X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fflacenc.c;h=58961b8dfa143248e81f1a6f8bf273b700ce19ab;hb=49623f531972be5dc2dd8c1b4b8748cad7c424ff;hp=47fcca1b40dbe26c4b4c218cd93dc2835a597950;hpb=0dc6bab092f207f237b166a543cd9ea118825272;p=ffmpeg diff --git a/libavcodec/flacenc.c b/libavcodec/flacenc.c index 47fcca1b40d..58961b8dfa1 100644 --- a/libavcodec/flacenc.c +++ b/libavcodec/flacenc.c @@ -1,4 +1,4 @@ -/** +/* * FLAC audio encoder * Copyright (c) 2006 Justin Ruggles * @@ -20,14 +20,18 @@ */ #include "libavutil/crc.h" +#include "libavutil/intmath.h" #include "libavutil/md5.h" #include "libavutil/opt.h" #include "avcodec.h" +#include "bswapdsp.h" #include "get_bits.h" #include "golomb.h" +#include "internal.h" #include "lpc.h" #include "flac.h" #include "flacdata.h" +#include "flacdsp.h" #define FLAC_SUBFRAME_CONSTANT 0 #define FLAC_SUBFRAME_VERBATIM 1 @@ -39,7 +43,11 @@ #define MAX_PARTITIONS (1 << MAX_PARTITION_ORDER) #define MAX_LPC_PRECISION 15 #define MAX_LPC_SHIFT 15 -#define MAX_RICE_PARAM 14 + +enum CodingMode { + CODING_MODE_RICE = 4, + CODING_MODE_RICE2 = 5, +}; typedef struct CompressionOptions { int compression_level; @@ -52,17 +60,21 @@ typedef struct CompressionOptions { int prediction_order_method; int min_partition_order; int max_partition_order; + int ch_mode; } CompressionOptions; typedef struct RiceContext { + enum CodingMode coding_mode; int porder; int params[MAX_PARTITIONS]; + uint32_t udata[FLAC_MAX_BLOCKSIZE]; } RiceContext; typedef struct FlacSubframe { int type; int type_code; int obits; + int wasted; int order; int32_t coefs[MAX_LPC_ORDER]; int shift; @@ -86,6 +98,7 @@ typedef struct FlacEncodeContext { int channels; int samplerate; int sr_code[2]; + int bps_code; int max_blocksize; int min_framesize; int max_framesize; @@ -98,6 +111,13 @@ typedef struct FlacEncodeContext { AVCodecContext *avctx; LPCContext lpc_ctx; struct AVMD5 *md5ctx; + uint8_t *md5_buffer; + unsigned int md5_buffer_size; + BswapDSPContext bdsp; + FLACDSPContext flac_dsp; + + int flushed; + int64_t next_pts; } FlacEncodeContext; @@ -118,7 +138,7 @@ static void write_streaminfo(FlacEncodeContext *s, uint8_t *header) 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 */ + put_bits(&pb, 5, s->avctx->bits_per_raw_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); @@ -218,8 +238,18 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) s->avctx = avctx; - if (avctx->sample_fmt != AV_SAMPLE_FMT_S16) - return -1; + switch (avctx->sample_fmt) { + case AV_SAMPLE_FMT_S16: + avctx->bits_per_raw_sample = 16; + s->bps_code = 4; + break; + case AV_SAMPLE_FMT_S32: + if (avctx->bits_per_raw_sample != 24) + av_log(avctx, AV_LOG_WARNING, "encoding as 24 bits-per-sample\n"); + avctx->bits_per_raw_sample = 24; + s->bps_code = 6; + break; + } if (channels < 1 || channels > FLAC_MAX_CHANNELS) return -1; @@ -349,10 +379,11 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) /* set maximum encoded frame size in verbatim mode */ s->max_framesize = ff_flac_get_max_frame_size(s->avctx->frame_size, - s->channels, 16); + s->channels, + s->avctx->bits_per_raw_sample); /* initialize MD5 context */ - s->md5ctx = av_malloc(av_md5_size); + s->md5ctx = av_md5_alloc(); if (!s->md5ctx) return AVERROR(ENOMEM); av_md5_init(s->md5ctx); @@ -367,20 +398,20 @@ static av_cold int flac_encode_init(AVCodecContext *avctx) s->frame_count = 0; s->min_framesize = s->max_framesize; - avctx->coded_frame = avcodec_alloc_frame(); - 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); + ff_bswapdsp_init(&s->bdsp); + ff_flacdsp_init(&s->flac_dsp, avctx->sample_fmt, + avctx->bits_per_raw_sample); + dprint_compression_options(s); return ret; } -static void init_frame(FlacEncodeContext *s) +static void init_frame(FlacEncodeContext *s, int nb_samples) { int i, ch; FlacFrame *frame; @@ -388,7 +419,7 @@ static void init_frame(FlacEncodeContext *s) frame = &s->frame; for (i = 0; i < 16; i++) { - if (s->avctx->frame_size == ff_flac_blocksize_table[i]) { + if (nb_samples == ff_flac_blocksize_table[i]) { frame->blocksize = ff_flac_blocksize_table[i]; frame->bs_code[0] = i; frame->bs_code[1] = 0; @@ -396,7 +427,7 @@ static void init_frame(FlacEncodeContext *s) } } if (i == 16) { - frame->blocksize = s->avctx->frame_size; + frame->blocksize = nb_samples; if (frame->blocksize <= 256) { frame->bs_code[0] = 6; frame->bs_code[1] = frame->blocksize-1; @@ -406,8 +437,17 @@ static void init_frame(FlacEncodeContext *s) } } - for (ch = 0; ch < s->channels; ch++) - frame->subframes[ch].obits = 16; + for (ch = 0; ch < s->channels; ch++) { + FlacSubframe *sub = &frame->subframes[ch]; + + sub->wasted = 0; + sub->obits = s->avctx->bits_per_raw_sample; + + if (sub->obits > 16) + sub->rc.coding_mode = CODING_MODE_RICE2; + else + sub->rc.coding_mode = CODING_MODE_RICE; + } frame->verbatim_only = 0; } @@ -416,22 +456,32 @@ static void init_frame(FlacEncodeContext *s) /** * Copy channel-interleaved input samples into separate subframes. */ -static void copy_samples(FlacEncodeContext *s, const int16_t *samples) +static void copy_samples(FlacEncodeContext *s, const void *samples) { int i, j, ch; FlacFrame *frame; - - frame = &s->frame; - for (i = 0, j = 0; i < frame->blocksize; i++) - for (ch = 0; ch < s->channels; ch++, j++) - frame->subframes[ch].samples[i] = samples[j]; + int shift = av_get_bytes_per_sample(s->avctx->sample_fmt) * 8 - + s->avctx->bits_per_raw_sample; + +#define COPY_SAMPLES(bits) do { \ + const int ## bits ## _t *samples0 = samples; \ + frame = &s->frame; \ + for (i = 0, j = 0; i < frame->blocksize; i++) \ + for (ch = 0; ch < s->channels; ch++, j++) \ + frame->subframes[ch].samples[i] = samples0[j] >> shift; \ +} while (0) + + if (s->avctx->sample_fmt == AV_SAMPLE_FMT_S16) + COPY_SAMPLES(16); + else + COPY_SAMPLES(32); } -static int rice_count_exact(int32_t *res, int n, int k) +static uint64_t rice_count_exact(int32_t *res, int n, int k) { int i; - int count = 0; + uint64_t count = 0; for (i = 0; i < n; i++) { int32_t v = -2 * res[i] - 1; @@ -442,12 +492,12 @@ static int rice_count_exact(int32_t *res, int n, int k) } -static int subframe_count_exact(FlacEncodeContext *s, FlacSubframe *sub, - int pred_order) +static uint64_t subframe_count_exact(FlacEncodeContext *s, FlacSubframe *sub, + int pred_order) { int p, porder, psize; int i, part_end; - int count = 0; + uint64_t count = 0; /* subframe header */ count += 8; @@ -478,7 +528,7 @@ static int subframe_count_exact(FlacEncodeContext *s, FlacSubframe *sub, part_end = psize; for (p = 0; p < 1 << porder; p++) { int k = sub->rc.params[p]; - count += 4; + count += sub->rc.coding_mode; count += rice_count_exact(&sub->residual[i], part_end - i, k); i = part_end; part_end = FFMIN(s->frame.blocksize, part_end + psize); @@ -494,32 +544,34 @@ static int subframe_count_exact(FlacEncodeContext *s, FlacSubframe *sub, /** * Solve for d/dk(rice_encode_count) = n-((sum-(n>>1))>>(k+1)) = 0. */ -static int find_optimal_param(uint32_t sum, int n) +static int find_optimal_param(uint64_t sum, int n, int max_param) { int k; - uint32_t sum2; + uint64_t sum2; if (sum <= n >> 1) return 0; sum2 = sum - (n >> 1); - k = av_log2(n < 256 ? FASTDIV(sum2, n) : sum2 / n); - return FFMIN(k, MAX_RICE_PARAM); + k = av_log2(av_clipl_int32(sum2 / n)); + return FFMIN(k, max_param); } -static uint32_t calc_optimal_rice_params(RiceContext *rc, int porder, - uint32_t *sums, int n, int pred_order) +static uint64_t calc_optimal_rice_params(RiceContext *rc, int porder, + uint64_t *sums, int n, int pred_order) { int i; - int k, cnt, part; - uint32_t all_bits; + int k, cnt, part, max_param; + uint64_t all_bits; + + max_param = (1 << rc->coding_mode) - 2; part = (1 << porder); all_bits = 4 * part; cnt = (n >> porder) - pred_order; for (i = 0; i < part; i++) { - k = find_optimal_param(sums[i], cnt); + k = find_optimal_param(sums[i], cnt, max_param); rc->params[i] = k; all_bits += rice_encode_count(sums[i], cnt, k); cnt = n >> porder; @@ -532,7 +584,7 @@ static uint32_t calc_optimal_rice_params(RiceContext *rc, int porder, static void calc_sums(int pmin, int pmax, uint32_t *data, int n, int pred_order, - uint32_t sums[][MAX_PARTITIONS]) + uint64_t sums[][MAX_PARTITIONS]) { int i, j; int parts; @@ -543,7 +595,7 @@ static void calc_sums(int pmin, int pmax, uint32_t *data, int n, int pred_order, res = &data[pred_order]; res_end = &data[n >> pmax]; for (i = 0; i < parts; i++) { - uint32_t sum = 0; + uint64_t sum = 0; while (res < res_end) sum += *(res++); sums[pmax][i] = sum; @@ -558,25 +610,25 @@ static void calc_sums(int pmin, int pmax, uint32_t *data, int n, int pred_order, } -static uint32_t calc_rice_params(RiceContext *rc, int pmin, int pmax, +static uint64_t calc_rice_params(RiceContext *rc, int pmin, int pmax, int32_t *data, int n, int pred_order) { int i; - uint32_t bits[MAX_PARTITION_ORDER+1]; + uint64_t bits[MAX_PARTITION_ORDER+1]; int opt_porder; RiceContext tmp_rc; - uint32_t *udata; - uint32_t sums[MAX_PARTITION_ORDER+1][MAX_PARTITIONS]; + uint64_t sums[MAX_PARTITION_ORDER + 1][MAX_PARTITIONS] = { { 0 } }; assert(pmin >= 0 && pmin <= MAX_PARTITION_ORDER); assert(pmax >= 0 && pmax <= MAX_PARTITION_ORDER); assert(pmin <= pmax); - udata = av_malloc(n * sizeof(uint32_t)); + tmp_rc.coding_mode = rc->coding_mode; + for (i = 0; i < n; i++) - udata[i] = (2*data[i]) ^ (data[i]>>31); + rc->udata[i] = (2 * data[i]) ^ (data[i] >> 31); - calc_sums(pmin, pmax, udata, n, pred_order, sums); + calc_sums(pmin, pmax, rc->udata, n, pred_order, sums); opt_porder = pmin; bits[pmin] = UINT32_MAX; @@ -588,7 +640,6 @@ static uint32_t calc_rice_params(RiceContext *rc, int pmin, int pmax, } } - av_freep(&udata); return bits[opt_porder]; } @@ -602,7 +653,7 @@ static int get_max_p_order(int max_porder, int n, int order) } -static uint32_t find_subframe_rice_params(FlacEncodeContext *s, +static uint64_t find_subframe_rice_params(FlacEncodeContext *s, FlacSubframe *sub, int pred_order) { int pmin = get_max_p_order(s->options.min_partition_order, @@ -610,7 +661,7 @@ static uint32_t find_subframe_rice_params(FlacEncodeContext *s, 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; + uint64_t bits = 8 + pred_order * sub->obits + 2 + sub->rc.coding_mode; 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, @@ -670,110 +721,6 @@ static void encode_residual_fixed(int32_t *res, const int32_t *smp, int n, } -#define LPC1(x) {\ - int c = coefs[(x)-1];\ - p0 += c * s;\ - s = smp[i-(x)+1];\ - p1 += c * s;\ -} - -static av_always_inline void encode_residual_lpc_unrolled(int32_t *res, - const int32_t *smp, int n, int order, - const int32_t *coefs, int shift, int big) -{ - int i; - for (i = order; i < n; i += 2) { - int s = smp[i-order]; - int p0 = 0, p1 = 0; - if (big) { - switch (order) { - case 32: LPC1(32) - case 31: LPC1(31) - case 30: LPC1(30) - case 29: LPC1(29) - case 28: LPC1(28) - case 27: LPC1(27) - case 26: LPC1(26) - case 25: LPC1(25) - case 24: LPC1(24) - case 23: LPC1(23) - case 22: LPC1(22) - case 21: LPC1(21) - case 20: LPC1(20) - case 19: LPC1(19) - case 18: LPC1(18) - case 17: LPC1(17) - case 16: LPC1(16) - case 15: LPC1(15) - case 14: LPC1(14) - case 13: LPC1(13) - case 12: LPC1(12) - case 11: LPC1(11) - case 10: LPC1(10) - case 9: LPC1( 9) - LPC1( 8) - LPC1( 7) - LPC1( 6) - LPC1( 5) - LPC1( 4) - LPC1( 3) - LPC1( 2) - LPC1( 1) - } - } else { - switch (order) { - case 8: LPC1( 8) - case 7: LPC1( 7) - case 6: LPC1( 6) - case 5: LPC1( 5) - case 4: LPC1( 4) - case 3: LPC1( 3) - case 2: LPC1( 2) - case 1: LPC1( 1) - } - } - res[i ] = smp[i ] - (p0 >> 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 < order; i++) - res[i] = smp[i]; -#if CONFIG_SMALL - for (i = order; i < n; i += 2) { - int j; - int s = smp[i]; - int p0 = 0, p1 = 0; - for (j = 0; j < order; j++) { - int c = coefs[j]; - p1 += c * s; - s = smp[i-j-1]; - p0 += c * s; - } - 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; - } -#endif -} - - static int encode_residual_ch(FlacEncodeContext *s, int ch) { int i, n; @@ -815,7 +762,7 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch) 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]; + uint64_t bits[MAX_FIXED_ORDER+1]; if (max_order > MAX_FIXED_ORDER) max_order = MAX_FIXED_ORDER; opt_order = 0; @@ -846,16 +793,19 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch) omethod == ORDER_METHOD_4LEVEL || omethod == ORDER_METHOD_8LEVEL) { int levels = 1 << omethod; - uint32_t bits[1 << ORDER_METHOD_8LEVEL]; - int order; + uint64_t bits[1 << ORDER_METHOD_8LEVEL]; + int order = -1; int opt_index = levels-1; opt_order = max_order-1; bits[opt_index] = UINT32_MAX; for (i = levels-1; i >= 0; i--) { + int last_order = order; order = min_order + (((max_order-min_order+1) * (i+1)) / levels)-1; - if (order < 0) - order = 0; - encode_residual_lpc(res, smp, n, order+1, coefs[order], shift[order]); + order = av_clip(order, min_order - 1, max_order - 1); + if (order == last_order) + continue; + s->flac_dsp.lpc_encode(res, smp, n, order+1, coefs[order], + shift[order]); bits[i] = find_subframe_rice_params(s, sub, order+1); if (bits[i] < bits[opt_index]) { opt_index = i; @@ -865,18 +815,18 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch) opt_order++; } else if (omethod == ORDER_METHOD_SEARCH) { // brute-force optimal order search - uint32_t bits[MAX_LPC_ORDER]; + uint64_t bits[MAX_LPC_ORDER]; opt_order = 0; 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]); + s->flac_dsp.lpc_encode(res, smp, n, i+1, coefs[i], shift[i]); 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) { - uint32_t bits[MAX_LPC_ORDER]; + uint64_t bits[MAX_LPC_ORDER]; int step; opt_order = min_order - 1 + (max_order-min_order)/3; @@ -887,7 +837,7 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch) 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]); + s->flac_dsp.lpc_encode(res, smp, n, i+1, coefs[i], shift[i]); bits[i] = find_subframe_rice_params(s, sub, i+1); if (bits[i] < bits[opt_order]) opt_order = i; @@ -902,7 +852,7 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch) 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); + s->flac_dsp.lpc_encode(res, smp, n, sub->order, sub->coefs, sub->shift); find_subframe_rice_params(s, sub, sub->order); @@ -948,7 +898,8 @@ static int count_frame_header(FlacEncodeContext *s) static int encode_frame(FlacEncodeContext *s) { - int ch, count; + int ch; + uint64_t count; count = count_frame_header(s); @@ -958,11 +909,47 @@ static int encode_frame(FlacEncodeContext *s) count += (8 - (count & 7)) & 7; // byte alignment count += 16; // CRC-16 - return count >> 3; + count >>= 3; + if (count > INT_MAX) + return AVERROR_BUG; + return count; } -static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n) +static void remove_wasted_bits(FlacEncodeContext *s) +{ + int ch, i; + + for (ch = 0; ch < s->channels; ch++) { + FlacSubframe *sub = &s->frame.subframes[ch]; + int32_t v = 0; + + for (i = 0; i < s->frame.blocksize; i++) { + v |= sub->samples[i]; + if (v & 1) + break; + } + + if (v && !(v & 1)) { + v = av_ctz(v); + + for (i = 0; i < s->frame.blocksize; i++) + sub->samples[i] >>= v; + + sub->wasted = v; + sub->obits -= v; + + /* for 24-bit, check if removing wasted bits makes the range better + suited for using RICE instead of RICE2 for entropy coding */ + if (sub->obits <= 17) + sub->rc.coding_mode = CODING_MODE_RICE; + } + } +} + + +static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n, + int max_rice_param) { int i, best; int32_t lt, rt; @@ -982,7 +969,7 @@ static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n) } /* estimate bit counts */ for (i = 0; i < 4; i++) { - k = find_optimal_param(2 * sum[i], n); + k = find_optimal_param(2 * sum[i], n, max_rice_param); sum[i] = rice_encode_count( 2 * sum[i], n, k); } @@ -997,15 +984,8 @@ static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n) for (i = 1; i < 4; i++) if (score[i] < score[best]) best = i; - if (best == 0) { - return FLAC_CHMODE_INDEPENDENT; - } else if (best == 1) { - return FLAC_CHMODE_LEFT_SIDE; - } else if (best == 2) { - return FLAC_CHMODE_RIGHT_SIDE; - } else { - return FLAC_CHMODE_MID_SIDE; - } + + return best; } @@ -1028,7 +1008,11 @@ static void channel_decorrelation(FlacEncodeContext *s) return; } - frame->ch_mode = estimate_stereo_mode(left, right, n); + if (s->options.ch_mode < 0) { + int max_rice_param = (1 << frame->subframes[0].rc.coding_mode) - 2; + frame->ch_mode = estimate_stereo_mode(left, right, n, max_rice_param); + } else + frame->ch_mode = s->options.ch_mode; /* perform decorrelation and adjust bits-per-sample */ if (frame->ch_mode == FLAC_CHMODE_INDEPENDENT) @@ -1074,9 +1058,9 @@ static void write_frame_header(FlacEncodeContext *s) 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, 4, frame->ch_mode + FLAC_MAX_CHANNELS - 1); - put_bits(&s->pb, 3, 4); /* bits-per-sample code */ + put_bits(&s->pb, 3, s->bps_code); put_bits(&s->pb, 1, 0); write_utf8(&s->pb, s->frame_count); @@ -1111,7 +1095,9 @@ static void write_subframes(FlacEncodeContext *s) /* subframe header */ put_bits(&s->pb, 1, 0); put_bits(&s->pb, 6, sub->type_code); - put_bits(&s->pb, 1, 0); /* no wasted bits */ + put_bits(&s->pb, 1, !!sub->wasted); + if (sub->wasted) + put_bits(&s->pb, sub->wasted, 1); /* subframe */ if (sub->type == FLAC_SUBFRAME_CONSTANT) { @@ -1134,7 +1120,7 @@ static void write_subframes(FlacEncodeContext *s) } /* rice-encoded block */ - put_bits(&s->pb, 2, 0); + put_bits(&s->pb, 2, sub->rc.coding_mode - 4); /* partition order */ porder = sub->rc.porder; @@ -1145,7 +1131,7 @@ static void write_subframes(FlacEncodeContext *s) part_end = &sub->residual[psize]; for (p = 0; p < 1 << porder; p++) { int k = sub->rc.params[p]; - put_bits(&s->pb, 4, k); + put_bits(&s->pb, sub->rc.coding_mode, k); while (res < part_end) set_sr_golomb_flac(&s->pb, *res++, k, INT32_MAX, 0); part_end = FFMIN(frame_end, part_end + psize); @@ -1166,9 +1152,9 @@ static void write_frame_footer(FlacEncodeContext *s) } -static int write_frame(FlacEncodeContext *s, uint8_t *frame, int buf_size) +static int write_frame(FlacEncodeContext *s, AVPacket *avpkt) { - init_put_bits(&s->pb, frame, buf_size); + init_put_bits(&s->pb, avpkt->data, avpkt->size); write_frame_header(s); write_subframes(s); write_frame_footer(s); @@ -1176,74 +1162,134 @@ static int write_frame(FlacEncodeContext *s, uint8_t *frame, int buf_size) } -static void update_md5_sum(FlacEncodeContext *s, const int16_t *samples) +static int update_md5_sum(FlacEncodeContext *s, const void *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); + const uint8_t *buf; + int buf_size = s->frame.blocksize * s->channels * + ((s->avctx->bits_per_raw_sample + 7) / 8); + + if (s->avctx->bits_per_raw_sample > 16 || HAVE_BIGENDIAN) { + av_fast_malloc(&s->md5_buffer, &s->md5_buffer_size, buf_size); + if (!s->md5_buffer) + return AVERROR(ENOMEM); } -#else - av_md5_update(s->md5ctx, (const uint8_t *)samples, s->frame.blocksize*s->channels*2); + + if (s->avctx->bits_per_raw_sample <= 16) { + buf = (const uint8_t *)samples; +#if HAVE_BIGENDIAN + s->bdsp.bswap16_buf((uint16_t *) s->md5_buffer, + (const uint16_t *) samples, buf_size / 2); + buf = s->md5_buffer; #endif + } else { + int i; + const int32_t *samples0 = samples; + uint8_t *tmp = s->md5_buffer; + + for (i = 0; i < s->frame.blocksize * s->channels; i++) { + int32_t v = samples0[i] >> 8; + *tmp++ = (v ) & 0xFF; + *tmp++ = (v >> 8) & 0xFF; + *tmp++ = (v >> 16) & 0xFF; + } + buf = s->md5_buffer; + } + av_md5_update(s->md5ctx, buf, buf_size); + + return 0; } -static int flac_encode_frame(AVCodecContext *avctx, uint8_t *frame, - int buf_size, void *data) +static int flac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, + const AVFrame *frame, int *got_packet_ptr) { FlacEncodeContext *s; - const int16_t *samples = data; - int frame_bytes, out_bytes; + int frame_bytes, out_bytes, ret; s = avctx->priv_data; /* when the last block is reached, update the header in extradata */ - if (!data) { + if (!frame) { s->max_framesize = s->max_encoded_framesize; av_md5_final(s->md5ctx, s->md5sum); write_streaminfo(s, avctx->extradata); + +#if FF_API_SIDEDATA_ONLY_PKT +FF_DISABLE_DEPRECATION_WARNINGS + if (avctx->side_data_only_packets && !s->flushed) { +FF_ENABLE_DEPRECATION_WARNINGS +#else + if (!s->flushed) { +#endif + uint8_t *side_data = av_packet_new_side_data(avpkt, AV_PKT_DATA_NEW_EXTRADATA, + avctx->extradata_size); + if (!side_data) + return AVERROR(ENOMEM); + memcpy(side_data, avctx->extradata, avctx->extradata_size); + + avpkt->pts = s->next_pts; + + *got_packet_ptr = 1; + s->flushed = 1; + } + 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); + if (frame->nb_samples < s->frame.blocksize) { + s->max_framesize = ff_flac_get_max_frame_size(frame->nb_samples, + s->channels, + avctx->bits_per_raw_sample); } - init_frame(s); + init_frame(s, frame->nb_samples); - copy_samples(s, samples); + copy_samples(s, frame->data[0]); channel_decorrelation(s); + remove_wasted_bits(s); + frame_bytes = encode_frame(s); - /* fallback to verbatim mode if the compressed frame is larger than it + /* Fall back on verbatim mode if the compressed frame is larger than it would be if encoded uncompressed. */ - if (frame_bytes > s->max_framesize) { + if (frame_bytes < 0 || frame_bytes > s->max_framesize) { s->frame.verbatim_only = 1; frame_bytes = encode_frame(s); + if (frame_bytes < 0) { + av_log(avctx, AV_LOG_ERROR, "Bad frame count\n"); + return frame_bytes; + } } - if (buf_size < frame_bytes) { - av_log(avctx, AV_LOG_ERROR, "output buffer too small\n"); - return 0; + if ((ret = ff_alloc_packet(avpkt, frame_bytes))) { + av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n"); + return ret; } - out_bytes = write_frame(s, frame, buf_size); + + out_bytes = write_frame(s, avpkt); s->frame_count++; - avctx->coded_frame->pts = s->sample_count; - s->sample_count += avctx->frame_size; - update_md5_sum(s, samples); + s->sample_count += frame->nb_samples; + if ((ret = update_md5_sum(s, frame->data[0])) < 0) { + av_log(avctx, AV_LOG_ERROR, "Error updating MD5 checksum\n"); + return ret; + } 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; + avpkt->pts = frame->pts; + avpkt->duration = ff_samples_to_time_base(avctx, frame->nb_samples); + avpkt->size = out_bytes; + + s->next_pts = avpkt->pts + avpkt->duration; + + *got_packet_ptr = 1; + return 0; } @@ -1252,32 +1298,38 @@ static av_cold int flac_encode_close(AVCodecContext *avctx) if (avctx->priv_data) { FlacEncodeContext *s = avctx->priv_data; av_freep(&s->md5ctx); + av_freep(&s->md5_buffer); ff_lpc_end(&s->lpc_ctx); } av_freep(&avctx->extradata); avctx->extradata_size = 0; - av_freep(&avctx->coded_frame); 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), AV_OPT_TYPE_INT, {.dbl = 15 }, 0, MAX_LPC_PRECISION, FLAGS }, -{ "lpc_type", "LPC algorithm", offsetof(FlacEncodeContext, options.lpc_type), AV_OPT_TYPE_INT, {.dbl = FF_LPC_TYPE_DEFAULT }, FF_LPC_TYPE_DEFAULT, FF_LPC_TYPE_NB-1, FLAGS, "lpc_type" }, -{ "none", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = FF_LPC_TYPE_NONE }, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, -{ "fixed", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = FF_LPC_TYPE_FIXED }, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, -{ "levinson", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = FF_LPC_TYPE_LEVINSON }, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, -{ "cholesky", NULL, 0, AV_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), AV_OPT_TYPE_INT, {.dbl = -1 }, INT_MIN, INT_MAX, FLAGS }, -{ "min_partition_order", NULL, offsetof(FlacEncodeContext, options.min_partition_order), AV_OPT_TYPE_INT, {.dbl = -1 }, -1, MAX_PARTITION_ORDER, FLAGS }, -{ "max_partition_order", NULL, offsetof(FlacEncodeContext, options.max_partition_order), AV_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), AV_OPT_TYPE_INT, {.dbl = -1 }, -1, ORDER_METHOD_LOG, FLAGS, "predm" }, -{ "estimation", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_EST }, INT_MIN, INT_MAX, FLAGS, "predm" }, -{ "2level", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_2LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" }, -{ "4level", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_4LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" }, -{ "8level", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_8LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" }, -{ "search", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_SEARCH }, INT_MIN, INT_MAX, FLAGS, "predm" }, -{ "log", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = ORDER_METHOD_LOG }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "lpc_coeff_precision", "LPC coefficient precision", offsetof(FlacEncodeContext, options.lpc_coeff_precision), AV_OPT_TYPE_INT, {.i64 = 15 }, 0, MAX_LPC_PRECISION, FLAGS }, +{ "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" }, +{ "none", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_LPC_TYPE_NONE }, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, +{ "fixed", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_LPC_TYPE_FIXED }, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, +{ "levinson", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_LPC_TYPE_LEVINSON }, INT_MIN, INT_MAX, FLAGS, "lpc_type" }, +{ "cholesky", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 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), AV_OPT_TYPE_INT, {.i64 = 1 }, 1, INT_MAX, FLAGS }, +{ "min_partition_order", NULL, offsetof(FlacEncodeContext, options.min_partition_order), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, MAX_PARTITION_ORDER, FLAGS }, +{ "max_partition_order", NULL, offsetof(FlacEncodeContext, options.max_partition_order), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, MAX_PARTITION_ORDER, FLAGS }, +{ "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" }, +{ "estimation", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_EST }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "2level", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_2LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "4level", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_4LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "8level", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_8LEVEL }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "search", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_SEARCH }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "log", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = ORDER_METHOD_LOG }, INT_MIN, INT_MAX, FLAGS, "predm" }, +{ "ch_mode", "Stereo decorrelation mode", offsetof(FlacEncodeContext, options.ch_mode), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, FLAC_CHMODE_MID_SIDE, FLAGS, "ch_mode" }, +{ "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = -1 }, INT_MIN, INT_MAX, FLAGS, "ch_mode" }, +{ "indep", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FLAC_CHMODE_INDEPENDENT }, INT_MIN, INT_MAX, FLAGS, "ch_mode" }, +{ "left_side", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FLAC_CHMODE_LEFT_SIDE }, INT_MIN, INT_MAX, FLAGS, "ch_mode" }, +{ "right_side", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FLAC_CHMODE_RIGHT_SIDE }, INT_MIN, INT_MAX, FLAGS, "ch_mode" }, +{ "mid_side", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FLAC_CHMODE_MID_SIDE }, INT_MIN, INT_MAX, FLAGS, "ch_mode" }, { NULL }, }; @@ -1290,14 +1342,16 @@ static const AVClass flac_encoder_class = { AVCodec ff_flac_encoder = { .name = "flac", + .long_name = NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"), .type = AVMEDIA_TYPE_AUDIO, - .id = CODEC_ID_FLAC, + .id = AV_CODEC_ID_FLAC, .priv_data_size = sizeof(FlacEncodeContext), .init = flac_encode_init, - .encode = flac_encode_frame, + .encode2 = flac_encode_frame, .close = flac_encode_close, - .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, + .capabilities = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY, + .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16, + AV_SAMPLE_FMT_S32, + AV_SAMPLE_FMT_NONE }, + .priv_class = &flac_encoder_class, };