xref: /aosp_15_r20/external/sonic/sonic.c (revision b290403dc9d28f89f133eb7e190ea8185d440ecd)

/* Sonic library
   Copyright 2010
   Bill Cox
   This file is part of the Sonic Library.

   This file is licensed under the Apache 2.0 license.
*/

#include "sonic.h"

#include <limits.h>
#include <math.h>
#include <stdlib.h>
#include <string.h>

/*
    The following code was used to generate the following sinc lookup table.

    #include <limits.h>
    #include <math.h>
    #include <stdio.h>

    double findHannWeight(int N, double x) {
        return 0.5*(1.0 - cos(2*M_PI*x/N));
    }

    double findSincCoefficient(int N, double x) {
        double hannWindowWeight = findHannWeight(N, x);
        double sincWeight;

        x -= N/2.0;
        if (x > 1e-9 || x < -1e-9) {
            sincWeight = sin(M_PI*x)/(M_PI*x);
        } else {
            sincWeight = 1.0;
        }
        return hannWindowWeight*sincWeight;
    }

    int main() {
        double x;
        int i;
        int N = 12;

        for (i = 0, x = 0.0; x <= N; x += 0.02, i++) {
            printf("%u %d\n", i, (int)(SHRT_MAX*findSincCoefficient(N, x)));
        }
        return 0;
    }
*/

/* The number of points to use in the sinc FIR filter for resampling. */
#define SINC_FILTER_POINTS \
  12 /* I am not able to hear improvement with higher N. */
#define SINC_TABLE_SIZE 601

/* Lookup table for windowed sinc function of SINC_FILTER_POINTS points. */
static short sincTable[SINC_TABLE_SIZE] = {
    0,     0,     0,     0,     0,     0,     0,     -1,    -1,    -2,    -2,
    -3,    -4,    -6,    -7,    -9,    -10,   -12,   -14,   -17,   -19,   -21,
    -24,   -26,   -29,   -32,   -34,   -37,   -40,   -42,   -44,   -47,   -48,
    -50,   -51,   -52,   -53,   -53,   -53,   -52,   -50,   -48,   -46,   -43,
    -39,   -34,   -29,   -22,   -16,   -8,    0,     9,     19,    29,    41,
    53,    65,    79,    92,    107,   121,   137,   152,   168,   184,   200,
    215,   231,   247,   262,   276,   291,   304,   317,   328,   339,   348,
    357,   363,   369,   372,   374,   375,   373,   369,   363,   355,   345,
    332,   318,   300,   281,   259,   234,   208,   178,   147,   113,   77,
    39,    0,     -41,   -85,   -130,  -177,  -225,  -274,  -324,  -375,  -426,
    -478,  -530,  -581,  -632,  -682,  -731,  -779,  -825,  -870,  -912,  -951,
    -989,  -1023, -1053, -1080, -1104, -1123, -1138, -1149, -1154, -1155, -1151,
    -1141, -1125, -1105, -1078, -1046, -1007, -963,  -913,  -857,  -796,  -728,
    -655,  -576,  -492,  -403,  -309,  -210,  -107,  0,     111,   225,   342,
    462,   584,   708,   833,   958,   1084,  1209,  1333,  1455,  1575,  1693,
    1807,  1916,  2022,  2122,  2216,  2304,  2384,  2457,  2522,  2579,  2625,
    2663,  2689,  2706,  2711,  2705,  2687,  2657,  2614,  2559,  2491,  2411,
    2317,  2211,  2092,  1960,  1815,  1658,  1489,  1308,  1115,  912,   698,
    474,   241,   0,     -249,  -506,  -769,  -1037, -1310, -1586, -1864, -2144,
    -2424, -2703, -2980, -3254, -3523, -3787, -4043, -4291, -4529, -4757, -4972,
    -5174, -5360, -5531, -5685, -5819, -5935, -6029, -6101, -6150, -6175, -6175,
    -6149, -6096, -6015, -5905, -5767, -5599, -5401, -5172, -4912, -4621, -4298,
    -3944, -3558, -3141, -2693, -2214, -1705, -1166, -597,  0,     625,   1277,
    1955,  2658,  3386,  4135,  4906,  5697,  6506,  7332,  8173,  9027,  9893,
    10769, 11654, 12544, 13439, 14335, 15232, 16128, 17019, 17904, 18782, 19649,
    20504, 21345, 22170, 22977, 23763, 24527, 25268, 25982, 26669, 27327, 27953,
    28547, 29107, 29632, 30119, 30569, 30979, 31349, 31678, 31964, 32208, 32408,
    32565, 32677, 32744, 32767, 32744, 32677, 32565, 32408, 32208, 31964, 31678,
    31349, 30979, 30569, 30119, 29632, 29107, 28547, 27953, 27327, 26669, 25982,
    25268, 24527, 23763, 22977, 22170, 21345, 20504, 19649, 18782, 17904, 17019,
    16128, 15232, 14335, 13439, 12544, 11654, 10769, 9893,  9027,  8173,  7332,
    6506,  5697,  4906,  4135,  3386,  2658,  1955,  1277,  625,   0,     -597,
    -1166, -1705, -2214, -2693, -3141, -3558, -3944, -4298, -4621, -4912, -5172,
    -5401, -5599, -5767, -5905, -6015, -6096, -6149, -6175, -6175, -6150, -6101,
    -6029, -5935, -5819, -5685, -5531, -5360, -5174, -4972, -4757, -4529, -4291,
    -4043, -3787, -3523, -3254, -2980, -2703, -2424, -2144, -1864, -1586, -1310,
    -1037, -769,  -506,  -249,  0,     241,   474,   698,   912,   1115,  1308,
    1489,  1658,  1815,  1960,  2092,  2211,  2317,  2411,  2491,  2559,  2614,
    2657,  2687,  2705,  2711,  2706,  2689,  2663,  2625,  2579,  2522,  2457,
    2384,  2304,  2216,  2122,  2022,  1916,  1807,  1693,  1575,  1455,  1333,
    1209,  1084,  958,   833,   708,   584,   462,   342,   225,   111,   0,
    -107,  -210,  -309,  -403,  -492,  -576,  -655,  -728,  -796,  -857,  -913,
    -963,  -1007, -1046, -1078, -1105, -1125, -1141, -1151, -1155, -1154, -1149,
    -1138, -1123, -1104, -1080, -1053, -1023, -989,  -951,  -912,  -870,  -825,
    -779,  -731,  -682,  -632,  -581,  -530,  -478,  -426,  -375,  -324,  -274,
    -225,  -177,  -130,  -85,   -41,   0,     39,    77,    113,   147,   178,
    208,   234,   259,   281,   300,   318,   332,   345,   355,   363,   369,
    373,   375,   374,   372,   369,   363,   357,   348,   339,   328,   317,
    304,   291,   276,   262,   247,   231,   215,   200,   184,   168,   152,
    137,   121,   107,   92,    79,    65,    53,    41,    29,    19,    9,
    0,     -8,    -16,   -22,   -29,   -34,   -39,   -43,   -46,   -48,   -50,
    -52,   -53,   -53,   -53,   -52,   -51,   -50,   -48,   -47,   -44,   -42,
    -40,   -37,   -34,   -32,   -29,   -26,   -24,   -21,   -19,   -17,   -14,
    -12,   -10,   -9,    -7,    -6,    -4,    -3,    -2,    -2,    -1,    -1,
    0,     0,     0,     0,     0,     0,     0};

/* These functions allocate out of a static array rather than calling
   calloc/realloc/free if the NO_MALLOC flag is defined.  Otherwise, call
   calloc/realloc/free as usual.  This is useful for running on small
   microcontrollers. */
#ifndef SONIC_NO_MALLOC

/* Just call calloc. */
static void *sonicCalloc(int num, int size) {
  return calloc(num, size);
}

/* Just call realloc */
static void *sonicRealloc(void *p, int oldNum, int newNum, int size) {
  return realloc(p, newNum * size);
}

/* Just call free. */
static void sonicFree(void *p) {
  free(p);
}

#else

#ifndef SONIC_MAX_MEMORY
/* Large enough for speedup/slowdown at 8KHz, 16-bit mono samples/second. */
#define SONIC_MAX_MEMORY (16 * 1024)
#endif

/* This static buffer is used to hold data allocated for the sonicStream struct
   and its buffers.  There should never be more than one sonicStream in use at a
   time when using SONIC_NO_MALLOC mode.  Calls to realloc move the data to the
   end of memoryBuffer.  Calls to free reset the memory buffer to empty. */
static void*
    memoryBufferAligned[(SONIC_MAX_MEMORY + sizeof(void) - 1) / sizeof(void*)];
static unsigned char* memoryBuffer = (unsigned char*)memoryBufferAligned;
static int memoryBufferPos = 0;

/* Allocate elements from a static memory buffer. */
static void *sonicCalloc(int num, int size) {
  int len = num * size;

  if (memoryBufferPos + len > SONIC_MAX_MEMORY) {
    return 0;
  }
  unsigned char *p = memoryBuffer + memoryBufferPos;
  memoryBufferPos += len;
  memset(p, 0, len);
  return p;
}

/* Preferably, SONIC_MAX_MEMORY has been set large enough that this is never
 * called. */
static void *sonicRealloc(void *p, int oldNum, int newNum, int size) {
  if (newNum <= oldNum) {
    return p;
  }
  void *newBuffer = sonicCalloc(newNum, size);
  if (newBuffer == NULL) {
    return NULL;
  }
  memcpy(newBuffer, p, oldNum * size);
  return newBuffer;
}

/* Reset memoryBufferPos to 0.  We asssume all data is freed at the same time. */
static void sonicFree(void *p) {
  memoryBufferPos = 0;
}

#endif

struct sonicStreamStruct {
#ifdef SONIC_SPECTROGRAM
  sonicSpectrogram spectrogram;
#endif  /* SONIC_SPECTROGRAM */
  short* inputBuffer;
  short* outputBuffer;
  short* pitchBuffer;
  short* downSampleBuffer;
  void* userData;
  float speed;
  float volume;
  float pitch;
  float rate;
  /* The point of the following 3 new variables is to gracefully handle rapidly
     changing input speed.

     samplePeriod is just 1.0/sampleRate.  It is used in accumulating
     inputPlayTime, which is how long we expect the total time should be to play
     the current input samples in the input buffer.  timeError keeps track of
     the error in play time created when playing < 2.0X speed, where we either
     insert or delete a whole pitch period.  This can cause the output generated
     from the input to be off in play time by up to a pitch period.  timeError
     replaces PICOLA's concept of the number of samples to play unmodified after
     a pitch period insertion or deletion.  If speeding up, and the error is >=
     0.0, then remove a pitch period, and play samples unmodified until
     timeError is >= 0 again.  If slowing down, and the error is <= 0.0,
     then add a pitch period, and play samples unmodified until timeError is <=
     0 again. */
  float samplePeriod;  /* How long each output sample takes to play. */
  /* How long we expect the entire input buffer to take to play. */
  float inputPlayTime;
  /* The difference in when the latest output sample was played vs when we wanted.  */
  float timeError;
  int oldRatePosition;
  int newRatePosition;
  int quality;
  int numChannels;
  int inputBufferSize;
  int pitchBufferSize;
  int outputBufferSize;
  int numInputSamples;
  int numOutputSamples;
  int numPitchSamples;
  int minPeriod;
  int maxPeriod;
  int maxRequired;
  int remainingInputToCopy;
  int sampleRate;
  int prevPeriod;
  int prevMinDiff;
};

/* Attach user data to the stream. */
void sonicSetUserData(sonicStream stream, void *userData) {
  stream->userData = userData;
}

/* Retrieve user data attached to the stream. */
void *sonicGetUserData(sonicStream stream) {
  return stream->userData;
}

#ifdef SONIC_SPECTROGRAM

/* Compute a spectrogram on the fly. */
void sonicComputeSpectrogram(sonicStream stream) {
  stream->spectrogram = sonicCreateSpectrogram(stream->sampleRate);
  /* Force changeSpeed to be called to compute the spectrogram. */
  sonicSetSpeed(stream, 2.0);
}

/* Get the spectrogram. */
sonicSpectrogram sonicGetSpectrogram(sonicStream stream) {
  return stream->spectrogram;
}

#endif

/* Scale the samples by the factor. */
static void scaleSamples(short* samples, int numSamples, float volume) {
  /* This is 24-bit integer and 8-bit fraction fixed-point representation. */
  int fixedPointVolume = volume * 256.0f;
  int value;

  while (numSamples--) {
    value = (*samples * fixedPointVolume) >> 8;
    if (value > 32767) {
      value = 32767;
    } else if (value < -32767) {
      value = -32767;
    }
    *samples++ = value;
  }
}

/* Get the speed of the stream. */
float sonicGetSpeed(sonicStream stream) { return stream->speed; }

/* Set the speed of the stream. */
void sonicSetSpeed(sonicStream stream, float speed) { stream->speed = speed; }

/* Get the pitch of the stream. */
float sonicGetPitch(sonicStream stream) { return stream->pitch; }

/* Set the pitch of the stream. */
void sonicSetPitch(sonicStream stream, float pitch) { stream->pitch = pitch; }

/* Get the rate of the stream. */
float sonicGetRate(sonicStream stream) { return stream->rate; }

/* Set the playback rate of the stream. This scales pitch and speed at the same
   time. */
void sonicSetRate(sonicStream stream, float rate) {
  stream->rate = rate;

  stream->oldRatePosition = 0;
  stream->newRatePosition = 0;
}

/* DEPRECATED.  Get the vocal chord pitch setting. */
int sonicGetChordPitch(sonicStream stream) {
  return 0;
}

/* DEPRECATED. Set the vocal chord mode for pitch computation.  Default is off. */
void sonicSetChordPitch(sonicStream stream, int useChordPitch) {
}

/* Get the quality setting. */
int sonicGetQuality(sonicStream stream) { return stream->quality; }

/* Set the "quality".  Default 0 is virtually as good as 1, but very much
   faster. */
void sonicSetQuality(sonicStream stream, int quality) {
  stream->quality = quality;
}

/* Get the scaling factor of the stream. */
float sonicGetVolume(sonicStream stream) { return stream->volume; }

/* Set the scaling factor of the stream. */
void sonicSetVolume(sonicStream stream, float volume) {
  stream->volume = volume;
}

/* Free stream buffers. */
static void freeStreamBuffers(sonicStream stream) {
  if (stream->inputBuffer != NULL) {
    sonicFree(stream->inputBuffer);
  }
  if (stream->outputBuffer != NULL) {
    sonicFree(stream->outputBuffer);
  }
  if (stream->pitchBuffer != NULL) {
    sonicFree(stream->pitchBuffer);
  }
  if (stream->downSampleBuffer != NULL) {
    sonicFree(stream->downSampleBuffer);
  }
}

/* Destroy the sonic stream. */
void sonicDestroyStream(sonicStream stream) {
#ifdef SONIC_SPECTROGRAM
  if (stream->spectrogram != NULL) {
    sonicDestroySpectrogram(stream->spectrogram);
  }
#endif  /* SONIC_SPECTROGRAM */
  freeStreamBuffers(stream);
  sonicFree(stream);
}

/* Compute the number of samples to skip to down-sample the input. */
static int computeSkip(sonicStream stream) {
  int skip = 1;
  if (stream->sampleRate > SONIC_AMDF_FREQ && stream->quality == 0) {
    skip = stream->sampleRate / SONIC_AMDF_FREQ;
  }
  return skip;
}

/* Allocate stream buffers. */
static int allocateStreamBuffers(sonicStream stream, int sampleRate,
                                 int numChannels) {
  int minPeriod = sampleRate / SONIC_MAX_PITCH;
  int maxPeriod = sampleRate / SONIC_MIN_PITCH;
  int maxRequired = 2 * maxPeriod;
  int skip = computeSkip(stream);

  /* Allocate 25% more than needed so we hopefully won't grow. */
  stream->inputBufferSize = maxRequired + (maxRequired >> 2);;
  stream->inputBuffer =
      (short*)sonicCalloc(stream->inputBufferSize, sizeof(short) * numChannels);
  if (stream->inputBuffer == NULL) {
    sonicDestroyStream(stream);
    return 0;
  }
  /* Allocate 25% more than needed so we hopefully won't grow. */
  stream->outputBufferSize = maxRequired + (maxRequired >> 2);
  stream->outputBuffer =
      (short*)sonicCalloc(stream->outputBufferSize, sizeof(short) * numChannels);
  if (stream->outputBuffer == NULL) {
    sonicDestroyStream(stream);
    return 0;
  }
  /* Allocate 25% more than needed so we hopefully won't grow. */
  stream->pitchBufferSize = maxRequired + (maxRequired >> 2);
  stream->pitchBuffer =
      (short*)sonicCalloc(maxRequired, sizeof(short) * numChannels);
  if (stream->pitchBuffer == NULL) {
    sonicDestroyStream(stream);
    return 0;
  }
  int downSampleBufferSize = (maxRequired + skip - 1)/ skip;
  stream->downSampleBuffer = (short*)sonicCalloc(downSampleBufferSize, sizeof(short));
  if (stream->downSampleBuffer == NULL) {
    sonicDestroyStream(stream);
    return 0;
  }
  stream->sampleRate = sampleRate;
  stream->samplePeriod = 1.0 / sampleRate;
  stream->numChannels = numChannels;
  stream->oldRatePosition = 0;
  stream->newRatePosition = 0;
  stream->minPeriod = minPeriod;
  stream->maxPeriod = maxPeriod;
  stream->maxRequired = maxRequired;
  stream->prevPeriod = 0;
  return 1;
}

/* Create a sonic stream.  Return NULL only if we are out of memory and cannot
   allocate the stream. */
sonicStream sonicCreateStream(int sampleRate, int numChannels) {
  sonicStream stream = (sonicStream)sonicCalloc(
      1, sizeof(struct sonicStreamStruct));

  if (stream == NULL) {
    return NULL;
  }
  if (!allocateStreamBuffers(stream, sampleRate, numChannels)) {
    return NULL;
  }
  stream->speed = 1.0f;
  stream->pitch = 1.0f;
  stream->volume = 1.0f;
  stream->rate = 1.0f;
  stream->oldRatePosition = 0;
  stream->newRatePosition = 0;
  stream->quality = 0;
  return stream;
}

/* Get the sample rate of the stream. */
int sonicGetSampleRate(sonicStream stream) { return stream->sampleRate; }

/* Set the sample rate of the stream.  This will cause samples buffered in the
   stream to be lost. */
void sonicSetSampleRate(sonicStream stream, int sampleRate) {
  freeStreamBuffers(stream);
  allocateStreamBuffers(stream, sampleRate, stream->numChannels);
}

/* Get the number of channels. */
int sonicGetNumChannels(sonicStream stream) { return stream->numChannels; }

/* Set the num channels of the stream.  This will cause samples buffered in the
   stream to be lost. */
void sonicSetNumChannels(sonicStream stream, int numChannels) {
  freeStreamBuffers(stream);
  allocateStreamBuffers(stream, stream->sampleRate, numChannels);
}

/* Enlarge the output buffer if needed. */
static int enlargeOutputBufferIfNeeded(sonicStream stream, int numSamples) {
  int outputBufferSize = stream->outputBufferSize;

  if (stream->numOutputSamples + numSamples > outputBufferSize) {
    stream->outputBufferSize += (outputBufferSize >> 1) + numSamples;
    stream->outputBuffer = (short*)sonicRealloc(
        stream->outputBuffer,
        outputBufferSize,
        stream->outputBufferSize,
        sizeof(short) * stream->numChannels);
    if (stream->outputBuffer == NULL) {
      return 0;
    }
  }
  return 1;
}

/* Enlarge the input buffer if needed. */
static int enlargeInputBufferIfNeeded(sonicStream stream, int numSamples) {
  int inputBufferSize = stream->inputBufferSize;

  if (stream->numInputSamples + numSamples > inputBufferSize) {
    stream->inputBufferSize += (inputBufferSize >> 1) + numSamples;
    stream->inputBuffer = (short*)sonicRealloc(
        stream->inputBuffer,
        inputBufferSize,
        stream->inputBufferSize,
        sizeof(short) * stream->numChannels);
    if (stream->inputBuffer == NULL) {
      return 0;
    }
  }
  return 1;
}

/* Update stream->numInputSamples, and update stream->inputPlayTime.  Call this
   whenever adding samples to the input buffer, to keep track of total expected
   input play time accounting. */
static void updateNumInputSamples(sonicStream stream, int numSamples) {
  float speed = stream->speed / stream->pitch;

  stream->numInputSamples += numSamples;
  stream->inputPlayTime += numSamples * stream->samplePeriod / speed;
}

/* Add the input samples to the input buffer. */
static int addFloatSamplesToInputBuffer(sonicStream stream, const float* samples,
                                        int numSamples) {
  short* buffer;
  int count = numSamples * stream->numChannels;

  if (numSamples == 0) {
    return 1;
  }
  if (!enlargeInputBufferIfNeeded(stream, numSamples)) {
    return 0;
  }
  buffer = stream->inputBuffer + stream->numInputSamples * stream->numChannels;
  while (count--) {
    *buffer++ = (*samples++) * 32767.0f;
  }
  updateNumInputSamples(stream, numSamples);
  return 1;
}

/* Add the input samples to the input buffer. */
static int addShortSamplesToInputBuffer(sonicStream stream, const short* samples,
                                        int numSamples) {
  if (numSamples == 0) {
    return 1;
  }
  if (!enlargeInputBufferIfNeeded(stream, numSamples)) {
    return 0;
  }
  memcpy(stream->inputBuffer + stream->numInputSamples * stream->numChannels,
         samples, numSamples * sizeof(short) * stream->numChannels);
  updateNumInputSamples(stream, numSamples);
  return 1;
}

/* Add the input samples to the input buffer. */
static int addUnsignedCharSamplesToInputBuffer(sonicStream stream,
                                               const unsigned char* samples,
                                               int numSamples) {
  short* buffer;
  int count = numSamples * stream->numChannels;

  if (numSamples == 0) {
    return 1;
  }
  if (!enlargeInputBufferIfNeeded(stream, numSamples)) {
    return 0;
  }
  buffer = stream->inputBuffer + stream->numInputSamples * stream->numChannels;
  while (count--) {
    *buffer++ = (*samples++ - 128) << 8;
  }
  updateNumInputSamples(stream, numSamples);
  return 1;
}

/* Remove input samples that we have already processed. */
static void removeInputSamples(sonicStream stream, int position) {
  int remainingSamples = stream->numInputSamples - position;

  if (remainingSamples > 0) {
    memmove(stream->inputBuffer,
            stream->inputBuffer + position * stream->numChannels,
            remainingSamples * sizeof(short) * stream->numChannels);
  }
  /* If we play 3/4ths of the samples, then the expected play time of the
     remaining samples is 1/4th of the original expected play time. */
  stream->inputPlayTime =
      (stream->inputPlayTime * remainingSamples) / stream->numInputSamples;
  stream->numInputSamples = remainingSamples;
}

/* Copy from the input buffer to the output buffer, and remove the samples from
   the input buffer. */
static int copyInputToOutput(sonicStream stream, int numSamples) {
  if (!enlargeOutputBufferIfNeeded(stream, numSamples)) {
    return 0;
  }
  memcpy(stream->outputBuffer + stream->numOutputSamples * stream->numChannels,
         stream->inputBuffer, numSamples * sizeof(short) * stream->numChannels);
  stream->numOutputSamples += numSamples;
  removeInputSamples(stream, numSamples);
  return 1;
}

/* Copy from samples to the output buffer */
static int copyToOutput(sonicStream stream, short* samples, int numSamples) {
  if (!enlargeOutputBufferIfNeeded(stream, numSamples)) {
    return 0;
  }
  memcpy(stream->outputBuffer + stream->numOutputSamples * stream->numChannels,
         samples, numSamples * sizeof(short) * stream->numChannels);
  stream->numOutputSamples += numSamples;
  return 1;
}

/* Read data out of the stream.  Sometimes no data will be available, and zero
   is returned, which is not an error condition. */
int sonicReadFloatFromStream(sonicStream stream, float* samples,
                             int maxSamples) {
  int numSamples = stream->numOutputSamples;
  int remainingSamples = 0;
  short* buffer;
  int count;

  if (numSamples == 0) {
    return 0;
  }
  if (numSamples > maxSamples) {
    remainingSamples = numSamples - maxSamples;
    numSamples = maxSamples;
  }
  buffer = stream->outputBuffer;
  count = numSamples * stream->numChannels;
  while (count--) {
    *samples++ = (*buffer++) / 32767.0f;
  }
  if (remainingSamples > 0) {
    memmove(stream->outputBuffer,
            stream->outputBuffer + numSamples * stream->numChannels,
            remainingSamples * sizeof(short) * stream->numChannels);
  }
  stream->numOutputSamples = remainingSamples;
  return numSamples;
}

/* Read short data out of the stream.  Sometimes no data will be available, and
   zero is returned, which is not an error condition. */
int sonicReadShortFromStream(sonicStream stream, short* samples,
                             int maxSamples) {
  int numSamples = stream->numOutputSamples;
  int remainingSamples = 0;

  if (numSamples == 0) {
    return 0;
  }
  if (numSamples > maxSamples) {
    remainingSamples = numSamples - maxSamples;
    numSamples = maxSamples;
  }
  memcpy(samples, stream->outputBuffer,
         numSamples * sizeof(short) * stream->numChannels);
  if (remainingSamples > 0) {
    memmove(stream->outputBuffer,
            stream->outputBuffer + numSamples * stream->numChannels,
            remainingSamples * sizeof(short) * stream->numChannels);
  }
  stream->numOutputSamples = remainingSamples;
  return numSamples;
}

/* Read unsigned char data out of the stream.  Sometimes no data will be
   available, and zero is returned, which is not an error condition. */
int sonicReadUnsignedCharFromStream(sonicStream stream, unsigned char* samples,
                                    int maxSamples) {
  int numSamples = stream->numOutputSamples;
  int remainingSamples = 0;
  short* buffer;
  int count;

  if (numSamples == 0) {
    return 0;
  }
  if (numSamples > maxSamples) {
    remainingSamples = numSamples - maxSamples;
    numSamples = maxSamples;
  }
  buffer = stream->outputBuffer;
  count = numSamples * stream->numChannels;
  while (count--) {
    *samples++ = (char)((*buffer++) >> 8) + 128;
  }
  if (remainingSamples > 0) {
    memmove(stream->outputBuffer,
            stream->outputBuffer + numSamples * stream->numChannels,
            remainingSamples * sizeof(short) * stream->numChannels);
  }
  stream->numOutputSamples = remainingSamples;
  return numSamples;
}

/* Force the sonic stream to generate output using whatever data it currently
   has.  No extra delay will be added to the output, but flushing in the middle
   of words could introduce distortion. */
int sonicFlushStream(sonicStream stream) {
  int maxRequired = stream->maxRequired;
  int remainingSamples = stream->numInputSamples;
  float speed = stream->speed / stream->pitch;
  float rate = stream->rate * stream->pitch;
  int expectedOutputSamples =
      stream->numOutputSamples +
      (int)((remainingSamples / speed + stream->numPitchSamples) / rate + 0.5f);

  /* Add enough silence to flush both input and pitch buffers. */
  if (!enlargeInputBufferIfNeeded(stream, remainingSamples + 2 * maxRequired)) {
    return 0;
  }
  memset(stream->inputBuffer + remainingSamples * stream->numChannels, 0,
         2 * maxRequired * sizeof(short) * stream->numChannels);
  stream->numInputSamples += 2 * maxRequired;
  if (!sonicWriteShortToStream(stream, NULL, 0)) {
    return 0;
  }
  /* Throw away any extra samples we generated due to the silence we added */
  if (stream->numOutputSamples > expectedOutputSamples) {
    stream->numOutputSamples = expectedOutputSamples;
  }
  /* Empty input and pitch buffers */
  stream->numInputSamples = 0;
  stream->inputPlayTime = 0.0f;
  stream->timeError = 0.0f;
  stream->numPitchSamples = 0;
  return 1;
}

/* Return the number of samples in the output buffer */
int sonicSamplesAvailable(sonicStream stream) {
  return stream->numOutputSamples;
}

/* If skip is greater than one, average skip samples together and write them to
   the down-sample buffer.  If numChannels is greater than one, mix the channels
   together as we down sample. */
static void downSampleInput(sonicStream stream, short* samples, int skip) {
  int numSamples = stream->maxRequired / skip;
  int samplesPerValue = stream->numChannels * skip;
  int i, j;
  int value;
  short* downSamples = stream->downSampleBuffer;

  for (i = 0; i < numSamples; i++) {
    value = 0;
    for (j = 0; j < samplesPerValue; j++) {
      value += *samples++;
    }
    value /= samplesPerValue;
    *downSamples++ = value;
  }
}

/* Find the best frequency match in the range, and given a sample skip multiple.
   For now, just find the pitch of the first channel. */
static int findPitchPeriodInRange(short* samples, int minPeriod, int maxPeriod,
                                  int* retMinDiff, int* retMaxDiff) {
  int period, bestPeriod = 0, worstPeriod = 255;
  short* s;
  short* p;
  short sVal, pVal;
  unsigned long diff, minDiff = 1, maxDiff = 0;
  int i;

  for (period = minPeriod; period <= maxPeriod; period++) {
    diff = 0;
    s = samples;
    p = samples + period;
    for (i = 0; i < period; i++) {
      sVal = *s++;
      pVal = *p++;
      diff += sVal >= pVal ? (unsigned short)(sVal - pVal)
                           : (unsigned short)(pVal - sVal);
    }
    /* Note that the highest number of samples we add into diff will be less
       than 256, since we skip samples.  Thus, diff is a 24 bit number, and
       we can safely multiply by numSamples without overflow */
    if (bestPeriod == 0 || diff * bestPeriod < minDiff * period) {
      minDiff = diff;
      bestPeriod = period;
    }
    if (diff * worstPeriod > maxDiff * period) {
      maxDiff = diff;
      worstPeriod = period;
    }
  }
  *retMinDiff = minDiff / bestPeriod;
  *retMaxDiff = maxDiff / worstPeriod;
  return bestPeriod;
}

/* At abrupt ends of voiced words, we can have pitch periods that are better
   approximated by the previous pitch period estimate.  Try to detect this case.
 */
static int prevPeriodBetter(sonicStream stream, int minDiff,
                            int maxDiff, int preferNewPeriod) {
  if (minDiff == 0 || stream->prevPeriod == 0) {
    return 0;
  }
  if (preferNewPeriod) {
    if (maxDiff > minDiff * 3) {
      /* Got a reasonable match this period */
      return 0;
    }
    if (minDiff * 2 <= stream->prevMinDiff * 3) {
      /* Mismatch is not that much greater this period */
      return 0;
    }
  } else {
    if (minDiff <= stream->prevMinDiff) {
      return 0;
    }
  }
  return 1;
}

/* Find the pitch period.  This is a critical step, and we may have to try
   multiple ways to get a good answer.  This version uses Average Magnitude
   Difference Function (AMDF).  To improve speed, we down sample by an integer
   factor get in the 11KHz range, and then do it again with a narrower
   frequency range without down sampling */
static int findPitchPeriod(sonicStream stream, short* samples,
                           int preferNewPeriod) {
  int minPeriod = stream->minPeriod;
  int maxPeriod = stream->maxPeriod;
  int minDiff, maxDiff, retPeriod;
  int skip = computeSkip(stream);
  int period;

  if (stream->numChannels == 1 && skip == 1) {
    period = findPitchPeriodInRange(samples, minPeriod, maxPeriod, &minDiff,
                                    &maxDiff);
  } else {
    downSampleInput(stream, samples, skip);
    period = findPitchPeriodInRange(stream->downSampleBuffer, minPeriod / skip,
                                    maxPeriod / skip, &minDiff, &maxDiff);
    if (skip != 1) {
      period *= skip;
      minPeriod = period - (skip << 2);
      maxPeriod = period + (skip << 2);
      if (minPeriod < stream->minPeriod) {
        minPeriod = stream->minPeriod;
      }
      if (maxPeriod > stream->maxPeriod) {
        maxPeriod = stream->maxPeriod;
      }
      if (stream->numChannels == 1) {
        period = findPitchPeriodInRange(samples, minPeriod, maxPeriod, &minDiff,
                                        &maxDiff);
      } else {
        downSampleInput(stream, samples, 1);
        period = findPitchPeriodInRange(stream->downSampleBuffer, minPeriod,
                                        maxPeriod, &minDiff, &maxDiff);
      }
    }
  }
  if (prevPeriodBetter(stream, minDiff, maxDiff, preferNewPeriod)) {
    retPeriod = stream->prevPeriod;
  } else {
    retPeriod = period;
  }
  stream->prevMinDiff = minDiff;
  stream->prevPeriod = period;
  return retPeriod;
}

/* Overlap two sound segments, ramp the volume of one down, while ramping the
   other one from zero up, and add them, storing the result at the output. */
static void overlapAdd(int numSamples, int numChannels, short* out,
                       short* rampDown, short* rampUp) {
  short* o;
  short* u;
  short* d;
  int i, t;

  for (i = 0; i < numChannels; i++) {
    o = out + i;
    u = rampUp + i;
    d = rampDown + i;
    for (t = 0; t < numSamples; t++) {
#ifdef SONIC_USE_SIN
      float ratio = sin(t * M_PI / (2 * numSamples));
      *o = *d * (1.0f - ratio) + *u * ratio;
#else
      *o = (*d * (numSamples - t) + *u * t) / numSamples;
#endif
      o += numChannels;
      d += numChannels;
      u += numChannels;
    }
  }
}

/* Just move the new samples in the output buffer to the pitch buffer */
static int moveNewSamplesToPitchBuffer(sonicStream stream,
                                       int originalNumOutputSamples) {
  int numSamples = stream->numOutputSamples - originalNumOutputSamples;
  int numChannels = stream->numChannels;

  if (stream->numPitchSamples + numSamples > stream->pitchBufferSize) {
    int pitchBufferSize = stream->pitchBufferSize;
    stream->pitchBufferSize += (pitchBufferSize >> 1) + numSamples;
    stream->pitchBuffer = (short*)sonicRealloc(
        stream->pitchBuffer,
        pitchBufferSize,
        stream->pitchBufferSize,
        sizeof(short) * numChannels);
  }
  memcpy(stream->pitchBuffer + stream->numPitchSamples * numChannels,
         stream->outputBuffer + originalNumOutputSamples * numChannels,
         numSamples * sizeof(short) * numChannels);
  stream->numOutputSamples = originalNumOutputSamples;
  stream->numPitchSamples += numSamples;
  return 1;
}

/* Remove processed samples from the pitch buffer. */
static void removePitchSamples(sonicStream stream, int numSamples) {
  int numChannels = stream->numChannels;
  short* source = stream->pitchBuffer + numSamples * numChannels;

  if (numSamples == 0) {
    return;
  }
  if (numSamples != stream->numPitchSamples) {
    memmove(
        stream->pitchBuffer, source,
        (stream->numPitchSamples - numSamples) * sizeof(short) * numChannels);
  }
  stream->numPitchSamples -= numSamples;
}

/* Approximate the sinc function times a Hann window from the sinc table. */
static int findSincCoefficient(int i, int ratio, int width) {
  int lobePoints = (SINC_TABLE_SIZE - 1) / SINC_FILTER_POINTS;
  int left = i * lobePoints + (ratio * lobePoints) / width;
  int right = left + 1;
  int position = i * lobePoints * width + ratio * lobePoints - left * width;
  int leftVal = sincTable[left];
  int rightVal = sincTable[right];

  return ((leftVal * (width - position) + rightVal * position) << 1) / width;
}

/* Return 1 if value >= 0, else -1.  This represents the sign of value. */
static int getSign(int value) { return value >= 0 ? 1 : -1; }

/* Interpolate the new output sample. */
static short interpolate(sonicStream stream, short* in, int oldSampleRate,
                         int newSampleRate) {
  /* Compute N-point sinc FIR-filter here.  Clip rather than overflow. */
  int i;
  int total = 0;
  int position = stream->newRatePosition * oldSampleRate;
  int leftPosition = stream->oldRatePosition * newSampleRate;
  int rightPosition = (stream->oldRatePosition + 1) * newSampleRate;
  int ratio = rightPosition - position - 1;
  int width = rightPosition - leftPosition;
  int weight, value;
  int oldSign;
  int overflowCount = 0;

  for (i = 0; i < SINC_FILTER_POINTS; i++) {
    weight = findSincCoefficient(i, ratio, width);
    value = in[i * stream->numChannels] * weight;
    oldSign = getSign(total);
    total += value;
    if (oldSign != getSign(total) && getSign(value) == oldSign) {
      /* We must have overflowed.  This can happen with a sinc filter. */
      overflowCount += oldSign;
    }
  }
  /* It is better to clip than to wrap if there was a overflow. */
  if (overflowCount > 0) {
    return SHRT_MAX;
  } else if (overflowCount < 0) {
    return SHRT_MIN;
  }
  return total >> 16;
}

/* Change the rate.  Interpolate with a sinc FIR filter using a Hann window. */
static int adjustRate(sonicStream stream, float rate,
                      int originalNumOutputSamples) {
  int newSampleRate = stream->sampleRate / rate;
  int oldSampleRate = stream->sampleRate;
  int numChannels = stream->numChannels;
  int position;
  short *in, *out;
  int i;
  int N = SINC_FILTER_POINTS;

  /* Set these values to help with the integer math */
  while (newSampleRate > (1 << 14) || oldSampleRate > (1 << 14)) {
    newSampleRate >>= 1;
    oldSampleRate >>= 1;
  }
  if (stream->numOutputSamples == originalNumOutputSamples) {
    return 1;
  }
  if (!moveNewSamplesToPitchBuffer(stream, originalNumOutputSamples)) {
    return 0;
  }
  /* Leave at least N pitch sample in the buffer */
  for (position = 0; position < stream->numPitchSamples - N; position++) {
    while ((stream->oldRatePosition + 1) * newSampleRate >
           stream->newRatePosition * oldSampleRate) {
      if (!enlargeOutputBufferIfNeeded(stream, 1)) {
        return 0;
      }
      out = stream->outputBuffer + stream->numOutputSamples * numChannels;
      in = stream->pitchBuffer + position * numChannels;
      for (i = 0; i < numChannels; i++) {
        *out++ = interpolate(stream, in, oldSampleRate, newSampleRate);
        in++;
      }
      stream->newRatePosition++;
      stream->numOutputSamples++;
    }
    stream->oldRatePosition++;
    if (stream->oldRatePosition == oldSampleRate) {
      stream->oldRatePosition = 0;
      stream->newRatePosition = 0;
    }
  }
  removePitchSamples(stream, position);
  return 1;
}

/* Skip over a pitch period.  Return the number of output samples. */
static int skipPitchPeriod(sonicStream stream, short* samples, float speed,
                           int period) {
  long newSamples;
  int numChannels = stream->numChannels;

  if (speed >= 2.0f) {
    /* For speeds >= 2.0, we skip over a portion of each pitch period rather
       than dropping whole pitch periods. */
    newSamples = period / (speed - 1.0f);
  } else {
    newSamples = period;
  }
  if (!enlargeOutputBufferIfNeeded(stream, newSamples)) {
    return 0;
  }
  overlapAdd(newSamples, numChannels,
             stream->outputBuffer + stream->numOutputSamples * numChannels,
             samples, samples + period * numChannels);
  stream->numOutputSamples += newSamples;
  return newSamples;
}

/* Insert a pitch period, and determine how much input to copy directly. */
static int insertPitchPeriod(sonicStream stream, short* samples, float speed,
                             int period) {
  long newSamples;
  short* out;
  int numChannels = stream->numChannels;

  if (speed <= 0.5f) {
    newSamples = period * speed / (1.0f - speed);
  } else {
    newSamples = period;
  }
  if (!enlargeOutputBufferIfNeeded(stream, period + newSamples)) {
    return 0;
  }
  out = stream->outputBuffer + stream->numOutputSamples * numChannels;
  memcpy(out, samples, period * sizeof(short) * numChannels);
  out =
      stream->outputBuffer + (stream->numOutputSamples + period) * numChannels;
  overlapAdd(newSamples, numChannels, out, samples + period * numChannels,
             samples);
  stream->numOutputSamples += period + newSamples;
  return newSamples;
}

/* PICOLA copies input to output until the total output samples == consumed
   input samples * speed. */
static int copyUnmodifiedSamples(sonicStream stream, short* samples,
                                 float speed, int position, int* newSamples) {
  int availableSamples = stream->numInputSamples - position;
  float inputToCopyFloat =
      1 - stream->timeError * speed / (stream->samplePeriod * (speed - 1.0));

  *newSamples = inputToCopyFloat > availableSamples ? availableSamples
                                                    : (int)inputToCopyFloat;
  if (!copyToOutput(stream, samples, *newSamples)) {
    return 0;
  }
  stream->timeError +=
      *newSamples * stream->samplePeriod * (speed - 1.0) / speed;
  return 1;
}

/* Resample as many pitch periods as we have buffered on the input.  Return 0 if
   we fail to resize an input or output buffer. */
static int changeSpeed(sonicStream stream, float speed) {
  short* samples;
  int numSamples = stream->numInputSamples;
  int position = 0, period, newSamples;
  int maxRequired = stream->maxRequired;

  if (stream->numInputSamples < maxRequired) {
    return 1;
  }
  do {
    samples = stream->inputBuffer + position * stream->numChannels;
    if ((speed > 1.0f && speed < 2.0f && stream->timeError < 0.0f) ||
        (speed < 1.0f && speed > 0.5f && stream->timeError > 0.0f)) {
      /* Deal with the case where PICOLA is still copying input samples to
         output unmodified, */
      if (!copyUnmodifiedSamples(stream, samples, speed, position,
                                 &newSamples)) {
        return 0;
      }
      position += newSamples;
    } else {
      /* We are in the remaining cases, either inserting/removing a pitch period
         for speed < 2.0X, or a portion of one for speed >= 2.0X. */
      period = findPitchPeriod(stream, samples, 1);
#ifdef SONIC_SPECTROGRAM
      if (stream->spectrogram != NULL) {
        sonicAddPitchPeriodToSpectrogram(stream->spectrogram, samples, period,
                                         stream->numChannels);
        newSamples = period;
        position += period;
      } else
#endif /* SONIC_SPECTROGRAM */
        if (speed > 1.0) {
          newSamples = skipPitchPeriod(stream, samples, speed, period);
          position += period + newSamples;
          if (speed < 2.0) {
            stream->timeError += newSamples * stream->samplePeriod -
                                 (period + newSamples) * stream->inputPlayTime /
                                     stream->numInputSamples;
          }
        } else {
          newSamples = insertPitchPeriod(stream, samples, speed, period);
          position += newSamples;
          if (speed > 0.5) {
            stream->timeError +=
                (period + newSamples) * stream->samplePeriod -
                newSamples * stream->inputPlayTime / stream->numInputSamples;
          }
        }
      if (newSamples == 0) {
        return 0; /* Failed to resize output buffer */
      }
    }
  } while (position + maxRequired <= numSamples);
  removeInputSamples(stream, position);
  return 1;
}

/* Resample as many pitch periods as we have buffered on the input.  Return 0 if
   we fail to resize an input or output buffer.  Also scale the output by the
   volume. */
static int processStreamInput(sonicStream stream) {
  int originalNumOutputSamples = stream->numOutputSamples;
  float rate = stream->rate * stream->pitch;
  float localSpeed;

  if (stream->numInputSamples == 0) {
    return 1;
  }
  localSpeed =
      stream->numInputSamples * stream->samplePeriod / stream->inputPlayTime;
  if (localSpeed > 1.00001 || localSpeed < 0.99999) {
    changeSpeed(stream, localSpeed);
  } else {
    if (!copyInputToOutput(stream, stream->numInputSamples)) {
      return 0;
    }
  }
  if (rate != 1.0f) {
    if (!adjustRate(stream, rate, originalNumOutputSamples)) {
      return 0;
    }
  }
  if (stream->volume != 1.0f) {
    /* Adjust output volume. */
    scaleSamples(
        stream->outputBuffer + originalNumOutputSamples * stream->numChannels,
        (stream->numOutputSamples - originalNumOutputSamples) *
            stream->numChannels,
        stream->volume);
  }
  return 1;
}

/* Write floating point data to the input buffer and process it. */
int sonicWriteFloatToStream(sonicStream stream, const float* samples,
                            int numSamples) {
  if (!addFloatSamplesToInputBuffer(stream, samples, numSamples)) {
    return 0;
  }
  return processStreamInput(stream);
}

/* Simple wrapper around sonicWriteFloatToStream that does the short to float
   conversion for you. */
int sonicWriteShortToStream(sonicStream stream, const short* samples,
                            int numSamples) {
  if (!addShortSamplesToInputBuffer(stream, samples, numSamples)) {
    return 0;
  }
  return processStreamInput(stream);
}

/* Simple wrapper around sonicWriteFloatToStream that does the unsigned char to
   float conversion for you. */
int sonicWriteUnsignedCharToStream(sonicStream stream, const unsigned char* samples,
                                   int numSamples) {
  if (!addUnsignedCharSamplesToInputBuffer(stream, samples, numSamples)) {
    return 0;
  }
  return processStreamInput(stream);
}

/* This is a non-stream oriented interface to just change the speed of a sound
 * sample */
int sonicChangeFloatSpeed(float* samples, int numSamples, float speed,
                          float pitch, float rate, float volume,
                          int useChordPitch, int sampleRate, int numChannels) {
  sonicStream stream = sonicCreateStream(sampleRate, numChannels);

  sonicSetSpeed(stream, speed);
  sonicSetPitch(stream, pitch);
  sonicSetRate(stream, rate);
  sonicSetVolume(stream, volume);
  sonicWriteFloatToStream(stream, samples, numSamples);
  sonicFlushStream(stream);
  numSamples = sonicSamplesAvailable(stream);
  sonicReadFloatFromStream(stream, samples, numSamples);
  sonicDestroyStream(stream);
  return numSamples;
}

/* This is a non-stream oriented interface to just change the speed of a sound
 * sample */
int sonicChangeShortSpeed(short* samples, int numSamples, float speed,
                          float pitch, float rate, float volume,
                          int useChordPitch, int sampleRate, int numChannels) {
  sonicStream stream = sonicCreateStream(sampleRate, numChannels);

  sonicSetSpeed(stream, speed);
  sonicSetPitch(stream, pitch);
  sonicSetRate(stream, rate);
  sonicSetVolume(stream, volume);
  sonicWriteShortToStream(stream, samples, numSamples);
  sonicFlushStream(stream);
  numSamples = sonicSamplesAvailable(stream);
  sonicReadShortFromStream(stream, samples, numSamples);
  sonicDestroyStream(stream);
  return numSamples;
}