bump version

This is required because odd enc_delay cannot be exactly expressed in
downsampled scale, and HE-AACv2 FDK encoder actually has odd enc_delay.

ChangeLog

@@ -1,6 +1,26 @@
+2013-10-25  nu774 <honeycomb77@gmail.com>
+
+  * update ChangeLog [HEAD]
+
+  * bump version [v0.4.0]
+
+  * update README
+
+2013-10-24  nu774 <honeycomb77@gmail.com>
+
+  * caf input support
+
+  * refactor pcm io routines
+
+2013-10-23  nu774 <honeycomb77@gmail.com>
+
+  * cleanup metadata handling
+
+  * --tag-from-json: properly support number/total format in json track field
+
 2013-10-22  nu774 <honeycomb77@gmail.com>
 
-  * bump version [HEAD]
+  * bump version [v0.3.3]
 
   * fixed bogus sgpd written on --gapless-mode=1 and 2
 

MSVC/fdkaac.vcxproj

@@ -98,6 +98,8 @@ copy ..\fdk-aac\libSYS\include\machine_type.h include\fdk-aac\ </Command>
     <ClCompile Include="..\src\aacenc.c" />
     <ClCompile Include="..\src\caf_reader.c" />
     <ClCompile Include="..\src\compat_win32.c" />
+    <ClCompile Include="..\src\extrapolater.c" />
+    <ClCompile Include="..\src\lpc.c" />
     <ClCompile Include="..\src\lpcm.c" />
     <ClCompile Include="..\src\m4af.c" />
     <ClCompile Include="..\src\main.c" />
@@ -111,7 +113,10 @@ copy ..\fdk-aac\libSYS\include\machine_type.h include\fdk-aac\ </Command>
   <ItemGroup>
     <ClInclude Include="..\missings\getopt.h" />
     <ClInclude Include="..\src\aacenc.h" />
+    <ClInclude Include="..\src\caf_reader.h" />
+    <ClInclude Include="..\src\catypes.h" />
     <ClInclude Include="..\src\compat.h" />
+    <ClInclude Include="..\src\lpc.h" />
     <ClInclude Include="..\src\lpcm.h" />
     <ClInclude Include="..\src\m4af.h" />
     <ClInclude Include="..\src\m4af_endian.h" />

MSVC/fdkaac.vcxproj.filters

@@ -24,6 +24,12 @@
     <ClCompile Include="..\src\compat_win32.c">
       <Filter>Source Files</Filter>
     </ClCompile>
+    <ClCompile Include="..\src\extrapolater.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\src\lpc.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
     <ClCompile Include="..\src\lpcm.c">
       <Filter>Source Files</Filter>
     </ClCompile>
@@ -62,6 +68,9 @@
     <ClInclude Include="..\src\compat.h">
       <Filter>Header Files</Filter>
     </ClInclude>
+    <ClInclude Include="..\src\lpc.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
     <ClInclude Include="..\src\lpcm.h">
       <Filter>Header Files</Filter>
     </ClInclude>

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                 \

src/aacenc.c

@@ -10,6 +10,7 @@
 #endif
 #include <stdio.h>
 #include <stdlib.h>
+#include <string.h>
 #include "aacenc.h"
 
 int aacenc_is_sbr_active(const aacenc_param_t *params)
@@ -25,6 +26,68 @@ int aacenc_is_sbr_active(const aacenc_param_t *params)
     return 0;
 }
 
+static const unsigned aacenc_sampling_freq_tab[] = {
+    96000, 88200, 64000, 48000, 44100, 32000, 24000, 22050, 
+    16000, 12000, 11025, 8000, 7350, 0, 0, 0
+};
+
+static
+unsigned sampling_freq_index(unsigned rate)
+{
+    unsigned i;
+    for (i = 0; aacenc_sampling_freq_tab[i]; ++i)
+        if (aacenc_sampling_freq_tab[i] == rate)
+            return i;
+    return 0xf;
+}
+
+/*
+ * Append backward compatible SBR/PS signaling to implicit signaling ASC,
+ * if SBR/PS is present.
+ */
+int aacenc_mp4asc(const aacenc_param_t *params,
+                  const uint8_t *asc, uint32_t ascsize,
+                  uint8_t *outasc, uint32_t *outsize)
+{
+    unsigned asc_sfreq = aacenc_sampling_freq_tab[(asc[0]&0x7)<<1 |asc[1]>>7];
+
+    switch (params->profile) {
+    case AOT_SBR:
+    case AOT_PS:
+        if (*outsize < ascsize + 3)
+            return -1;
+        memcpy(outasc, asc, ascsize);
+        /* syncExtensionType:11 (value:0x2b7) */
+        outasc[ascsize+0] = 0x2b << 1;
+        outasc[ascsize+1] = 0x7 << 5;
+        /* extensionAudioObjectType:5 (value:5)*/
+        outasc[ascsize+1] |= 5;
+        /* sbrPresentFlag:1 (value:1) */
+        outasc[ascsize+2] = 0x80;
+        /* extensionSamplingFrequencyIndex:4 */
+        outasc[ascsize+2] |= sampling_freq_index(asc_sfreq << 1) << 3;
+        if (params->profile == AOT_SBR) {
+            *outsize = ascsize + 3;
+            break;
+        }
+        if (*outsize < ascsize + 5)
+            return -1;
+        /* syncExtensionType:11 (value:0x548) */
+        outasc[ascsize+2] |= 0x5;
+        outasc[ascsize+3] = 0x48;
+        /* 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;
+}
+
 static
 int aacenc_channel_mode(const pcm_sample_description_t *format)
 {
@@ -124,7 +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_frame_t *output)
 {
     uint32_t ilen = iframes * format->channels_per_frame;
     AACENC_BufDesc ibdesc = { 0 }, obdesc = { 0 };
@@ -142,12 +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;
+    obytes = 6144 / 8 * channel_mode;
-    if (!*output || *osize < obytes) {
+    if (!output->data || output->capacity < obytes) {
-        *osize = obytes;
+        uint8_t *p = realloc(output->data, obytes);
-        *output = realloc(*output, obytes);
+        if (!p) return -1;
+        output->capacity = obytes;
+        output->data = p;
     }
-    obufs[0] = *output;
+    obufs[0] = output->data;
     obuf_sizes[0] = obytes;
 
     iargs.numInSamples = ilen ? ilen : -1; /* -1 for signaling EOF */
@@ -167,6 +232,6 @@ int aac_encode_frame(HANDLE_AACENCODER encoder,
         fprintf(stderr, "ERROR: aacEncEncode() failed\n");
         return -1;
     }
-    *olen = oargs.numOutBytes;
+    output->size = oargs.numOutBytes;
-    return oargs.numInSamples;
+    return oargs.numInSamples / format->channels_per_frame;
 }

src/aacenc.h

@@ -24,8 +24,17 @@ typedef struct aacenc_param_t {
     AACENC_PARAMS
 } aacenc_param_t;
 
+typedef struct aacenc_frame_t {
+    uint8_t *data;
+    uint32_t size, capacity;
+} aacenc_frame_t;
+
 int aacenc_is_sbr_active(const aacenc_param_t *params);
 
+int aacenc_mp4asc(const aacenc_param_t *params,
+                  const uint8_t *asc, uint32_t ascsize,
+                  uint8_t *outasc, uint32_t *outsize);
+
 int aacenc_init(HANDLE_AACENCODER *encoder, const aacenc_param_t *params,
                 const pcm_sample_description_t *format,
                 AACENC_InfoStruct *info);
@@ -33,6 +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_frame_t *output);
 
 #endif

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'));

src/extrapolater.c

@@ -0,0 +1,208 @@
+/* 
+ * Copyright (C) 2013 nu774
+ * For conditions of distribution and use, see copyright notice in COPYING
+ */
+#if HAVE_CONFIG_H
+#  include "config.h"
+#endif
+#if HAVE_STDINT_H
+#  include <stdint.h>
+#endif
+#include <stdlib.h>
+#include <string.h>
+#include <assert.h>
+#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;
+}

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 <stdint.h>
+#endif
+
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+#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<n;i++)d+=(double)data[i*stride]*data[(i-j)*stride]/1073741824.0;
+    aut[j]=d;
+  }
+
+  /* Generate lpc coefficients from autocorr values */
+
+  /* set our noise floor to about -100dB */
+  error=aut[0] * (1. + 1e-10);
+  epsilon=1e-9*aut[0]+1e-10;
+
+  for(i=0;i<m;i++){
+    double r= -aut[i+1];
+
+    if(error<epsilon){
+      memset(lpc+i,0,(m-i)*sizeof(*lpc));
+      goto done;
+    }
+
+    /* Sum up this iteration's reflection coefficient; note that in
+       Vorbis we don't save it.  If anyone wants to recycle this code
+       and needs reflection coefficients, save the results of 'r' from
+       each iteration. */
+
+    for(j=0;j<i;j++)r-=lpc[j]*aut[i-j];
+    r/=error;
+
+    /* Update LPC coefficients and total error */
+
+    lpc[i]=r;
+    for(j=0;j<i/2;j++){
+      double tmp=lpc[j];
+
+      lpc[j]+=r*lpc[i-1-j];
+      lpc[i-1-j]+=r*tmp;
+    }
+    if(i&1)lpc[j]+=lpc[j]*r;
+
+    error*=1.-r*r;
+
+  }
+
+ done:
+
+  /* slightly damp the filter */
+  {
+    double g = .99;
+    double damp = g;
+    for(j=0;j<m;j++){
+      lpc[j]*=damp;
+      damp*=g;
+    }
+  }
+
+  for(j=0;j<m;j++)lpci[j]=(float)lpc[j];
+
+  /* we need the error value to know how big an impulse to hit the
+     filter with later */
+
+  free(aut);
+  free(lpc);
+  return error;
+}
+
+void vorbis_lpc_predict(float *coeff,short *prime,int m,
+                        short *data,long n,int stride){
+
+  /* in: coeff[0...m-1] LPC coefficients
+         prime[0...m-1] initial values (allocated size of n+m-1)
+    out: data[0...n-1] data samples */
+
+  long i,j,o,p;
+  float y;
+  float *work=malloc(sizeof(*work)*(m+n));
+
+  if(!prime)
+    for(i=0;i<m;i++)
+      work[i]=0.f;
+  else
+    for(i=0;i<m;i++)
+      work[i]=prime[i*stride]/32768.0f;
+
+  for(i=0;i<n;i++){
+    y=0;
+    o=i;
+    p=m;
+    for(j=0;j<m;j++)
+      y-=work[o++]*coeff[--p];
+
+    work[o]=y;
+    data[i*stride]=lrint(pcm_clip(y*32768.0,-32768.0,32767.0));
+  }
+  free(work);
+}

src/lpc.h

@@ -0,0 +1,27 @@
+/********************************************************************
+ *                                                                  *
+ * 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 1994-2007             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+  function: LPC low level routines
+  last mod: $Id: lpc.h 16037 2009-05-26 21:10:58Z xiphmont $
+
+ ********************************************************************/
+
+#ifndef _V_LPC_H_
+#define _V_LPC_H_
+
+/* simple linear scale LPC code */
+extern float vorbis_lpc_from_data(short *data,float *lpc,int n,int m,int stride);
+
+extern void vorbis_lpc_predict(float *coeff,short *prime,int m,
+                               short *data,long n,int stride);
+
+#endif

src/lpcm.c

@@ -13,36 +13,6 @@
 #include "lpcm.h"
 #include "m4af_endian.h"
 
-#ifdef _MSC_VER
-#  define inline __inline
-#  ifdef _M_IX86
-inline int lrint(double x)
-{
-    int n;
-    _asm {
-        fld x
-        fistp n
-    }
-    return n;
-}
-#  else
-#    include <emmintrin.h>
-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)
 {

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 <emmintrin.h>
+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);

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);

src/main.c

@@ -133,10 +133,6 @@ PROGNAME " %s\n"
 "                                 0: Off\n"
 "                                 1: On(default)\n"
 " -L, --lowdelay-sbr            Enable ELD-SBR (AAC ELD only)\n"
-" -s, --sbr-signaling <n>       SBR signaling mode\n"
-"                                 0: Implicit, backward compatible(default)\n"
-"                                 1: Explicit SBR and implicit PS\n"
-"                                 2: Explicit hierarchical signaling\n"
 " -f, --transport-format <n>    Transport format\n"
 "                                 0: RAW (default, muxed into M4A)\n"
 "                                 1: ADIF\n"
@@ -153,7 +149,10 @@ PROGNAME " %s\n"
 "                                 0: iTunSMPB (default)\n"
 "                                 1: ISO standard (edts + sgpd)\n"
 "                                 2: Both\n"
-" --ignorelength                Ignore length of WAV header\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"
 "\n"
@@ -208,6 +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;
@@ -230,6 +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')
@@ -247,12 +248,12 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params)
         { "bandwidth",        required_argument, 0, 'w' },
         { "afterburner",      required_argument, 0, 'a' },
         { "lowdelay-sbr",     no_argument,       0, 'L' },
-        { "sbr-signaling",    required_argument, 0, 's' },
         { "transport-format", required_argument, 0, 'f' },
         { "adts-crc-check",   no_argument,       0, 'C' },
         { "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   },
@@ -326,13 +327,6 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params)
         case 'L':
             params->lowdelay_sbr = 1;
             break;
-        case 's':
-            if (sscanf(optarg, "%u", &n) != 1 || n > 2) {
-                fprintf(stderr, "invalid arg for sbr-signaling\n");
-                return -1;
-            }
-            params->sbr_signaling = n;
-            break;
         case 'f':
             if (sscanf(optarg, "%u", &n) != 1) {
                 fprintf(stderr, "invalid arg for transport-format\n");
@@ -360,6 +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;
@@ -475,16 +472,15 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params)
 };
 
 static
-int write_sample(FILE *ofp, m4af_ctx_t *m4af,
+int write_sample(FILE *ofp, m4af_ctx_t *m4af, aacenc_frame_t *frame)
-                 const void *data, uint32_t size, uint32_t duration)
 {
     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;
     }
@@ -492,51 +488,76 @@ int write_sample(FILE *ofp, m4af_ctx_t *m4af,
 }
 
 static
-int encode(pcm_reader_t *reader, HANDLE_AACENCODER encoder,
+int encode(aacenc_param_ex_t *params, pcm_reader_t *reader,
-           uint32_t frame_length, FILE *ofp, m4af_ctx_t *m4af,
+           HANDLE_AACENCODER encoder, uint32_t frame_length, 
-           int show_progress)
+           m4af_ctx_t *m4af)
 {
-    int16_t *ibuf = 0;
+    int16_t *ibuf = 0, *ip;
-    uint8_t *obuf = 0;
+    aacenc_frame_t obuf[2] = {{ 0 }}, *obp;
-    uint32_t olen;
+    unsigned flip = 0;
-    uint32_t osize = 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,
+        ip = ibuf;
-                                         &obuf, &olen, &osize)) < 0)
+        remaining = nread;
-            goto END;
+        do {
-        if (olen > 0) {
+            obp = &obuf[flip];
-            if (write_sample(ofp, m4af, obuf, olen, frame_length) < 0)
+            consumed = aac_encode_frame(encoder, fmt, ip, remaining, obp);
+            if (consumed < 0) goto END;
+            if (consumed == 0 && obp->size == 0) goto DONE;
+            if (obp->size == 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) < 0)
                 goto END;
             ++frames_written;
+        } while (remaining > 0);
     }
-    } while (nread > 0 || olen > 0);
+DONE:
-
+    if (!params->silent)
-    if (show_progress)
         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;
 }
 
@@ -610,7 +631,7 @@ 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;
 }
@@ -654,7 +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)
@@ -669,7 +690,7 @@ pcm_reader_t *open_input(aacenc_param_ex_t *params)
         goto END;
     }
     io.cookie = params->input_fp;
-    if (fstat(fileno(io.cookie), &stb) == 0
+    if (fstat(fileno(params->input_fp), &stb) == 0
             && (stb.st_mode & S_IFMT) == S_IFREG)
         io.vtbl = &pcm_io_vtbl;
     else
@@ -712,10 +733,13 @@ 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;
 }
@@ -736,6 +760,7 @@ int main(int argc, char **argv)
     const pcm_sample_description_t *sample_format;
     int downsampled_timescale = 0;
     int frame_count = 0;
+    int sbr_mode = 0;
 
     setlocale(LC_CTYPE, "");
     setbuf(stderr, 0);
@@ -748,6 +773,17 @@ int main(int argc, char **argv)
 
     sample_format = pcm_get_format(reader);
 
+    /*
+     * We use explicit/hierarchical signaling for LOAS.
+     * Other than that, we request implicit signaling to FDK library, then 
+     * append explicit/backward-compatible signaling to ASC in case of MP4FF.
+     *
+     * Explicit/backward-compatible signaling of SBR is the most recommended
+     * 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.
+     */
+    params.sbr_signaling = (params.transport_format == TT_MP4_LOAS) ? 2 : 0;
+
     if (aacenc_init(&encoder, (aacenc_param_t*)&params, sample_format,
                     &aacinfo) < 0)
         goto END;
@@ -764,34 +800,51 @@ int main(int argc, char **argv)
         goto END;
     }
     handle_signals();
+    sbr_mode = aacenc_is_sbr_active((aacenc_param_t*)&params);
     if (!params.transport_format) {
         uint32_t scale;
+        uint8_t mp4asc[32];
+        uint32_t ascsize = sizeof(mp4asc);
         unsigned framelen = aacinfo.frameLength;
-        int sbr_mode = aacenc_is_sbr_active((aacenc_param_t*)&params);
+        if (sbr_mode)
-        int sig_mode = aacEncoder_GetParam(encoder, AACENC_SIGNALING_MODE);
-        if (sbr_mode && !sig_mode)
             downsampled_timescale = 1;
         scale = sample_format->sample_rate >> downsampled_timescale;
         if ((m4af = m4af_create(M4AF_CODEC_MP4A, scale, &m4af_io,
                                 params.output_fp)) < 0)
             goto END;
-        m4af_set_decoder_specific_info(m4af, 0, aacinfo.confBuf,
+        aacenc_mp4asc((aacenc_param_t*)&params, aacinfo.confBuf,
-                                       aacinfo.confSize);
+                      aacinfo.confSize, mp4asc, &ascsize);
+        m4af_set_decoder_specific_info(m4af, 0, mp4asc, ascsize);
         m4af_set_fixed_frame_duration(m4af, 0,
                                       framelen >> downsampled_timescale);
         m4af_set_vbr_mode(m4af, 0, params.bitrate_mode);
         m4af_set_priming_mode(m4af, params.gapless_mode + 1);
         m4af_begin_write(m4af);
     }
-    frame_count = encode(reader, encoder, aacinfo.frameLength,
+    if (sbr_mode && (aacinfo.encoderDelay & 1)) {
-                         params.output_fp, m4af, !params.silent);
+        /*
+         * 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_frame_t frame = { 0 };
+        aac_encode_frame(encoder, sample_format, zero, 1, &frame);
+        free(frame.data);
+    }
+    frame_count = encode(&params, reader, encoder, aacinfo.frameLength, m4af);
     if (frame_count < 0)
         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, &params, encoder) < 0)

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
@@ -56,7 +60,8 @@ static tag_key_mapping_t tag_mapping_table[] = {
     { "grouping",                   M4AF_TAG_GROUPING            },
     { "itunescompilation",          M4AF_TAG_COMPILATION         },
     { "lyrics",                     M4AF_TAG_LYRICS              },
-    { "performer",                  M4AF_TAG_ARTIST              },
+    { "tempo",                      M4AF_TAG_TEMPO               },
+    { "recordeddate",               M4AF_TAG_DATE                },
     { "title",                      M4AF_TAG_TITLE               },
     { "titlesort",                  M4AF_FOURCC('s','o','n','m') },
     { "titlesortorder",             M4AF_FOURCC('s','o','n','m') },

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
 

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
 
@@ -85,7 +89,7 @@ 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;
 }
@@ -109,4 +113,6 @@ 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

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

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;

version.h

@@ -1,4 +1,4 @@
 #ifndef VERSION_H
 #define VERSION_H
-const char *fdkaac_version = "0.4.0";
+const char *fdkaac_version = "0.4.2";
 #endif