X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fflacenc.c;h=7685ff6ea0d81fbca25796f13ca5eea8d4d46040;hb=2caf19e90f270abe1e80a3e85acaf0eb5c9d0aac;hp=5f9d78ef123f99d9de95addc613b19e721ec1734;hpb=6810b93a81a8c0e845dc2bb90f9e2d35eac17cd1;p=ffmpeg diff --git a/libavcodec/flacenc.c b/libavcodec/flacenc.c index 5f9d78ef123..7685ff6ea0d 100644 --- a/libavcodec/flacenc.c +++ b/libavcodec/flacenc.c @@ -1,57 +1,39 @@ /** * FLAC audio encoder - * Copyright (c) 2006 Justin Ruggles + * Copyright (c) 2006 Justin Ruggles * - * This file is part of FFmpeg. + * This file is part of Libav. * - * FFmpeg is free software; you can redistribute it and/or + * Libav is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * - * FFmpeg is distributed in the hope that it will be useful, + * Libav is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public - * License along with FFmpeg; if not, write to the Free Software + * License along with Libav; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ +#include "libavutil/crc.h" +#include "libavutil/md5.h" +#include "libavutil/opt.h" #include "avcodec.h" -#include "bitstream.h" -#include "crc.h" -#include "dsputil.h" +#include "get_bits.h" #include "golomb.h" -#include "lls.h" - -#define FLAC_MAX_CH 8 -#define FLAC_MIN_BLOCKSIZE 16 -#define FLAC_MAX_BLOCKSIZE 65535 +#include "lpc.h" +#include "flac.h" +#include "flacdata.h" #define FLAC_SUBFRAME_CONSTANT 0 #define FLAC_SUBFRAME_VERBATIM 1 #define FLAC_SUBFRAME_FIXED 8 #define FLAC_SUBFRAME_LPC 32 -#define FLAC_CHMODE_NOT_STEREO 0 -#define FLAC_CHMODE_LEFT_RIGHT 1 -#define FLAC_CHMODE_LEFT_SIDE 8 -#define FLAC_CHMODE_RIGHT_SIDE 9 -#define FLAC_CHMODE_MID_SIDE 10 - -#define ORDER_METHOD_EST 0 -#define ORDER_METHOD_2LEVEL 1 -#define ORDER_METHOD_4LEVEL 2 -#define ORDER_METHOD_8LEVEL 3 -#define ORDER_METHOD_SEARCH 4 -#define ORDER_METHOD_LOG 5 - -#define FLAC_STREAMINFO_SIZE 34 - -#define MIN_LPC_ORDER 1 -#define MAX_LPC_ORDER 32 #define MAX_FIXED_ORDER 4 #define MAX_PARTITION_ORDER 8 #define MAX_PARTITIONS (1 << MAX_PARTITION_ORDER) @@ -62,7 +44,8 @@ typedef struct CompressionOptions { int compression_level; int block_time_ms; - int use_lpc; + enum FFLPCType lpc_type; + int lpc_passes; int lpc_coeff_precision; int min_prediction_order; int max_prediction_order; @@ -89,44 +72,37 @@ typedef struct FlacSubframe { } FlacSubframe; typedef struct FlacFrame { - FlacSubframe subframes[FLAC_MAX_CH]; + FlacSubframe subframes[FLAC_MAX_CHANNELS]; int blocksize; int bs_code[2]; uint8_t crc8; int ch_mode; + int verbatim_only; } FlacFrame; typedef struct FlacEncodeContext { + AVClass *class; PutBitContext pb; int channels; - int ch_code; int samplerate; int sr_code[2]; - int blocksize; + int max_blocksize; + int min_framesize; int max_framesize; + int max_encoded_framesize; uint32_t frame_count; + uint64_t sample_count; + uint8_t md5sum[16]; FlacFrame frame; CompressionOptions options; AVCodecContext *avctx; - DSPContext dsp; + LPCContext lpc_ctx; + struct AVMD5 *md5ctx; } FlacEncodeContext; -static const int flac_samplerates[16] = { - 0, 0, 0, 0, - 8000, 16000, 22050, 24000, 32000, 44100, 48000, 96000, - 0, 0, 0, 0 -}; - -static const int flac_blocksizes[16] = { - 0, - 192, - 576, 1152, 2304, 4608, - 0, 0, - 256, 512, 1024, 2048, 4096, 8192, 16384, 32768 -}; /** - * Writes streaminfo metadata block to byte array + * Write streaminfo metadata block to byte array. */ static void write_streaminfo(FlacEncodeContext *s, uint8_t *header) { @@ -136,21 +112,24 @@ static void write_streaminfo(FlacEncodeContext *s, uint8_t *header) init_put_bits(&pb, header, FLAC_STREAMINFO_SIZE); /* streaminfo metadata block */ - put_bits(&pb, 16, s->blocksize); - put_bits(&pb, 16, s->blocksize); - put_bits(&pb, 24, 0); + put_bits(&pb, 16, s->max_blocksize); + put_bits(&pb, 16, s->max_blocksize); + put_bits(&pb, 24, s->min_framesize); put_bits(&pb, 24, s->max_framesize); put_bits(&pb, 20, s->samplerate); put_bits(&pb, 3, s->channels-1); put_bits(&pb, 5, 15); /* bits per sample - 1 */ + /* write 36-bit sample count in 2 put_bits() calls */ + put_bits(&pb, 24, (s->sample_count & 0xFFFFFF000LL) >> 12); + put_bits(&pb, 12, s->sample_count & 0x000000FFFLL); flush_put_bits(&pb); - /* total samples = 0 */ - /* MD5 signature = 0 */ + memcpy(&header[18], s->md5sum, 16); } + /** - * Sets blocksize based on samplerate - * Chooses the closest predefined blocksize >= BLOCK_TIME_MS milliseconds + * Set blocksize based on samplerate. + * Choose the closest predefined blocksize >= BLOCK_TIME_MS milliseconds. */ static int select_blocksize(int samplerate, int block_time_ms) { @@ -159,58 +138,113 @@ static int select_blocksize(int samplerate, int block_time_ms) int blocksize; assert(samplerate > 0); - blocksize = flac_blocksizes[1]; - target = (samplerate * block_time_ms) / 1000; - for(i=0; i<16; i++) { - if(target >= flac_blocksizes[i] && flac_blocksizes[i] > blocksize) { - blocksize = flac_blocksizes[i]; + blocksize = ff_flac_blocksize_table[1]; + target = (samplerate * block_time_ms) / 1000; + for (i = 0; i < 16; i++) { + if (target >= ff_flac_blocksize_table[i] && + ff_flac_blocksize_table[i] > blocksize) { + blocksize = ff_flac_blocksize_table[i]; } } return blocksize; } -static int flac_encode_init(AVCodecContext *avctx) + +static av_cold void dprint_compression_options(FlacEncodeContext *s) +{ + AVCodecContext *avctx = s->avctx; + CompressionOptions *opt = &s->options; + + av_log(avctx, AV_LOG_DEBUG, " compression: %d\n", opt->compression_level); + + switch (opt->lpc_type) { + case FF_LPC_TYPE_NONE: + av_log(avctx, AV_LOG_DEBUG, " lpc type: None\n"); + break; + case FF_LPC_TYPE_FIXED: + av_log(avctx, AV_LOG_DEBUG, " lpc type: Fixed pre-defined coefficients\n"); + break; + case FF_LPC_TYPE_LEVINSON: + av_log(avctx, AV_LOG_DEBUG, " lpc type: Levinson-Durbin recursion with Welch window\n"); + break; + case FF_LPC_TYPE_CHOLESKY: + av_log(avctx, AV_LOG_DEBUG, " lpc type: Cholesky factorization, %d pass%s\n", + opt->lpc_passes, opt->lpc_passes == 1 ? "" : "es"); + break; + } + + av_log(avctx, AV_LOG_DEBUG, " prediction order: %d, %d\n", + opt->min_prediction_order, opt->max_prediction_order); + + switch (opt->prediction_order_method) { + case ORDER_METHOD_EST: + av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "estimate"); + break; + case ORDER_METHOD_2LEVEL: + av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "2-level"); + break; + case ORDER_METHOD_4LEVEL: + av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "4-level"); + break; + case ORDER_METHOD_8LEVEL: + av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "8-level"); + break; + case ORDER_METHOD_SEARCH: + av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "full search"); + break; + case ORDER_METHOD_LOG: + av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", "log search"); + break; + } + + + av_log(avctx, AV_LOG_DEBUG, " partition order: %d, %d\n", + opt->min_partition_order, opt->max_partition_order); + + av_log(avctx, AV_LOG_DEBUG, " block size: %d\n", avctx->frame_size); + + av_log(avctx, AV_LOG_DEBUG, " lpc precision: %d\n", + opt->lpc_coeff_precision); +} + + +static av_cold int flac_encode_init(AVCodecContext *avctx) { int freq = avctx->sample_rate; int channels = avctx->channels; FlacEncodeContext *s = avctx->priv_data; - int i, level; + int i, level, ret; uint8_t *streaminfo; s->avctx = avctx; - dsputil_init(&s->dsp, avctx); - - if(avctx->sample_fmt != SAMPLE_FMT_S16) { + if (avctx->sample_fmt != AV_SAMPLE_FMT_S16) return -1; - } - if(channels < 1 || channels > FLAC_MAX_CH) { + if (channels < 1 || channels > FLAC_MAX_CHANNELS) return -1; - } s->channels = channels; - s->ch_code = s->channels-1; /* find samplerate in table */ - if(freq < 1) + if (freq < 1) return -1; - for(i=4; i<12; i++) { - if(freq == flac_samplerates[i]) { - s->samplerate = flac_samplerates[i]; + for (i = 4; i < 12; i++) { + if (freq == ff_flac_sample_rate_table[i]) { + s->samplerate = ff_flac_sample_rate_table[i]; s->sr_code[0] = i; s->sr_code[1] = 0; break; } } /* if not in table, samplerate is non-standard */ - if(i == 12) { - if(freq % 1000 == 0 && freq < 255000) { + if (i == 12) { + if (freq % 1000 == 0 && freq < 255000) { s->sr_code[0] = 12; s->sr_code[1] = freq / 1000; - } else if(freq % 10 == 0 && freq < 655350) { + } else if (freq % 10 == 0 && freq < 655350) { s->sr_code[0] = 14; s->sr_code[1] = freq / 10; - } else if(freq < 65535) { + } else if (freq < 65535) { s->sr_code[0] = 13; s->sr_code[1] = freq; } else { @@ -220,187 +254,200 @@ static int flac_encode_init(AVCodecContext *avctx) } /* set compression option defaults based on avctx->compression_level */ - if(avctx->compression_level < 0) { + if (avctx->compression_level < 0) s->options.compression_level = 5; - } else { + else s->options.compression_level = avctx->compression_level; - } - av_log(avctx, AV_LOG_DEBUG, " compression: %d\n", s->options.compression_level); - level= s->options.compression_level; - if(level > 12) { + level = s->options.compression_level; + if (level > 12) { av_log(avctx, AV_LOG_ERROR, "invalid compression level: %d\n", s->options.compression_level); return -1; } - s->options.block_time_ms = ((int[]){ 27, 27, 27,105,105,105,105,105,105,105,105,105,105})[level]; - s->options.use_lpc = ((int[]){ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1})[level]; - s->options.min_prediction_order= ((int[]){ 2, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1})[level]; - s->options.max_prediction_order= ((int[]){ 3, 4, 4, 6, 8, 8, 8, 8, 12, 12, 12, 32, 32})[level]; - s->options.prediction_order_method = ((int[]){ ORDER_METHOD_EST, ORDER_METHOD_EST, ORDER_METHOD_EST, - ORDER_METHOD_EST, ORDER_METHOD_EST, ORDER_METHOD_EST, - ORDER_METHOD_4LEVEL, ORDER_METHOD_LOG, ORDER_METHOD_4LEVEL, - ORDER_METHOD_LOG, ORDER_METHOD_SEARCH, ORDER_METHOD_LOG, - ORDER_METHOD_SEARCH})[level]; - s->options.min_partition_order = ((int[]){ 2, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0})[level]; - s->options.max_partition_order = ((int[]){ 2, 2, 3, 3, 3, 8, 8, 8, 8, 8, 8, 8, 8})[level]; + s->options.block_time_ms = ((int[]){ 27, 27, 27,105,105,105,105,105,105,105,105,105,105})[level]; - /* set compression option overrides from AVCodecContext */ - if(avctx->use_lpc >= 0) { - s->options.use_lpc = av_clip(avctx->use_lpc, 0, 11); + if (s->options.lpc_type == FF_LPC_TYPE_DEFAULT) + s->options.lpc_type = ((int[]){ FF_LPC_TYPE_FIXED, FF_LPC_TYPE_FIXED, FF_LPC_TYPE_FIXED, + FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON, + FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON, + FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON, FF_LPC_TYPE_LEVINSON, + FF_LPC_TYPE_LEVINSON})[level]; + + s->options.min_prediction_order = ((int[]){ 2, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1})[level]; + s->options.max_prediction_order = ((int[]){ 3, 4, 4, 6, 8, 8, 8, 8, 12, 12, 12, 32, 32})[level]; + + if (s->options.prediction_order_method < 0) + s->options.prediction_order_method = ((int[]){ ORDER_METHOD_EST, ORDER_METHOD_EST, ORDER_METHOD_EST, + ORDER_METHOD_EST, ORDER_METHOD_EST, ORDER_METHOD_EST, + ORDER_METHOD_4LEVEL, ORDER_METHOD_LOG, ORDER_METHOD_4LEVEL, + ORDER_METHOD_LOG, ORDER_METHOD_SEARCH, ORDER_METHOD_LOG, + ORDER_METHOD_SEARCH})[level]; + + if (s->options.min_partition_order > s->options.max_partition_order) { + av_log(avctx, AV_LOG_ERROR, "invalid partition orders: min=%d max=%d\n", + s->options.min_partition_order, s->options.max_partition_order); + return AVERROR(EINVAL); } - if(s->options.use_lpc == 1) - av_log(avctx, AV_LOG_DEBUG, " use lpc: Levinson-Durbin recursion with Welch window\n"); - else if(s->options.use_lpc > 1) - av_log(avctx, AV_LOG_DEBUG, " use lpc: Cholesky factorization\n"); - - if(avctx->min_prediction_order >= 0) { - if(s->options.use_lpc) { - if(avctx->min_prediction_order < MIN_LPC_ORDER || - avctx->min_prediction_order > MAX_LPC_ORDER) { - av_log(avctx, AV_LOG_ERROR, "invalid min prediction order: %d\n", - avctx->min_prediction_order); + if (s->options.min_partition_order < 0) + s->options.min_partition_order = ((int[]){ 2, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0})[level]; + if (s->options.max_partition_order < 0) + s->options.max_partition_order = ((int[]){ 2, 2, 3, 3, 3, 8, 8, 8, 8, 8, 8, 8, 8})[level]; + + /* set compression option overrides from AVCodecContext */ +#if FF_API_FLAC_GLOBAL_OPTS + if (avctx->lpc_type > FF_LPC_TYPE_DEFAULT) { + if (avctx->lpc_type > FF_LPC_TYPE_CHOLESKY) { + av_log(avctx, AV_LOG_ERROR, "unknown lpc type: %d\n", avctx->lpc_type); + return -1; + } + s->options.lpc_type = avctx->lpc_type; + if (s->options.lpc_type == FF_LPC_TYPE_CHOLESKY) { + if (avctx->lpc_passes < 0) { + // default number of passes for Cholesky + s->options.lpc_passes = 2; + } else if (avctx->lpc_passes == 0) { + av_log(avctx, AV_LOG_ERROR, "invalid number of lpc passes: %d\n", + avctx->lpc_passes); return -1; + } else { + s->options.lpc_passes = avctx->lpc_passes; } - } else { - if(avctx->min_prediction_order > MAX_FIXED_ORDER) { + } + } +#endif + + if (s->options.lpc_type == FF_LPC_TYPE_NONE) { + s->options.min_prediction_order = 0; + } else if (avctx->min_prediction_order >= 0) { + if (s->options.lpc_type == FF_LPC_TYPE_FIXED) { + if (avctx->min_prediction_order > MAX_FIXED_ORDER) { av_log(avctx, AV_LOG_ERROR, "invalid min prediction order: %d\n", avctx->min_prediction_order); return -1; } + } else if (avctx->min_prediction_order < MIN_LPC_ORDER || + avctx->min_prediction_order > MAX_LPC_ORDER) { + av_log(avctx, AV_LOG_ERROR, "invalid min prediction order: %d\n", + avctx->min_prediction_order); + return -1; } s->options.min_prediction_order = avctx->min_prediction_order; } - if(avctx->max_prediction_order >= 0) { - if(s->options.use_lpc) { - if(avctx->max_prediction_order < MIN_LPC_ORDER || - avctx->max_prediction_order > MAX_LPC_ORDER) { - av_log(avctx, AV_LOG_ERROR, "invalid max prediction order: %d\n", - avctx->max_prediction_order); - return -1; - } - } else { - if(avctx->max_prediction_order > MAX_FIXED_ORDER) { + if (s->options.lpc_type == FF_LPC_TYPE_NONE) { + s->options.max_prediction_order = 0; + } else if (avctx->max_prediction_order >= 0) { + if (s->options.lpc_type == FF_LPC_TYPE_FIXED) { + if (avctx->max_prediction_order > MAX_FIXED_ORDER) { av_log(avctx, AV_LOG_ERROR, "invalid max prediction order: %d\n", avctx->max_prediction_order); return -1; } + } else if (avctx->max_prediction_order < MIN_LPC_ORDER || + avctx->max_prediction_order > MAX_LPC_ORDER) { + av_log(avctx, AV_LOG_ERROR, "invalid max prediction order: %d\n", + avctx->max_prediction_order); + return -1; } s->options.max_prediction_order = avctx->max_prediction_order; } - if(s->options.max_prediction_order < s->options.min_prediction_order) { + if (s->options.max_prediction_order < s->options.min_prediction_order) { av_log(avctx, AV_LOG_ERROR, "invalid prediction orders: min=%d max=%d\n", s->options.min_prediction_order, s->options.max_prediction_order); return -1; } - av_log(avctx, AV_LOG_DEBUG, " prediction order: %d, %d\n", - s->options.min_prediction_order, s->options.max_prediction_order); - if(avctx->prediction_order_method >= 0) { - if(avctx->prediction_order_method > ORDER_METHOD_LOG) { +#if FF_API_FLAC_GLOBAL_OPTS + if (avctx->prediction_order_method >= 0) { + if (avctx->prediction_order_method > ORDER_METHOD_LOG) { av_log(avctx, AV_LOG_ERROR, "invalid prediction order method: %d\n", avctx->prediction_order_method); return -1; } s->options.prediction_order_method = avctx->prediction_order_method; } - switch(s->options.prediction_order_method) { - case ORDER_METHOD_EST: av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", - "estimate"); break; - case ORDER_METHOD_2LEVEL: av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", - "2-level"); break; - case ORDER_METHOD_4LEVEL: av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", - "4-level"); break; - case ORDER_METHOD_8LEVEL: av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", - "8-level"); break; - case ORDER_METHOD_SEARCH: av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", - "full search"); break; - case ORDER_METHOD_LOG: av_log(avctx, AV_LOG_DEBUG, " order method: %s\n", - "log search"); break; - } - if(avctx->min_partition_order >= 0) { - if(avctx->min_partition_order > MAX_PARTITION_ORDER) { + if (avctx->min_partition_order >= 0) { + if (avctx->min_partition_order > MAX_PARTITION_ORDER) { av_log(avctx, AV_LOG_ERROR, "invalid min partition order: %d\n", avctx->min_partition_order); return -1; } s->options.min_partition_order = avctx->min_partition_order; } - if(avctx->max_partition_order >= 0) { - if(avctx->max_partition_order > MAX_PARTITION_ORDER) { + if (avctx->max_partition_order >= 0) { + if (avctx->max_partition_order > MAX_PARTITION_ORDER) { av_log(avctx, AV_LOG_ERROR, "invalid max partition order: %d\n", avctx->max_partition_order); return -1; } s->options.max_partition_order = avctx->max_partition_order; } - if(s->options.max_partition_order < s->options.min_partition_order) { + if (s->options.max_partition_order < s->options.min_partition_order) { av_log(avctx, AV_LOG_ERROR, "invalid partition orders: min=%d max=%d\n", s->options.min_partition_order, s->options.max_partition_order); return -1; } - av_log(avctx, AV_LOG_DEBUG, " partition order: %d, %d\n", - s->options.min_partition_order, s->options.max_partition_order); +#endif - if(avctx->frame_size > 0) { - if(avctx->frame_size < FLAC_MIN_BLOCKSIZE || + if (avctx->frame_size > 0) { + if (avctx->frame_size < FLAC_MIN_BLOCKSIZE || avctx->frame_size > FLAC_MAX_BLOCKSIZE) { av_log(avctx, AV_LOG_ERROR, "invalid block size: %d\n", avctx->frame_size); return -1; } - s->blocksize = avctx->frame_size; } else { - s->blocksize = select_blocksize(s->samplerate, s->options.block_time_ms); - avctx->frame_size = s->blocksize; + s->avctx->frame_size = select_blocksize(s->samplerate, s->options.block_time_ms); } - av_log(avctx, AV_LOG_DEBUG, " block size: %d\n", s->blocksize); + s->max_blocksize = s->avctx->frame_size; +#if FF_API_FLAC_GLOBAL_OPTS /* set LPC precision */ - if(avctx->lpc_coeff_precision > 0) { - if(avctx->lpc_coeff_precision > MAX_LPC_PRECISION) { + if (avctx->lpc_coeff_precision > 0) { + if (avctx->lpc_coeff_precision > MAX_LPC_PRECISION) { av_log(avctx, AV_LOG_ERROR, "invalid lpc coeff precision: %d\n", avctx->lpc_coeff_precision); return -1; } s->options.lpc_coeff_precision = avctx->lpc_coeff_precision; - } else { - /* select LPC precision based on block size */ - if( s->blocksize <= 192) s->options.lpc_coeff_precision = 7; - else if(s->blocksize <= 384) s->options.lpc_coeff_precision = 8; - else if(s->blocksize <= 576) s->options.lpc_coeff_precision = 9; - else if(s->blocksize <= 1152) s->options.lpc_coeff_precision = 10; - else if(s->blocksize <= 2304) s->options.lpc_coeff_precision = 11; - else if(s->blocksize <= 4608) s->options.lpc_coeff_precision = 12; - else if(s->blocksize <= 8192) s->options.lpc_coeff_precision = 13; - else if(s->blocksize <= 16384) s->options.lpc_coeff_precision = 14; - else s->options.lpc_coeff_precision = 15; } - av_log(avctx, AV_LOG_DEBUG, " lpc precision: %d\n", - s->options.lpc_coeff_precision); +#endif /* set maximum encoded frame size in verbatim mode */ - if(s->channels == 2) { - s->max_framesize = 14 + ((s->blocksize * 33 + 7) >> 3); - } else { - s->max_framesize = 14 + (s->blocksize * s->channels * 2); - } + s->max_framesize = ff_flac_get_max_frame_size(s->avctx->frame_size, + s->channels, 16); + + /* initialize MD5 context */ + s->md5ctx = av_malloc(av_md5_size); + if (!s->md5ctx) + return AVERROR(ENOMEM); + av_md5_init(s->md5ctx); streaminfo = av_malloc(FLAC_STREAMINFO_SIZE); + if (!streaminfo) + return AVERROR(ENOMEM); write_streaminfo(s, streaminfo); avctx->extradata = streaminfo; avctx->extradata_size = FLAC_STREAMINFO_SIZE; - s->frame_count = 0; + s->frame_count = 0; + s->min_framesize = s->max_framesize; avctx->coded_frame = avcodec_alloc_frame(); - avctx->coded_frame->key_frame = 1; + if (!avctx->coded_frame) + return AVERROR(ENOMEM); - return 0; + ret = ff_lpc_init(&s->lpc_ctx, avctx->frame_size, + s->options.max_prediction_order, FF_LPC_TYPE_LEVINSON); + + dprint_compression_options(s); + + return ret; } + static void init_frame(FlacEncodeContext *s) { int i, ch; @@ -408,17 +455,17 @@ static void init_frame(FlacEncodeContext *s) frame = &s->frame; - for(i=0; i<16; i++) { - if(s->blocksize == flac_blocksizes[i]) { - frame->blocksize = flac_blocksizes[i]; + for (i = 0; i < 16; i++) { + if (s->avctx->frame_size == ff_flac_blocksize_table[i]) { + frame->blocksize = ff_flac_blocksize_table[i]; frame->bs_code[0] = i; frame->bs_code[1] = 0; break; } } - if(i == 16) { - frame->blocksize = s->blocksize; - if(frame->blocksize <= 256) { + if (i == 16) { + frame->blocksize = s->avctx->frame_size; + if (frame->blocksize <= 256) { frame->bs_code[0] = 6; frame->bs_code[1] = frame->blocksize-1; } else { @@ -427,45 +474,107 @@ static void init_frame(FlacEncodeContext *s) } } - for(ch=0; chchannels; ch++) { + for (ch = 0; ch < s->channels; ch++) frame->subframes[ch].obits = 16; - } + + frame->verbatim_only = 0; } + /** - * Copy channel-interleaved input samples into separate subframes + * Copy channel-interleaved input samples into separate subframes. */ -static void copy_samples(FlacEncodeContext *s, int16_t *samples) +static void copy_samples(FlacEncodeContext *s, const int16_t *samples) { int i, j, ch; FlacFrame *frame; frame = &s->frame; - for(i=0,j=0; iblocksize; i++) { - for(ch=0; chchannels; ch++,j++) { + for (i = 0, j = 0; i < frame->blocksize; i++) + for (ch = 0; ch < s->channels; ch++, j++) frame->subframes[ch].samples[i] = samples[j]; +} + + +static int rice_count_exact(int32_t *res, int n, int k) +{ + int i; + int count = 0; + + for (i = 0; i < n; i++) { + int32_t v = -2 * res[i] - 1; + v ^= v >> 31; + count += (v >> k) + 1 + k; + } + return count; +} + + +static int subframe_count_exact(FlacEncodeContext *s, FlacSubframe *sub, + int pred_order) +{ + int p, porder, psize; + int i, part_end; + int count = 0; + + /* subframe header */ + count += 8; + + /* subframe */ + if (sub->type == FLAC_SUBFRAME_CONSTANT) { + count += sub->obits; + } else if (sub->type == FLAC_SUBFRAME_VERBATIM) { + count += s->frame.blocksize * sub->obits; + } else { + /* warm-up samples */ + count += pred_order * sub->obits; + + /* LPC coefficients */ + if (sub->type == FLAC_SUBFRAME_LPC) + count += 4 + 5 + pred_order * s->options.lpc_coeff_precision; + + /* rice-encoded block */ + count += 2; + + /* partition order */ + porder = sub->rc.porder; + psize = s->frame.blocksize >> porder; + count += 4; + + /* residual */ + i = pred_order; + part_end = psize; + for (p = 0; p < 1 << porder; p++) { + int k = sub->rc.params[p]; + count += 4; + count += rice_count_exact(&sub->residual[i], part_end - i, k); + i = part_end; + part_end = FFMIN(s->frame.blocksize, part_end + psize); } } + + return count; } #define rice_encode_count(sum, n, k) (((n)*((k)+1))+((sum-(n>>1))>>(k))) /** - * Solve for d/dk(rice_encode_count) = n-((sum-(n>>1))>>(k+1)) = 0 + * Solve for d/dk(rice_encode_count) = n-((sum-(n>>1))>>(k+1)) = 0. */ static int find_optimal_param(uint32_t sum, int n) { int k; uint32_t sum2; - if(sum <= n>>1) + if (sum <= n >> 1) return 0; - sum2 = sum-(n>>1); - k = av_log2(n<256 ? FASTDIV(sum2,n) : sum2/n); + sum2 = sum - (n >> 1); + k = av_log2(n < 256 ? FASTDIV(sum2, n) : sum2 / n); return FFMIN(k, MAX_RICE_PARAM); } + static uint32_t calc_optimal_rice_params(RiceContext *rc, int porder, uint32_t *sums, int n, int pred_order) { @@ -473,23 +582,23 @@ static uint32_t calc_optimal_rice_params(RiceContext *rc, int porder, int k, cnt, part; uint32_t all_bits; - part = (1 << porder); - all_bits = 0; + part = (1 << porder); + all_bits = 4 * part; cnt = (n >> porder) - pred_order; - for(i=0; i> porder); + for (i = 0; i < part; i++) { k = find_optimal_param(sums[i], cnt); rc->params[i] = k; all_bits += rice_encode_count(sums[i], cnt, k); + cnt = n >> porder; } - all_bits += (4 * part); rc->porder = porder; return all_bits; } + static void calc_sums(int pmin, int pmax, uint32_t *data, int n, int pred_order, uint32_t sums[][MAX_PARTITIONS]) { @@ -498,26 +607,25 @@ static void calc_sums(int pmin, int pmax, uint32_t *data, int n, int pred_order, uint32_t *res, *res_end; /* sums for highest level */ - parts = (1 << pmax); - res = &data[pred_order]; + parts = (1 << pmax); + res = &data[pred_order]; res_end = &data[n >> pmax]; - for(i=0; i> pmax; + res_end += n >> pmax; } /* sums for lower levels */ - for(i=pmax-1; i>=pmin; i--) { + for (i = pmax - 1; i >= pmin; i--) { parts = (1 << i); - for(j=0; j>31); - } calc_sums(pmin, pmax, udata, n, pred_order, sums); opt_porder = pmin; bits[pmin] = UINT32_MAX; - for(i=pmin; i<=pmax; i++) { + for (i = pmin; i <= pmax; i++) { bits[i] = calc_optimal_rice_params(&tmp_rc, i, sums[i], n, pred_order); - if(bits[i] <= bits[opt_porder]) { + if (bits[i] <= bits[opt_porder]) { opt_porder = i; - *rc= tmp_rc; + *rc = tmp_rc; } } @@ -553,575 +660,376 @@ static uint32_t calc_rice_params(RiceContext *rc, int pmin, int pmax, return bits[opt_porder]; } + static int get_max_p_order(int max_porder, int n, int order) { int porder = FFMIN(max_porder, av_log2(n^(n-1))); - if(order > 0) + if (order > 0) porder = FFMIN(porder, av_log2(n/order)); return porder; } -static uint32_t calc_rice_params_fixed(RiceContext *rc, int pmin, int pmax, - int32_t *data, int n, int pred_order, - int bps) -{ - uint32_t bits; - pmin = get_max_p_order(pmin, n, pred_order); - pmax = get_max_p_order(pmax, n, pred_order); - bits = pred_order*bps + 6; - bits += calc_rice_params(rc, pmin, pmax, data, n, pred_order); - return bits; -} -static uint32_t calc_rice_params_lpc(RiceContext *rc, int pmin, int pmax, - int32_t *data, int n, int pred_order, - int bps, int precision) +static uint32_t find_subframe_rice_params(FlacEncodeContext *s, + FlacSubframe *sub, int pred_order) { - uint32_t bits; - pmin = get_max_p_order(pmin, n, pred_order); - pmax = get_max_p_order(pmax, n, pred_order); - bits = pred_order*bps + 4 + 5 + pred_order*precision + 6; - bits += calc_rice_params(rc, pmin, pmax, data, n, pred_order); + int pmin = get_max_p_order(s->options.min_partition_order, + s->frame.blocksize, pred_order); + int pmax = get_max_p_order(s->options.max_partition_order, + s->frame.blocksize, pred_order); + + uint32_t bits = 8 + pred_order * sub->obits + 2 + 4; + if (sub->type == FLAC_SUBFRAME_LPC) + bits += 4 + 5 + pred_order * s->options.lpc_coeff_precision; + bits += calc_rice_params(&sub->rc, pmin, pmax, sub->residual, + s->frame.blocksize, pred_order); return bits; } -/** - * Apply Welch window function to audio block - */ -static void apply_welch_window(const int32_t *data, int len, double *w_data) -{ - int i, n2; - double w; - double c; - - n2 = (len >> 1); - c = 2.0 / (len - 1.0); - for(i=0; i> 1); - lpc_tmp[i] = r; - for(j=0; j qmax) && (sh > 0)) { - sh--; - } - - /* since negative shift values are unsupported in decoder, scale down - coefficients instead */ - if(sh == 0 && cmax > qmax) { - double scale = ((double)qmax) / cmax; - for(i=0; i=0; i--) { - if(ref[i] > 0.10) { - est = i+1; - break; + if (order == 0) { + for (i = order; i < n; i++) + res[i] = smp[i]; + } else if (order == 1) { + for (i = order; i < n; i++) + res[i] = smp[i] - smp[i-1]; + } else if (order == 2) { + int a = smp[order-1] - smp[order-2]; + for (i = order; i < n; i += 2) { + int b = smp[i ] - smp[i-1]; + res[i] = b - a; + a = smp[i+1] - smp[i ]; + res[i+1] = a - b; } - } - return est; -} - -/** - * Calculate LPC coefficients for multiple orders - */ -static int lpc_calc_coefs(FlacEncodeContext *s, - const int32_t *samples, int blocksize, int max_order, - int precision, int32_t coefs[][MAX_LPC_ORDER], - int *shift, int use_lpc, int omethod) -{ - double autoc[MAX_LPC_ORDER+1]; - double ref[MAX_LPC_ORDER]; - double lpc[MAX_LPC_ORDER][MAX_LPC_ORDER]; - int i, j, pass; - int opt_order; - - assert(max_order >= MIN_LPC_ORDER && max_order <= MAX_LPC_ORDER); - - if(use_lpc == 1){ - s->dsp.flac_compute_autocorr(samples, blocksize, max_order, autoc); - - compute_lpc_coefs(autoc, max_order, lpc, ref); - }else{ - LLSModel m[2]; - double var[MAX_LPC_ORDER+1], eval, weight; - - for(pass=0; pass>pass) + fabs(eval - var[0]); - for(j=0; j<=max_order; j++) - var[j]/= sqrt(eval); - weight += 1/eval; - }else - weight++; - - av_update_lls(&m[pass&1], var, 1.0); - } - av_solve_lls(&m[pass&1], 0.001, 0); + } else if (order == 3) { + int a = smp[order-1] - smp[order-2]; + int c = smp[order-1] - 2*smp[order-2] + smp[order-3]; + for (i = order; i < n; i += 2) { + int b = smp[i ] - smp[i-1]; + int d = b - a; + res[i] = d - c; + a = smp[i+1] - smp[i ]; + c = a - b; + res[i+1] = c - d; } - - for(i=0; i0; i--) - ref[i] = ref[i-1] - ref[i]; - } - opt_order = max_order; - - if(omethod == ORDER_METHOD_EST) { - opt_order = estimate_best_order(ref, max_order); - i = opt_order-1; - quantize_lpc_coefs(lpc[i], i+1, precision, coefs[i], &shift[i]); } else { - for(i=0; i 0); - memcpy(res, smp, n * sizeof(int32_t)); -} - -static void encode_residual_fixed(int32_t *res, const int32_t *smp, int n, - int order) -{ - int i; - - for(i=0; i> shift); res[i+1] = smp[i+1] - (p1 >> shift); } } + static void encode_residual_lpc(int32_t *res, const int32_t *smp, int n, int order, const int32_t *coefs, int shift) { int i; - for(i=0; i> shift); + res[i ] = smp[i ] - (p0 >> shift); res[i+1] = smp[i+1] - (p1 >> shift); } #else - switch(order) { - case 1: encode_residual_lpc_unrolled(res, smp, n, 1, coefs, shift, 0); break; - case 2: encode_residual_lpc_unrolled(res, smp, n, 2, coefs, shift, 0); break; - case 3: encode_residual_lpc_unrolled(res, smp, n, 3, coefs, shift, 0); break; - case 4: encode_residual_lpc_unrolled(res, smp, n, 4, coefs, shift, 0); break; - case 5: encode_residual_lpc_unrolled(res, smp, n, 5, coefs, shift, 0); break; - case 6: encode_residual_lpc_unrolled(res, smp, n, 6, coefs, shift, 0); break; - case 7: encode_residual_lpc_unrolled(res, smp, n, 7, coefs, shift, 0); break; - case 8: encode_residual_lpc_unrolled(res, smp, n, 8, coefs, shift, 0); break; - default: encode_residual_lpc_unrolled(res, smp, n, order, coefs, shift, 1); break; + switch (order) { + case 1: encode_residual_lpc_unrolled(res, smp, n, 1, coefs, shift, 0); break; + case 2: encode_residual_lpc_unrolled(res, smp, n, 2, coefs, shift, 0); break; + case 3: encode_residual_lpc_unrolled(res, smp, n, 3, coefs, shift, 0); break; + case 4: encode_residual_lpc_unrolled(res, smp, n, 4, coefs, shift, 0); break; + case 5: encode_residual_lpc_unrolled(res, smp, n, 5, coefs, shift, 0); break; + case 6: encode_residual_lpc_unrolled(res, smp, n, 6, coefs, shift, 0); break; + case 7: encode_residual_lpc_unrolled(res, smp, n, 7, coefs, shift, 0); break; + case 8: encode_residual_lpc_unrolled(res, smp, n, 8, coefs, shift, 0); break; + default: encode_residual_lpc_unrolled(res, smp, n, order, coefs, shift, 1); break; } #endif } -static int encode_residual(FlacEncodeContext *ctx, int ch) + +static int encode_residual_ch(FlacEncodeContext *s, int ch) { int i, n; - int min_order, max_order, opt_order, precision, omethod; - int min_porder, max_porder; + int min_order, max_order, opt_order, omethod; FlacFrame *frame; FlacSubframe *sub; int32_t coefs[MAX_LPC_ORDER][MAX_LPC_ORDER]; int shift[MAX_LPC_ORDER]; int32_t *res, *smp; - frame = &ctx->frame; - sub = &frame->subframes[ch]; - res = sub->residual; - smp = sub->samples; - n = frame->blocksize; + frame = &s->frame; + sub = &frame->subframes[ch]; + res = sub->residual; + smp = sub->samples; + n = frame->blocksize; /* CONSTANT */ - for(i=1; itype = sub->type_code = FLAC_SUBFRAME_CONSTANT; res[0] = smp[0]; - return sub->obits; + return subframe_count_exact(s, sub, 0); } /* VERBATIM */ - if(n < 5) { + if (frame->verbatim_only || n < 5) { sub->type = sub->type_code = FLAC_SUBFRAME_VERBATIM; - encode_residual_verbatim(res, smp, n); - return sub->obits * n; + memcpy(res, smp, n * sizeof(int32_t)); + return subframe_count_exact(s, sub, 0); } - min_order = ctx->options.min_prediction_order; - max_order = ctx->options.max_prediction_order; - min_porder = ctx->options.min_partition_order; - max_porder = ctx->options.max_partition_order; - precision = ctx->options.lpc_coeff_precision; - omethod = ctx->options.prediction_order_method; + min_order = s->options.min_prediction_order; + max_order = s->options.max_prediction_order; + omethod = s->options.prediction_order_method; /* FIXED */ - if(!ctx->options.use_lpc || max_order == 0 || (n <= max_order)) { + sub->type = FLAC_SUBFRAME_FIXED; + if (s->options.lpc_type == FF_LPC_TYPE_NONE || + s->options.lpc_type == FF_LPC_TYPE_FIXED || n <= max_order) { uint32_t bits[MAX_FIXED_ORDER+1]; - if(max_order > MAX_FIXED_ORDER) max_order = MAX_FIXED_ORDER; + if (max_order > MAX_FIXED_ORDER) + max_order = MAX_FIXED_ORDER; opt_order = 0; - bits[0] = UINT32_MAX; - for(i=min_order; i<=max_order; i++) { + bits[0] = UINT32_MAX; + for (i = min_order; i <= max_order; i++) { encode_residual_fixed(res, smp, n, i); - bits[i] = calc_rice_params_fixed(&sub->rc, min_porder, max_porder, res, - n, i, sub->obits); - if(bits[i] < bits[opt_order]) { + bits[i] = find_subframe_rice_params(s, sub, i); + if (bits[i] < bits[opt_order]) opt_order = i; - } } - sub->order = opt_order; - sub->type = FLAC_SUBFRAME_FIXED; + sub->order = opt_order; sub->type_code = sub->type | sub->order; - if(sub->order != max_order) { + if (sub->order != max_order) { encode_residual_fixed(res, smp, n, sub->order); - return calc_rice_params_fixed(&sub->rc, min_porder, max_porder, res, n, - sub->order, sub->obits); + find_subframe_rice_params(s, sub, sub->order); } - return bits[sub->order]; + return subframe_count_exact(s, sub, sub->order); } /* LPC */ - opt_order = lpc_calc_coefs(ctx, smp, n, max_order, precision, coefs, shift, ctx->options.use_lpc, omethod); - - if(omethod == ORDER_METHOD_2LEVEL || - omethod == ORDER_METHOD_4LEVEL || - omethod == ORDER_METHOD_8LEVEL) { + sub->type = FLAC_SUBFRAME_LPC; + opt_order = ff_lpc_calc_coefs(&s->lpc_ctx, smp, n, min_order, max_order, + s->options.lpc_coeff_precision, coefs, shift, s->options.lpc_type, + s->options.lpc_passes, omethod, + MAX_LPC_SHIFT, 0); + + if (omethod == ORDER_METHOD_2LEVEL || + omethod == ORDER_METHOD_4LEVEL || + omethod == ORDER_METHOD_8LEVEL) { int levels = 1 << omethod; - uint32_t bits[levels]; + uint32_t bits[1 << ORDER_METHOD_8LEVEL]; int order; - int opt_index = levels-1; - opt_order = max_order-1; + int opt_index = levels-1; + opt_order = max_order-1; bits[opt_index] = UINT32_MAX; - for(i=levels-1; i>=0; i--) { + for (i = levels-1; i >= 0; i--) { order = min_order + (((max_order-min_order+1) * (i+1)) / levels)-1; - if(order < 0) order = 0; + if (order < 0) + order = 0; encode_residual_lpc(res, smp, n, order+1, coefs[order], shift[order]); - bits[i] = calc_rice_params_lpc(&sub->rc, min_porder, max_porder, - res, n, order+1, sub->obits, precision); - if(bits[i] < bits[opt_index]) { + bits[i] = find_subframe_rice_params(s, sub, order+1); + if (bits[i] < bits[opt_index]) { opt_index = i; opt_order = order; } } opt_order++; - } else if(omethod == ORDER_METHOD_SEARCH) { + } else if (omethod == ORDER_METHOD_SEARCH) { // brute-force optimal order search uint32_t bits[MAX_LPC_ORDER]; opt_order = 0; - bits[0] = UINT32_MAX; - for(i=min_order-1; irc, min_porder, max_porder, - res, n, i+1, sub->obits, precision); - if(bits[i] < bits[opt_order]) { + bits[i] = find_subframe_rice_params(s, sub, i+1); + if (bits[i] < bits[opt_order]) opt_order = i; - } } opt_order++; - } else if(omethod == ORDER_METHOD_LOG) { + } else if (omethod == ORDER_METHOD_LOG) { uint32_t bits[MAX_LPC_ORDER]; int step; - opt_order= min_order - 1 + (max_order-min_order)/3; + opt_order = min_order - 1 + (max_order-min_order)/3; memset(bits, -1, sizeof(bits)); - for(step=16 ;step; step>>=1){ - int last= opt_order; - for(i=last-step; i<=last+step; i+= step){ - if(i=max_order || bits[i] < UINT32_MAX) + for (step = 16; step; step >>= 1) { + int last = opt_order; + for (i = last-step; i <= last+step; i += step) { + if (i < min_order-1 || i >= max_order || bits[i] < UINT32_MAX) continue; encode_residual_lpc(res, smp, n, i+1, coefs[i], shift[i]); - bits[i] = calc_rice_params_lpc(&sub->rc, min_porder, max_porder, - res, n, i+1, sub->obits, precision); - if(bits[i] < bits[opt_order]) - opt_order= i; + bits[i] = find_subframe_rice_params(s, sub, i+1); + if (bits[i] < bits[opt_order]) + opt_order = i; } } opt_order++; } - sub->order = opt_order; - sub->type = FLAC_SUBFRAME_LPC; + sub->order = opt_order; sub->type_code = sub->type | (sub->order-1); - sub->shift = shift[sub->order-1]; - for(i=0; iorder; i++) { + sub->shift = shift[sub->order-1]; + for (i = 0; i < sub->order; i++) sub->coefs[i] = coefs[sub->order-1][i]; - } + encode_residual_lpc(res, smp, n, sub->order, sub->coefs, sub->shift); - return calc_rice_params_lpc(&sub->rc, min_porder, max_porder, res, n, sub->order, - sub->obits, precision); + + find_subframe_rice_params(s, sub, sub->order); + + return subframe_count_exact(s, sub, sub->order); } -static int encode_residual_v(FlacEncodeContext *ctx, int ch) + +static int count_frame_header(FlacEncodeContext *s) { - int i, n; - FlacFrame *frame; - FlacSubframe *sub; - int32_t *res, *smp; + uint8_t tmp; + int count; + + /* + <14> Sync code + <1> Reserved + <1> Blocking strategy + <4> Block size in inter-channel samples + <4> Sample rate + <4> Channel assignment + <3> Sample size in bits + <1> Reserved + */ + count = 32; + + /* coded frame number */ + PUT_UTF8(s->frame_count, tmp, count += 8;) + + /* explicit block size */ + if (s->frame.bs_code[0] == 6) + count += 8; + else if (s->frame.bs_code[0] == 7) + count += 16; + + /* explicit sample rate */ + count += ((s->sr_code[0] == 12) + (s->sr_code[0] > 12)) * 8; + + /* frame header CRC-8 */ + count += 8; + + return count; +} - frame = &ctx->frame; - sub = &frame->subframes[ch]; - res = sub->residual; - smp = sub->samples; - n = frame->blocksize; - /* CONSTANT */ - for(i=1; itype = sub->type_code = FLAC_SUBFRAME_CONSTANT; - res[0] = smp[0]; - return sub->obits; - } +static int encode_frame(FlacEncodeContext *s) +{ + int ch, count; - /* VERBATIM */ - sub->type = sub->type_code = FLAC_SUBFRAME_VERBATIM; - encode_residual_verbatim(res, smp, n); - return sub->obits * n; + count = count_frame_header(s); + + for (ch = 0; ch < s->channels; ch++) + count += encode_residual_ch(s, ch); + + count += (8 - (count & 7)) & 7; // byte alignment + count += 16; // CRC-16 + + return count >> 3; } + static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n) { int i, best; @@ -1132,8 +1040,8 @@ static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n) /* calculate sum of 2nd order residual for each channel */ sum[0] = sum[1] = sum[2] = sum[3] = 0; - for(i=2; i> 1); sum[3] += FFABS(lt - rt); @@ -1141,9 +1049,9 @@ static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n) sum[1] += FFABS(rt); } /* estimate bit counts */ - for(i=0; i<4; i++) { - k = find_optimal_param(2*sum[i], n); - sum[i] = rice_encode_count(2*sum[i], n, k); + for (i = 0; i < 4; i++) { + k = find_optimal_param(2 * sum[i], n); + sum[i] = rice_encode_count( 2 * sum[i], n, k); } /* calculate score for each mode */ @@ -1154,74 +1062,64 @@ static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n) /* return mode with lowest score */ best = 0; - for(i=1; i<4; i++) { - if(score[i] < score[best]) { + for (i = 1; i < 4; i++) + if (score[i] < score[best]) best = i; - } - } - if(best == 0) { - return FLAC_CHMODE_LEFT_RIGHT; - } else if(best == 1) { + if (best == 0) { + return FLAC_CHMODE_INDEPENDENT; + } else if (best == 1) { return FLAC_CHMODE_LEFT_SIDE; - } else if(best == 2) { + } else if (best == 2) { return FLAC_CHMODE_RIGHT_SIDE; } else { return FLAC_CHMODE_MID_SIDE; } } + /** - * Perform stereo channel decorrelation + * Perform stereo channel decorrelation. */ -static void channel_decorrelation(FlacEncodeContext *ctx) +static void channel_decorrelation(FlacEncodeContext *s) { FlacFrame *frame; int32_t *left, *right; int i, n; - frame = &ctx->frame; - n = frame->blocksize; + frame = &s->frame; + n = frame->blocksize; left = frame->subframes[0].samples; right = frame->subframes[1].samples; - if(ctx->channels != 2) { - frame->ch_mode = FLAC_CHMODE_NOT_STEREO; + if (s->channels != 2) { + frame->ch_mode = FLAC_CHMODE_INDEPENDENT; return; } frame->ch_mode = estimate_stereo_mode(left, right, n); /* perform decorrelation and adjust bits-per-sample */ - if(frame->ch_mode == FLAC_CHMODE_LEFT_RIGHT) { + if (frame->ch_mode == FLAC_CHMODE_INDEPENDENT) return; - } - if(frame->ch_mode == FLAC_CHMODE_MID_SIDE) { + if (frame->ch_mode == FLAC_CHMODE_MID_SIDE) { int32_t tmp; - for(i=0; i> 1; - right[i] = tmp - right[i]; + for (i = 0; i < n; i++) { + tmp = left[i]; + left[i] = (tmp + right[i]) >> 1; + right[i] = tmp - right[i]; } frame->subframes[1].obits++; - } else if(frame->ch_mode == FLAC_CHMODE_LEFT_SIDE) { - for(i=0; ich_mode == FLAC_CHMODE_LEFT_SIDE) { + for (i = 0; i < n; i++) right[i] = left[i] - right[i]; - } frame->subframes[1].obits++; } else { - for(i=0; isubframes[0].obits++; } } -static void put_sbits(PutBitContext *pb, int bits, int32_t val) -{ - assert(bits >= 0 && bits <= 31); - - put_bits(pb, bits, val & ((1<pb, 16, 0xFFF8); put_bits(&s->pb, 4, frame->bs_code[0]); put_bits(&s->pb, 4, s->sr_code[0]); - if(frame->ch_mode == FLAC_CHMODE_NOT_STEREO) { - put_bits(&s->pb, 4, s->ch_code); - } else { + + if (frame->ch_mode == FLAC_CHMODE_INDEPENDENT) + put_bits(&s->pb, 4, s->channels-1); + else put_bits(&s->pb, 4, frame->ch_mode); - } + put_bits(&s->pb, 3, 4); /* bits-per-sample code */ put_bits(&s->pb, 1, 0); write_utf8(&s->pb, s->frame_count); - if(frame->bs_code[0] == 6) { + + if (frame->bs_code[0] == 6) put_bits(&s->pb, 8, frame->bs_code[1]); - } else if(frame->bs_code[0] == 7) { + else if (frame->bs_code[0] == 7) put_bits(&s->pb, 16, frame->bs_code[1]); - } - if(s->sr_code[0] == 12) { + + if (s->sr_code[0] == 12) put_bits(&s->pb, 8, s->sr_code[1]); - } else if(s->sr_code[0] > 12) { + else if (s->sr_code[0] > 12) put_bits(&s->pb, 16, s->sr_code[1]); - } + flush_put_bits(&s->pb); - crc = av_crc(av_crc07, 0, s->pb.buf, put_bits_count(&s->pb)>>3); + crc = av_crc(av_crc_get_table(AV_CRC_8_ATM), 0, s->pb.buf, + put_bits_count(&s->pb) >> 3); put_bits(&s->pb, 8, crc); } -static void output_subframe_constant(FlacEncodeContext *s, int ch) -{ - FlacSubframe *sub; - int32_t res; - - sub = &s->frame.subframes[ch]; - res = sub->residual[0]; - put_sbits(&s->pb, sub->obits, res); -} -static void output_subframe_verbatim(FlacEncodeContext *s, int ch) +static void write_subframes(FlacEncodeContext *s) { - int i; - FlacFrame *frame; - FlacSubframe *sub; - int32_t res; - - frame = &s->frame; - sub = &frame->subframes[ch]; - - for(i=0; iblocksize; i++) { - res = sub->residual[i]; - put_sbits(&s->pb, sub->obits, res); - } -} - -static void output_residual(FlacEncodeContext *ctx, int ch) -{ - int i, j, p, n, parts; - int k, porder, psize, res_cnt; - FlacFrame *frame; - FlacSubframe *sub; - int32_t *res; - - frame = &ctx->frame; - sub = &frame->subframes[ch]; - res = sub->residual; - n = frame->blocksize; - - /* rice-encoded block */ - put_bits(&ctx->pb, 2, 0); - - /* partition order */ - porder = sub->rc.porder; - psize = n >> porder; - parts = (1 << porder); - put_bits(&ctx->pb, 4, porder); - res_cnt = psize - sub->order; - - /* residual */ - j = sub->order; - for(p=0; prc.params[p]; - put_bits(&ctx->pb, 4, k); - if(p == 1) res_cnt = psize; - for(i=0; ipb, res[j], k, INT32_MAX, 0); - } - } -} - -static void output_subframe_fixed(FlacEncodeContext *ctx, int ch) -{ - int i; - FlacFrame *frame; - FlacSubframe *sub; - - frame = &ctx->frame; - sub = &frame->subframes[ch]; - - /* warm-up samples */ - for(i=0; iorder; i++) { - put_sbits(&ctx->pb, sub->obits, sub->residual[i]); - } - - /* residual */ - output_residual(ctx, ch); -} - -static void output_subframe_lpc(FlacEncodeContext *ctx, int ch) -{ - int i, cbits; - FlacFrame *frame; - FlacSubframe *sub; - - frame = &ctx->frame; - sub = &frame->subframes[ch]; - - /* warm-up samples */ - for(i=0; iorder; i++) { - put_sbits(&ctx->pb, sub->obits, sub->residual[i]); - } - - /* LPC coefficients */ - cbits = ctx->options.lpc_coeff_precision; - put_bits(&ctx->pb, 4, cbits-1); - put_sbits(&ctx->pb, 5, sub->shift); - for(i=0; iorder; i++) { - put_sbits(&ctx->pb, cbits, sub->coefs[i]); - } - - /* residual */ - output_residual(ctx, ch); -} - -static void output_subframes(FlacEncodeContext *s) -{ - FlacFrame *frame; - FlacSubframe *sub; int ch; - frame = &s->frame; - - for(ch=0; chchannels; ch++) { - sub = &frame->subframes[ch]; + for (ch = 0; ch < s->channels; ch++) { + FlacSubframe *sub = &s->frame.subframes[ch]; + int i, p, porder, psize; + int32_t *part_end; + int32_t *res = sub->residual; + int32_t *frame_end = &sub->residual[s->frame.blocksize]; /* subframe header */ put_bits(&s->pb, 1, 0); @@ -1384,91 +1182,191 @@ static void output_subframes(FlacEncodeContext *s) put_bits(&s->pb, 1, 0); /* no wasted bits */ /* subframe */ - if(sub->type == FLAC_SUBFRAME_CONSTANT) { - output_subframe_constant(s, ch); - } else if(sub->type == FLAC_SUBFRAME_VERBATIM) { - output_subframe_verbatim(s, ch); - } else if(sub->type == FLAC_SUBFRAME_FIXED) { - output_subframe_fixed(s, ch); - } else if(sub->type == FLAC_SUBFRAME_LPC) { - output_subframe_lpc(s, ch); + if (sub->type == FLAC_SUBFRAME_CONSTANT) { + put_sbits(&s->pb, sub->obits, res[0]); + } else if (sub->type == FLAC_SUBFRAME_VERBATIM) { + while (res < frame_end) + put_sbits(&s->pb, sub->obits, *res++); + } else { + /* warm-up samples */ + for (i = 0; i < sub->order; i++) + put_sbits(&s->pb, sub->obits, *res++); + + /* LPC coefficients */ + if (sub->type == FLAC_SUBFRAME_LPC) { + int cbits = s->options.lpc_coeff_precision; + put_bits( &s->pb, 4, cbits-1); + put_sbits(&s->pb, 5, sub->shift); + for (i = 0; i < sub->order; i++) + put_sbits(&s->pb, cbits, sub->coefs[i]); + } + + /* rice-encoded block */ + put_bits(&s->pb, 2, 0); + + /* partition order */ + porder = sub->rc.porder; + psize = s->frame.blocksize >> porder; + put_bits(&s->pb, 4, porder); + + /* residual */ + part_end = &sub->residual[psize]; + for (p = 0; p < 1 << porder; p++) { + int k = sub->rc.params[p]; + put_bits(&s->pb, 4, k); + while (res < part_end) + set_sr_golomb_flac(&s->pb, *res++, k, INT32_MAX, 0); + part_end = FFMIN(frame_end, part_end + psize); + } } } } -static void output_frame_footer(FlacEncodeContext *s) + +static void write_frame_footer(FlacEncodeContext *s) { int crc; flush_put_bits(&s->pb); - crc = bswap_16(av_crc(av_crc8005, 0, s->pb.buf, put_bits_count(&s->pb)>>3)); + crc = av_bswap16(av_crc(av_crc_get_table(AV_CRC_16_ANSI), 0, s->pb.buf, + put_bits_count(&s->pb)>>3)); put_bits(&s->pb, 16, crc); flush_put_bits(&s->pb); } + +static int write_frame(FlacEncodeContext *s, uint8_t *frame, int buf_size) +{ + init_put_bits(&s->pb, frame, buf_size); + write_frame_header(s); + write_subframes(s); + write_frame_footer(s); + return put_bits_count(&s->pb) >> 3; +} + + +static void update_md5_sum(FlacEncodeContext *s, const int16_t *samples) +{ +#if HAVE_BIGENDIAN + int i; + for (i = 0; i < s->frame.blocksize * s->channels; i++) { + int16_t smp = av_le2ne16(samples[i]); + av_md5_update(s->md5ctx, (uint8_t *)&smp, 2); + } +#else + av_md5_update(s->md5ctx, (const uint8_t *)samples, s->frame.blocksize*s->channels*2); +#endif +} + + static int flac_encode_frame(AVCodecContext *avctx, uint8_t *frame, int buf_size, void *data) { - int ch; FlacEncodeContext *s; - int16_t *samples = data; - int out_bytes; + const int16_t *samples = data; + int frame_bytes, out_bytes; s = avctx->priv_data; - s->blocksize = avctx->frame_size; + /* when the last block is reached, update the header in extradata */ + if (!data) { + s->max_framesize = s->max_encoded_framesize; + av_md5_final(s->md5ctx, s->md5sum); + write_streaminfo(s, avctx->extradata); + return 0; + } + + /* change max_framesize for small final frame */ + if (avctx->frame_size < s->frame.blocksize) { + s->max_framesize = ff_flac_get_max_frame_size(avctx->frame_size, + s->channels, 16); + } + init_frame(s); copy_samples(s, samples); channel_decorrelation(s); - for(ch=0; chchannels; ch++) { - encode_residual(s, ch); + frame_bytes = encode_frame(s); + + /* fallback to verbatim mode if the compressed frame is larger than it + would be if encoded uncompressed. */ + if (frame_bytes > s->max_framesize) { + s->frame.verbatim_only = 1; + frame_bytes = encode_frame(s); } - init_put_bits(&s->pb, frame, buf_size); - output_frame_header(s); - output_subframes(s); - output_frame_footer(s); - out_bytes = put_bits_count(&s->pb) >> 3; - - if(out_bytes > s->max_framesize || out_bytes >= buf_size) { - /* frame too large. use verbatim mode */ - for(ch=0; chchannels; ch++) { - encode_residual_v(s, ch); - } - init_put_bits(&s->pb, frame, buf_size); - output_frame_header(s); - output_subframes(s); - output_frame_footer(s); - out_bytes = put_bits_count(&s->pb) >> 3; - - if(out_bytes > s->max_framesize || out_bytes >= buf_size) { - /* still too large. must be an error. */ - av_log(avctx, AV_LOG_ERROR, "error encoding frame\n"); - return -1; - } + + if (buf_size < frame_bytes) { + av_log(avctx, AV_LOG_ERROR, "output buffer too small\n"); + return 0; } + out_bytes = write_frame(s, frame, buf_size); s->frame_count++; + avctx->coded_frame->pts = s->sample_count; + s->sample_count += avctx->frame_size; + update_md5_sum(s, samples); + if (out_bytes > s->max_encoded_framesize) + s->max_encoded_framesize = out_bytes; + if (out_bytes < s->min_framesize) + s->min_framesize = out_bytes; + return out_bytes; } -static int flac_encode_close(AVCodecContext *avctx) + +static av_cold int flac_encode_close(AVCodecContext *avctx) { + if (avctx->priv_data) { + FlacEncodeContext *s = avctx->priv_data; + av_freep(&s->md5ctx); + ff_lpc_end(&s->lpc_ctx); + } av_freep(&avctx->extradata); avctx->extradata_size = 0; av_freep(&avctx->coded_frame); return 0; } -AVCodec flac_encoder = { +#define FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM +static const AVOption options[] = { +{ "lpc_coeff_precision", "LPC coefficient precision", offsetof(FlacEncodeContext, options.lpc_coeff_precision), FF_OPT_TYPE_INT, 15, 0, MAX_LPC_PRECISION, FLAGS }, +{ "lpc_type", "LPC algorithm", offsetof(FlacEncodeContext, options.lpc_type), FF_OPT_TYPE_INT, FF_LPC_TYPE_DEFAULT, FF_LPC_TYPE_DEFAULT, FF_LPC_TYPE_NB-1, FLAGS, "lpc_type" }, +{ "none", NULL, 0, FF_OPT_TYPE_CONST, FF_LPC_TYPE_NONE, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, +{ "fixed", NULL, 0, FF_OPT_TYPE_CONST, FF_LPC_TYPE_FIXED, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, +{ "levinson", NULL, 0, FF_OPT_TYPE_CONST, FF_LPC_TYPE_LEVINSON, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, +{ "cholesky", NULL, 0, FF_OPT_TYPE_CONST, FF_LPC_TYPE_CHOLESKY, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, +{ "lpc_passes", "Number of passes to use for Cholesky factorization during LPC analysis", offsetof(FlacEncodeContext, options.lpc_passes), FF_OPT_TYPE_INT, -1, INT_MIN, INT_MAX, FLAGS }, +{ "min_partition_order", NULL, offsetof(FlacEncodeContext, options.min_partition_order), FF_OPT_TYPE_INT, -1, -1, MAX_PARTITION_ORDER, FLAGS }, +{ "max_partition_order", NULL, offsetof(FlacEncodeContext, options.max_partition_order), FF_OPT_TYPE_INT, -1, -1, MAX_PARTITION_ORDER, FLAGS }, +{ "prediction_order_method", "Search method for selecting prediction order", offsetof(FlacEncodeContext, options.prediction_order_method), FF_OPT_TYPE_INT, -1, -1, ORDER_METHOD_LOG, FLAGS, "predm" }, +{ "estimation", NULL, 0, FF_OPT_TYPE_CONST, ORDER_METHOD_EST, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "2level", NULL, 0, FF_OPT_TYPE_CONST, ORDER_METHOD_2LEVEL, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "4level", NULL, 0, FF_OPT_TYPE_CONST, ORDER_METHOD_4LEVEL, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "8level", NULL, 0, FF_OPT_TYPE_CONST, ORDER_METHOD_8LEVEL, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "search", NULL, 0, FF_OPT_TYPE_CONST, ORDER_METHOD_SEARCH, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "log", NULL, 0, FF_OPT_TYPE_CONST, ORDER_METHOD_LOG, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ NULL }, +}; + +static const AVClass flac_encoder_class = { + "FLAC encoder", + av_default_item_name, + options, + LIBAVUTIL_VERSION_INT, +}; + +AVCodec ff_flac_encoder = { "flac", - CODEC_TYPE_AUDIO, + AVMEDIA_TYPE_AUDIO, CODEC_ID_FLAC, sizeof(FlacEncodeContext), flac_encode_init, flac_encode_frame, flac_encode_close, NULL, - .capabilities = CODEC_CAP_SMALL_LAST_FRAME, + .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY, + .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE}, + .long_name = NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"), + .priv_class = &flac_encoder_class, };