
670 int a = smp[order-1] - smp[order-2];
671 int c = smp[order-1] - 2*smp[order-2] + smp[order-3];
672 int e = smp[order-1] - 3*smp[order-2] + 3*smp[order-3] - smp[order-4];
673 for (i = order; i < n; i += 2) {
674 int b = smp[i ] - smp[i-1];
678 a = smp[i+1] - smp[i ];
688 int c = coefs[(x)-1];\
694 static av_always_inline void encode_residual_lpc_unrolled(int32_t *res,
695 const int32_t *smp, int n, int order,
696 const int32_t *coefs, int shift, int big)
699 for (i = order; i < n; i += 2) {
700 int s = smp[i-order];
749 res[i ] = smp[i ] - (p0 >> shift);
750 res[i+1] = smp[i+1] - (p1 >> shift);
755 static void encode_residual_lpc(int32_t *res, const int32_t *smp, int n,
756 int order, const int32_t *coefs, int shift)
759 for (i = 0; i < order; i++)
762 for (i = order; i < n; i += 2) {
766 for (j = 0; j < order; j++) {
772 res[i ] = smp[i ] - (p0 >> shift);
773 res[i+1] = smp[i+1] - (p1 >> shift);
777 case 1: encode_residual_lpc_unrolled(res, smp, n, 1, coefs, shift, 0); break;
778 case 2: encode_residual_lpc_unrolled(res, smp, n, 2, coefs, shift, 0); break;
779 case 3: encode_residual_lpc_unrolled(res, smp, n, 3, coefs, shift, 0); break;
780 case 4: encode_residual_lpc_unrolled(res, smp, n, 4, coefs, shift, 0); break;
781 case 5: encode_residual_lpc_unrolled(res, smp, n, 5, coefs, shift, 0); break;
782 case 6: encode_residual_lpc_unrolled(res, smp, n, 6, coefs, shift, 0); break;
783 case 7: encode_residual_lpc_unrolled(res, smp, n, 7, coefs, shift, 0); break;
784 case 8: encode_residual_lpc_unrolled(res, smp, n, 8, coefs, shift, 0); break;
785 default: encode_residual_lpc_unrolled(res, smp, n, order, coefs, shift, 1); break;
791 static int encode_residual_ch(FlacEncodeContext *s, int ch)
794 int min_order, max_order, opt_order, omethod;
797 int32_t coefs[MAX_LPC_ORDER][MAX_LPC_ORDER];
798 int shift[MAX_LPC_ORDER];
802 sub = &frame->subframes[ch];
805 n = frame->blocksize;
808 for (i = 1; i < n; i++)
812 sub->type = sub->type_code = FLAC_SUBFRAME_CONSTANT;
814 return subframe_count_exact(s, sub, 0);
818 if (frame->verbatim_only || n < 5) {
819 sub->type = sub->type_code = FLAC_SUBFRAME_VERBATIM;
820 memcpy(res, smp, n * sizeof(int32_t));
821 return subframe_count_exact(s, sub, 0);
824 min_order = s->options.min_prediction_order;
825 max_order = s->options.max_prediction_order;
826 omethod = s->options.prediction_order_method;
829 sub->type = FLAC_SUBFRAME_FIXED;
830 if (s->options.lpc_type == FF_LPC_TYPE_NONE ||
831 s->options.lpc_type == FF_LPC_TYPE_FIXED || n <= max_order) {
832 uint64_t bits[MAX_FIXED_ORDER+1];
833 if (max_order > MAX_FIXED_ORDER)
834 max_order = MAX_FIXED_ORDER;
836 bits[0] = UINT32_MAX;
837 for (i = min_order; i <= max_order; i++) {
838 encode_residual_fixed(res, smp, n, i);
839 bits[i] = find_subframe_rice_params(s, sub, i);
840 if (bits[i] < bits[opt_order])
843 sub->order = opt_order;
844 sub->type_code = sub->type | sub->order;
845 if (sub->order != max_order) {
846 encode_residual_fixed(res, smp, n, sub->order);
847 find_subframe_rice_params(s, sub, sub->order);
849 return subframe_count_exact(s, sub, sub->order);
853 sub->type = FLAC_SUBFRAME_LPC;
854 opt_order = ff_lpc_calc_coefs(&s->lpc_ctx, smp, n, min_order, max_order,
855 s->options.lpc_coeff_precision, coefs, shift, s->options.lpc_type,
856 s->options.lpc_passes, omethod,
859 if (omethod == ORDER_METHOD_2LEVEL ||
860 omethod == ORDER_METHOD_4LEVEL ||
861 omethod == ORDER_METHOD_8LEVEL) {
862 int levels = 1 << omethod;
863 uint64_t bits[1 << ORDER_METHOD_8LEVEL];
865 int opt_index = levels-1;
866 opt_order = max_order-1;
867 bits[opt_index] = UINT32_MAX;
868 for (i = levels-1; i >= 0; i--) {
869 order = min_order + (((max_order-min_order+1) * (i+1)) / levels)-1;
872 encode_residual_lpc(res, smp, n, order+1, coefs[order], shift[order]);
873 bits[i] = find_subframe_rice_params(s, sub, order+1);
874 if (bits[i] < bits[opt_index]) {
880 } else if (omethod == ORDER_METHOD_SEARCH) {
881 // brute-force optimal order search
882 uint64_t bits[MAX_LPC_ORDER];
884 bits[0] = UINT32_MAX;
885 for (i = min_order-1; i < max_order; i++) {
886 encode_residual_lpc(res, smp, n, i+1, coefs[i], shift[i]);
887 bits[i] = find_subframe_rice_params(s, sub, i+1);
888 if (bits[i] < bits[opt_order])
892 } else if (omethod == ORDER_METHOD_LOG) {
893 uint64_t bits[MAX_LPC_ORDER];
896 opt_order = min_order - 1 + (max_order-min_order)/3;
897 memset(bits, -1, sizeof(bits));
899 for (step = 16; step; step >>= 1) {
900 int last = opt_order;
901 for (i = last-step; i <= last+step; i += step) {
902 if (i < min_order-1 || i >= max_order || bits[i] < UINT32_MAX)
904 encode_residual_lpc(res, smp, n, i+1, coefs[i], shift[i]);
905 bits[i] = find_subframe_rice_params(s, sub, i+1);
906 if (bits[i] < bits[opt_order])
913 sub->order = opt_order;
914 sub->type_code = sub->type | (sub->order-1);
915 sub->shift = shift[sub->order-1];
916 for (i = 0; i < sub->order; i++)
917 sub->coefs[i] = coefs[sub->order-1][i];
919 encode_residual_lpc(res, smp, n, sub->order, sub->coefs, sub->shift);
921 find_subframe_rice_params(s, sub, sub->order);
923 return subframe_count_exact(s, sub, sub->order);
927 static int count_frame_header(FlacEncodeContext *s)
929 uint8_t av_unused tmp;
935 <1> Blocking strategy
936 <4> Block size in inter-channel samples
938 <4> Channel assignment
939 <3> Sample size in bits
944 /* coded frame number */
945 PUT_UTF8(s->frame_count, tmp, count += 8;)
947 /* explicit block size */
948 if (s->frame.bs_code[0] == 6)
950 else if (s->frame.bs_code[0] == 7)
953 /* explicit sample rate */
954 count += ((s->sr_code[0] == 12) + (s->sr_code[0] > 12)) * 8;
956 /* frame header CRC-8 */
963 static int encode_frame(FlacEncodeContext *s)
968 count = count_frame_header(s);
970 for (ch = 0; ch < s->channels; ch++)
971 count += encode_residual_ch(s, ch);
973 count += (8 - (count & 7)) & 7; // byte alignment
974 count += 16; // CRC-16
983 static void remove_wasted_bits(FlacEncodeContext *s)
987 for (ch = 0; ch < s->channels; ch++) {
988 FlacSubframe *sub = &s->frame.subframes[ch];
991 for (i = 0; i < s->frame.blocksize; i++) {
992 v |= sub->samples[i];
1000 for (i = 0; i < s->frame.blocksize; i++)
1001 sub->samples[i] >>= v;
1010 static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n)
1018 /* calculate sum of 2nd order residual for each channel */
1019 sum[0] = sum[1] = sum[2] = sum[3] = 0;
1020 for (i = 2; i < n; i++) {
1021 lt = left_ch[i] - 2*left_ch[i-1] + left_ch[i-2];
1022 rt = right_ch[i] - 2*right_ch[i-1] + right_ch[i-2];
1023 sum[2] += FFABS((lt + rt) >> 1);
1024 sum[3] += FFABS(lt - rt);
1025 sum[0] += FFABS(lt);
1026 sum[1] += FFABS(rt);
1028 /* estimate bit counts */
1029 for (i = 0; i < 4; i++) {
1030 k = find_optimal_param(2 * sum[i], n);
1031 sum[i] = rice_encode_count( 2 * sum[i], n, k);
1034 /* calculate score for each mode */
1035 score[0] = sum[0] + sum[1];
1036 score[1] = sum[0] + sum[3];
1037 score[2] = sum[1] + sum[3];
1038 score[3] = sum[2] + sum[3];
1040 /* return mode with lowest score */
1042 for (i = 1; i < 4; i++)
1043 if (score[i] < score[best])
1051 * Perform stereo channel decorrelation.
1053 static void channel_decorrelation(FlacEncodeContext *s)
1056 int32_t *left, *right;
1060 n = frame->blocksize;
1061 left = frame->subframes[0].samples;
1062 right = frame->subframes[1].samples;
1064 if (s->channels != 2) {
1065 frame->ch_mode = FLAC_CHMODE_INDEPENDENT;
1069 if (s->options.ch_mode < 0)
1070 frame->ch_mode = estimate_stereo_mode(left, right, n);
1072 frame->ch_mode = s->options.ch_mode;
1074 /* perform decorrelation and adjust bits-per-sample */
1075 if (frame->ch_mode == FLAC_CHMODE_INDEPENDENT)
1077 if (frame->ch_mode == FLAC_CHMODE_MID_SIDE) {
1079 for (i = 0; i < n; i++) {
1081 left[i] = (tmp + right[i]) >> 1;
1082 right[i] = tmp - right[i];
1084 frame->subframes[1].obits++;
1085 } else if (frame->ch_mode == FLAC_CHMODE_LEFT_SIDE) {
1086 for (i = 0; i < n; i++)
1087 right[i] = left[i] - right[i];
1088 frame->subframes[1].obits++;
1090 for (i = 0; i < n; i++)
1091 left[i] -= right[i];
1092 frame->subframes[0].obits++;
1097 static void write_utf8(PutBitContext *pb, uint32_t val)
1100 PUT_UTF8(val, tmp, put_bits(pb, 8, tmp);)
1104 static void write_frame_header(FlacEncodeContext *s)
1111 put_bits(&s->pb, 16, 0xFFF8);
1112 put_bits(&s->pb, 4, frame->bs_code[0]);
1113 put_bits(&s->pb, 4, s->sr_code[0]);
1115 if (frame->ch_mode == FLAC_CHMODE_INDEPENDENT)
1116 put_bits(&s->pb, 4, s->channels-1);
1118 put_bits(&s->pb, 4, frame->ch_mode + FLAC_MAX_CHANNELS - 1);
1120 put_bits(&s->pb, 3, 4); /* bits-per-sample code */
1121 put_bits(&s->pb, 1, 0);
1122 write_utf8(&s->pb, s->frame_count);
1124 if (frame->bs_code[0] == 6)
1125 put_bits(&s->pb, 8, frame->bs_code[1]);
1126 else if (frame->bs_code[0] == 7)
1127 put_bits(&s->pb, 16, frame->bs_code[1]);
1129 if (s->sr_code[0] == 12)
1130 put_bits(&s->pb, 8, s->sr_code[1]);
1131 else if (s->sr_code[0] > 12)
1132 put_bits(&s->pb, 16, s->sr_code[1]);
1134 flush_put_bits(&s->pb);
1135 crc = av_crc(av_crc_get_table(AV_CRC_8_ATM), 0, s->pb.buf,
1136 put_bits_count(&s->pb) >> 3);
1137 put_bits(&s->pb, 8, crc);
1141 static void write_subframes(FlacEncodeContext *s)
1145 for (ch = 0; ch < s->channels; ch++) {
1146 FlacSubframe *sub = &s->frame.subframes[ch];
1147 int i, p, porder, psize;
1149 int32_t *res = sub->residual;
1150 int32_t *frame_end = &sub->residual[s->frame.blocksize];
1152 /* subframe header */
1153 put_bits(&s->pb, 1, 0);
1154 put_bits(&s->pb, 6, sub->type_code);
1155 put_bits(&s->pb, 1, !!sub->wasted);
1157 put_bits(&s->pb, sub->wasted, 1);
1160 if (sub->type == FLAC_SUBFRAME_CONSTANT) {
1161 put_sbits(&s->pb, sub->obits, res[0]);
1162 } else if (sub->type == FLAC_SUBFRAME_VERBATIM) {
1163 while (res < frame_end)
1164 put_sbits(&s->pb, sub->obits, *res++);
1166 /* warm-up samples */
1167 for (i = 0; i < sub->order; i++)
1168 put_sbits(&s->pb, sub->obits, *res++);
1170 /* LPC coefficients */
1171 if (sub->type == FLAC_SUBFRAME_LPC) {
1172 int cbits = s->options.lpc_coeff_precision;
1173 put_bits( &s->pb, 4, cbits-1);
1174 put_sbits(&s->pb, 5, sub->shift);
1175 for (i = 0; i < sub->order; i++)
1176 put_sbits(&s->pb, cbits, sub->coefs[i]);
1179 /* rice-encoded block */
1180 put_bits(&s->pb, 2, 0);
1182 /* partition order */
1183 porder = sub->rc.porder;
1184 psize = s->frame.blocksize >> porder;
1185 put_bits(&s->pb, 4, porder);
1188 part_end = &sub->residual[psize];
1189 for (p = 0; p < 1 << porder; p++) {
1190 int k = sub->rc.params[p];
1191 put_bits(&s->pb, 4, k);
1192 while (res < part_end)
1193 set_sr_golomb_flac(&s->pb, *res++, k, INT32_MAX, 0);
1194 part_end = FFMIN(frame_end, part_end + psize);
1201 static void write_frame_footer(FlacEncodeContext *s)
1204 flush_put_bits(&s->pb);
1205 crc = av_bswap16(av_crc(av_crc_get_table(AV_CRC_16_ANSI), 0, s->pb.buf,
1206 put_bits_count(&s->pb)>>3));
1207 put_bits(&s->pb, 16, crc);
1208 flush_put_bits(&s->pb);
1212 static int write_frame(FlacEncodeContext *s, AVPacket *avpkt)
1214 init_put_bits(&s->pb, avpkt->data, avpkt->size);
1215 write_frame_header(s);
1217 write_frame_footer(s);
1218 return put_bits_count(&s->pb) >> 3;
1222 static int update_md5_sum(FlacEncodeContext *s, const int16_t *samples)
1225 int buf_size = s->frame.blocksize * s->channels * 2;
1227 if (HAVE_BIGENDIAN) {
1228 av_fast_malloc(&s->md5_buffer, &s->md5_buffer_size, buf_size);
1230 return AVERROR(ENOMEM);
1233 buf = (const uint8_t *)samples;
1235 s->dsp.bswap16_buf((uint16_t *)s->md5_buffer,
1236 (const uint16_t *)samples, buf_size / 2);
1237 buf = s->md5_buffer;
1239 av_md5_update(s->md5ctx, buf, buf_size);
1245 static int flac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
1246 const AVFrame *frame, int *got_packet_ptr)
1248 FlacEncodeContext *s;
1249 const int16_t *samples;
1250 int frame_bytes, out_bytes, ret;
1252 s = avctx->priv_data;
1254 /* when the last block is reached, update the header in extradata */
1256 s->max_framesize = s->max_encoded_framesize;
1257 av_md5_final(s->md5ctx, s->md5sum);
1258 write_streaminfo(s, avctx->extradata);
1261 samples = (const int16_t *)frame->data[0];
1263 /* change max_framesize for small final frame */
1264 if (frame->nb_samples < s->frame.blocksize) {
1265 s->max_framesize = ff_flac_get_max_frame_size(frame->nb_samples,
1269 init_frame(s, frame->nb_samples);
1271 copy_samples(s, samples);
1273 channel_decorrelation(s);
1275 remove_wasted_bits(s);
1277 frame_bytes = encode_frame(s);
1279 /* fallback to verbatim mode if the compressed frame is larger than it
1280 would be if encoded uncompressed. */
1281 if (frame_bytes < 0 || frame_bytes > s->max_framesize) {
1282 s->frame.verbatim_only = 1;
1283 frame_bytes = encode_frame(s);
1284 if (frame_bytes < 0) {
1285 av_log(avctx, AV_LOG_ERROR, "Bad frame count\n");
1290 if ((ret = ff_alloc_packet(avpkt, frame_bytes))) {
1291 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
1295 out_bytes = write_frame(s, avpkt);
1298 s->sample_count += frame->nb_samples;
1299 if ((ret = update_md5_sum(s, samples)) < 0) {
1300 av_log(avctx, AV_LOG_ERROR, "Error updating MD5 checksum\n");
1303 if (out_bytes > s->max_encoded_framesize)
1304 s->max_encoded_framesize = out_bytes;
1305 if (out_bytes < s->min_framesize)
1306 s->min_framesize = out_bytes;
1308 avpkt->pts = frame->pts;
1309 avpkt->duration = ff_samples_to_time_base(avctx, frame->nb_samples);
1310 avpkt->size = out_bytes;
1311 *got_packet_ptr = 1;
1316 static av_cold int flac_encode_close(AVCodecContext *avctx)
1318 if (avctx->priv_data) {
1319 FlacEncodeContext *s = avctx->priv_data;
1320 av_freep(&s->md5ctx);
1321 av_freep(&s->md5_buffer);
1322 ff_lpc_end(&s->lpc_ctx);
1324 av_freep(&avctx->extradata);
1325 avctx->extradata_size = 0;
1326 #if FF_API_OLD_ENCODE_AUDIO
1327 av_freep(&avctx->coded_frame);
1332 #define FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
1333 static const AVOption options[] = {
1334 { "lpc_coeff_precision", "LPC coefficient precision", offsetof(FlacEncodeContext, options.lpc_coeff_precision), AV_OPT_TYPE_INT, {.i64 = 15 }, 0, MAX_LPC_PRECISION, FLAGS },
1335 { "lpc_type", "LPC algorithm", offsetof(FlacEncodeContext, options.lpc_type), AV_OPT_TYPE_INT, {.i64 = FF_LPC_TYPE_DEFAULT }, FF_LPC_TYPE_DEFAULT, FF_LPC_TYPE_NB-1, FLAGS, "lpc_type" },
1336 { "none", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_LPC_TYPE_NONE }, INT_MIN, INT_MAX, FLAGS, "lpc_type" },
1337 { "fixed", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_LPC_TYPE_FIXED }, INT_MIN, INT_MAX, FLAGS, "lpc_type" },
1338 { "levinson", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_LPC_TYPE_LEVINSON }, INT_MIN, INT_MAX, FLAGS, "lpc_type" },
1339 { "cholesky", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_LPC_TYPE_CHOLESKY }, INT_MIN, INT_MAX, FLAGS, "lpc_type" },
1340 { "lpc_passes", "Number of passes to use for Cholesky factorization during LPC analysis", offsetof(FlacEncodeContext, options.lpc_passes), AV_OPT_TYPE_INT, {.i64 = -1 }, INT_MIN, INT_MAX, FLAGS },
1341 { "min_partition_order", NULL, offsetof(FlacEncodeContext, options.min_partition_order), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, MAX_PARTITION_ORDER, FLAGS },
1342 { "max_partition_order", NULL, offsetof(FlacEncodeContext, options.max_partition_order), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, MAX_PARTITION_ORDER, FLAGS },
1343 { "prediction_order_method", "Search method for selecting prediction order", offsetof(FlacEncodeContext, options.prediction_order_method), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, ORDER_METHOD_LOG, FLAGS, "predm" },
1344 { "estimation", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_EST }, INT_MIN, INT_MAX, FLAGS, "predm" },
1345 { "2level", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_2LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" },
1346 { "4level", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_4LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" },
1347 { "8level", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_8LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" },
1348 { "search", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_SEARCH }, INT_MIN, INT_MAX, FLAGS, "predm" },
1349 { "log", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_LOG }, INT_MIN, INT_MAX, FLAGS, "predm" },
1350 { "ch_mode", "Stereo decorrelation mode", offsetof(FlacEncodeContext, options.ch_mode), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, FLAC_CHMODE_MID_SIDE, FLAGS, "ch_mode" },
1351 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = -1 }, INT_MIN, INT_MAX, FLAGS, "ch_mode" },
1352 { "indep", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FLAC_CHMODE_INDEPENDENT }, INT_MIN, INT_MAX, FLAGS, "ch_mode" },
1353 { "left_side", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FLAC_CHMODE_LEFT_SIDE }, INT_MIN, INT_MAX, FLAGS, "ch_mode" },
1354 { "right_side", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FLAC_CHMODE_RIGHT_SIDE }, INT_MIN, INT_MAX, FLAGS, "ch_mode" },
1355 { "mid_side", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FLAC_CHMODE_MID_SIDE }, INT_MIN, INT_MAX, FLAGS, "ch_mode" },
1359 static const AVClass flac_encoder_class = {
1361 av_default_item_name,
1363 LIBAVUTIL_VERSION_INT,
1366 AVCodec ff_flac_encoder = {
1368 .type = AVMEDIA_TYPE_AUDIO,
1369 .id = AV_CODEC_ID_FLAC,
1370 .priv_data_size = sizeof(FlacEncodeContext),
1371 .init = flac_encode_init,
1372 .encode2 = flac_encode_frame,
1373 .close = flac_encode_close,
1374 .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY,
1375 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
1376 AV_SAMPLE_FMT_NONE },
1377 .long_name = NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),
1378 .priv_class = &flac_encoder_class,