From be234dc464159da366ec770fe8e5a958f317f7a9 Mon Sep 17 00:00:00 2001 From: nu774 Date: Sun, 27 Oct 2013 21:31:35 +0900 Subject: [PATCH 01/16] add --include-sbr-delay --- src/main.c | 19 +++++++++++++++++-- 1 file changed, 17 insertions(+), 2 deletions(-) diff --git a/src/main.c b/src/main.c index ec0b14d..5fe46ad 100644 --- a/src/main.c +++ b/src/main.c @@ -149,6 +149,9 @@ PROGNAME " %s\n" " 0: iTunSMPB (default)\n" " 1: ISO standard (edts + sgpd)\n" " 2: Both\n" +" --include-sbr-delay Count SBR decoder delay in encoder delay\n" +" This is not iTunes compatible, but is default\n" +" behavior of FDK library.\n" " -I, --ignorelength Ignore length of WAV header\n" " -S, --silent Don't print progress messages\n" " --moov-before-mdat Place moov box before mdat box on m4a output\n" @@ -204,6 +207,7 @@ typedef struct aacenc_param_ex_t { char *output_filename; FILE *output_fp; unsigned gapless_mode; + unsigned include_sbr_delay; unsigned ignore_length; int silent; int moov_before_mdat; @@ -226,6 +230,7 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params) int ch; unsigned n; +#define OPT_INCLUDE_SBR_DELAY M4AF_FOURCC('s','d','l','y') #define OPT_MOOV_BEFORE_MDAT M4AF_FOURCC('m','o','o','v') #define OPT_RAW_CHANNELS M4AF_FOURCC('r','c','h','n') #define OPT_RAW_RATE M4AF_FOURCC('r','r','a','t') @@ -248,6 +253,7 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params) { "header-period", required_argument, 0, 'P' }, { "gapless-mode", required_argument, 0, 'G' }, + { "include-sbr-delay", no_argument, 0, OPT_INCLUDE_SBR_DELAY }, { "ignorelength", no_argument, 0, 'I' }, { "silent", no_argument, 0, 'S' }, { "moov-before-mdat", no_argument, 0, OPT_MOOV_BEFORE_MDAT }, @@ -348,6 +354,9 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params) } params->gapless_mode = n; break; + case OPT_INCLUDE_SBR_DELAY: + params->include_sbr_delay = 1; + break; case 'I': params->ignore_length = 1; break; @@ -791,9 +800,15 @@ int main(int argc, char **argv) goto END; if (m4af) { uint32_t delay = aacinfo.encoderDelay; + uint32_t padding; int64_t frames_read = pcm_get_position(reader); - uint32_t padding = frame_count * aacinfo.frameLength - - frames_read - aacinfo.encoderDelay; + + if (sbr_mode && params.profile != AOT_ER_AAC_ELD && + !params.include_sbr_delay) + delay -= 481 << 1; + if (sbr_mode && (delay & 1)) + ++delay; + padding = frame_count * aacinfo.frameLength - frames_read - delay; m4af_set_priming(m4af, 0, delay >> downsampled_timescale, padding >> downsampled_timescale); if (finalize_m4a(m4af, ¶ms, encoder) < 0) -- 2.39.2 From b159a7b0957cad750f601ba14ac5c9d5a88c6327 Mon Sep 17 00:00:00 2001 From: nu774 Date: Sun, 27 Oct 2013 22:43:18 +0900 Subject: [PATCH 02/16] bump version --- version.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/version.h b/version.h index 6c59fa3..b4ebd10 100644 --- a/version.h +++ b/version.h @@ -1,4 +1,4 @@ #ifndef VERSION_H #define VERSION_H -const char *fdkaac_version = "0.4.0"; +const char *fdkaac_version = "0.4.1"; #endif -- 2.39.2 From 3aa2787e343c84c2404aecb71a3123784e8fb85d Mon Sep 17 00:00:00 2001 From: nu774 Date: Tue, 29 Oct 2013 19:21:34 +0900 Subject: [PATCH 03/16] fix pcm_seek() to inline --- src/pcm_reader.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/pcm_reader.h b/src/pcm_reader.h index b9490a4..ee062aa 100644 --- a/src/pcm_reader.h +++ b/src/pcm_reader.h @@ -85,7 +85,7 @@ uint32_t bitcount(uint32_t bits) int pcm_read(pcm_io_context_t *io, void *buffer, uint32_t size); int pcm_skip(pcm_io_context_t *io, int64_t count); -static int pcm_seek(pcm_io_context_t *io, int64_t off, int whence) +static inline int pcm_seek(pcm_io_context_t *io, int64_t off, int whence) { return io->vtbl->seek ? io->vtbl->seek(io->cookie, off, whence) : -1; } -- 2.39.2 From 7259767b094c543f41bfab045b0cceb715879e4a Mon Sep 17 00:00:00 2001 From: nu774 Date: Tue, 29 Oct 2013 19:24:03 +0900 Subject: [PATCH 04/16] fix vcxproj --- MSVC/fdkaac.vcxproj | 2 ++ 1 file changed, 2 insertions(+) diff --git a/MSVC/fdkaac.vcxproj b/MSVC/fdkaac.vcxproj index a07b91e..03591a9 100644 --- a/MSVC/fdkaac.vcxproj +++ b/MSVC/fdkaac.vcxproj @@ -111,6 +111,8 @@ copy ..\fdk-aac\libSYS\include\machine_type.h include\fdk-aac\ + + -- 2.39.2 From 6709cf694c2bdab475488397196520d9cc274dd8 Mon Sep 17 00:00:00 2001 From: nu774 Date: Tue, 29 Oct 2013 19:32:54 +0900 Subject: [PATCH 05/16] fix warning: cast size_t as sprintf() arg to int --- src/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/main.c b/src/main.c index 5fe46ad..0d07690 100644 --- a/src/main.c +++ b/src/main.c @@ -607,7 +607,7 @@ char *generate_output_filename(const char *filename, const char *ext) const char *ext_org = strrchr(base, '.'); if (ext_org) ilen = ext_org - base; p = malloc(ilen + ext_len + 1); - sprintf(p, "%.*s%s", ilen, base, ext); + sprintf(p, "%.*s%s", (int)ilen, base, ext); } return p; } -- 2.39.2 From d11b0441317d05811b30490b778d1051bafb2570 Mon Sep 17 00:00:00 2001 From: nu774 Date: Tue, 29 Oct 2013 19:35:57 +0900 Subject: [PATCH 06/16] fix unused variable warning --- src/m4af.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/src/m4af.c b/src/m4af.c index 261ce71..bdcae5f 100644 --- a/src/m4af.c +++ b/src/m4af.c @@ -864,7 +864,6 @@ void m4af_write_sbgp_box(m4af_ctx_t *ctx, uint32_t track_idx) static void m4af_write_sgpd_box(m4af_ctx_t *ctx, uint32_t track_idx) { - m4af_track_t *track = &ctx->track[track_idx]; m4af_write(ctx, "\0\0\0\026" /* size: 22 */ "sgpd" /* type */ @@ -1050,7 +1049,6 @@ void m4af_write_elst_box(m4af_ctx_t *ctx, uint32_t track_idx) static void m4af_write_edts_box(m4af_ctx_t *ctx, uint32_t track_idx) { - m4af_track_t *track = &ctx->track[track_idx]; int64_t pos = m4af_tell(ctx); m4af_write(ctx, "\0\0\0\0edts", 8); m4af_write_elst_box(ctx, track_idx); -- 2.39.2 From 4d48b091d49818772a47559ba7fd7ab58fdd7682 Mon Sep 17 00:00:00 2001 From: nu774 Date: Tue, 29 Oct 2013 17:34:28 +0900 Subject: [PATCH 07/16] smart padding for better gapless playback Taken smart padding code using LPC extrapolation from vorbis/opus. Padding is done on both beginning and ending, but enc_delay and padding remains the same (we discard extra padding frame introduced on our side after encoding). --- MSVC/fdkaac.vcxproj | 3 + MSVC/fdkaac.vcxproj.filters | 9 ++ Makefile.am | 2 + src/aacenc.c | 2 +- src/extrapolater.c | 204 ++++++++++++++++++++++++++++++++++++ src/lpc.c | 169 +++++++++++++++++++++++++++++ src/lpc.h | 27 +++++ src/lpcm.c | 29 ----- src/lpcm.h | 30 ++++++ src/main.c | 83 ++++++++++----- src/pcm_reader.h | 2 + 11 files changed, 505 insertions(+), 55 deletions(-) create mode 100644 src/extrapolater.c create mode 100644 src/lpc.c create mode 100644 src/lpc.h diff --git a/MSVC/fdkaac.vcxproj b/MSVC/fdkaac.vcxproj index 03591a9..fc9a7ef 100644 --- a/MSVC/fdkaac.vcxproj +++ b/MSVC/fdkaac.vcxproj @@ -98,6 +98,8 @@ copy ..\fdk-aac\libSYS\include\machine_type.h include\fdk-aac\ + + @@ -114,6 +116,7 @@ copy ..\fdk-aac\libSYS\include\machine_type.h include\fdk-aac\ + diff --git a/MSVC/fdkaac.vcxproj.filters b/MSVC/fdkaac.vcxproj.filters index feae3ab..3ad1065 100644 --- a/MSVC/fdkaac.vcxproj.filters +++ b/MSVC/fdkaac.vcxproj.filters @@ -24,6 +24,12 @@ Source Files + + Source Files + + + Source Files + Source Files @@ -62,6 +68,9 @@ Header Files + + Header Files + Header Files diff --git a/Makefile.am b/Makefile.am index 94a1c7c..f410b67 100644 --- a/Makefile.am +++ b/Makefile.am @@ -6,6 +6,8 @@ bin_PROGRAMS = fdkaac fdkaac_SOURCES = \ src/aacenc.c \ src/caf_reader.c \ + src/extrapolater.c \ + src/lpc.c \ src/lpcm.c \ src/m4af.c \ src/main.c \ diff --git a/src/aacenc.c b/src/aacenc.c index d25be8a..bf975d0 100644 --- a/src/aacenc.c +++ b/src/aacenc.c @@ -231,5 +231,5 @@ int aac_encode_frame(HANDLE_AACENCODER encoder, return -1; } *olen = oargs.numOutBytes; - return oargs.numInSamples; + return oargs.numInSamples / format->channels_per_frame; } diff --git a/src/extrapolater.c b/src/extrapolater.c new file mode 100644 index 0000000..22b0f12 --- /dev/null +++ b/src/extrapolater.c @@ -0,0 +1,204 @@ +#if HAVE_CONFIG_H +# include "config.h" +#endif +#if HAVE_STDINT_H +# include +#endif +#include +#include +#include +#include "pcm_reader.h" +#include "lpc.h" + +typedef int16_t sample_t; + +typedef struct buffer_t { + sample_t *data; + unsigned count; /* count in frames */ + unsigned capacity; /* size in bytes */ +} buffer_t; + +typedef struct extrapolater_t { + pcm_reader_vtbl_t *vtbl; + pcm_reader_t *src; + pcm_sample_description_t format; + buffer_t buffer[2]; + unsigned nbuffer; + int (*process)(struct extrapolater_t *, void *, unsigned); +} extrapolater_t; + +#define LPC_ORDER 32 + +static inline pcm_reader_t *get_source(pcm_reader_t *reader) +{ + return ((extrapolater_t *)reader)->src; +} + +static const +pcm_sample_description_t *get_format(pcm_reader_t *reader) +{ + return pcm_get_format(get_source(reader)); +} + +static int64_t get_length(pcm_reader_t *reader) +{ + return pcm_get_length(get_source(reader)); +} + +static int64_t get_position(pcm_reader_t *reader) +{ + return pcm_get_position(get_source(reader)); +} + +static int realloc_buffer(buffer_t *bp, size_t size) +{ + if (bp->capacity < size) { + void *p = realloc(bp->data, size); + if (!p) return -1; + bp->data = p; + bp->capacity = size; + } + return 0; +} + +static void reverse_buffer(sample_t *data, unsigned nframes, unsigned nchannels) +{ + unsigned i = 0, j = nchannels * (nframes - 1), n; + + for (; i < j; i += nchannels, j -= nchannels) { + for (n = 0; n < nchannels; ++n) { + sample_t tmp = data[i + n]; + data[i + n] = data[j + n]; + data[j + n] = tmp; + } + } +} + +static int fetch(extrapolater_t *self, unsigned nframes) +{ + const pcm_sample_description_t *sfmt = pcm_get_format(self->src); + buffer_t *bp = &self->buffer[self->nbuffer]; + int rc = 0; + + if (realloc_buffer(bp, nframes * sfmt->bytes_per_frame) == 0) { + rc = pcm_read_frames(self->src, bp->data, nframes); + bp->count = rc > 0 ? rc : 0; + } + if (rc > 0) + self->nbuffer ^= 1; + return bp->count; +} + +static int extrapolate(extrapolater_t *self, const buffer_t *bp, + void *dst, unsigned nframes) +{ + const pcm_sample_description_t *sfmt = pcm_get_format(self->src); + unsigned i, n = sfmt->channels_per_frame; + float lpc[LPC_ORDER]; + + for (i = 0; i < n; ++i) { + vorbis_lpc_from_data(bp->data + i, lpc, bp->count, LPC_ORDER, n); + vorbis_lpc_predict(lpc, &bp->data[i + n * (bp->count - LPC_ORDER)], + LPC_ORDER, (sample_t*)dst + i, nframes, n); + } + return nframes; +} + +static int process1(extrapolater_t *self, void *buffer, unsigned nframes); +static int process2(extrapolater_t *self, void *buffer, unsigned nframes); +static int process3(extrapolater_t *self, void *buffer, unsigned nframes); + +static int process0(extrapolater_t *self, void *buffer, unsigned nframes) +{ + const pcm_sample_description_t *sfmt = pcm_get_format(self->src); + unsigned nchannels = sfmt->channels_per_frame; + buffer_t *bp = &self->buffer[self->nbuffer]; + + if (fetch(self, nframes) < 2 * LPC_ORDER) + memset(buffer, 0, nframes * sfmt->bytes_per_frame); + else { + reverse_buffer(bp->data, bp->count, nchannels); + extrapolate(self, bp, buffer, nframes); + reverse_buffer(buffer, nframes, nchannels); + reverse_buffer(bp->data, bp->count, nchannels); + } + self->process = bp->count ? process1 : process2; + return nframes; +} + +static int process1(extrapolater_t *self, void *buffer, unsigned nframes) +{ + const pcm_sample_description_t *sfmt = pcm_get_format(self->src); + buffer_t *bp = &self->buffer[self->nbuffer ^ 1]; + + assert(bp->count <= nframes); + memcpy(buffer, bp->data, bp->count * sfmt->bytes_per_frame); + if (!fetch(self, nframes)) + self->process = process2; + return bp->count; +} + +static int process2(extrapolater_t *self, void *buffer, unsigned nframes) +{ + const pcm_sample_description_t *sfmt = pcm_get_format(self->src); + buffer_t *bp = &self->buffer[self->nbuffer]; + buffer_t *bbp = &self->buffer[self->nbuffer ^ 1]; + + if (bp->count < 2 * LPC_ORDER) { + size_t total = bp->count + bbp->count; + if (bbp->count && + realloc_buffer(bbp, total * sfmt->bytes_per_frame) == 0) + { + memcpy(bbp->data + bbp->count * sfmt->channels_per_frame, + bp->data, bp->count * sfmt->bytes_per_frame); + bbp->count = total; + bp->count = 0; + bp = bbp; + self->nbuffer ^= 1; + } + } + self->process = process3; + + if (bp->count >= 2 * LPC_ORDER) + extrapolate(self, bp, buffer, nframes); + else + memset(buffer, 0, nframes * sfmt->bytes_per_frame); + return nframes; +} + +static int process3(extrapolater_t *self, void *buffer, unsigned nframes) +{ + return 0; +} + +static int read_frames(pcm_reader_t *reader, void *buffer, unsigned nframes) +{ + extrapolater_t *self = (extrapolater_t *)reader; + return self->process(self, buffer, nframes); +} + +static void teardown(pcm_reader_t **reader) +{ + extrapolater_t *self = (extrapolater_t *)*reader; + pcm_teardown(&self->src); + free(self->buffer[0].data); + free(self->buffer[1].data); + free(self); + *reader = 0; +} + +static pcm_reader_vtbl_t my_vtable = { + get_format, get_length, get_position, read_frames, teardown +}; + +pcm_reader_t *extrapolater_open(pcm_reader_t *reader) +{ + extrapolater_t *self = 0; + + if ((self = calloc(1, sizeof(extrapolater_t))) == 0) + return 0; + self->src = reader; + self->vtbl = &my_vtable; + self->process = process0; + return (pcm_reader_t *)self; +} diff --git a/src/lpc.c b/src/lpc.c new file mode 100644 index 0000000..dbd8d90 --- /dev/null +++ b/src/lpc.c @@ -0,0 +1,169 @@ +/******************************************************************** + * * + * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE. * + * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS * + * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE * + * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. * + * * + * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2009 * + * by the Xiph.Org Foundation http://www.xiph.org/ * + * * + ******************************************************************** + + function: LPC low level routines + last mod: $Id: lpc.c 16227 2009-07-08 06:58:46Z xiphmont $ + + ********************************************************************/ + +/* Some of these routines (autocorrelator, LPC coefficient estimator) + are derived from code written by Jutta Degener and Carsten Bormann; + thus we include their copyright below. The entirety of this file + is freely redistributable on the condition that both of these + copyright notices are preserved without modification. */ + +/* Preserved Copyright: *********************************************/ + +/* Copyright 1992, 1993, 1994 by Jutta Degener and Carsten Bormann, +Technische Universita"t Berlin + +Any use of this software is permitted provided that this notice is not +removed and that neither the authors nor the Technische Universita"t +Berlin are deemed to have made any representations as to the +suitability of this software for any purpose nor are held responsible +for any defects of this software. THERE IS ABSOLUTELY NO WARRANTY FOR +THIS SOFTWARE. + +As a matter of courtesy, the authors request to be informed about uses +this software has found, about bugs in this software, and about any +improvements that may be of general interest. + +Berlin, 28.11.1994 +Jutta Degener +Carsten Bormann + +*********************************************************************/ + +#ifdef HAVE_CONFIG_H +# include "config.h" +#endif + +#if HAVE_STDINT_H +# include +#endif + +#include +#include +#include +#include "lpc.h" +#include "lpcm.h" + +/* Autocorrelation LPC coeff generation algorithm invented by + N. Levinson in 1947, modified by J. Durbin in 1959. */ + +/* Input : n elements of time doamin data + Output: m lpc coefficients, excitation energy */ + +float vorbis_lpc_from_data(short *data,float *lpci,int n,int m,int stride){ + double *aut=malloc(sizeof(*aut)*(m+1)); + double *lpc=malloc(sizeof(*lpc)*(m)); + double error; + double epsilon; + int i,j; + + /* autocorrelation, p+1 lag coefficients */ + j=m+1; + while(j--){ + double d=0; /* double needed for accumulator depth */ + for(i=j;i -inline int lrint(double x) -{ - return _mm_cvtsd_si32(_mm_load_sd(&x)); -} -# endif -#endif - -static -inline double pcm_clip(double n, double min_value, double max_value) -{ - if (n < min_value) - return min_value; - else if (n > max_value) - return max_value; - return n; -} static inline float pcm_i2f(int32_t n) { diff --git a/src/lpcm.h b/src/lpcm.h index 8e896ae..2ac93ab 100644 --- a/src/lpcm.h +++ b/src/lpcm.h @@ -31,6 +31,36 @@ typedef struct pcm_sample_description_t { #define PCM_BYTES_PER_CHANNEL(desc) \ ((desc)->bytes_per_frame / (desc)->channels_per_frame) +#if defined(_MSC_VER) && _MSC_VER < 1800 +# ifdef _M_IX86 +static inline int lrint(double x) +{ + int n; + _asm { + fld x + fistp n + } + return n; +} +# else +# include +static inline int lrint(double x) +{ + return _mm_cvtsd_si32(_mm_load_sd(&x)); +} +# endif +#endif + +static +inline double pcm_clip(double n, double min_value, double max_value) +{ + if (n < min_value) + return min_value; + else if (n > max_value) + return max_value; + return n; +} + int pcm_convert_to_native_sint16(const pcm_sample_description_t *format, const void *input, uint32_t nframes, int16_t *result); diff --git a/src/main.c b/src/main.c index 0d07690..5e0dc0c 100644 --- a/src/main.c +++ b/src/main.c @@ -489,51 +489,82 @@ int write_sample(FILE *ofp, m4af_ctx_t *m4af, } static -int encode(pcm_reader_t *reader, HANDLE_AACENCODER encoder, - uint32_t frame_length, FILE *ofp, m4af_ctx_t *m4af, - int show_progress) +int encode(aacenc_param_ex_t *params, pcm_reader_t *reader, + HANDLE_AACENCODER encoder, uint32_t frame_length, + m4af_ctx_t *m4af) { - int16_t *ibuf = 0; - uint8_t *obuf = 0; - uint32_t olen; - uint32_t osize = 0; + struct buffer_t { + uint8_t *data; + uint32_t len, size; + }; + int16_t *ibuf = 0, *ip; + struct buffer_t obuf[2] = {{ 0 }}, *obp; + unsigned flip = 0; int nread = 1; - int consumed; int rc = -1; - int frames_written = 0; + int remaining, consumed; + int frames_written = 0, encoded = 0; aacenc_progress_t progress = { 0 }; const pcm_sample_description_t *fmt = pcm_get_format(reader); ibuf = malloc(frame_length * fmt->bytes_per_frame); aacenc_progress_init(&progress, pcm_get_length(reader), fmt->sample_rate); - do { + + for (;;) { + /* + * Since we delay the write, we cannot just exit loop when interrupted. + * Instead, we regard it as EOF. + */ if (g_interrupted) nread = 0; - else if (nread) { + if (nread > 0) { if ((nread = pcm_read_frames(reader, ibuf, frame_length)) < 0) { fprintf(stderr, "ERROR: read failed\n"); goto END; } - if (show_progress) + if (!params->silent) aacenc_progress_update(&progress, pcm_get_position(reader), fmt->sample_rate * 2); } - if ((consumed = aac_encode_frame(encoder, fmt, ibuf, nread, - &obuf, &olen, &osize)) < 0) - goto END; - if (olen > 0) { - if (write_sample(ofp, m4af, obuf, olen, frame_length) < 0) + ip = ibuf; + remaining = nread; + do { + obp = &obuf[flip]; + consumed = aac_encode_frame(encoder, fmt, ip, remaining, + &obp->data, &obp->len, &obp->size); + if (consumed < 0) goto END; + if (consumed == 0 && obp->len == 0) goto DONE; + if (obp->len == 0) break; + + remaining -= consumed; + ip += consumed * fmt->channels_per_frame; + flip ^= 1; + /* + * As we pad 1 frame at beginning and ending by our extrapolator, + * we want to drop them. + * We delay output by 1 frame by double buffering, and discard + * second frame and final frame from the encoder. + * Since sbr_header is included in the first frame (in case of + * SBR), we cannot discard first frame. So we pick second instead. + */ + ++encoded; + if (encoded == 1 || encoded == 3) + continue; + obp = &obuf[flip]; + if (write_sample(params->output_fp, m4af, obp->data, obp->len, + frame_length) < 0) goto END; ++frames_written; - } - } while (nread > 0 || olen > 0); - - if (show_progress) + } while (remaining > 0); + } +DONE: + if (!params->silent) aacenc_progress_finish(&progress, pcm_get_position(reader)); rc = frames_written; END: if (ibuf) free(ibuf); - if (obuf) free(obuf); + if (obuf[0].data) free(obuf[0].data); + if (obuf[1].data) free(obuf[1].data); return rc; } @@ -709,10 +740,13 @@ pcm_reader_t *open_input(aacenc_param_ex_t *params) } break; default: + fprintf(stderr, "ERROR: unsupported input file\n"); goto END; } } - return pcm_open_sint16_converter(reader); + if ((reader = pcm_open_sint16_converter(reader)) != 0) + reader = extrapolater_open(reader); + return reader; END: return 0; } @@ -794,8 +828,7 @@ int main(int argc, char **argv) m4af_set_priming_mode(m4af, params.gapless_mode + 1); m4af_begin_write(m4af); } - frame_count = encode(reader, encoder, aacinfo.frameLength, - params.output_fp, m4af, !params.silent); + frame_count = encode(¶ms, reader, encoder, aacinfo.frameLength, m4af); if (frame_count < 0) goto END; if (m4af) { diff --git a/src/pcm_reader.h b/src/pcm_reader.h index ee062aa..ba998e3 100644 --- a/src/pcm_reader.h +++ b/src/pcm_reader.h @@ -109,4 +109,6 @@ int apple_chan_chunk(pcm_io_context_t *io, uint32_t chunk_size, pcm_reader_t *pcm_open_sint16_converter(pcm_reader_t *reader); +pcm_reader_t *extrapolater_open(pcm_reader_t *reader); + #endif -- 2.39.2 From 556a3db11b01f05a3895537270a6b1b4aa816113 Mon Sep 17 00:00:00 2001 From: nu774 Date: Wed, 30 Oct 2013 00:27:50 +0900 Subject: [PATCH 08/16] add some copyright notice --- src/extrapolater.c | 4 ++++ src/metadata.c | 4 ++++ src/metadata.h | 4 ++++ src/pcm_reader.h | 4 ++++ src/pcm_sint16_converter.c | 4 ++++ 5 files changed, 20 insertions(+) diff --git a/src/extrapolater.c b/src/extrapolater.c index 22b0f12..c976be5 100644 --- a/src/extrapolater.c +++ b/src/extrapolater.c @@ -1,3 +1,7 @@ +/* + * Copyright (C) 2013 nu774 + * For conditions of distribution and use, see copyright notice in COPYING + */ #if HAVE_CONFIG_H # include "config.h" #endif diff --git a/src/metadata.c b/src/metadata.c index 9781b37..49781b2 100644 --- a/src/metadata.c +++ b/src/metadata.c @@ -1,3 +1,7 @@ +/* + * Copyright (C) 2013 nu774 + * For conditions of distribution and use, see copyright notice in COPYING + */ #if HAVE_CONFIG_H # include "config.h" #endif diff --git a/src/metadata.h b/src/metadata.h index 0438b79..a832d23 100644 --- a/src/metadata.h +++ b/src/metadata.h @@ -1,3 +1,7 @@ +/* + * Copyright (C) 2013 nu774 + * For conditions of distribution and use, see copyright notice in COPYING + */ #ifndef METADATA_H #define METADATA_H diff --git a/src/pcm_reader.h b/src/pcm_reader.h index ba998e3..1dfef34 100644 --- a/src/pcm_reader.h +++ b/src/pcm_reader.h @@ -1,3 +1,7 @@ +/* + * Copyright (C) 2013 nu774 + * For conditions of distribution and use, see copyright notice in COPYING + */ #ifndef PCM_READER_H #define PCM_READER_H diff --git a/src/pcm_sint16_converter.c b/src/pcm_sint16_converter.c index da31239..f4a7ad6 100644 --- a/src/pcm_sint16_converter.c +++ b/src/pcm_sint16_converter.c @@ -1,3 +1,7 @@ +/* + * Copyright (C) 2013 nu774 + * For conditions of distribution and use, see copyright notice in COPYING + */ #if HAVE_CONFIG_H # include "config.h" #endif -- 2.39.2 From 3b518efd31691737ad2cca7764554208198f59a3 Mon Sep 17 00:00:00 2001 From: nu774 Date: Wed, 30 Oct 2013 00:46:11 +0900 Subject: [PATCH 09/16] cleanup interface of aac_encode_frame() --- src/aacenc.c | 16 +++++++++------- src/aacenc.h | 7 ++++++- src/main.c | 9 ++------- 3 files changed, 17 insertions(+), 15 deletions(-) diff --git a/src/aacenc.c b/src/aacenc.c index bf975d0..99c288a 100644 --- a/src/aacenc.c +++ b/src/aacenc.c @@ -187,7 +187,7 @@ FAIL: int aac_encode_frame(HANDLE_AACENCODER encoder, const pcm_sample_description_t *format, const int16_t *input, unsigned iframes, - uint8_t **output, uint32_t *olen, uint32_t *osize) + aacenc_result_t *output) { uint32_t ilen = iframes * format->channels_per_frame; AACENC_BufDesc ibdesc = { 0 }, obdesc = { 0 }; @@ -205,12 +205,14 @@ int aac_encode_frame(HANDLE_AACENCODER encoder, unsigned channel_mode, obytes; channel_mode = aacEncoder_GetParam(encoder, AACENC_CHANNELMODE); - obytes = 6144 / 8 * channel_mode + 7; - if (!*output || *osize < obytes) { - *osize = obytes; - *output = realloc(*output, obytes); + obytes = 6144 / 8 * channel_mode; + if (!output->data || output->size < obytes) { + uint8_t *p = realloc(output->data, obytes); + if (!p) return -1; + output->size = obytes; + output->data = p; } - obufs[0] = *output; + obufs[0] = output->data; obuf_sizes[0] = obytes; iargs.numInSamples = ilen ? ilen : -1; /* -1 for signaling EOF */ @@ -230,6 +232,6 @@ int aac_encode_frame(HANDLE_AACENCODER encoder, fprintf(stderr, "ERROR: aacEncEncode() failed\n"); return -1; } - *olen = oargs.numOutBytes; + output->len = oargs.numOutBytes; return oargs.numInSamples / format->channels_per_frame; } diff --git a/src/aacenc.h b/src/aacenc.h index 2db945f..dc6751b 100644 --- a/src/aacenc.h +++ b/src/aacenc.h @@ -24,6 +24,11 @@ typedef struct aacenc_param_t { AACENC_PARAMS } aacenc_param_t; +typedef struct aacenc_result_t { + uint8_t *data; + uint32_t len, size; +} aacenc_result_t; + int aacenc_is_sbr_active(const aacenc_param_t *params); int aacenc_mp4asc(const aacenc_param_t *params, @@ -37,6 +42,6 @@ int aacenc_init(HANDLE_AACENCODER *encoder, const aacenc_param_t *params, int aac_encode_frame(HANDLE_AACENCODER encoder, const pcm_sample_description_t *format, const int16_t *input, unsigned iframes, - uint8_t **output, uint32_t *olen, uint32_t *osize); + aacenc_result_t *output); #endif diff --git a/src/main.c b/src/main.c index 5e0dc0c..d06c683 100644 --- a/src/main.c +++ b/src/main.c @@ -493,12 +493,8 @@ int encode(aacenc_param_ex_t *params, pcm_reader_t *reader, HANDLE_AACENCODER encoder, uint32_t frame_length, m4af_ctx_t *m4af) { - struct buffer_t { - uint8_t *data; - uint32_t len, size; - }; int16_t *ibuf = 0, *ip; - struct buffer_t obuf[2] = {{ 0 }}, *obp; + aacenc_result_t obuf[2] = {{ 0 }}, *obp; unsigned flip = 0; int nread = 1; int rc = -1; @@ -530,8 +526,7 @@ int encode(aacenc_param_ex_t *params, pcm_reader_t *reader, remaining = nread; do { obp = &obuf[flip]; - consumed = aac_encode_frame(encoder, fmt, ip, remaining, - &obp->data, &obp->len, &obp->size); + consumed = aac_encode_frame(encoder, fmt, ip, remaining, obp); if (consumed < 0) goto END; if (consumed == 0 && obp->len == 0) goto DONE; if (obp->len == 0) break; -- 2.39.2 From c5eec1554942ef99bcf6b2065008fd3730121ec9 Mon Sep 17 00:00:00 2001 From: nu774 Date: Wed, 30 Oct 2013 01:32:52 +0900 Subject: [PATCH 10/16] prepend 1 sample zero padding in case of SBR and enc_delay is odd This is required because odd enc_delay cannot be exactly expressed in downsampled scale, and HE-AACv2 FDK encoder actually has odd enc_delay. --- src/main.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/src/main.c b/src/main.c index d06c683..7d1a68f 100644 --- a/src/main.c +++ b/src/main.c @@ -823,6 +823,16 @@ int main(int argc, char **argv) m4af_set_priming_mode(m4af, params.gapless_mode + 1); m4af_begin_write(m4af); } + if (sbr_mode && (aacinfo.encoderDelay & 1)) { + /* + * Since odd delay cannot be exactly expressed in downsampled scale, + * we push one zero frame to the encoder here, to make delay even + */ + int16_t zero[8] = { 0 }; + aacenc_result_t obuf = { 0 }; + aac_encode_frame(encoder, sample_format, zero, 1, &obuf); + free(obuf.data); + } frame_count = encode(¶ms, reader, encoder, aacinfo.frameLength, m4af); if (frame_count < 0) goto END; -- 2.39.2 From d533c8e002bd64237ed54ca6b24b55b05578f41f Mon Sep 17 00:00:00 2001 From: nu774 Date: Wed, 30 Oct 2013 11:58:30 +0900 Subject: [PATCH 11/16] rename aacenc_result_t -> aacenc_frame_t, simplify write_sample() --- src/aacenc.c | 8 ++++---- src/aacenc.h | 8 ++++---- src/main.c | 22 ++++++++++------------ 3 files changed, 18 insertions(+), 20 deletions(-) diff --git a/src/aacenc.c b/src/aacenc.c index 99c288a..3640498 100644 --- a/src/aacenc.c +++ b/src/aacenc.c @@ -187,7 +187,7 @@ FAIL: int aac_encode_frame(HANDLE_AACENCODER encoder, const pcm_sample_description_t *format, const int16_t *input, unsigned iframes, - aacenc_result_t *output) + aacenc_frame_t *output) { uint32_t ilen = iframes * format->channels_per_frame; AACENC_BufDesc ibdesc = { 0 }, obdesc = { 0 }; @@ -206,10 +206,10 @@ int aac_encode_frame(HANDLE_AACENCODER encoder, channel_mode = aacEncoder_GetParam(encoder, AACENC_CHANNELMODE); obytes = 6144 / 8 * channel_mode; - if (!output->data || output->size < obytes) { + if (!output->data || output->capacity < obytes) { uint8_t *p = realloc(output->data, obytes); if (!p) return -1; - output->size = obytes; + output->capacity = obytes; output->data = p; } obufs[0] = output->data; @@ -232,6 +232,6 @@ int aac_encode_frame(HANDLE_AACENCODER encoder, fprintf(stderr, "ERROR: aacEncEncode() failed\n"); return -1; } - output->len = oargs.numOutBytes; + output->size = oargs.numOutBytes; return oargs.numInSamples / format->channels_per_frame; } diff --git a/src/aacenc.h b/src/aacenc.h index dc6751b..e622054 100644 --- a/src/aacenc.h +++ b/src/aacenc.h @@ -24,10 +24,10 @@ typedef struct aacenc_param_t { AACENC_PARAMS } aacenc_param_t; -typedef struct aacenc_result_t { +typedef struct aacenc_frame_t { uint8_t *data; - uint32_t len, size; -} aacenc_result_t; + uint32_t size, capacity; +} aacenc_frame_t; int aacenc_is_sbr_active(const aacenc_param_t *params); @@ -42,6 +42,6 @@ int aacenc_init(HANDLE_AACENCODER *encoder, const aacenc_param_t *params, int aac_encode_frame(HANDLE_AACENCODER encoder, const pcm_sample_description_t *format, const int16_t *input, unsigned iframes, - aacenc_result_t *output); + aacenc_frame_t *output); #endif diff --git a/src/main.c b/src/main.c index 7d1a68f..c3a9cca 100644 --- a/src/main.c +++ b/src/main.c @@ -472,16 +472,15 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params) }; static -int write_sample(FILE *ofp, m4af_ctx_t *m4af, - const void *data, uint32_t size, uint32_t duration) +int write_sample(FILE *ofp, m4af_ctx_t *m4af, aacenc_frame_t *frame) { if (!m4af) { - fwrite(data, 1, size, ofp); + fwrite(frame->data, 1, frame->size, ofp); if (ferror(ofp)) { fprintf(stderr, "ERROR: fwrite(): %s\n", strerror(errno)); return -1; } - } else if (m4af_write_sample(m4af, 0, data, size, duration) < 0) { + } else if (m4af_write_sample(m4af, 0, frame->data, frame->size, 0) < 0) { fprintf(stderr, "ERROR: failed to write m4a sample\n"); return -1; } @@ -494,7 +493,7 @@ int encode(aacenc_param_ex_t *params, pcm_reader_t *reader, m4af_ctx_t *m4af) { int16_t *ibuf = 0, *ip; - aacenc_result_t obuf[2] = {{ 0 }}, *obp; + aacenc_frame_t obuf[2] = {{ 0 }}, *obp; unsigned flip = 0; int nread = 1; int rc = -1; @@ -528,8 +527,8 @@ int encode(aacenc_param_ex_t *params, pcm_reader_t *reader, obp = &obuf[flip]; consumed = aac_encode_frame(encoder, fmt, ip, remaining, obp); if (consumed < 0) goto END; - if (consumed == 0 && obp->len == 0) goto DONE; - if (obp->len == 0) break; + if (consumed == 0 && obp->size == 0) goto DONE; + if (obp->size == 0) break; remaining -= consumed; ip += consumed * fmt->channels_per_frame; @@ -546,8 +545,7 @@ int encode(aacenc_param_ex_t *params, pcm_reader_t *reader, if (encoded == 1 || encoded == 3) continue; obp = &obuf[flip]; - if (write_sample(params->output_fp, m4af, obp->data, obp->len, - frame_length) < 0) + if (write_sample(params->output_fp, m4af, obp) < 0) goto END; ++frames_written; } while (remaining > 0); @@ -829,9 +827,9 @@ int main(int argc, char **argv) * we push one zero frame to the encoder here, to make delay even */ int16_t zero[8] = { 0 }; - aacenc_result_t obuf = { 0 }; - aac_encode_frame(encoder, sample_format, zero, 1, &obuf); - free(obuf.data); + aacenc_frame_t frame = { 0 }; + aac_encode_frame(encoder, sample_format, zero, 1, &frame); + free(frame.data); } frame_count = encode(¶ms, reader, encoder, aacinfo.frameLength, m4af); if (frame_count < 0) -- 2.39.2 From 3de0e22dcc176fb7f34be00d3406235870ae7db4 Mon Sep 17 00:00:00 2001 From: nu774 Date: Wed, 30 Oct 2013 14:27:28 +0900 Subject: [PATCH 12/16] use tell() to obtain data chunk offset --- src/caf_reader.c | 4 +--- src/main.c | 2 +- src/wav_reader.c | 5 +---- 3 files changed, 3 insertions(+), 8 deletions(-) diff --git a/src/caf_reader.c b/src/caf_reader.c index ede919a..40162b4 100644 --- a/src/caf_reader.c +++ b/src/caf_reader.c @@ -193,12 +193,10 @@ int caf_parse(caf_reader_t *reader, int64_t *data_length) } else if (fcc == M4AF_FOURCC('d','a','t','a')) { TRY_IO(pcm_skip(&reader->io, 4)); /* mEditCount */ *data_length = (chunk_size == ~0ULL) ? chunk_size : chunk_size - 4; - reader->data_offset += 12; + reader->data_offset = pcm_tell(&reader->io); break; } else TRY_IO(pcm_skip(&reader->io, chunk_size)); - - reader->data_offset += (chunk_size + 8); } ENSURE(reader->sample_format.channels_per_frame); ENSURE(fcc == M4AF_FOURCC('d','a','t','a')); diff --git a/src/main.c b/src/main.c index c3a9cca..ee07e35 100644 --- a/src/main.c +++ b/src/main.c @@ -675,7 +675,7 @@ int parse_raw_spec(const char *spec, pcm_sample_description_t *desc) static pcm_io_vtbl_t pcm_io_vtbl = { read_callback, seek_callback, tell_callback }; -static pcm_io_vtbl_t pcm_io_vtbl_noseek = { read_callback, 0, 0 }; +static pcm_io_vtbl_t pcm_io_vtbl_noseek = { read_callback, 0, tell_callback }; static pcm_reader_t *open_input(aacenc_param_ex_t *params) diff --git a/src/wav_reader.c b/src/wav_reader.c index e69cba7..bb0e93e 100644 --- a/src/wav_reader.c +++ b/src/wav_reader.c @@ -93,7 +93,6 @@ int riff_ds64(wav_reader_t *reader, int64_t *length) TRY_IO(pcm_scanl(&reader->io, "QQQL", &riff_size, length, &sample_count, &table_size) != 4); TRY_IO(pcm_skip(&reader->io, (chunk_size - 27) & ~1)); - reader->data_offset += (chunk_size + 9) & ~1; FAIL: return -1; } @@ -163,7 +162,6 @@ int wav_parse(wav_reader_t *reader, int64_t *data_length) container == RIFF_FOURCC('R','F','6','4')); TRY_IO(pcm_read32le(&reader->io, &fcc)); ENSURE(fcc == RIFF_FOURCC('W','A','V','E')); - reader->data_offset = 12; if (container == RIFF_FOURCC('R','F','6','4')) riff_ds64(reader, data_length); @@ -174,12 +172,11 @@ int wav_parse(wav_reader_t *reader, int64_t *data_length) } else if (fcc == RIFF_FOURCC('d','a','t','a')) { if (container == RIFF_FOURCC('R','I','F','F')) *data_length = chunk_size; - reader->data_offset += 8; + reader->data_offset = pcm_tell(&reader->io); break; } else { TRY_IO(pcm_skip(&reader->io, (chunk_size + 1) & ~1)); } - reader->data_offset += (chunk_size + 9) & ~1; } if (fcc == RIFF_FOURCC('d','a','t','a')) return 0; -- 2.39.2 From a3271927b857614e65efe1da56ce01bd100ff205 Mon Sep 17 00:00:00 2001 From: nu774 Date: Wed, 30 Oct 2013 19:02:00 +0900 Subject: [PATCH 13/16] bump version --- version.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/version.h b/version.h index b4ebd10..fa20fd1 100644 --- a/version.h +++ b/version.h @@ -1,4 +1,4 @@ #ifndef VERSION_H #define VERSION_H -const char *fdkaac_version = "0.4.1"; +const char *fdkaac_version = "0.4.2"; #endif -- 2.39.2 From 5732f1f6c59ac59bf1daf6baafa65ba1d756fa99 Mon Sep 17 00:00:00 2001 From: nu774 Date: Wed, 30 Oct 2013 19:04:56 +0900 Subject: [PATCH 14/16] update ChangeLog --- ChangeLog | 50 +++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 49 insertions(+), 1 deletion(-) diff --git a/ChangeLog b/ChangeLog index 133167f..4bc0fe0 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,7 +1,55 @@ -2013-10-25 nu774 +2013-10-30 nu774 * update ChangeLog [HEAD] + * bump version [v0.4.2] + + * use tell() to obtain data chunk offset + + * rename aacenc_result_t -> aacenc_frame_t, simplify write_sample() + + * prepend 1 sample zero padding in case of SBR and enc_delay is odd + + * cleanup interface of aac_encode_frame() + + * add some copyright notice + +2013-10-29 nu774 + + * smart padding for better gapless playback + + * fix unused variable warning + + * fix warning: cast size_t as sprintf() arg to int + + * fix vcxproj + + * fix pcm_seek() to inline + +2013-10-27 nu774 + + * bump version [v0.4.1] + + * add --include-sbr-delay + + * fix help message: show -I as shorthand for --ignorelength + + * remove --sbr-signaling + +2013-10-26 nu774 + + * re-fix #ifdef cond for lrint() + + * tag mapping: add recorded date and tempo, remove performer->artist + +2013-10-25 nu774 + + * fix MSVC12 build issue + + * fix build issue on platform where fileno is a naive macro + + * update ChangeLog + * bump version [v0.4.0] * update README -- 2.39.2 From 3c0f152d39a935dbabba2e32db23a6c954f5854f Mon Sep 17 00:00:00 2001 From: nu774 Date: Sun, 3 Nov 2013 11:26:12 +0900 Subject: [PATCH 15/16] support 7.1 channel mode added on FDK 3.4.12 --- src/aacenc.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/src/aacenc.c b/src/aacenc.c index 3640498..00c37f3 100644 --- a/src/aacenc.c +++ b/src/aacenc.c @@ -93,10 +93,10 @@ int aacenc_channel_mode(const pcm_sample_description_t *format) { uint32_t chanmask = format->channel_mask; - if (format->channels_per_frame > 6) + if (format->channels_per_frame > 8) return 0; if (!chanmask) { - static uint32_t defaults[] = { 0x4, 0x3, 0x7, 0, 0x37, 0x3f }; + static uint32_t defaults[] = { 0x4, 0x3, 0x7, 0, 0x37, 0x3f, 0, 0x63f }; chanmask = defaults[format->channels_per_frame - 1]; } switch (chanmask) { @@ -108,6 +108,10 @@ int aacenc_channel_mode(const pcm_sample_description_t *format) case 0x107: return MODE_1_2_1; case 0x607: return MODE_1_2_2; case 0x60f: return MODE_1_2_2_1; +#if AACENCODER_LIB_VL0 > 3 || (AACENCODER_LIB_VL0==3 && AACENCODER_LIB_VL1>=4) + case 0xff: return MODE_1_2_2_2_1; + case 0x63f: return MODE_7_1_REAR_SURROUND; +#endif } return 0; } -- 2.39.2 From e1adc1783585421fdbd32d9a1d3edf5337e9af4f Mon Sep 17 00:00:00 2001 From: nu774 Date: Sun, 3 Nov 2013 13:14:41 +0900 Subject: [PATCH 16/16] add --sbr-ratio to support AACENC_SBR_RATIO appeared on libFDK 3.4.12 --- src/aacenc.c | 73 ++++++++++++++++++++++++++++++++++++++++++++-------- src/aacenc.h | 7 +++++ src/main.c | 69 +++++++++++++++++++++++++++++++------------------ 3 files changed, 113 insertions(+), 36 deletions(-) diff --git a/src/aacenc.c b/src/aacenc.c index 00c37f3..44c611a 100644 --- a/src/aacenc.c +++ b/src/aacenc.c @@ -13,6 +13,17 @@ #include #include "aacenc.h" +int aacenc_is_sbr_ratio_available() +{ +#if AACENCODER_LIB_VL0 < 3 || (AACENCODER_LIB_VL0==3 && AACENCODER_LIB_VL1<4) + return 0; +#else + LIB_INFO lib_info; + aacenc_get_lib_info(&lib_info); + return lib_info.version > 0x03040000; +#endif +} + int aacenc_is_sbr_active(const aacenc_param_t *params) { switch (params->profile) { @@ -26,6 +37,33 @@ int aacenc_is_sbr_active(const aacenc_param_t *params) return 0; } +int aacenc_is_dual_rate_sbr(const aacenc_param_t *params) +{ + if (params->profile == AOT_PS || params->profile == AOT_MP2_PS) + return 1; + else if (params->profile == AOT_SBR || params->profile == AOT_MP2_SBR) + return params->sbr_ratio == 0 || params->sbr_ratio == 2; + else if (params->profile == AOT_ER_AAC_ELD && params->lowdelay_sbr) + return params->sbr_ratio == 2; + return 0; +} + +void aacenc_get_lib_info(LIB_INFO *info) +{ + LIB_INFO *lib_info = 0; + lib_info = calloc(FDK_MODULE_LAST, sizeof(LIB_INFO)); + if (aacEncGetLibInfo(lib_info) == AACENC_OK) { + int i; + for (i = 0; i < FDK_MODULE_LAST; ++i) { + if (lib_info[i].module_id == FDK_AACENC) { + memcpy(info, &lib_info[i], sizeof(LIB_INFO)); + break; + } + } + } + free(lib_info); +} + static const unsigned aacenc_sampling_freq_tab[] = { 96000, 88200, 64000, 48000, 44100, 32000, 24000, 22050, 16000, 12000, 11025, 8000, 7350, 0, 0, 0 @@ -50,10 +88,13 @@ int aacenc_mp4asc(const aacenc_param_t *params, uint8_t *outasc, uint32_t *outsize) { unsigned asc_sfreq = aacenc_sampling_freq_tab[(asc[0]&0x7)<<1 |asc[1]>>7]; + unsigned shift = aacenc_is_dual_rate_sbr(params); switch (params->profile) { case AOT_SBR: case AOT_PS: + if (!shift) + break; if (*outsize < ascsize + 3) return -1; memcpy(outasc, asc, ascsize); @@ -65,10 +106,10 @@ int aacenc_mp4asc(const aacenc_param_t *params, /* sbrPresentFlag:1 (value:1) */ outasc[ascsize+2] = 0x80; /* extensionSamplingFrequencyIndex:4 */ - outasc[ascsize+2] |= sampling_freq_index(asc_sfreq << 1) << 3; + outasc[ascsize+2] |= sampling_freq_index(asc_sfreq << shift) << 3; if (params->profile == AOT_SBR) { *outsize = ascsize + 3; - break; + return 0; } if (*outsize < ascsize + 5) return -1; @@ -78,13 +119,12 @@ int aacenc_mp4asc(const aacenc_param_t *params, /* psPresentFlag:1 (value:1) */ outasc[ascsize+4] = 0x80; *outsize = ascsize + 5; - break; - default: - if (*outsize < ascsize) - return -1; - memcpy(outasc, asc, ascsize); - *outsize = ascsize; + return 0; } + if (*outsize < ascsize) + return -1; + memcpy(outasc, asc, ascsize); + *outsize = ascsize; return 0; } @@ -122,8 +162,11 @@ int aacenc_init(HANDLE_AACENCODER *encoder, const aacenc_param_t *params, { int channel_mode; int aot; + LIB_INFO lib_info; *encoder = 0; + aacenc_get_lib_info(&lib_info); + if ((channel_mode = aacenc_channel_mode(format)) == 0) { fprintf(stderr, "ERROR: unsupported channel layout\n"); goto FAIL; @@ -149,13 +192,21 @@ int aacenc_init(HANDLE_AACENCODER *encoder, const aacenc_param_t *params, fprintf(stderr, "ERROR: unsupported sample rate\n"); goto FAIL; } - aacEncoder_SetParam(*encoder, AACENC_CHANNELMODE, channel_mode); + if (aacEncoder_SetParam(*encoder, AACENC_CHANNELMODE, + channel_mode) != AACENC_OK) { + fprintf(stderr, "ERROR: unsupported channel mode\n"); + goto FAIL; + } aacEncoder_SetParam(*encoder, AACENC_BANDWIDTH, params->bandwidth); aacEncoder_SetParam(*encoder, AACENC_CHANNELORDER, 1); aacEncoder_SetParam(*encoder, AACENC_AFTERBURNER, !!params->afterburner); - if (aot == AOT_ER_AAC_ELD && params->lowdelay_sbr) - aacEncoder_SetParam(*encoder, AACENC_SBR_MODE, 1); + aacEncoder_SetParam(*encoder, AACENC_SBR_MODE, params->lowdelay_sbr); + +#if AACENCODER_LIB_VL0 > 3 || (AACENCODER_LIB_VL0==3 && AACENCODER_LIB_VL1>=4) + if (lib_info.version > 0x03040000) + aacEncoder_SetParam(*encoder, AACENC_SBR_RATIO, params->sbr_ratio); +#endif if (aacEncoder_SetParam(*encoder, AACENC_TRANSMUX, params->transport_format) != AACENC_OK) { diff --git a/src/aacenc.h b/src/aacenc.h index e622054..368a440 100644 --- a/src/aacenc.h +++ b/src/aacenc.h @@ -15,6 +15,7 @@ unsigned bandwidth; \ unsigned afterburner; \ unsigned lowdelay_sbr; \ + unsigned sbr_ratio; \ unsigned sbr_signaling; \ unsigned transport_format; \ unsigned adts_crc_check; \ @@ -29,8 +30,14 @@ typedef struct aacenc_frame_t { uint32_t size, capacity; } aacenc_frame_t; +int aacenc_is_sbr_ratio_available(); + int aacenc_is_sbr_active(const aacenc_param_t *params); +int aacenc_is_dual_rate_sbr(const aacenc_param_t *params); + +void aacenc_get_lib_info(LIB_INFO *info); + int aacenc_mp4asc(const aacenc_param_t *params, const uint8_t *asc, uint32_t ascsize, uint8_t *outasc, uint32_t *outsize); diff --git a/src/main.c b/src/main.c index ee07e35..8521a33 100644 --- a/src/main.c +++ b/src/main.c @@ -132,7 +132,14 @@ PROGNAME " %s\n" " -a, --afterburner Afterburner\n" " 0: Off\n" " 1: On(default)\n" -" -L, --lowdelay-sbr Enable ELD-SBR (AAC ELD only)\n" +" -L, --lowdelay-sbr <-1|0|1> Configure SBR activity on AAC ELD\n" +" -1: Use ELD SBR auto configurator\n" +" 0: Disable SBR on ELD (default)\n" +" 1: Enable SBR on ELD\n" +" -s, --sbr-ratio <0|1|2> Controls activation of downsampled SBR\n" +" 0: Use lib default (default)\n" +" 1: downsampled SBR (default for ELD+SBR)\n" +" 2: dual-rate SBR (default for HE-AAC)\n" " -f, --transport-format Transport format\n" " 0: RAW (default, muxed into M4A)\n" " 1: ADIF\n" @@ -228,7 +235,7 @@ static int parse_options(int argc, char **argv, aacenc_param_ex_t *params) { int ch; - unsigned n; + int n; #define OPT_INCLUDE_SBR_DELAY M4AF_FOURCC('s','d','l','y') #define OPT_MOOV_BEFORE_MDAT M4AF_FOURCC('m','o','o','v') @@ -247,7 +254,8 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params) { "bitrate-mode", required_argument, 0, 'm' }, { "bandwidth", required_argument, 0, 'w' }, { "afterburner", required_argument, 0, 'a' }, - { "lowdelay-sbr", no_argument, 0, 'L' }, + { "lowdelay-sbr", required_argument, 0, 'L' }, + { "sbr-ratio", required_argument, 0, 's' }, { "transport-format", required_argument, 0, 'f' }, { "adts-crc-check", no_argument, 0, 'C' }, { "header-period", required_argument, 0, 'P' }, @@ -325,7 +333,18 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params) params->afterburner = n; break; case 'L': - params->lowdelay_sbr = 1; + if (sscanf(optarg, "%d", &n) != 1 || n < -1 || n > 1) { + fprintf(stderr, "invalid arg for lowdelay-sbr\n"); + return -1; + } + params->lowdelay_sbr = n; + break; + case 's': + if (sscanf(optarg, "%u", &n) != 1 || n > 2) { + fprintf(stderr, "invalid arg for sbr-ratio\n"); + return -1; + } + params->sbr_ratio = n; break; case 'f': if (sscanf(optarg, "%u", &n) != 1) { @@ -567,19 +586,11 @@ void put_tool_tag(m4af_ctx_t *m4af, const aacenc_param_ex_t *params, { char tool_info[256]; char *p = tool_info; - LIB_INFO *lib_info = 0; + LIB_INFO lib_info; p += sprintf(p, PROGNAME " %s, ", fdkaac_version); - - lib_info = calloc(FDK_MODULE_LAST, sizeof(LIB_INFO)); - if (aacEncGetLibInfo(lib_info) == AACENC_OK) { - int i; - for (i = 0; i < FDK_MODULE_LAST; ++i) - if (lib_info[i].module_id == FDK_AACENC) - break; - p += sprintf(p, "libfdk-aac %s, ", lib_info[i].versionStr); - } - free(lib_info); + aacenc_get_lib_info(&lib_info); + p += sprintf(p, "libfdk-aac %s, ", lib_info.versionStr); if (params->bitrate_mode) sprintf(p, "VBR mode %d", params->bitrate_mode); else @@ -756,11 +767,13 @@ int main(int argc, char **argv) pcm_reader_t *reader = 0; HANDLE_AACENCODER encoder = 0; AACENC_InfoStruct aacinfo = { 0 }; + LIB_INFO lib_info = { 0 }; m4af_ctx_t *m4af = 0; const pcm_sample_description_t *sample_format; int downsampled_timescale = 0; int frame_count = 0; int sbr_mode = 0; + unsigned scale_shift = 0; setlocale(LC_CTYPE, ""); setbuf(stderr, 0); @@ -782,7 +795,16 @@ int main(int argc, char **argv) * way in MPEG4 part3 spec, and seems the only way supported by iTunes. * Since FDK library does not support it, we have to do it on our side. */ + sbr_mode = aacenc_is_sbr_active((aacenc_param_t*)¶ms); + if (sbr_mode && !aacenc_is_sbr_ratio_available()) { + fprintf(stderr, "WARNING: Only dual-rate SBR is available " + "for this version\n"); + params.sbr_ratio = 2; + } + scale_shift = aacenc_is_dual_rate_sbr((aacenc_param_t*)¶ms); params.sbr_signaling = (params.transport_format == TT_MP4_LOAS) ? 2 : 0; + if (sbr_mode && !scale_shift) + params.sbr_signaling = 2; if (aacenc_init(&encoder, (aacenc_param_t*)¶ms, sample_format, &aacinfo) < 0) @@ -800,15 +822,13 @@ int main(int argc, char **argv) goto END; } handle_signals(); - sbr_mode = aacenc_is_sbr_active((aacenc_param_t*)¶ms); + if (!params.transport_format) { uint32_t scale; uint8_t mp4asc[32]; uint32_t ascsize = sizeof(mp4asc); unsigned framelen = aacinfo.frameLength; - if (sbr_mode) - downsampled_timescale = 1; - scale = sample_format->sample_rate >> downsampled_timescale; + scale = sample_format->sample_rate >> scale_shift; if ((m4af = m4af_create(M4AF_CODEC_MP4A, scale, &m4af_io, params.output_fp)) < 0) goto END; @@ -816,12 +836,12 @@ int main(int argc, char **argv) aacinfo.confSize, mp4asc, &ascsize); m4af_set_decoder_specific_info(m4af, 0, mp4asc, ascsize); m4af_set_fixed_frame_duration(m4af, 0, - framelen >> downsampled_timescale); + framelen >> scale_shift); m4af_set_vbr_mode(m4af, 0, params.bitrate_mode); m4af_set_priming_mode(m4af, params.gapless_mode + 1); m4af_begin_write(m4af); } - if (sbr_mode && (aacinfo.encoderDelay & 1)) { + if (scale_shift && (aacinfo.encoderDelay & 1)) { /* * Since odd delay cannot be exactly expressed in downsampled scale, * we push one zero frame to the encoder here, to make delay even @@ -841,12 +861,11 @@ int main(int argc, char **argv) if (sbr_mode && params.profile != AOT_ER_AAC_ELD && !params.include_sbr_delay) - delay -= 481 << 1; - if (sbr_mode && (delay & 1)) + delay -= 481 << scale_shift; + if (scale_shift && (delay & 1)) ++delay; padding = frame_count * aacinfo.frameLength - frames_read - delay; - m4af_set_priming(m4af, 0, delay >> downsampled_timescale, - padding >> downsampled_timescale); + m4af_set_priming(m4af, 0, delay >> scale_shift, padding >> scale_shift); if (finalize_m4a(m4af, ¶ms, encoder) < 0) goto END; } -- 2.39.2