X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fflacenc.c;h=c39a721642e1eea96ee5ecc30fd0fe2b6ae74fd0;hb=2c6cf1394096d08396faadc6e7c0b404fd6df006;hp=b23d3a6c9797c5abe02a974fe5c5b70a669ea6ee;hpb=27e3418844aea51e9d8a8c145871f422504a6598;p=ffmpeg diff --git a/libavcodec/flacenc.c b/libavcodec/flacenc.c index b23d3a6c979..c39a721642e 100644 --- a/libavcodec/flacenc.c +++ b/libavcodec/flacenc.c @@ -2,28 +2,28 @@ * FLAC audio encoder * 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 "get_bits.h" -#include "dsputil.h" #include "golomb.h" #include "lpc.h" #include "flac.h" @@ -44,7 +44,7 @@ typedef struct CompressionOptions { int compression_level; int block_time_ms; - enum AVLPCType lpc_type; + enum FFLPCType lpc_type; int lpc_passes; int lpc_coeff_precision; int min_prediction_order; @@ -77,9 +77,11 @@ typedef struct FlacFrame { int bs_code[2]; uint8_t crc8; int ch_mode; + int verbatim_only; } FlacFrame; typedef struct FlacEncodeContext { + AVClass *class; PutBitContext pb; int channels; int samplerate; @@ -94,7 +96,7 @@ typedef struct FlacEncodeContext { FlacFrame frame; CompressionOptions options; AVCodecContext *avctx; - DSPContext dsp; + LPCContext lpc_ctx; struct AVMD5 *md5ctx; } FlacEncodeContext; @@ -156,16 +158,16 @@ static av_cold void dprint_compression_options(FlacEncodeContext *s) av_log(avctx, AV_LOG_DEBUG, " compression: %d\n", opt->compression_level); switch (opt->lpc_type) { - case AV_LPC_TYPE_NONE: + case FF_LPC_TYPE_NONE: av_log(avctx, AV_LOG_DEBUG, " lpc type: None\n"); break; - case AV_LPC_TYPE_FIXED: + case FF_LPC_TYPE_FIXED: av_log(avctx, AV_LOG_DEBUG, " lpc type: Fixed pre-defined coefficients\n"); break; - case AV_LPC_TYPE_LEVINSON: + case FF_LPC_TYPE_LEVINSON: av_log(avctx, AV_LOG_DEBUG, " lpc type: Levinson-Durbin recursion with Welch window\n"); break; - case AV_LPC_TYPE_CHOLESKY: + 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; @@ -211,14 +213,12 @@ 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_CHANNELS) @@ -268,43 +268,42 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) s->options.block_time_ms = ((int[]){ 27, 27, 27,105,105,105,105,105,105,105,105,105,105})[level]; - s->options.lpc_type = ((int[]){ AV_LPC_TYPE_FIXED, AV_LPC_TYPE_FIXED, AV_LPC_TYPE_FIXED, - AV_LPC_TYPE_LEVINSON, AV_LPC_TYPE_LEVINSON, AV_LPC_TYPE_LEVINSON, - AV_LPC_TYPE_LEVINSON, AV_LPC_TYPE_LEVINSON, AV_LPC_TYPE_LEVINSON, - AV_LPC_TYPE_LEVINSON, AV_LPC_TYPE_LEVINSON, AV_LPC_TYPE_LEVINSON, - AV_LPC_TYPE_LEVINSON})[level]; + 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]; - 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.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]; - 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]; + 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.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 LIBAVCODEC_VERSION_MAJOR < 53 - /* for compatibility with deprecated AVCodecContext.use_lpc */ - if (avctx->use_lpc == 0) { - s->options.lpc_type = AV_LPC_TYPE_FIXED; - } else if (avctx->use_lpc == 1) { - s->options.lpc_type = AV_LPC_TYPE_LEVINSON; - } else if (avctx->use_lpc > 1) { - s->options.lpc_type = AV_LPC_TYPE_CHOLESKY; - s->options.lpc_passes = avctx->use_lpc - 1; - } -#endif - if (avctx->lpc_type > AV_LPC_TYPE_DEFAULT) { - if (avctx->lpc_type > AV_LPC_TYPE_CHOLESKY) { +#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 == AV_LPC_TYPE_CHOLESKY) { + 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; @@ -317,11 +316,12 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) } } } +#endif - if (s->options.lpc_type == AV_LPC_TYPE_NONE) { + 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 == AV_LPC_TYPE_FIXED) { + 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); @@ -335,10 +335,10 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) } s->options.min_prediction_order = avctx->min_prediction_order; } - if (s->options.lpc_type == AV_LPC_TYPE_NONE) { + 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 == AV_LPC_TYPE_FIXED) { + 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); @@ -358,6 +358,7 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) return -1; } +#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", @@ -388,6 +389,7 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) s->options.min_partition_order, s->options.max_partition_order); return -1; } +#endif if (avctx->frame_size > 0) { if (avctx->frame_size < FLAC_MIN_BLOCKSIZE || @@ -401,6 +403,7 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) } 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) { @@ -409,10 +412,8 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) return -1; } s->options.lpc_coeff_precision = avctx->lpc_coeff_precision; - } else { - /* default LPC precision */ - s->options.lpc_coeff_precision = 15; } +#endif /* set maximum encoded frame size in verbatim mode */ s->max_framesize = ff_flac_get_max_frame_size(s->avctx->frame_size, @@ -438,9 +439,12 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) if (!avctx->coded_frame) return AVERROR(ENOMEM); + ret = ff_lpc_init(&s->lpc_ctx, avctx->frame_size, + s->options.max_prediction_order, FF_LPC_TYPE_LEVINSON); + dprint_compression_options(s); - return 0; + return ret; } @@ -472,6 +476,8 @@ static void init_frame(FlacEncodeContext *s) for (ch = 0; ch < s->channels; ch++) frame->subframes[ch].obits = 16; + + frame->verbatim_only = 0; } @@ -490,6 +496,67 @@ static void copy_samples(FlacEncodeContext *s, const int16_t *samples) } +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))) /** @@ -603,39 +670,23 @@ static int get_max_p_order(int max_porder, int n, int order) } -static uint32_t calc_rice_params_fixed(RiceContext *rc, int pmin, int pmax, - int32_t *data, int n, int pred_order, - int bps) +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 + 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; } -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) -{ - 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); - return bits; -} - - -static void encode_residual_verbatim(int32_t *res, int32_t *smp, int n) -{ - assert(n > 0); - memcpy(res, smp, n * sizeof(int32_t)); -} - - static void encode_residual_fixed(int32_t *res, const int32_t *smp, int n, int order) { @@ -791,18 +842,17 @@ static void encode_residual_lpc(int32_t *res, const int32_t *smp, int n, } -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; + frame = &s->frame; sub = &frame->subframes[ch]; res = sub->residual; smp = sub->samples; @@ -815,26 +865,24 @@ static int encode_residual(FlacEncodeContext *ctx, int ch) if (i == n) { sub->type = 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.lpc_type == AV_LPC_TYPE_NONE || - ctx->options.lpc_type == AV_LPC_TYPE_FIXED || 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; @@ -842,26 +890,24 @@ static int encode_residual(FlacEncodeContext *ctx, int ch) 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); + 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->type_code = sub->type | sub->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 = ff_lpc_calc_coefs(&ctx->dsp, smp, n, min_order, max_order, - precision, coefs, shift, ctx->options.lpc_type, - ctx->options.lpc_passes, omethod, + 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 || @@ -878,8 +924,7 @@ static int encode_residual(FlacEncodeContext *ctx, int ch) 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); + bits[i] = find_subframe_rice_params(s, sub, order+1); if (bits[i] < bits[opt_index]) { opt_index = i; opt_order = order; @@ -893,8 +938,7 @@ static int encode_residual(FlacEncodeContext *ctx, int ch) bits[0] = UINT32_MAX; for (i = min_order-1; i < max_order; i++) { 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); + bits[i] = find_subframe_rice_params(s, sub, i+1); if (bits[i] < bits[opt_order]) opt_order = i; } @@ -912,9 +956,7 @@ static int encode_residual(FlacEncodeContext *ctx, int ch) 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); + bits[i] = find_subframe_rice_params(s, sub, i+1); if (bits[i] < bits[opt_order]) opt_order = i; } @@ -923,7 +965,6 @@ static int encode_residual(FlacEncodeContext *ctx, int ch) } sub->order = opt_order; - sub->type = FLAC_SUBFRAME_LPC; sub->type_code = sub->type | (sub->order-1); sub->shift = shift[sub->order-1]; for (i = 0; i < sub->order; i++) @@ -931,38 +972,61 @@ static int encode_residual(FlacEncodeContext *ctx, int ch) 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 av_unused 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; i < n; i++) - if (smp[i] != smp[0]) - break; - if (i == n) { - sub->type = 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; } @@ -1016,18 +1080,18 @@ static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n) /** * 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; + frame = &s->frame; n = frame->blocksize; left = frame->subframes[0].samples; right = frame->subframes[1].samples; - if (ctx->channels != 2) { + if (s->channels != 2) { frame->ch_mode = FLAC_CHMODE_INDEPENDENT; return; } @@ -1064,7 +1128,7 @@ static void write_utf8(PutBitContext *pb, uint32_t val) } -static void output_frame_header(FlacEncodeContext *s) +static void write_frame_header(FlacEncodeContext *s) { FlacFrame *frame; int crc; @@ -1101,123 +1165,16 @@ static void output_frame_header(FlacEncodeContext *s) } -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) -{ - int i; - FlacFrame *frame; - FlacSubframe *sub; - int32_t res; - - frame = &s->frame; - sub = &frame->subframes[ch]; - - for (i = 0; i < frame->blocksize; i++) { - res = sub->residual[i]; - put_sbits(&s->pb, sub->obits, res); - } -} - - -static void output_residual(FlacEncodeContext *ctx, int ch) +static void write_subframes(FlacEncodeContext *s) { - 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; p < parts; p++) { - k = sub->rc.params[p]; - put_bits(&ctx->pb, 4, k); - if (p == 1) - res_cnt = psize; - for (i = 0; i < res_cnt && j < n; i++, j++) - set_sr_golomb_flac(&ctx->pb, 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; i < sub->order; 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; i < sub->order; 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; i < sub->order; 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; ch < s->channels; ch++) { - sub = &frame->subframes[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); @@ -1225,19 +1182,48 @@ 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); @@ -1248,6 +1234,16 @@ static void output_frame_footer(FlacEncodeContext *s) } +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 @@ -1265,19 +1261,12 @@ static void update_md5_sum(FlacEncodeContext *s, const int16_t *samples) static int flac_encode_frame(AVCodecContext *avctx, uint8_t *frame, int buf_size, void *data) { - int ch; FlacEncodeContext *s; const int16_t *samples = data; - int out_bytes; - int reencoded=0; + int frame_bytes, out_bytes; s = avctx->priv_data; - if (buf_size < s->max_framesize * 2) { - av_log(avctx, AV_LOG_ERROR, "output buffer too small\n"); - return 0; - } - /* when the last block is reached, update the header in extradata */ if (!data) { s->max_framesize = s->max_encoded_framesize; @@ -1286,37 +1275,35 @@ static int flac_encode_frame(AVCodecContext *avctx, uint8_t *frame, 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; ch < s->channels; ch++) - encode_residual(s, ch); + frame_bytes = encode_frame(s); -write_frame: - 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) { - if (reencoded) { - /* still too large. must be an error. */ - av_log(avctx, AV_LOG_ERROR, "error encoding frame\n"); - return -1; - } + /* 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); + } - /* frame too large. use verbatim mode */ - for (ch = 0; ch < s->channels; ch++) - encode_residual_v(s, ch); - reencoded = 1; - goto write_frame; + 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) @@ -1333,6 +1320,7 @@ 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; @@ -1340,17 +1328,44 @@ static av_cold int flac_encode_close(AVCodecContext *avctx) return 0; } +#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, {.dbl = 15 }, 0, MAX_LPC_PRECISION, FLAGS }, +{ "lpc_type", "LPC algorithm", offsetof(FlacEncodeContext, options.lpc_type), FF_OPT_TYPE_INT, {.dbl = FF_LPC_TYPE_DEFAULT }, FF_LPC_TYPE_DEFAULT, FF_LPC_TYPE_NB-1, FLAGS, "lpc_type" }, +{ "none", NULL, 0, FF_OPT_TYPE_CONST, {.dbl = FF_LPC_TYPE_NONE }, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, +{ "fixed", NULL, 0, FF_OPT_TYPE_CONST, {.dbl = FF_LPC_TYPE_FIXED }, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, +{ "levinson", NULL, 0, FF_OPT_TYPE_CONST, {.dbl = FF_LPC_TYPE_LEVINSON }, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, +{ "cholesky", NULL, 0, FF_OPT_TYPE_CONST, {.dbl = 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, {.dbl = -1 }, INT_MIN, INT_MAX, FLAGS }, +{ "min_partition_order", NULL, offsetof(FlacEncodeContext, options.min_partition_order), FF_OPT_TYPE_INT, {.dbl = -1 }, -1, MAX_PARTITION_ORDER, FLAGS }, +{ "max_partition_order", NULL, offsetof(FlacEncodeContext, options.max_partition_order), FF_OPT_TYPE_INT, {.dbl = -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, {.dbl = -1 }, -1, ORDER_METHOD_LOG, FLAGS, "predm" }, +{ "estimation", NULL, 0, FF_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_EST }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "2level", NULL, 0, FF_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_2LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "4level", NULL, 0, FF_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_4LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "8level", NULL, 0, FF_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_8LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "search", NULL, 0, FF_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_SEARCH }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "log", NULL, 0, FF_OPT_TYPE_CONST, {.dbl = 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 flac_encoder = { - "flac", - AVMEDIA_TYPE_AUDIO, - CODEC_ID_FLAC, - sizeof(FlacEncodeContext), - flac_encode_init, - flac_encode_frame, - flac_encode_close, - NULL, +AVCodec ff_flac_encoder = { + .name = "flac", + .type = AVMEDIA_TYPE_AUDIO, + .id = CODEC_ID_FLAC, + .priv_data_size = sizeof(FlacEncodeContext), + .init = flac_encode_init, + .encode = flac_encode_frame, + .close = flac_encode_close, .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY, - .sample_fmts = (const enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE}, + .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, };