RPCS3/soundtouch
1
0
Fork 0
mirror of https://github.com/RPCS3/soundtouch.git synced 2024-09-16 14:32:29 +02:00
Code Issues Projects Releases Wiki Activity

dos2unix:ify line endings, source code formatter

Browse Source
This commit is contained in:
Olli 2024-03-29 19:32:39 +02:00
parent 290b0b13e2
commit 63002027de
49 changed files with 10809 additions and 10806 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
Makefile.am
View File

@ -1,16 +1,16 @@
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA

3
README.html
View File

@ -147,6 +147,9 @@
cmake .
make -j
make install</pre>
<p>To list available build options:</p>
<pre>
cmake -LH</pre>
<p>To compile the additional portable Shared Library / DLL version with the native C-language API:</p>
<pre>
cmake . -DSOUNDTOUCH_DLL=ON

2
bootstrap
View File

@ -10,7 +10,7 @@ then
elif [ -a configure ]
then
configure && $0 --clean
else
else
bootstrap && configure && $0 --clean
fi

410
include/BPMDetect.h
View File

@ -1,205 +1,205 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Beats-per-minute (BPM) detection routine.
///
/// The beat detection algorithm works as follows:
/// - Use function 'inputSamples' to input a chunks of samples to the class for
/// analysis. It's a good idea to enter a large sound file or stream in smallish
/// chunks of around few kilosamples in order not to extinguish too much RAM memory.
/// - Input sound data is decimated to approx 500 Hz to reduce calculation burden,
/// which is basically ok as low (bass) frequencies mostly determine the beat rate.
/// Simple averaging is used for anti-alias filtering because the resulting signal
/// quality isn't of that high importance.
/// - Decimated sound data is enveloped, i.e. the amplitude shape is detected by
/// taking absolute value that's smoothed by sliding average. Signal levels that
/// are below a couple of times the general RMS amplitude level are cut away to
/// leave only notable peaks there.
/// - Repeating sound patterns (e.g. beats) are detected by calculating short-term
/// autocorrelation function of the enveloped signal.
/// - After whole sound data file has been analyzed as above, the bpm level is
/// detected by function 'getBpm' that finds the highest peak of the autocorrelation
/// function, calculates it's precise location and converts this reading to bpm's.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _BPMDetect_H_
#define _BPMDetect_H_
#include <vector>
#include "STTypes.h"
#include "FIFOSampleBuffer.h"
namespace soundtouch
{
/// Minimum allowed BPM rate. Used to restrict accepted result above a reasonable limit.
#define MIN_BPM 45
/// Maximum allowed BPM rate range. Used for calculating algorithm parametrs
#define MAX_BPM_RANGE 200
/// Maximum allowed BPM rate range. Used to restrict accepted result below a reasonable limit.
#define MAX_BPM_VALID 190
////////////////////////////////////////////////////////////////////////////////
typedef struct
{
float pos;
float strength;
} BEAT;
class IIR2_filter
{
double coeffs[5];
double prev[5];
public:
IIR2_filter(const double *lpf_coeffs);
float update(float x);
};
/// Class for calculating BPM rate for audio data.
class BPMDetect
{
protected:
/// Auto-correlation accumulator bins.
float *xcorr;
/// Sample average counter.
int decimateCount;
/// Sample average accumulator for FIFO-like decimation.
soundtouch::LONG_SAMPLETYPE decimateSum;
/// Decimate sound by this coefficient to reach approx. 500 Hz.
int decimateBy;
/// Auto-correlation window length
int windowLen;
/// Number of channels (1 = mono, 2 = stereo)
int channels;
/// sample rate
int sampleRate;
/// Beginning of auto-correlation window: Autocorrelation isn't being updated for
/// the first these many correlation bins.
int windowStart;
/// window functions for data preconditioning
float *hamw;
float *hamw2;
// beat detection variables
int pos;
int peakPos;
int beatcorr_ringbuffpos;
int init_scaler;
float peakVal;
float *beatcorr_ringbuff;
/// FIFO-buffer for decimated processing samples.
soundtouch::FIFOSampleBuffer *buffer;
/// Collection of detected beat positions
//BeatCollection beats;
std::vector<BEAT> beats;
// 2nd order low-pass-filter
IIR2_filter beat_lpf;
/// Updates auto-correlation function for given number of decimated samples that
/// are read from the internal 'buffer' pipe (samples aren't removed from the pipe
/// though).
void updateXCorr(int process_samples /// How many samples are processed.
);
/// Decimates samples to approx. 500 Hz.
///
/// \return Number of output samples.
int decimate(soundtouch::SAMPLETYPE *dest, ///< Destination buffer
const soundtouch::SAMPLETYPE *src, ///< Source sample buffer
int numsamples ///< Number of source samples.
);
/// Calculates amplitude envelope for the buffer of samples.
/// Result is output to 'samples'.
void calcEnvelope(soundtouch::SAMPLETYPE *samples, ///< Pointer to input/output data buffer
int numsamples ///< Number of samples in buffer
);
/// remove constant bias from xcorr data
void removeBias();
// Detect individual beat positions
void updateBeatPos(int process_samples);
public:
/// Constructor.
BPMDetect(int numChannels, ///< Number of channels in sample data.
int sampleRate ///< Sample rate in Hz.
);
/// Destructor.
virtual ~BPMDetect();
/// Inputs a block of samples for analyzing: Envelopes the samples and then
/// updates the autocorrelation estimation. When whole song data has been input
/// in smaller blocks using this function, read the resulting bpm with 'getBpm'
/// function.
///
/// Notice that data in 'samples' array can be disrupted in processing.
void inputSamples(const soundtouch::SAMPLETYPE *samples, ///< Pointer to input/working data buffer
int numSamples ///< Number of samples in buffer
);
/// Analyzes the results and returns the BPM rate. Use this function to read result
/// after whole song data has been input to the class by consecutive calls of
/// 'inputSamples' function.
///
/// \return Beats-per-minute rate, or zero if detection failed.
float getBpm();
/// Get beat position arrays. Note: The array includes also really low beat detection values
/// in absence of clear strong beats. Consumer may wish to filter low values away.
/// - "pos" receive array of beat positions
/// - "values" receive array of beat detection strengths
/// - max_num indicates max.size of "pos" and "values" array.
///
/// You can query a suitable array sized by calling this with nullptr in "pos" & "values".
///
/// \return number of beats in the arrays.
int getBeats(float *pos, float *strength, int max_num);
};
}
#endif // _BPMDetect_H_
////////////////////////////////////////////////////////////////////////////////
///
/// Beats-per-minute (BPM) detection routine.
///
/// The beat detection algorithm works as follows:
/// - Use function 'inputSamples' to input a chunks of samples to the class for
/// analysis. It's a good idea to enter a large sound file or stream in smallish
/// chunks of around few kilosamples in order not to extinguish too much RAM memory.
/// - Input sound data is decimated to approx 500 Hz to reduce calculation burden,
/// which is basically ok as low (bass) frequencies mostly determine the beat rate.
/// Simple averaging is used for anti-alias filtering because the resulting signal
/// quality isn't of that high importance.
/// - Decimated sound data is enveloped, i.e. the amplitude shape is detected by
/// taking absolute value that's smoothed by sliding average. Signal levels that
/// are below a couple of times the general RMS amplitude level are cut away to
/// leave only notable peaks there.
/// - Repeating sound patterns (e.g. beats) are detected by calculating short-term
/// autocorrelation function of the enveloped signal.
/// - After whole sound data file has been analyzed as above, the bpm level is
/// detected by function 'getBpm' that finds the highest peak of the autocorrelation
/// function, calculates it's precise location and converts this reading to bpm's.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _BPMDetect_H_
#define _BPMDetect_H_
#include <vector>
#include "STTypes.h"
#include "FIFOSampleBuffer.h"
namespace soundtouch
{
/// Minimum allowed BPM rate. Used to restrict accepted result above a reasonable limit.
#define MIN_BPM 45
/// Maximum allowed BPM rate range. Used for calculating algorithm parametrs
#define MAX_BPM_RANGE 200
/// Maximum allowed BPM rate range. Used to restrict accepted result below a reasonable limit.
#define MAX_BPM_VALID 190
////////////////////////////////////////////////////////////////////////////////
typedef struct
{
float pos;
float strength;
} BEAT;
class IIR2_filter
{
double coeffs[5];
double prev[5];
public:
IIR2_filter(const double *lpf_coeffs);
float update(float x);
};
/// Class for calculating BPM rate for audio data.
class BPMDetect
{
protected:
/// Auto-correlation accumulator bins.
float *xcorr;
/// Sample average counter.
int decimateCount;
/// Sample average accumulator for FIFO-like decimation.
soundtouch::LONG_SAMPLETYPE decimateSum;
/// Decimate sound by this coefficient to reach approx. 500 Hz.
int decimateBy;
/// Auto-correlation window length
int windowLen;
/// Number of channels (1 = mono, 2 = stereo)
int channels;
/// sample rate
int sampleRate;
/// Beginning of auto-correlation window: Autocorrelation isn't being updated for
/// the first these many correlation bins.
int windowStart;
/// window functions for data preconditioning
float *hamw;
float *hamw2;
// beat detection variables
int pos;
int peakPos;
int beatcorr_ringbuffpos;
int init_scaler;
float peakVal;
float *beatcorr_ringbuff;
/// FIFO-buffer for decimated processing samples.
soundtouch::FIFOSampleBuffer *buffer;
/// Collection of detected beat positions
//BeatCollection beats;
std::vector<BEAT> beats;
// 2nd order low-pass-filter
IIR2_filter beat_lpf;
/// Updates auto-correlation function for given number of decimated samples that
/// are read from the internal 'buffer' pipe (samples aren't removed from the pipe
/// though).
void updateXCorr(int process_samples /// How many samples are processed.
);
/// Decimates samples to approx. 500 Hz.
///
/// \return Number of output samples.
int decimate(soundtouch::SAMPLETYPE *dest, ///< Destination buffer
const soundtouch::SAMPLETYPE *src, ///< Source sample buffer
int numsamples ///< Number of source samples.
);
/// Calculates amplitude envelope for the buffer of samples.
/// Result is output to 'samples'.
void calcEnvelope(soundtouch::SAMPLETYPE *samples, ///< Pointer to input/output data buffer
int numsamples ///< Number of samples in buffer
);
/// remove constant bias from xcorr data
void removeBias();
// Detect individual beat positions
void updateBeatPos(int process_samples);
public:
/// Constructor.
BPMDetect(int numChannels, ///< Number of channels in sample data.
int sampleRate ///< Sample rate in Hz.
);
/// Destructor.
virtual ~BPMDetect();
/// Inputs a block of samples for analyzing: Envelopes the samples and then
/// updates the autocorrelation estimation. When whole song data has been input
/// in smaller blocks using this function, read the resulting bpm with 'getBpm'
/// function.
///
/// Notice that data in 'samples' array can be disrupted in processing.
void inputSamples(const soundtouch::SAMPLETYPE *samples, ///< Pointer to input/working data buffer
int numSamples ///< Number of samples in buffer
);
/// Analyzes the results and returns the BPM rate. Use this function to read result
/// after whole song data has been input to the class by consecutive calls of
/// 'inputSamples' function.
///
/// \return Beats-per-minute rate, or zero if detection failed.
float getBpm();
/// Get beat position arrays. Note: The array includes also really low beat detection values
/// in absence of clear strong beats. Consumer may wish to filter low values away.
/// - "pos" receive array of beat positions
/// - "values" receive array of beat detection strengths
/// - max_num indicates max.size of "pos" and "values" array.
///
/// You can query a suitable array sized by calling this with nullptr in "pos" & "values".
///
/// \return number of beats in the arrays.
int getBeats(float *pos, float *strength, int max_num);
};
}
#endif // _BPMDetect_H_

360
include/FIFOSampleBuffer.h
View File

@ -1,180 +1,180 @@
////////////////////////////////////////////////////////////////////////////////
///
/// A buffer class for temporarily storaging sound samples, operates as a
/// first-in-first-out pipe.
///
/// Samples are added to the end of the sample buffer with the 'putSamples'
/// function, and are received from the beginning of the buffer by calling
/// the 'receiveSamples' function. The class automatically removes the
/// output samples from the buffer as well as grows the storage size
/// whenever necessary.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef FIFOSampleBuffer_H
#define FIFOSampleBuffer_H
#include "FIFOSamplePipe.h"
namespace soundtouch
{
/// Sample buffer working in FIFO (first-in-first-out) principle. The class takes
/// care of storage size adjustment and data moving during input/output operations.
///
/// Notice that in case of stereo audio, one sample is considered to consist of
/// both channel data.
class FIFOSampleBuffer : public FIFOSamplePipe
{
private:
/// Sample buffer.
SAMPLETYPE *buffer;
// Raw unaligned buffer memory. 'buffer' is made aligned by pointing it to first
// 16-byte aligned location of this buffer
SAMPLETYPE *bufferUnaligned;
/// Sample buffer size in bytes
uint sizeInBytes;
/// How many samples are currently in buffer.
uint samplesInBuffer;
/// Channels, 1=mono, 2=stereo.
uint channels;
/// Current position pointer to the buffer. This pointer is increased when samples are
/// removed from the pipe so that it's necessary to actually rewind buffer (move data)
/// only new data when is put to the pipe.
uint bufferPos;
/// Rewind the buffer by moving data from position pointed by 'bufferPos' to real
/// beginning of the buffer.
void rewind();
/// Ensures that the buffer has capacity for at least this many samples.
void ensureCapacity(uint capacityRequirement);
/// Returns current capacity.
uint getCapacity() const;
public:
/// Constructor
FIFOSampleBuffer(int numChannels = 2 ///< Number of channels, 1=mono, 2=stereo.
///< Default is stereo.
);
/// destructor
~FIFOSampleBuffer() override;
/// Returns a pointer to the beginning of the output samples.
/// This function is provided for accessing the output samples directly.
/// Please be careful for not to corrupt the book-keeping!
///
/// When using this function to output samples, also remember to 'remove' the
/// output samples from the buffer by calling the
/// 'receiveSamples(numSamples)' function
virtual SAMPLETYPE *ptrBegin() override;
/// Returns a pointer to the end of the used part of the sample buffer (i.e.
/// where the new samples are to be inserted). This function may be used for
/// inserting new samples into the sample buffer directly. Please be careful
/// not corrupt the book-keeping!
///
/// When using this function as means for inserting new samples, also remember
/// to increase the sample count afterwards, by calling the
/// 'putSamples(numSamples)' function.
SAMPLETYPE *ptrEnd(
uint slackCapacity ///< How much free capacity (in samples) there _at least_
///< should be so that the caller can successfully insert the
///< desired samples to the buffer. If necessary, the function
///< grows the buffer size to comply with this requirement.
);
/// Adds 'numSamples' pcs of samples from the 'samples' memory position to
/// the sample buffer.
virtual void putSamples(const SAMPLETYPE *samples, ///< Pointer to samples.
uint numSamples ///< Number of samples to insert.
) override;
/// Adjusts the book-keeping to increase number of samples in the buffer without
/// copying any actual samples.
///
/// This function is used to update the number of samples in the sample buffer
/// when accessing the buffer directly with 'ptrEnd' function. Please be
/// careful though!
virtual void putSamples(uint numSamples ///< Number of samples been inserted.
);
/// Output samples from beginning of the sample buffer. Copies requested samples to
/// output buffer and removes them from the sample buffer. If there are less than
/// 'numsample' samples in the buffer, returns all that available.
///
/// \return Number of samples returned.
virtual uint receiveSamples(SAMPLETYPE *output, ///< Buffer where to copy output samples.
uint maxSamples ///< How many samples to receive at max.
) override;
/// Adjusts book-keeping so that given number of samples are removed from beginning of the
/// sample buffer without copying them anywhere.
///
/// Used to reduce the number of samples in the buffer when accessing the sample buffer directly
/// with 'ptrBegin' function.
virtual uint receiveSamples(uint maxSamples ///< Remove this many samples from the beginning of pipe.
) override;
/// Returns number of samples currently available.
virtual uint numSamples() const override;
/// Sets number of channels, 1 = mono, 2 = stereo.
void setChannels(int numChannels);
/// Get number of channels
int getChannels()
{
return channels;
}
/// Returns nonzero if there aren't any samples available for outputting.
virtual int isEmpty() const override;
/// Clears all the samples.
virtual void clear() override;
/// allow trimming (downwards) amount of samples in pipeline.
/// Returns adjusted amount of samples
uint adjustAmountOfSamples(uint numSamples) override;
/// Add silence to end of buffer
void addSilent(uint nSamples);
};
}
#endif
////////////////////////////////////////////////////////////////////////////////
///
/// A buffer class for temporarily storaging sound samples, operates as a
/// first-in-first-out pipe.
///
/// Samples are added to the end of the sample buffer with the 'putSamples'
/// function, and are received from the beginning of the buffer by calling
/// the 'receiveSamples' function. The class automatically removes the
/// output samples from the buffer as well as grows the storage size
/// whenever necessary.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef FIFOSampleBuffer_H
#define FIFOSampleBuffer_H
#include "FIFOSamplePipe.h"
namespace soundtouch
{
/// Sample buffer working in FIFO (first-in-first-out) principle. The class takes
/// care of storage size adjustment and data moving during input/output operations.
///
/// Notice that in case of stereo audio, one sample is considered to consist of
/// both channel data.
class FIFOSampleBuffer : public FIFOSamplePipe
{
private:
/// Sample buffer.
SAMPLETYPE *buffer;
// Raw unaligned buffer memory. 'buffer' is made aligned by pointing it to first
// 16-byte aligned location of this buffer
SAMPLETYPE *bufferUnaligned;
/// Sample buffer size in bytes
uint sizeInBytes;
/// How many samples are currently in buffer.
uint samplesInBuffer;
/// Channels, 1=mono, 2=stereo.
uint channels;
/// Current position pointer to the buffer. This pointer is increased when samples are
/// removed from the pipe so that it's necessary to actually rewind buffer (move data)
/// only new data when is put to the pipe.
uint bufferPos;
/// Rewind the buffer by moving data from position pointed by 'bufferPos' to real
/// beginning of the buffer.
void rewind();
/// Ensures that the buffer has capacity for at least this many samples.
void ensureCapacity(uint capacityRequirement);
/// Returns current capacity.
uint getCapacity() const;
public:
/// Constructor
FIFOSampleBuffer(int numChannels = 2 ///< Number of channels, 1=mono, 2=stereo.
///< Default is stereo.
);
/// destructor
~FIFOSampleBuffer() override;
/// Returns a pointer to the beginning of the output samples.
/// This function is provided for accessing the output samples directly.
/// Please be careful for not to corrupt the book-keeping!
///
/// When using this function to output samples, also remember to 'remove' the
/// output samples from the buffer by calling the
/// 'receiveSamples(numSamples)' function
virtual SAMPLETYPE *ptrBegin() override;
/// Returns a pointer to the end of the used part of the sample buffer (i.e.
/// where the new samples are to be inserted). This function may be used for
/// inserting new samples into the sample buffer directly. Please be careful
/// not corrupt the book-keeping!
///
/// When using this function as means for inserting new samples, also remember
/// to increase the sample count afterwards, by calling the
/// 'putSamples(numSamples)' function.
SAMPLETYPE *ptrEnd(
uint slackCapacity ///< How much free capacity (in samples) there _at least_
///< should be so that the caller can successfully insert the
///< desired samples to the buffer. If necessary, the function
///< grows the buffer size to comply with this requirement.
);
/// Adds 'numSamples' pcs of samples from the 'samples' memory position to
/// the sample buffer.
virtual void putSamples(const SAMPLETYPE *samples, ///< Pointer to samples.
uint numSamples ///< Number of samples to insert.
) override;
/// Adjusts the book-keeping to increase number of samples in the buffer without
/// copying any actual samples.
///
/// This function is used to update the number of samples in the sample buffer
/// when accessing the buffer directly with 'ptrEnd' function. Please be
/// careful though!
virtual void putSamples(uint numSamples ///< Number of samples been inserted.
);
/// Output samples from beginning of the sample buffer. Copies requested samples to
/// output buffer and removes them from the sample buffer. If there are less than
/// 'numsample' samples in the buffer, returns all that available.
///
/// \return Number of samples returned.
virtual uint receiveSamples(SAMPLETYPE *output, ///< Buffer where to copy output samples.
uint maxSamples ///< How many samples to receive at max.
) override;
/// Adjusts book-keeping so that given number of samples are removed from beginning of the
/// sample buffer without copying them anywhere.
///
/// Used to reduce the number of samples in the buffer when accessing the sample buffer directly
/// with 'ptrBegin' function.
virtual uint receiveSamples(uint maxSamples ///< Remove this many samples from the beginning of pipe.
) override;
/// Returns number of samples currently available.
virtual uint numSamples() const override;
/// Sets number of channels, 1 = mono, 2 = stereo.
void setChannels(int numChannels);
/// Get number of channels
int getChannels()
{
return channels;
}
/// Returns nonzero if there aren't any samples available for outputting.
virtual int isEmpty() const override;
/// Clears all the samples.
virtual void clear() override;
/// allow trimming (downwards) amount of samples in pipeline.
/// Returns adjusted amount of samples
uint adjustAmountOfSamples(uint numSamples) override;
/// Add silence to end of buffer
void addSilent(uint nSamples);
};
}
#endif

460
include/FIFOSamplePipe.h
View File

@ -1,230 +1,230 @@
////////////////////////////////////////////////////////////////////////////////
///
/// 'FIFOSamplePipe' : An abstract base class for classes that manipulate sound
/// samples by operating like a first-in-first-out pipe: New samples are fed
/// into one end of the pipe with the 'putSamples' function, and the processed
/// samples are received from the other end with the 'receiveSamples' function.
///
/// 'FIFOProcessor' : A base class for classes the do signal processing with
/// the samples while operating like a first-in-first-out pipe. When samples
/// are input with the 'putSamples' function, the class processes them
/// and moves the processed samples to the given 'output' pipe object, which
/// may be either another processing stage, or a fifo sample buffer object.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef FIFOSamplePipe_H
#define FIFOSamplePipe_H
#include <assert.h>
#include <stdlib.h>
#include "STTypes.h"
namespace soundtouch
{
/// Abstract base class for FIFO (first-in-first-out) sample processing classes.
class FIFOSamplePipe
{
protected:
bool verifyNumberOfChannels(int nChannels) const
{
if ((nChannels > 0) && (nChannels <= SOUNDTOUCH_MAX_CHANNELS))
{
return true;
}
ST_THROW_RT_ERROR("Error: Illegal number of channels");
return false;
}
public:
// virtual default destructor
virtual ~FIFOSamplePipe() {}
/// Returns a pointer to the beginning of the output samples.
/// This function is provided for accessing the output samples directly.
/// Please be careful for not to corrupt the book-keeping!
///
/// When using this function to output samples, also remember to 'remove' the
/// output samples from the buffer by calling the
/// 'receiveSamples(numSamples)' function
virtual SAMPLETYPE *ptrBegin() = 0;
/// Adds 'numSamples' pcs of samples from the 'samples' memory position to
/// the sample buffer.
virtual void putSamples(const SAMPLETYPE *samples, ///< Pointer to samples.
uint numSamples ///< Number of samples to insert.
) = 0;
// Moves samples from the 'other' pipe instance to this instance.
void moveSamples(FIFOSamplePipe &other ///< Other pipe instance where from the receive the data.
)
{
const uint oNumSamples = other.numSamples();
putSamples(other.ptrBegin(), oNumSamples);
other.receiveSamples(oNumSamples);
}
/// Output samples from beginning of the sample buffer. Copies requested samples to
/// output buffer and removes them from the sample buffer. If there are less than
/// 'numsample' samples in the buffer, returns all that available.
///
/// \return Number of samples returned.
virtual uint receiveSamples(SAMPLETYPE *output, ///< Buffer where to copy output samples.
uint maxSamples ///< How many samples to receive at max.
) = 0;
/// Adjusts book-keeping so that given number of samples are removed from beginning of the
/// sample buffer without copying them anywhere.
///
/// Used to reduce the number of samples in the buffer when accessing the sample buffer directly
/// with 'ptrBegin' function.
virtual uint receiveSamples(uint maxSamples ///< Remove this many samples from the beginning of pipe.
) = 0;
/// Returns number of samples currently available.
virtual uint numSamples() const = 0;
// Returns nonzero if there aren't any samples available for outputting.
virtual int isEmpty() const = 0;
/// Clears all the samples.
virtual void clear() = 0;
/// allow trimming (downwards) amount of samples in pipeline.
/// Returns adjusted amount of samples
virtual uint adjustAmountOfSamples(uint numSamples) = 0;
};
/// Base-class for sound processing routines working in FIFO principle. With this base
/// class it's easy to implement sound processing stages that can be chained together,
/// so that samples that are fed into beginning of the pipe automatically go through
/// all the processing stages.
///
/// When samples are input to this class, they're first processed and then put to
/// the FIFO pipe that's defined as output of this class. This output pipe can be
/// either other processing stage or a FIFO sample buffer.
class FIFOProcessor :public FIFOSamplePipe
{
protected:
/// Internal pipe where processed samples are put.
FIFOSamplePipe *output;
/// Sets output pipe.
void setOutPipe(FIFOSamplePipe *pOutput)
{
assert(output == nullptr);
assert(pOutput != nullptr);
output = pOutput;
}
/// Constructor. Doesn't define output pipe; it has to be set be
/// 'setOutPipe' function.
FIFOProcessor()
{
output = nullptr;
}
/// Constructor. Configures output pipe.
FIFOProcessor(FIFOSamplePipe *pOutput ///< Output pipe.
)
{
output = pOutput;
}
/// Destructor.
virtual ~FIFOProcessor() override
{
}
/// Returns a pointer to the beginning of the output samples.
/// This function is provided for accessing the output samples directly.
/// Please be careful for not to corrupt the book-keeping!
///
/// When using this function to output samples, also remember to 'remove' the
/// output samples from the buffer by calling the
/// 'receiveSamples(numSamples)' function
virtual SAMPLETYPE *ptrBegin() override
{
return output->ptrBegin();
}
public:
/// Output samples from beginning of the sample buffer. Copies requested samples to
/// output buffer and removes them from the sample buffer. If there are less than
/// 'numsample' samples in the buffer, returns all that available.
///
/// \return Number of samples returned.
virtual uint receiveSamples(SAMPLETYPE *outBuffer, ///< Buffer where to copy output samples.
uint maxSamples ///< How many samples to receive at max.
) override
{
return output->receiveSamples(outBuffer, maxSamples);
}
/// Adjusts book-keeping so that given number of samples are removed from beginning of the
/// sample buffer without copying them anywhere.
///
/// Used to reduce the number of samples in the buffer when accessing the sample buffer directly
/// with 'ptrBegin' function.
virtual uint receiveSamples(uint maxSamples ///< Remove this many samples from the beginning of pipe.
) override
{
return output->receiveSamples(maxSamples);
}
/// Returns number of samples currently available.
virtual uint numSamples() const override
{
return output->numSamples();
}
/// Returns nonzero if there aren't any samples available for outputting.
virtual int isEmpty() const override
{
return output->isEmpty();
}
/// allow trimming (downwards) amount of samples in pipeline.
/// Returns adjusted amount of samples
virtual uint adjustAmountOfSamples(uint numSamples) override
{
return output->adjustAmountOfSamples(numSamples);
}
};
}
#endif
////////////////////////////////////////////////////////////////////////////////
///
/// 'FIFOSamplePipe' : An abstract base class for classes that manipulate sound
/// samples by operating like a first-in-first-out pipe: New samples are fed
/// into one end of the pipe with the 'putSamples' function, and the processed
/// samples are received from the other end with the 'receiveSamples' function.
///
/// 'FIFOProcessor' : A base class for classes the do signal processing with
/// the samples while operating like a first-in-first-out pipe. When samples
/// are input with the 'putSamples' function, the class processes them
/// and moves the processed samples to the given 'output' pipe object, which
/// may be either another processing stage, or a fifo sample buffer object.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef FIFOSamplePipe_H
#define FIFOSamplePipe_H
#include <assert.h>
#include <stdlib.h>
#include "STTypes.h"
namespace soundtouch
{
/// Abstract base class for FIFO (first-in-first-out) sample processing classes.
class FIFOSamplePipe
{
protected:
bool verifyNumberOfChannels(int nChannels) const
{
if ((nChannels > 0) && (nChannels <= SOUNDTOUCH_MAX_CHANNELS))
{
return true;
}
ST_THROW_RT_ERROR("Error: Illegal number of channels");
return false;
}
public:
// virtual default destructor
virtual ~FIFOSamplePipe() {}
/// Returns a pointer to the beginning of the output samples.
/// This function is provided for accessing the output samples directly.
/// Please be careful for not to corrupt the book-keeping!
///
/// When using this function to output samples, also remember to 'remove' the
/// output samples from the buffer by calling the
/// 'receiveSamples(numSamples)' function
virtual SAMPLETYPE *ptrBegin() = 0;
/// Adds 'numSamples' pcs of samples from the 'samples' memory position to
/// the sample buffer.
virtual void putSamples(const SAMPLETYPE *samples, ///< Pointer to samples.
uint numSamples ///< Number of samples to insert.
) = 0;
// Moves samples from the 'other' pipe instance to this instance.
void moveSamples(FIFOSamplePipe &other ///< Other pipe instance where from the receive the data.
)
{
const uint oNumSamples = other.numSamples();
putSamples(other.ptrBegin(), oNumSamples);
other.receiveSamples(oNumSamples);
}
/// Output samples from beginning of the sample buffer. Copies requested samples to
/// output buffer and removes them from the sample buffer. If there are less than
/// 'numsample' samples in the buffer, returns all that available.
///
/// \return Number of samples returned.
virtual uint receiveSamples(SAMPLETYPE *output, ///< Buffer where to copy output samples.
uint maxSamples ///< How many samples to receive at max.
) = 0;
/// Adjusts book-keeping so that given number of samples are removed from beginning of the
/// sample buffer without copying them anywhere.
///
/// Used to reduce the number of samples in the buffer when accessing the sample buffer directly
/// with 'ptrBegin' function.
virtual uint receiveSamples(uint maxSamples ///< Remove this many samples from the beginning of pipe.
) = 0;
/// Returns number of samples currently available.
virtual uint numSamples() const = 0;
// Returns nonzero if there aren't any samples available for outputting.
virtual int isEmpty() const = 0;
/// Clears all the samples.
virtual void clear() = 0;
/// allow trimming (downwards) amount of samples in pipeline.
/// Returns adjusted amount of samples
virtual uint adjustAmountOfSamples(uint numSamples) = 0;
};
/// Base-class for sound processing routines working in FIFO principle. With this base
/// class it's easy to implement sound processing stages that can be chained together,
/// so that samples that are fed into beginning of the pipe automatically go through
/// all the processing stages.
///
/// When samples are input to this class, they're first processed and then put to
/// the FIFO pipe that's defined as output of this class. This output pipe can be
/// either other processing stage or a FIFO sample buffer.
class FIFOProcessor :public FIFOSamplePipe
{
protected:
/// Internal pipe where processed samples are put.
FIFOSamplePipe *output;
/// Sets output pipe.
void setOutPipe(FIFOSamplePipe *pOutput)
{
assert(output == nullptr);
assert(pOutput != nullptr);
output = pOutput;
}
/// Constructor. Doesn't define output pipe; it has to be set be
/// 'setOutPipe' function.
FIFOProcessor()
{
output = nullptr;
}
/// Constructor. Configures output pipe.
FIFOProcessor(FIFOSamplePipe *pOutput ///< Output pipe.
)
{
output = pOutput;
}
/// Destructor.
virtual ~FIFOProcessor() override
{
}
/// Returns a pointer to the beginning of the output samples.
/// This function is provided for accessing the output samples directly.
/// Please be careful for not to corrupt the book-keeping!
///
/// When using this function to output samples, also remember to 'remove' the
/// output samples from the buffer by calling the
/// 'receiveSamples(numSamples)' function
virtual SAMPLETYPE *ptrBegin() override
{
return output->ptrBegin();
}
public:
/// Output samples from beginning of the sample buffer. Copies requested samples to
/// output buffer and removes them from the sample buffer. If there are less than
/// 'numsample' samples in the buffer, returns all that available.
///
/// \return Number of samples returned.
virtual uint receiveSamples(SAMPLETYPE *outBuffer, ///< Buffer where to copy output samples.
uint maxSamples ///< How many samples to receive at max.
) override
{
return output->receiveSamples(outBuffer, maxSamples);
}
/// Adjusts book-keeping so that given number of samples are removed from beginning of the
/// sample buffer without copying them anywhere.
///
/// Used to reduce the number of samples in the buffer when accessing the sample buffer directly
/// with 'ptrBegin' function.
virtual uint receiveSamples(uint maxSamples ///< Remove this many samples from the beginning of pipe.
) override
{
return output->receiveSamples(maxSamples);
}
/// Returns number of samples currently available.
virtual uint numSamples() const override
{
return output->numSamples();
}
/// Returns nonzero if there aren't any samples available for outputting.
virtual int isEmpty() const override
{
return output->isEmpty();
}
/// allow trimming (downwards) amount of samples in pipeline.
/// Returns adjusted amount of samples
virtual uint adjustAmountOfSamples(uint numSamples) override
{
return output->adjustAmountOfSamples(numSamples);
}
};
}
#endif

44
include/Makefile.am
View File

@ -1,22 +1,22 @@
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA
## I used config/am_include.mk for common definitions
include $(top_srcdir)/config/am_include.mk
pkginclude_HEADERS=FIFOSampleBuffer.h FIFOSamplePipe.h SoundTouch.h STTypes.h BPMDetect.h soundtouch_config.h
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA
## I used config/am_include.mk for common definitions
include $(top_srcdir)/config/am_include.mk
pkginclude_HEADERS=FIFOSampleBuffer.h FIFOSamplePipe.h SoundTouch.h STTypes.h BPMDetect.h soundtouch_config.h

380
include/STTypes.h
View File

@ -1,190 +1,190 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Common type definitions for SoundTouch audio processing library.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef STTypes_H
#define STTypes_H
typedef unsigned int uint;
typedef unsigned long ulong;
// Patch for MinGW: on Win64 long is 32-bit
#ifdef _WIN64
typedef unsigned long long ulongptr;
#else
typedef ulong ulongptr;
#endif
// Helper macro for aligning pointer up to next 16-byte boundary
#define SOUNDTOUCH_ALIGN_POINTER_16(x) ( ( (ulongptr)(x) + 15 ) & ~(ulongptr)15 )
#if (defined(__GNUC__) && !defined(ANDROID))
// In GCC, include soundtouch_config.h made by config scritps.
// Skip this in Android compilation that uses GCC but without configure scripts.
#include "soundtouch_config.h"
#endif
namespace soundtouch
{
/// Max allowed number of channels
#define SOUNDTOUCH_MAX_CHANNELS 16
/// Activate these undef's to overrule the possible sampletype
/// setting inherited from some other header file:
//#undef SOUNDTOUCH_INTEGER_SAMPLES
//#undef SOUNDTOUCH_FLOAT_SAMPLES
/// If following flag is defined, always uses multichannel processing
/// routines also for mono and stero sound. This is for routine testing
/// purposes; output should be same with either routines, yet disabling
/// the dedicated mono/stereo processing routines will result in slower
/// runtime performance so recommendation is to keep this off.
// #define USE_MULTICH_ALWAYS
#if (defined(__SOFTFP__) && defined(ANDROID))
// For Android compilation: Force use of Integer samples in case that
// compilation uses soft-floating point emulation - soft-fp is way too slow
#undef SOUNDTOUCH_FLOAT_SAMPLES
#define SOUNDTOUCH_INTEGER_SAMPLES 1
#endif
#if !(SOUNDTOUCH_INTEGER_SAMPLES || SOUNDTOUCH_FLOAT_SAMPLES)
/// Choose either 32bit floating point or 16bit integer sampletype
/// by choosing one of the following defines, unless this selection
/// has already been done in some other file.
////
/// Notes:
/// - In Windows environment, choose the sample format with the
/// following defines.
/// - In GNU environment, the floating point samples are used by
/// default, but integer samples can be chosen by giving the
/// following switch to the configure script:
/// ./configure --enable-integer-samples
/// However, if you still prefer to select the sample format here
/// also in GNU environment, then please #undef the INTEGER_SAMPLE
/// and FLOAT_SAMPLE defines first as in comments above.
//#define SOUNDTOUCH_INTEGER_SAMPLES 1 //< 16bit integer samples
#define SOUNDTOUCH_FLOAT_SAMPLES 1 //< 32bit float samples
#endif
#if (_M_IX86 || __i386__ || __x86_64__ || _M_X64)
/// Define this to allow X86-specific assembler/intrinsic optimizations.
/// Notice that library contains also usual C++ versions of each of these
/// these routines, so if you're having difficulties getting the optimized
/// routines compiled for whatever reason, you may disable these optimizations
/// to make the library compile.
#define SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS 1
/// In GNU environment, allow the user to override this setting by
/// giving the following switch to the configure script:
/// ./configure --disable-x86-optimizations
/// ./configure --enable-x86-optimizations=no
#ifdef SOUNDTOUCH_DISABLE_X86_OPTIMIZATIONS
#undef SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS
#endif
#else
/// Always disable optimizations when not using a x86 systems.
#undef SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS
#endif
// If defined, allows the SIMD-optimized routines to skip unevenly aligned
// memory offsets that can cause performance penalty in some SIMD implementations.
// Causes slight compromise in sound quality.
// #define SOUNDTOUCH_ALLOW_NONEXACT_SIMD_OPTIMIZATION 1
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
// 16bit integer sample type
typedef short SAMPLETYPE;
// data type for sample accumulation: Use 32bit integer to prevent overflows
typedef long LONG_SAMPLETYPE;
#ifdef SOUNDTOUCH_FLOAT_SAMPLES
// check that only one sample type is defined
#error "conflicting sample types defined"
#endif // SOUNDTOUCH_FLOAT_SAMPLES
#ifdef SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS
// Allow MMX optimizations (not available in X64 mode)
#if (!_M_X64)
#define SOUNDTOUCH_ALLOW_MMX 1
#endif
#endif
#else
// floating point samples
typedef float SAMPLETYPE;
// data type for sample accumulation: Use float also here to enable
// efficient autovectorization
typedef float LONG_SAMPLETYPE;
#ifdef SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS
// Allow SSE optimizations
#define SOUNDTOUCH_ALLOW_SSE 1
#endif
#endif // SOUNDTOUCH_INTEGER_SAMPLES
#if ((SOUNDTOUCH_ALLOW_SSE) || (__SSE__) || (SOUNDTOUCH_USE_NEON))
#if SOUNDTOUCH_ALLOW_NONEXACT_SIMD_OPTIMIZATION
#define ST_SIMD_AVOID_UNALIGNED
#endif
#endif
}
// define ST_NO_EXCEPTION_HANDLING switch to disable throwing std exceptions:
// #define ST_NO_EXCEPTION_HANDLING 1
#ifdef ST_NO_EXCEPTION_HANDLING
// Exceptions disabled. Throw asserts instead if enabled.
#include <assert.h>
#define ST_THROW_RT_ERROR(x) {assert((const char *)x);}
#else
// use c++ standard exceptions
#include <stdexcept>
#include <string>
#define ST_THROW_RT_ERROR(x) {throw std::runtime_error(x);}
#endif
// When this #define is active, eliminates a clicking sound when the "rate" or "pitch"
// parameter setting crosses from value <1 to >=1 or vice versa during processing.
// Default is off as such crossover is untypical case and involves a slight sound
// quality compromise.
//#define SOUNDTOUCH_PREVENT_CLICK_AT_RATE_CROSSOVER 1
#endif
////////////////////////////////////////////////////////////////////////////////
///
/// Common type definitions for SoundTouch audio processing library.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef STTypes_H
#define STTypes_H
typedef unsigned int uint;
typedef unsigned long ulong;
// Patch for MinGW: on Win64 long is 32-bit
#ifdef _WIN64
typedef unsigned long long ulongptr;
#else
typedef ulong ulongptr;
#endif
// Helper macro for aligning pointer up to next 16-byte boundary
#define SOUNDTOUCH_ALIGN_POINTER_16(x) ( ( (ulongptr)(x) + 15 ) & ~(ulongptr)15 )
#if (defined(__GNUC__) && !defined(ANDROID))
// In GCC, include soundtouch_config.h made by config scritps.
// Skip this in Android compilation that uses GCC but without configure scripts.
#include "soundtouch_config.h"
#endif
namespace soundtouch
{
/// Max allowed number of channels
#define SOUNDTOUCH_MAX_CHANNELS 16
/// Activate these undef's to overrule the possible sampletype
/// setting inherited from some other header file:
//#undef SOUNDTOUCH_INTEGER_SAMPLES
//#undef SOUNDTOUCH_FLOAT_SAMPLES
/// If following flag is defined, always uses multichannel processing
/// routines also for mono and stero sound. This is for routine testing
/// purposes; output should be same with either routines, yet disabling
/// the dedicated mono/stereo processing routines will result in slower
/// runtime performance so recommendation is to keep this off.
// #define USE_MULTICH_ALWAYS
#if (defined(__SOFTFP__) && defined(ANDROID))
// For Android compilation: Force use of Integer samples in case that
// compilation uses soft-floating point emulation - soft-fp is way too slow
#undef SOUNDTOUCH_FLOAT_SAMPLES
#define SOUNDTOUCH_INTEGER_SAMPLES 1
#endif
#if !(SOUNDTOUCH_INTEGER_SAMPLES || SOUNDTOUCH_FLOAT_SAMPLES)
/// Choose either 32bit floating point or 16bit integer sampletype
/// by choosing one of the following defines, unless this selection
/// has already been done in some other file.
////
/// Notes:
/// - In Windows environment, choose the sample format with the
/// following defines.
/// - In GNU environment, the floating point samples are used by
/// default, but integer samples can be chosen by giving the
/// following switch to the configure script:
/// ./configure --enable-integer-samples
/// However, if you still prefer to select the sample format here
/// also in GNU environment, then please #undef the INTEGER_SAMPLE
/// and FLOAT_SAMPLE defines first as in comments above.
//#define SOUNDTOUCH_INTEGER_SAMPLES 1 //< 16bit integer samples
#define SOUNDTOUCH_FLOAT_SAMPLES 1 //< 32bit float samples
#endif
#if (_M_IX86 || __i386__ || __x86_64__ || _M_X64)
/// Define this to allow X86-specific assembler/intrinsic optimizations.
/// Notice that library contains also usual C++ versions of each of these
/// these routines, so if you're having difficulties getting the optimized
/// routines compiled for whatever reason, you may disable these optimizations
/// to make the library compile.
#define SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS 1
/// In GNU environment, allow the user to override this setting by
/// giving the following switch to the configure script:
/// ./configure --disable-x86-optimizations
/// ./configure --enable-x86-optimizations=no
#ifdef SOUNDTOUCH_DISABLE_X86_OPTIMIZATIONS
#undef SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS
#endif
#else
/// Always disable optimizations when not using a x86 systems.
#undef SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS
#endif
// If defined, allows the SIMD-optimized routines to skip unevenly aligned
// memory offsets that can cause performance penalty in some SIMD implementations.
// Causes slight compromise in sound quality.
// #define SOUNDTOUCH_ALLOW_NONEXACT_SIMD_OPTIMIZATION 1
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
// 16bit integer sample type
typedef short SAMPLETYPE;
// data type for sample accumulation: Use 32bit integer to prevent overflows
typedef long LONG_SAMPLETYPE;
#ifdef SOUNDTOUCH_FLOAT_SAMPLES
// check that only one sample type is defined
#error "conflicting sample types defined"
#endif // SOUNDTOUCH_FLOAT_SAMPLES
#ifdef SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS
// Allow MMX optimizations (not available in X64 mode)
#if (!_M_X64)
#define SOUNDTOUCH_ALLOW_MMX 1
#endif
#endif
#else
// floating point samples
typedef float SAMPLETYPE;
// data type for sample accumulation: Use float also here to enable
// efficient autovectorization
typedef float LONG_SAMPLETYPE;
#ifdef SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS
// Allow SSE optimizations
#define SOUNDTOUCH_ALLOW_SSE 1
#endif
#endif // SOUNDTOUCH_INTEGER_SAMPLES
#if ((SOUNDTOUCH_ALLOW_SSE) || (__SSE__) || (SOUNDTOUCH_USE_NEON))
#if SOUNDTOUCH_ALLOW_NONEXACT_SIMD_OPTIMIZATION
#define ST_SIMD_AVOID_UNALIGNED
#endif
#endif
}
// define ST_NO_EXCEPTION_HANDLING switch to disable throwing std exceptions:
// #define ST_NO_EXCEPTION_HANDLING 1
#ifdef ST_NO_EXCEPTION_HANDLING
// Exceptions disabled. Throw asserts instead if enabled.
#include <assert.h>
#define ST_THROW_RT_ERROR(x) {assert((const char *)x);}
#else
// use c++ standard exceptions
#include <stdexcept>
#include <string>
#define ST_THROW_RT_ERROR(x) {throw std::runtime_error(x);}
#endif
// When this #define is active, eliminates a clicking sound when the "rate" or "pitch"
// parameter setting crosses from value <1 to >=1 or vice versa during processing.
// Default is off as such crossover is untypical case and involves a slight sound
// quality compromise.
//#define SOUNDTOUCH_PREVENT_CLICK_AT_RATE_CROSSOVER 1
#endif

696
include/SoundTouch.h
View File

@ -1,348 +1,348 @@
//////////////////////////////////////////////////////////////////////////////
///
/// SoundTouch - main class for tempo/pitch/rate adjusting routines.
///
/// Notes:
/// - Initialize the SoundTouch object instance by setting up the sound stream
/// parameters with functions 'setSampleRate' and 'setChannels', then set
/// desired tempo/pitch/rate settings with the corresponding functions.
///
/// - The SoundTouch class behaves like a first-in-first-out pipeline: The
/// samples that are to be processed are fed into one of the pipe by calling
/// function 'putSamples', while the ready processed samples can be read
/// from the other end of the pipeline with function 'receiveSamples'.
///
/// - The SoundTouch processing classes require certain sized 'batches' of
/// samples in order to process the sound. For this reason the classes buffer
/// incoming samples until there are enough of samples available for
/// processing, then they carry out the processing step and consequently
/// make the processed samples available for outputting.
///
/// - For the above reason, the processing routines introduce a certain
/// 'latency' between the input and output, so that the samples input to
/// SoundTouch may not be immediately available in the output, and neither
/// the amount of outputtable samples may not immediately be in direct
/// relationship with the amount of previously input samples.
///
/// - The tempo/pitch/rate control parameters can be altered during processing.
/// Please notice though that they aren't currently protected by semaphores,
/// so in multi-thread application external semaphore protection may be
/// required.
///
/// - This class utilizes classes 'TDStretch' for tempo change (without modifying
/// pitch) and 'RateTransposer' for changing the playback rate (that is, both
/// tempo and pitch in the same ratio) of the sound. The third available control
/// 'pitch' (change pitch but maintain tempo) is produced by a combination of
/// combining the two other controls.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef SoundTouch_H
#define SoundTouch_H
#include "FIFOSamplePipe.h"
#include "STTypes.h"
namespace soundtouch
{
/// Soundtouch library version string
#define SOUNDTOUCH_VERSION "2.3.3"
/// SoundTouch library version id
#define SOUNDTOUCH_VERSION_ID (20303)
//
// Available setting IDs for the 'setSetting' & 'get_setting' functions:
/// Enable/disable anti-alias filter in pitch transposer (0 = disable)
#define SETTING_USE_AA_FILTER 0
/// Pitch transposer anti-alias filter length (8 .. 128 taps, default = 32)
#define SETTING_AA_FILTER_LENGTH 1
/// Enable/disable quick seeking algorithm in tempo changer routine
/// (enabling quick seeking lowers CPU utilization but causes a minor sound
/// quality compromising)
#define SETTING_USE_QUICKSEEK 2
/// Time-stretch algorithm single processing sequence length in milliseconds. This determines
/// to how long sequences the original sound is chopped in the time-stretch algorithm.
/// See "STTypes.h" or README for more information.
#define SETTING_SEQUENCE_MS 3
/// Time-stretch algorithm seeking window length in milliseconds for algorithm that finds the
/// best possible overlapping location. This determines from how wide window the algorithm
/// may look for an optimal joining location when mixing the sound sequences back together.
/// See "STTypes.h" or README for more information.
#define SETTING_SEEKWINDOW_MS 4
/// Time-stretch algorithm overlap length in milliseconds. When the chopped sound sequences
/// are mixed back together, to form a continuous sound stream, this parameter defines over
/// how long period the two consecutive sequences are let to overlap each other.
/// See "STTypes.h" or README for more information.
#define SETTING_OVERLAP_MS 5
/// Call "getSetting" with this ID to query processing sequence size in samples.
/// This value gives approximate value of how many input samples you'll need to
/// feed into SoundTouch after initial buffering to get out a new batch of
/// output samples.
///
/// This value does not include initial buffering at beginning of a new processing
/// stream, use SETTING_INITIAL_LATENCY to get the initial buffering size.
///
/// Notices:
/// - This is read-only parameter, i.e. setSetting ignores this parameter
/// - This parameter value is not constant but change depending on
/// tempo/pitch/rate/samplerate settings.
#define SETTING_NOMINAL_INPUT_SEQUENCE 6
/// Call "getSetting" with this ID to query nominal average processing output
/// size in samples. This value tells approcimate value how many output samples
/// SoundTouch outputs once it does DSP processing run for a batch of input samples.
///
/// Notices:
/// - This is read-only parameter, i.e. setSetting ignores this parameter
/// - This parameter value is not constant but change depending on
/// tempo/pitch/rate/samplerate settings.
#define SETTING_NOMINAL_OUTPUT_SEQUENCE 7
/// Call "getSetting" with this ID to query initial processing latency, i.e.
/// approx. how many samples you'll need to enter to SoundTouch pipeline before
/// you can expect to get first batch of ready output samples out.
///
/// After the first output batch, you can then expect to get approx.
/// SETTING_NOMINAL_OUTPUT_SEQUENCE ready samples out for every
/// SETTING_NOMINAL_INPUT_SEQUENCE samples that you enter into SoundTouch.
///
/// Example:
/// processing with parameter -tempo=5
/// => initial latency = 5509 samples
/// input sequence = 4167 samples
/// output sequence = 3969 samples
///
/// Accordingly, you can expect to feed in approx. 5509 samples at beginning of
/// the stream, and then you'll get out the first 3969 samples. After that, for
/// every approx. 4167 samples that you'll put in, you'll receive again approx.
/// 3969 samples out.
///
/// This also means that average latency during stream processing is
/// INITIAL_LATENCY-OUTPUT_SEQUENCE/2, in the above example case 5509-3969/2
/// = 3524 samples
///
/// Notices:
/// - This is read-only parameter, i.e. setSetting ignores this parameter
/// - This parameter value is not constant but change depending on
/// tempo/pitch/rate/samplerate settings.
#define SETTING_INITIAL_LATENCY 8
class SoundTouch : public FIFOProcessor
{
private:
/// Rate transposer class instance
class RateTransposer *pRateTransposer;
/// Time-stretch class instance
class TDStretch *pTDStretch;
/// Virtual pitch parameter. Effective rate & tempo are calculated from these parameters.
double virtualRate;
/// Virtual pitch parameter. Effective rate & tempo are calculated from these parameters.
double virtualTempo;
/// Virtual pitch parameter. Effective rate & tempo are calculated from these parameters.
double virtualPitch;
/// Flag: Has sample rate been set?
bool bSrateSet;
/// Accumulator for how many samples in total will be expected as output vs. samples put in,
/// considering current processing settings.
double samplesExpectedOut;
/// Accumulator for how many samples in total have been read out from the processing so far
long samplesOutput;
/// Calculates effective rate & tempo valuescfrom 'virtualRate', 'virtualTempo' and
/// 'virtualPitch' parameters.
void calcEffectiveRateAndTempo();
protected :
/// Number of channels
uint channels;
/// Effective 'rate' value calculated from 'virtualRate', 'virtualTempo' and 'virtualPitch'
double rate;
/// Effective 'tempo' value calculated from 'virtualRate', 'virtualTempo' and 'virtualPitch'
double tempo;
public:
SoundTouch();
virtual ~SoundTouch() override;
/// Get SoundTouch library version string
static const char *getVersionString();
/// Get SoundTouch library version Id
static uint getVersionId();
/// Sets new rate control value. Normal rate = 1.0, smaller values
/// represent slower rate, larger faster rates.
void setRate(double newRate);
/// Sets new tempo control value. Normal tempo = 1.0, smaller values
/// represent slower tempo, larger faster tempo.
void setTempo(double newTempo);
/// Sets new rate control value as a difference in percents compared
/// to the original rate (-50 .. +100 %)
void setRateChange(double newRate);
/// Sets new tempo control value as a difference in percents compared
/// to the original tempo (-50 .. +100 %)
void setTempoChange(double newTempo);
/// Sets new pitch control value. Original pitch = 1.0, smaller values
/// represent lower pitches, larger values higher pitch.
void setPitch(double newPitch);
/// Sets pitch change in octaves compared to the original pitch
/// (-1.00 .. +1.00)
void setPitchOctaves(double newPitch);
/// Sets pitch change in semi-tones compared to the original pitch
/// (-12 .. +12)
void setPitchSemiTones(int newPitch);
void setPitchSemiTones(double newPitch);
/// Sets the number of channels, 1 = mono, 2 = stereo
void setChannels(uint numChannels);
/// Sets sample rate.
void setSampleRate(uint srate);
/// Get ratio between input and output audio durations, useful for calculating
/// processed output duration: if you'll process a stream of N samples, then
/// you can expect to get out N * getInputOutputSampleRatio() samples.
///
/// This ratio will give accurate target duration ratio for a full audio track,
/// given that the the whole track is processed with same processing parameters.
///
/// If this ratio is applied to calculate intermediate offsets inside a processing
/// stream, then this ratio is approximate and can deviate +- some tens of milliseconds
/// from ideal offset, yet by end of the audio stream the duration ratio will become
/// exact.
///
/// Example: if processing with parameters "-tempo=15 -pitch=-3", the function
/// will return value 0.8695652... Now, if processing an audio stream whose duration
/// is exactly one million audio samples, then you can expect the processed
/// output duration be 0.869565 * 1000000 = 869565 samples.
double getInputOutputSampleRatio();
/// Flushes the last samples from the processing pipeline to the output.
/// Clears also the internal processing buffers.
//
/// Note: This function is meant for extracting the last samples of a sound
/// stream. This function may introduce additional blank samples in the end
/// of the sound stream, and thus it's not recommended to call this function
/// in the middle of a sound stream.
void flush();
/// Adds 'numSamples' pcs of samples from the 'samples' memory position into
/// the input of the object. Notice that sample rate _has_to_ be set before
/// calling this function, otherwise throws a runtime_error exception.
virtual void putSamples(
const SAMPLETYPE *samples, ///< Pointer to sample buffer.
uint numSamples ///< Number of samples in buffer. Notice
///< that in case of stereo-sound a single sample
///< contains data for both channels.
) override;
/// Output samples from beginning of the sample buffer. Copies requested samples to
/// output buffer and removes them from the sample buffer. If there are less than
/// 'numsample' samples in the buffer, returns all that available.
///
/// \return Number of samples returned.
virtual uint receiveSamples(SAMPLETYPE *output, ///< Buffer where to copy output samples.
uint maxSamples ///< How many samples to receive at max.
) override;
/// Adjusts book-keeping so that given number of samples are removed from beginning of the
/// sample buffer without copying them anywhere.
///
/// Used to reduce the number of samples in the buffer when accessing the sample buffer directly
/// with 'ptrBegin' function.
virtual uint receiveSamples(uint maxSamples ///< Remove this many samples from the beginning of pipe.
) override;
/// Clears all the samples in the object's output and internal processing
/// buffers.
virtual void clear() override;
/// Changes a setting controlling the processing system behaviour. See the
/// 'SETTING_...' defines for available setting ID's.
///
/// \return 'true' if the setting was successfully changed
bool setSetting(int settingId, ///< Setting ID number. see SETTING_... defines.
int value ///< New setting value.
);
/// Reads a setting controlling the processing system behaviour. See the
/// 'SETTING_...' defines for available setting ID's.
///
/// \return the setting value.
int getSetting(int settingId ///< Setting ID number, see SETTING_... defines.
) const;
/// Returns number of samples currently unprocessed.
virtual uint numUnprocessedSamples() const;
/// Return number of channels
uint numChannels() const
{
return channels;
}
/// Other handy functions that are implemented in the ancestor classes (see
/// classes 'FIFOProcessor' and 'FIFOSamplePipe')
///
/// - receiveSamples() : Use this function to receive 'ready' processed samples from SoundTouch.
/// - numSamples() : Get number of 'ready' samples that can be received with
/// function 'receiveSamples()'
/// - isEmpty() : Returns nonzero if there aren't any 'ready' samples.
/// - clear() : Clears all samples from ready/processing buffers.
};
}
#endif
//////////////////////////////////////////////////////////////////////////////
///
/// SoundTouch - main class for tempo/pitch/rate adjusting routines.
///
/// Notes:
/// - Initialize the SoundTouch object instance by setting up the sound stream
/// parameters with functions 'setSampleRate' and 'setChannels', then set
/// desired tempo/pitch/rate settings with the corresponding functions.
///
/// - The SoundTouch class behaves like a first-in-first-out pipeline: The
/// samples that are to be processed are fed into one of the pipe by calling
/// function 'putSamples', while the ready processed samples can be read
/// from the other end of the pipeline with function 'receiveSamples'.
///
/// - The SoundTouch processing classes require certain sized 'batches' of
/// samples in order to process the sound. For this reason the classes buffer
/// incoming samples until there are enough of samples available for
/// processing, then they carry out the processing step and consequently
/// make the processed samples available for outputting.
///
/// - For the above reason, the processing routines introduce a certain
/// 'latency' between the input and output, so that the samples input to
/// SoundTouch may not be immediately available in the output, and neither
/// the amount of outputtable samples may not immediately be in direct
/// relationship with the amount of previously input samples.
///
/// - The tempo/pitch/rate control parameters can be altered during processing.
/// Please notice though that they aren't currently protected by semaphores,
/// so in multi-thread application external semaphore protection may be
/// required.
///
/// - This class utilizes classes 'TDStretch' for tempo change (without modifying
/// pitch) and 'RateTransposer' for changing the playback rate (that is, both
/// tempo and pitch in the same ratio) of the sound. The third available control
/// 'pitch' (change pitch but maintain tempo) is produced by a combination of
/// combining the two other controls.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef SoundTouch_H
#define SoundTouch_H
#include "FIFOSamplePipe.h"
#include "STTypes.h"
namespace soundtouch
{
/// Soundtouch library version string
#define SOUNDTOUCH_VERSION "2.3.3"
/// SoundTouch library version id
#define SOUNDTOUCH_VERSION_ID (20303)
//
// Available setting IDs for the 'setSetting' & 'get_setting' functions:
/// Enable/disable anti-alias filter in pitch transposer (0 = disable)
#define SETTING_USE_AA_FILTER 0
/// Pitch transposer anti-alias filter length (8 .. 128 taps, default = 32)
#define SETTING_AA_FILTER_LENGTH 1
/// Enable/disable quick seeking algorithm in tempo changer routine
/// (enabling quick seeking lowers CPU utilization but causes a minor sound
/// quality compromising)
#define SETTING_USE_QUICKSEEK 2
/// Time-stretch algorithm single processing sequence length in milliseconds. This determines
/// to how long sequences the original sound is chopped in the time-stretch algorithm.
/// See "STTypes.h" or README for more information.
#define SETTING_SEQUENCE_MS 3
/// Time-stretch algorithm seeking window length in milliseconds for algorithm that finds the
/// best possible overlapping location. This determines from how wide window the algorithm
/// may look for an optimal joining location when mixing the sound sequences back together.
/// See "STTypes.h" or README for more information.
#define SETTING_SEEKWINDOW_MS 4
/// Time-stretch algorithm overlap length in milliseconds. When the chopped sound sequences
/// are mixed back together, to form a continuous sound stream, this parameter defines over
/// how long period the two consecutive sequences are let to overlap each other.
/// See "STTypes.h" or README for more information.
#define SETTING_OVERLAP_MS 5
/// Call "getSetting" with this ID to query processing sequence size in samples.
/// This value gives approximate value of how many input samples you'll need to
/// feed into SoundTouch after initial buffering to get out a new batch of
/// output samples.
///
/// This value does not include initial buffering at beginning of a new processing
/// stream, use SETTING_INITIAL_LATENCY to get the initial buffering size.
///
/// Notices:
/// - This is read-only parameter, i.e. setSetting ignores this parameter
/// - This parameter value is not constant but change depending on
/// tempo/pitch/rate/samplerate settings.
#define SETTING_NOMINAL_INPUT_SEQUENCE 6
/// Call "getSetting" with this ID to query nominal average processing output
/// size in samples. This value tells approcimate value how many output samples
/// SoundTouch outputs once it does DSP processing run for a batch of input samples.
///
/// Notices:
/// - This is read-only parameter, i.e. setSetting ignores this parameter
/// - This parameter value is not constant but change depending on
/// tempo/pitch/rate/samplerate settings.
#define SETTING_NOMINAL_OUTPUT_SEQUENCE 7
/// Call "getSetting" with this ID to query initial processing latency, i.e.
/// approx. how many samples you'll need to enter to SoundTouch pipeline before
/// you can expect to get first batch of ready output samples out.
///
/// After the first output batch, you can then expect to get approx.
/// SETTING_NOMINAL_OUTPUT_SEQUENCE ready samples out for every
/// SETTING_NOMINAL_INPUT_SEQUENCE samples that you enter into SoundTouch.
///
/// Example:
/// processing with parameter -tempo=5
/// => initial latency = 5509 samples
/// input sequence = 4167 samples
/// output sequence = 3969 samples
///
/// Accordingly, you can expect to feed in approx. 5509 samples at beginning of
/// the stream, and then you'll get out the first 3969 samples. After that, for
/// every approx. 4167 samples that you'll put in, you'll receive again approx.
/// 3969 samples out.
///
/// This also means that average latency during stream processing is
/// INITIAL_LATENCY-OUTPUT_SEQUENCE/2, in the above example case 5509-3969/2
/// = 3524 samples
///
/// Notices:
/// - This is read-only parameter, i.e. setSetting ignores this parameter
/// - This parameter value is not constant but change depending on
/// tempo/pitch/rate/samplerate settings.
#define SETTING_INITIAL_LATENCY 8
class SoundTouch : public FIFOProcessor
{
private:
/// Rate transposer class instance
class RateTransposer *pRateTransposer;
/// Time-stretch class instance
class TDStretch *pTDStretch;
/// Virtual pitch parameter. Effective rate & tempo are calculated from these parameters.
double virtualRate;
/// Virtual pitch parameter. Effective rate & tempo are calculated from these parameters.
double virtualTempo;
/// Virtual pitch parameter. Effective rate & tempo are calculated from these parameters.
double virtualPitch;
/// Flag: Has sample rate been set?
bool bSrateSet;
/// Accumulator for how many samples in total will be expected as output vs. samples put in,
/// considering current processing settings.
double samplesExpectedOut;
/// Accumulator for how many samples in total have been read out from the processing so far
long samplesOutput;
/// Calculates effective rate & tempo valuescfrom 'virtualRate', 'virtualTempo' and
/// 'virtualPitch' parameters.
void calcEffectiveRateAndTempo();
protected :
/// Number of channels
uint channels;
/// Effective 'rate' value calculated from 'virtualRate', 'virtualTempo' and 'virtualPitch'
double rate;
/// Effective 'tempo' value calculated from 'virtualRate', 'virtualTempo' and 'virtualPitch'
double tempo;
public:
SoundTouch();
virtual ~SoundTouch() override;
/// Get SoundTouch library version string
static const char *getVersionString();
/// Get SoundTouch library version Id
static uint getVersionId();
/// Sets new rate control value. Normal rate = 1.0, smaller values
/// represent slower rate, larger faster rates.
void setRate(double newRate);
/// Sets new tempo control value. Normal tempo = 1.0, smaller values
/// represent slower tempo, larger faster tempo.
void setTempo(double newTempo);
/// Sets new rate control value as a difference in percents compared
/// to the original rate (-50 .. +100 %)
void setRateChange(double newRate);
/// Sets new tempo control value as a difference in percents compared
/// to the original tempo (-50 .. +100 %)
void setTempoChange(double newTempo);
/// Sets new pitch control value. Original pitch = 1.0, smaller values
/// represent lower pitches, larger values higher pitch.
void setPitch(double newPitch);
/// Sets pitch change in octaves compared to the original pitch
/// (-1.00 .. +1.00)
void setPitchOctaves(double newPitch);
/// Sets pitch change in semi-tones compared to the original pitch
/// (-12 .. +12)
void setPitchSemiTones(int newPitch);
void setPitchSemiTones(double newPitch);
/// Sets the number of channels, 1 = mono, 2 = stereo
void setChannels(uint numChannels);
/// Sets sample rate.
void setSampleRate(uint srate);
/// Get ratio between input and output audio durations, useful for calculating
/// processed output duration: if you'll process a stream of N samples, then
/// you can expect to get out N * getInputOutputSampleRatio() samples.
///
/// This ratio will give accurate target duration ratio for a full audio track,
/// given that the the whole track is processed with same processing parameters.
///
/// If this ratio is applied to calculate intermediate offsets inside a processing
/// stream, then this ratio is approximate and can deviate +- some tens of milliseconds
/// from ideal offset, yet by end of the audio stream the duration ratio will become
/// exact.
///
/// Example: if processing with parameters "-tempo=15 -pitch=-3", the function
/// will return value 0.8695652... Now, if processing an audio stream whose duration
/// is exactly one million audio samples, then you can expect the processed
/// output duration be 0.869565 * 1000000 = 869565 samples.
double getInputOutputSampleRatio();
/// Flushes the last samples from the processing pipeline to the output.
/// Clears also the internal processing buffers.
//
/// Note: This function is meant for extracting the last samples of a sound
/// stream. This function may introduce additional blank samples in the end
/// of the sound stream, and thus it's not recommended to call this function
/// in the middle of a sound stream.
void flush();
/// Adds 'numSamples' pcs of samples from the 'samples' memory position into
/// the input of the object. Notice that sample rate _has_to_ be set before
/// calling this function, otherwise throws a runtime_error exception.
virtual void putSamples(
const SAMPLETYPE *samples, ///< Pointer to sample buffer.
uint numSamples ///< Number of samples in buffer. Notice
///< that in case of stereo-sound a single sample
///< contains data for both channels.
) override;
/// Output samples from beginning of the sample buffer. Copies requested samples to
/// output buffer and removes them from the sample buffer. If there are less than
/// 'numsample' samples in the buffer, returns all that available.
///
/// \return Number of samples returned.
virtual uint receiveSamples(SAMPLETYPE *output, ///< Buffer where to copy output samples.
uint maxSamples ///< How many samples to receive at max.
) override;
/// Adjusts book-keeping so that given number of samples are removed from beginning of the
/// sample buffer without copying them anywhere.
///
/// Used to reduce the number of samples in the buffer when accessing the sample buffer directly
/// with 'ptrBegin' function.
virtual uint receiveSamples(uint maxSamples ///< Remove this many samples from the beginning of pipe.
) override;
/// Clears all the samples in the object's output and internal processing
/// buffers.
virtual void clear() override;
/// Changes a setting controlling the processing system behaviour. See the
/// 'SETTING_...' defines for available setting ID's.
///
/// \return 'true' if the setting was successfully changed
bool setSetting(int settingId, ///< Setting ID number. see SETTING_... defines.
int value ///< New setting value.
);
/// Reads a setting controlling the processing system behaviour. See the
/// 'SETTING_...' defines for available setting ID's.
///
/// \return the setting value.
int getSetting(int settingId ///< Setting ID number, see SETTING_... defines.
) const;
/// Returns number of samples currently unprocessed.
virtual uint numUnprocessedSamples() const;
/// Return number of channels
uint numChannels() const
{
return channels;
}
/// Other handy functions that are implemented in the ancestor classes (see
/// classes 'FIFOProcessor' and 'FIFOSamplePipe')
///
/// - receiveSamples() : Use this function to receive 'ready' processed samples from SoundTouch.
/// - numSamples() : Get number of 'ready' samples that can be received with
/// function 'receiveSamples()'
/// - isEmpty() : Returns nonzero if there aren't any 'ready' samples.
/// - clear() : Clears all samples from ready/processing buffers.
};
}
#endif

8
soundtouch.m4
View File

@ -8,7 +8,7 @@
# It also defines some flags to the configure script for specifying
# the location to search for libSoundTouch
#
# A user of libSoundTouch should add @SOUNDTOUCH_LIBS@ and
# A user of libSoundTouch should add @SOUNDTOUCH_LIBS@ and
# @SOUNDTOUCH_CXXFLAGS@ to the appropriate variables in his
# Makefile.am files
#
@ -32,10 +32,10 @@ AC_DEFUN([AM_PATH_SOUNDTOUCH],[
then
saved_CPPFLAGS="$CPPFLAGS"
saved_LDFLAGS="$LDFLAGS"
CPPFLAGS="$CPPFLAGS -I$soundtouch_prefix/include"
LDFLAGS="$LDFLAGS -L$soundtouch_prefix/lib"
dnl make sure SoundTouch.h header file exists
dnl could use AC_CHECK_HEADERS to check for all of them, but the supporting .h file names may change later
AC_CHECK_HEADER([soundtouch/SoundTouch.h],[
@ -49,7 +49,7 @@ AC_DEFUN([AM_PATH_SOUNDTOUCH],[
dnl run action-if-found
ifelse([$2], , :, [$2])
],[
],[
dnl run action-if-not-found
ifelse([$3], , :, [$3])
])

516
source/Android-lib/jni/soundtouch-jni.cpp
View File

@ -1,258 +1,258 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Example Interface class for SoundTouch native compilation
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// WWW : http://www.surina.net
///
////////////////////////////////////////////////////////////////////////////////
#include <jni.h>
#include <android/log.h>
#include <stdexcept>
#include <string>
using namespace std;
#include "../../../include/SoundTouch.h"
#include "../source/SoundStretch/WavFile.h"
#define LOGV(...) __android_log_print((int)ANDROID_LOG_INFO, "SOUNDTOUCH", __VA_ARGS__)
//#define LOGV(...)
// String for keeping possible c++ exception error messages. Notice that this isn't
// thread-safe but it's expected that exceptions are special situations that won't
// occur in several threads in parallel.
static string _errMsg = "";
#define DLL_PUBLIC __attribute__ ((visibility ("default")))
#define BUFF_SIZE 4096
using namespace soundtouch;
// Set error message to return
static void _setErrmsg(const char *msg)
{
_errMsg = msg;
}
#if 0 // apparently following workaround not needed any more with concurrent Android SDKs
#ifdef _OPENMP
#include <pthread.h>
extern pthread_key_t gomp_tls_key;
static void * _p_gomp_tls = nullptr;
/// Function to initialize threading for OpenMP.
///
/// This is a workaround for bug in Android NDK v10 regarding OpenMP: OpenMP works only if
/// called from the Android App main thread because in the main thread the gomp_tls storage is
/// properly set, however, Android does not properly initialize gomp_tls storage for other threads.
/// Thus if OpenMP routines are invoked from some other thread than the main thread,
/// the OpenMP routine will crash the application due to nullptr access on uninitialized storage.
///
/// This workaround stores the gomp_tls storage from main thread, and copies to other threads.
/// In order this to work, the Application main thread needws to call at least "getVersionString"
/// routine.
static int _init_threading(bool warn)
{
void *ptr = pthread_getspecific(gomp_tls_key);
LOGV("JNI thread-specific TLS storage %ld", (long)ptr);
if (ptr == nullptr)
{
LOGV("JNI set missing TLS storage to %ld", (long)_p_gomp_tls);
pthread_setspecific(gomp_tls_key, _p_gomp_tls);
}
else
{
LOGV("JNI store this TLS storage");
_p_gomp_tls = ptr;
}
// Where critical, show warning if storage still not properly initialized
if ((warn) && (_p_gomp_tls == nullptr))
{
_setErrmsg("Error - OpenMP threading not properly initialized: Call SoundTouch.getVersionString() from the App main thread!");
return -1;
}
return 0;
}
#else
static int _init_threading(bool warn)
{
// do nothing if not OpenMP build
return 0;
}
#endif
#endif
// Processes the sound file
static void _processFile(SoundTouch *pSoundTouch, const char *inFileName, const char *outFileName)
{
int nSamples;
int nChannels;
int buffSizeSamples;
SAMPLETYPE sampleBuffer[BUFF_SIZE];
// open input file
WavInFile inFile(inFileName);
int sampleRate = inFile.getSampleRate();
int bits = inFile.getNumBits();
nChannels = inFile.getNumChannels();
// create output file
WavOutFile outFile(outFileName, sampleRate, bits, nChannels);
pSoundTouch->setSampleRate(sampleRate);
pSoundTouch->setChannels(nChannels);
assert(nChannels > 0);
buffSizeSamples = BUFF_SIZE / nChannels;
// Process samples read from the input file
while (inFile.eof() == 0)
{
int num;
// Read a chunk of samples from the input file
num = inFile.read(sampleBuffer, BUFF_SIZE);
nSamples = num / nChannels;
// Feed the samples into SoundTouch processor
pSoundTouch->putSamples(sampleBuffer, nSamples);
// Read ready samples from SoundTouch processor & write them output file.
// NOTES:
// - 'receiveSamples' doesn't necessarily return any samples at all
// during some rounds!
// - On the other hand, during some round 'receiveSamples' may have more
// ready samples than would fit into 'sampleBuffer', and for this reason
// the 'receiveSamples' call is iterated for as many times as it
// outputs samples.
do
{
nSamples = pSoundTouch->receiveSamples(sampleBuffer, buffSizeSamples);
outFile.write(sampleBuffer, nSamples * nChannels);
} while (nSamples != 0);
}
// Now the input file is processed, yet 'flush' few last samples that are
// hiding in the SoundTouch's internal processing pipeline.
pSoundTouch->flush();
do
{
nSamples = pSoundTouch->receiveSamples(sampleBuffer, buffSizeSamples);
outFile.write(sampleBuffer, nSamples * nChannels);
} while (nSamples != 0);
}
extern "C" DLL_PUBLIC jstring Java_net_surina_soundtouch_SoundTouch_getVersionString(JNIEnv *env, jobject thiz)
{
const char *verStr;
LOGV("JNI call SoundTouch.getVersionString");
// Call example SoundTouch routine
verStr = SoundTouch::getVersionString();
// gomp_tls storage bug workaround - see comments in _init_threading() function!
// update: apparently this is not needed any more with concurrent Android SDKs
// _init_threading(false);
int threads = 0;
#pragma omp parallel
{
#pragma omp atomic
threads ++;
}
LOGV("JNI thread count %d", threads);
// return version as string
return env->NewStringUTF(verStr);
}
extern "C" DLL_PUBLIC jlong Java_net_surina_soundtouch_SoundTouch_newInstance(JNIEnv *env, jobject thiz)
{
return (jlong)(new SoundTouch());
}
extern "C" DLL_PUBLIC void Java_net_surina_soundtouch_SoundTouch_deleteInstance(JNIEnv *env, jobject thiz, jlong handle)
{
SoundTouch *ptr = (SoundTouch*)handle;
delete ptr;
}
extern "C" DLL_PUBLIC void Java_net_surina_soundtouch_SoundTouch_setTempo(JNIEnv *env, jobject thiz, jlong handle, jfloat tempo)
{
SoundTouch *ptr = (SoundTouch*)handle;
ptr->setTempo(tempo);
}
extern "C" DLL_PUBLIC void Java_net_surina_soundtouch_SoundTouch_setPitchSemiTones(JNIEnv *env, jobject thiz, jlong handle, jfloat pitch)
{
SoundTouch *ptr = (SoundTouch*)handle;
ptr->setPitchSemiTones(pitch);
}
extern "C" DLL_PUBLIC void Java_net_surina_soundtouch_SoundTouch_setSpeed(JNIEnv *env, jobject thiz, jlong handle, jfloat speed)
{
SoundTouch *ptr = (SoundTouch*)handle;
ptr->setRate(speed);
}
extern "C" DLL_PUBLIC jstring Java_net_surina_soundtouch_SoundTouch_getErrorString(JNIEnv *env, jobject thiz)
{
jstring result = env->NewStringUTF(_errMsg.c_str());
_errMsg.clear();
return result;
}
extern "C" DLL_PUBLIC int Java_net_surina_soundtouch_SoundTouch_processFile(JNIEnv *env, jobject thiz, jlong handle, jstring jinputFile, jstring joutputFile)
{
SoundTouch *ptr = (SoundTouch*)handle;
const char *inputFile = env->GetStringUTFChars(jinputFile, 0);
const char *outputFile = env->GetStringUTFChars(joutputFile, 0);
LOGV("JNI process file %s", inputFile);
/// gomp_tls storage bug workaround - see comments in _init_threading() function!
// update: apparently this is not needed any more with concurrent Android SDKs
// if (_init_threading(true)) return -1;
try
{
_processFile(ptr, inputFile, outputFile);
}
catch (const runtime_error &e)
{
const char *err = e.what();
// An exception occurred during processing, return the error message
LOGV("JNI exception in SoundTouch::processFile: %s", err);
_setErrmsg(err);
return -1;
}
env->ReleaseStringUTFChars(jinputFile, inputFile);
env->ReleaseStringUTFChars(joutputFile, outputFile);
return 0;
}
////////////////////////////////////////////////////////////////////////////////
///
/// Example Interface class for SoundTouch native compilation
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// WWW : http://www.surina.net
///
////////////////////////////////////////////////////////////////////////////////
#include <jni.h>
#include <android/log.h>
#include <stdexcept>
#include <string>
using namespace std;
#include "../../../include/SoundTouch.h"
#include "../source/SoundStretch/WavFile.h"
#define LOGV(...) __android_log_print((int)ANDROID_LOG_INFO, "SOUNDTOUCH", __VA_ARGS__)
//#define LOGV(...)
// String for keeping possible c++ exception error messages. Notice that this isn't
// thread-safe but it's expected that exceptions are special situations that won't
// occur in several threads in parallel.
static string _errMsg = "";
#define DLL_PUBLIC __attribute__ ((visibility ("default")))
#define BUFF_SIZE 4096
using namespace soundtouch;
// Set error message to return
static void _setErrmsg(const char *msg)
{
_errMsg = msg;
}
#if 0 // apparently following workaround not needed any more with concurrent Android SDKs
#ifdef _OPENMP
#include <pthread.h>
extern pthread_key_t gomp_tls_key;
static void * _p_gomp_tls = nullptr;
/// Function to initialize threading for OpenMP.
///
/// This is a workaround for bug in Android NDK v10 regarding OpenMP: OpenMP works only if
/// called from the Android App main thread because in the main thread the gomp_tls storage is
/// properly set, however, Android does not properly initialize gomp_tls storage for other threads.
/// Thus if OpenMP routines are invoked from some other thread than the main thread,
/// the OpenMP routine will crash the application due to nullptr access on uninitialized storage.
///
/// This workaround stores the gomp_tls storage from main thread, and copies to other threads.
/// In order this to work, the Application main thread needws to call at least "getVersionString"
/// routine.
static int _init_threading(bool warn)
{
void *ptr = pthread_getspecific(gomp_tls_key);
LOGV("JNI thread-specific TLS storage %ld", (long)ptr);
if (ptr == nullptr)
{
LOGV("JNI set missing TLS storage to %ld", (long)_p_gomp_tls);
pthread_setspecific(gomp_tls_key, _p_gomp_tls);
}
else
{
LOGV("JNI store this TLS storage");
_p_gomp_tls = ptr;
}
// Where critical, show warning if storage still not properly initialized
if ((warn) && (_p_gomp_tls == nullptr))
{
_setErrmsg("Error - OpenMP threading not properly initialized: Call SoundTouch.getVersionString() from the App main thread!");
return -1;
}
return 0;
}
#else
static int _init_threading(bool warn)
{
// do nothing if not OpenMP build
return 0;
}
#endif
#endif
// Processes the sound file
static void _processFile(SoundTouch *pSoundTouch, const char *inFileName, const char *outFileName)
{
int nSamples;
int nChannels;
int buffSizeSamples;
SAMPLETYPE sampleBuffer[BUFF_SIZE];
// open input file
WavInFile inFile(inFileName);
int sampleRate = inFile.getSampleRate();
int bits = inFile.getNumBits();
nChannels = inFile.getNumChannels();
// create output file
WavOutFile outFile(outFileName, sampleRate, bits, nChannels);
pSoundTouch->setSampleRate(sampleRate);
pSoundTouch->setChannels(nChannels);
assert(nChannels > 0);
buffSizeSamples = BUFF_SIZE / nChannels;
// Process samples read from the input file
while (inFile.eof() == 0)
{
int num;
// Read a chunk of samples from the input file
num = inFile.read(sampleBuffer, BUFF_SIZE);
nSamples = num / nChannels;
// Feed the samples into SoundTouch processor
pSoundTouch->putSamples(sampleBuffer, nSamples);
// Read ready samples from SoundTouch processor & write them output file.
// NOTES:
// - 'receiveSamples' doesn't necessarily return any samples at all
// during some rounds!
// - On the other hand, during some round 'receiveSamples' may have more
// ready samples than would fit into 'sampleBuffer', and for this reason
// the 'receiveSamples' call is iterated for as many times as it
// outputs samples.
do
{
nSamples = pSoundTouch->receiveSamples(sampleBuffer, buffSizeSamples);
outFile.write(sampleBuffer, nSamples * nChannels);
} while (nSamples != 0);
}
// Now the input file is processed, yet 'flush' few last samples that are
// hiding in the SoundTouch's internal processing pipeline.
pSoundTouch->flush();
do
{
nSamples = pSoundTouch->receiveSamples(sampleBuffer, buffSizeSamples);
outFile.write(sampleBuffer, nSamples * nChannels);
} while (nSamples != 0);
}
extern "C" DLL_PUBLIC jstring Java_net_surina_soundtouch_SoundTouch_getVersionString(JNIEnv *env, jobject thiz)
{
const char *verStr;
LOGV("JNI call SoundTouch.getVersionString");
// Call example SoundTouch routine
verStr = SoundTouch::getVersionString();
// gomp_tls storage bug workaround - see comments in _init_threading() function!
// update: apparently this is not needed any more with concurrent Android SDKs
// _init_threading(false);
int threads = 0;
#pragma omp parallel
{
#pragma omp atomic
threads ++;
}
LOGV("JNI thread count %d", threads);
// return version as string
return env->NewStringUTF(verStr);
}
extern "C" DLL_PUBLIC jlong Java_net_surina_soundtouch_SoundTouch_newInstance(JNIEnv *env, jobject thiz)
{
return (jlong)(new SoundTouch());
}
extern "C" DLL_PUBLIC void Java_net_surina_soundtouch_SoundTouch_deleteInstance(JNIEnv *env, jobject thiz, jlong handle)
{
SoundTouch *ptr = (SoundTouch*)handle;
delete ptr;
}
extern "C" DLL_PUBLIC void Java_net_surina_soundtouch_SoundTouch_setTempo(JNIEnv *env, jobject thiz, jlong handle, jfloat tempo)
{
SoundTouch *ptr = (SoundTouch*)handle;
ptr->setTempo(tempo);
}
extern "C" DLL_PUBLIC void Java_net_surina_soundtouch_SoundTouch_setPitchSemiTones(JNIEnv *env, jobject thiz, jlong handle, jfloat pitch)
{
SoundTouch *ptr = (SoundTouch*)handle;
ptr->setPitchSemiTones(pitch);
}
extern "C" DLL_PUBLIC void Java_net_surina_soundtouch_SoundTouch_setSpeed(JNIEnv *env, jobject thiz, jlong handle, jfloat speed)
{
SoundTouch *ptr = (SoundTouch*)handle;
ptr->setRate(speed);
}
extern "C" DLL_PUBLIC jstring Java_net_surina_soundtouch_SoundTouch_getErrorString(JNIEnv *env, jobject thiz)
{
jstring result = env->NewStringUTF(_errMsg.c_str());
_errMsg.clear();
return result;
}
extern "C" DLL_PUBLIC int Java_net_surina_soundtouch_SoundTouch_processFile(JNIEnv *env, jobject thiz, jlong handle, jstring jinputFile, jstring joutputFile)
{
SoundTouch *ptr = (SoundTouch*)handle;
const char *inputFile = env->GetStringUTFChars(jinputFile, 0);
const char *outputFile = env->GetStringUTFChars(joutputFile, 0);
LOGV("JNI process file %s", inputFile);
/// gomp_tls storage bug workaround - see comments in _init_threading() function!
// update: apparently this is not needed any more with concurrent Android SDKs
// if (_init_threading(true)) return -1;
try
{
_processFile(ptr, inputFile, outputFile);
}
catch (const runtime_error &e)
{
const char *err = e.what();
// An exception occurred during processing, return the error message
LOGV("JNI exception in SoundTouch::processFile: %s", err);
_setErrmsg(err);
return -1;
}
env->ReleaseStringUTFChars(jinputFile, inputFile);
env->ReleaseStringUTFChars(joutputFile, outputFile);
return 0;
}

48
source/Makefile.am
View File

@ -1,24 +1,24 @@
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA
include $(top_srcdir)/config/am_include.mk
SUBDIRS=SoundTouch SoundStretch SoundTouchDLL
# set to something if you want other stuff to be included in the distribution tarball
#EXTRA_DIST=
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA
include $(top_srcdir)/config/am_include.mk
SUBDIRS=SoundTouch SoundStretch SoundTouchDLL
# set to something if you want other stuff to be included in the distribution tarball
#EXTRA_DIST=

100
source/SoundStretch/Makefile.am
View File

@ -1,50 +1,50 @@
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA
include $(top_srcdir)/config/am_include.mk
## bin_PROGRAMS is the macro that tells automake the name of the programs to
## install in the bin directory (/usr/local/bin) by default. By setting
## --prefix= at configure time the user can change this (eg: ./configure
## --prefix=/usr will install soundstretch under /usr/bin/soundstretch )
bin_PROGRAMS=soundstretch
noinst_HEADERS=RunParameters.h WavFile.h
# extra files to include in distribution tarball
EXTRA_DIST=soundstretch.sln soundstretch.vcxproj
## for every name listed under bin_PROGRAMS, you have a <prog>_SOURCES. This lists
## all the sources in the current directory that are used to build soundstretch.
soundstretch_SOURCES=main.cpp RunParameters.cpp WavFile.cpp
## soundstretch_LDADD is a list of extras to pass at link time. All the objects
## created by the above soundstretch_SOURCES are automatically linked in, so here I
## list object files from other directories as well as flags passed to the
## linker.
soundstretch_LDADD=../SoundTouch/libSoundTouch.la -lm
## linker flags.
# Linker flag -s disabled to prevent stripping symbols by default
#soundstretch_LDFLAGS=-s
## additional compiler flags
soundstretch_CXXFLAGS=$(AM_CXXFLAGS)
#clean-local:
# -rm -f additional-files-to-remove-on-make-clean
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA
include $(top_srcdir)/config/am_include.mk
## bin_PROGRAMS is the macro that tells automake the name of the programs to
## install in the bin directory (/usr/local/bin) by default. By setting
## --prefix= at configure time the user can change this (eg: ./configure
## --prefix=/usr will install soundstretch under /usr/bin/soundstretch )
bin_PROGRAMS=soundstretch
noinst_HEADERS=RunParameters.h WavFile.h
# extra files to include in distribution tarball
EXTRA_DIST=soundstretch.sln soundstretch.vcxproj
## for every name listed under bin_PROGRAMS, you have a <prog>_SOURCES. This lists
## all the sources in the current directory that are used to build soundstretch.
soundstretch_SOURCES=main.cpp RunParameters.cpp WavFile.cpp
## soundstretch_LDADD is a list of extras to pass at link time. All the objects
## created by the above soundstretch_SOURCES are automatically linked in, so here I
## list object files from other directories as well as flags passed to the
## linker.
soundstretch_LDADD=../SoundTouch/libSoundTouch.la -lm
## linker flags.
# Linker flag -s disabled to prevent stripping symbols by default
#soundstretch_LDFLAGS=-s
## additional compiler flags
soundstretch_CXXFLAGS=$(AM_CXXFLAGS)
#clean-local:
# -rm -f additional-files-to-remove-on-make-clean

584
source/SoundStretch/RunParameters.cpp
View File

@ -1,292 +1,292 @@
////////////////////////////////////////////////////////////////////////////////
///
/// A class for parsing the 'soundstretch' application command line parameters
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <string>
#include <cstdlib>
#include "RunParameters.h"
using namespace std;
namespace soundstretch
{
// Program usage instructions
static const char licenseText[] =
" LICENSE:\n"
" ========\n"
" \n"
" SoundTouch sound processing library\n"
" Copyright (c) Olli Parviainen\n"
" \n"
" This library is free software; you can redistribute it and/or\n"
" modify it under the terms of the GNU Lesser General Public\n"
" License version 2.1 as published by the Free Software Foundation.\n"
" \n"
" This library is distributed in the hope that it will be useful,\n"
" but WITHOUT ANY WARRANTY; without even the implied warranty of\n"
" MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU\n"
" Lesser General Public License for more details.\n"
" \n"
" You should have received a copy of the GNU Lesser General Public\n"
" License along with this library; if not, write to the Free Software\n"
" Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA\n"
" \n"
"This application is distributed with full source codes; however, if you\n"
"didn't receive them, please visit the author's homepage (see the link above).";
static const char whatText[] =
"This application processes WAV audio files by modifying the sound tempo,\n"
"pitch and playback rate properties independently from each other.\n"
"\n";
static const char usage[] =
"Usage :\n"
" soundstretch infilename outfilename [switches]\n"
"\n"
"To use standard input/output pipes, give 'stdin' and 'stdout' as filenames.\n"
"\n"
"Available switches are:\n"
" -tempo=n : Change sound tempo by n percents (n=-95..+5000 %)\n"
" -pitch=n : Change sound pitch by n semitones (n=-60..+60 semitones)\n"
" -rate=n : Change sound rate by n percents (n=-95..+5000 %)\n"
" -bpm=n : Detect the BPM rate of sound and adjust tempo to meet 'n' BPMs.\n"
" If '=n' is omitted, just detects the BPM rate.\n"
" -quick : Use quicker tempo change algorithm (gain speed, lose quality)\n"
" -naa : Don't use anti-alias filtering (gain speed, lose quality)\n"
" -speech : Tune algorithm for speech processing (default is for music)\n"
" -license : Display the program license text (LGPL)\n";
// Converts a char into lower case
static int _toLowerCase(int c)
{
if (c >= 'A' && c <= 'Z')
{
c += 'a' - 'A';
}
return c;
}
// Constructor
RunParameters::RunParameters(int nParams, const CHARTYPE* paramStr[])
{
int i;
int nFirstParam;
if (nParams < 3)
{
// Too few parameters
if (nParams > 1 && paramStr[1][0] == '-' &&
_toLowerCase(paramStr[1][1]) == 'l')
{
// '-license' switch
throwLicense();
}
string msg = whatText;
msg += usage;
throw(msg);
}
// Get input & output file names
inFileName = paramStr[1];
outFileName = paramStr[2];
if (outFileName[0] == '-')
{
// outputfile name was omitted but other parameter switches given instead
outFileName = STRING_CONST("");
nFirstParam = 2;
}
else
{
nFirstParam = 3;
}
// parse switch parameters
for (i = nFirstParam; i < nParams; i ++)
{
parseSwitchParam(paramStr[i]);
}
checkLimits();
}
// Checks parameter limits
void RunParameters::checkLimits()
{
if (tempoDelta < -95.0f)
{
tempoDelta = -95.0f;
}
else if (tempoDelta > 5000.0f)
{
tempoDelta = 5000.0f;
}
if (pitchDelta < -60.0f)
{
pitchDelta = -60.0f;
}
else if (pitchDelta > 60.0f)
{
pitchDelta = 60.0f;
}
if (rateDelta < -95.0f)
{
rateDelta = -95.0f;
}
else if (rateDelta > 5000.0f)
{
rateDelta = 5000.0f;
}
}
// Convert STRING to std::string. Actually needed only if STRING is std::wstring, but conversion penalty is negligible
std::string convertString(const STRING& str)
{
std::string res;
for (auto c : str)
{
res += (char)c;
}
return res;
}
// Unknown switch parameter -- throws an exception with an error message
void RunParameters::throwIllegalParamExp(const STRING &str) const
{
string msg = "ERROR : Illegal parameter \"";
msg += convertString(str);
msg += "\".\n\n";
msg += usage;
ST_THROW_RT_ERROR(msg);
}
void RunParameters::throwLicense() const
{
ST_THROW_RT_ERROR(licenseText);
}
float RunParameters::parseSwitchValue(const STRING& str) const
{
int pos;
pos = (int)str.find_first_of('=');
if (pos < 0)
{
// '=' missing
throwIllegalParamExp(str);
}
// Read numerical parameter value after '='
return (float)stof(str.substr(pos + 1).c_str());
}
// Interprets a single switch parameter string of format "-switch=xx"
// Valid switches are "-tempo=xx", "-pitch=xx" and "-rate=xx". Stores
// switch values into 'params' structure.
void RunParameters::parseSwitchParam(const STRING& str)
{
int upS;
if (str[0] != '-')
{
// leading hyphen missing => not a valid parameter
throwIllegalParamExp(str);
}
// Take the first character of switch name & change to lower case
upS = _toLowerCase(str[1]);
// interpret the switch name & operate accordingly
switch (upS)
{
case 't' :
// switch '-tempo=xx'
tempoDelta = parseSwitchValue(str);
break;
case 'p' :
// switch '-pitch=xx'
pitchDelta = parseSwitchValue(str);
break;
case 'r' :
// switch '-rate=xx'
rateDelta = parseSwitchValue(str);
break;
case 'b' :
// switch '-bpm=xx'
detectBPM = true;
try
{
goalBPM = parseSwitchValue(str);
}
catch (const runtime_error &)
{
// illegal or missing bpm value => just calculate bpm
goalBPM = 0;
}
break;
case 'q' :
// switch '-quick'
quick = 1;
break;
case 'n' :
// switch '-naa'
noAntiAlias = 1;
break;
case 'l' :
// switch '-license'
throwLicense();
break;
case 's' :
// switch '-speech'
speech = true;
break;
default:
// unknown switch
throwIllegalParamExp(str);
}
}
}
////////////////////////////////////////////////////////////////////////////////
///
/// A class for parsing the 'soundstretch' application command line parameters
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <string>
#include <cstdlib>
#include "RunParameters.h"
using namespace std;
namespace soundstretch
{
// Program usage instructions
static const char licenseText[] =
" LICENSE:\n"
" ========\n"
" \n"
" SoundTouch sound processing library\n"
" Copyright (c) Olli Parviainen\n"
" \n"
" This library is free software; you can redistribute it and/or\n"
" modify it under the terms of the GNU Lesser General Public\n"
" License version 2.1 as published by the Free Software Foundation.\n"
" \n"
" This library is distributed in the hope that it will be useful,\n"
" but WITHOUT ANY WARRANTY; without even the implied warranty of\n"
" MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU\n"
" Lesser General Public License for more details.\n"
" \n"
" You should have received a copy of the GNU Lesser General Public\n"
" License along with this library; if not, write to the Free Software\n"
" Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA\n"
" \n"
"This application is distributed with full source codes; however, if you\n"
"didn't receive them, please visit the author's homepage (see the link above).";
static const char whatText[] =
"This application processes WAV audio files by modifying the sound tempo,\n"
"pitch and playback rate properties independently from each other.\n"
"\n";
static const char usage[] =
"Usage :\n"
" soundstretch infilename outfilename [switches]\n"
"\n"
"To use standard input/output pipes, give 'stdin' and 'stdout' as filenames.\n"
"\n"
"Available switches are:\n"
" -tempo=n : Change sound tempo by n percents (n=-95..+5000 %)\n"
" -pitch=n : Change sound pitch by n semitones (n=-60..+60 semitones)\n"
" -rate=n : Change sound rate by n percents (n=-95..+5000 %)\n"
" -bpm=n : Detect the BPM rate of sound and adjust tempo to meet 'n' BPMs.\n"
" If '=n' is omitted, just detects the BPM rate.\n"
" -quick : Use quicker tempo change algorithm (gain speed, lose quality)\n"
" -naa : Don't use anti-alias filtering (gain speed, lose quality)\n"
" -speech : Tune algorithm for speech processing (default is for music)\n"
" -license : Display the program license text (LGPL)\n";
// Converts a char into lower case
static int _toLowerCase(int c)
{
if (c >= 'A' && c <= 'Z')
{
c += 'a' - 'A';
}
return c;
}
// Constructor
RunParameters::RunParameters(int nParams, const CHARTYPE* paramStr[])
{
int i;
int nFirstParam;
if (nParams < 3)
{
// Too few parameters
if (nParams > 1 && paramStr[1][0] == '-' &&
_toLowerCase(paramStr[1][1]) == 'l')
{
// '-license' switch
throwLicense();
}
string msg = whatText;
msg += usage;
throw(msg);
}
// Get input & output file names
inFileName = paramStr[1];
outFileName = paramStr[2];
if (outFileName[0] == '-')
{
// outputfile name was omitted but other parameter switches given instead
outFileName = STRING_CONST("");
nFirstParam = 2;
}
else
{
nFirstParam = 3;
}
// parse switch parameters
for (i = nFirstParam; i < nParams; i ++)
{
parseSwitchParam(paramStr[i]);
}
checkLimits();
}
// Checks parameter limits
void RunParameters::checkLimits()
{
if (tempoDelta < -95.0f)
{
tempoDelta = -95.0f;
}
else if (tempoDelta > 5000.0f)
{
tempoDelta = 5000.0f;
}
if (pitchDelta < -60.0f)
{
pitchDelta = -60.0f;
}
else if (pitchDelta > 60.0f)
{
pitchDelta = 60.0f;
}
if (rateDelta < -95.0f)
{
rateDelta = -95.0f;
}
else if (rateDelta > 5000.0f)
{
rateDelta = 5000.0f;
}
}
// Convert STRING to std::string. Actually needed only if STRING is std::wstring, but conversion penalty is negligible
std::string convertString(const STRING& str)
{
std::string res;
for (auto c : str)
{
res += (char)c;
}
return res;
}
// Unknown switch parameter -- throws an exception with an error message
void RunParameters::throwIllegalParamExp(const STRING &str) const
{
string msg = "ERROR : Illegal parameter \"";
msg += convertString(str);
msg += "\".\n\n";
msg += usage;
ST_THROW_RT_ERROR(msg);
}
void RunParameters::throwLicense() const
{
ST_THROW_RT_ERROR(licenseText);
}
float RunParameters::parseSwitchValue(const STRING& str) const
{
int pos;
pos = (int)str.find_first_of('=');
if (pos < 0)
{
// '=' missing
throwIllegalParamExp(str);
}
// Read numerical parameter value after '='
return (float)stof(str.substr(pos + 1).c_str());
}
// Interprets a single switch parameter string of format "-switch=xx"
// Valid switches are "-tempo=xx", "-pitch=xx" and "-rate=xx". Stores
// switch values into 'params' structure.
void RunParameters::parseSwitchParam(const STRING& str)
{
int upS;
if (str[0] != '-')
{
// leading hyphen missing => not a valid parameter
throwIllegalParamExp(str);
}
// Take the first character of switch name & change to lower case
upS = _toLowerCase(str[1]);
// interpret the switch name & operate accordingly
switch (upS)
{
case 't' :
// switch '-tempo=xx'
tempoDelta = parseSwitchValue(str);
break;
case 'p' :
// switch '-pitch=xx'
pitchDelta = parseSwitchValue(str);
break;
case 'r' :
// switch '-rate=xx'
rateDelta = parseSwitchValue(str);
break;
case 'b' :
// switch '-bpm=xx'
detectBPM = true;
try
{
goalBPM = parseSwitchValue(str);
}
catch (const runtime_error &)
{
// illegal or missing bpm value => just calculate bpm
goalBPM = 0;
}
break;
case 'q' :
// switch '-quick'
quick = 1;
break;
case 'n' :
// switch '-naa'
noAntiAlias = 1;
break;
case 'l' :
// switch '-license'
throwLicense();
break;
case 's' :
// switch '-speech'
speech = true;
break;
default:
// unknown switch
throwIllegalParamExp(str);
}
}
}

140
source/SoundStretch/RunParameters.h
View File

@ -1,70 +1,70 @@
////////////////////////////////////////////////////////////////////////////////
///
/// A class for parsing the 'soundstretch' application command line parameters
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef RUNPARAMETERS_H
#define RUNPARAMETERS_H
#include <string>
#include "STTypes.h"
#include "SS_CharTypes.h"
#include "WavFile.h"
namespace soundstretch
{
/// Parses command line parameters into program parameters
class RunParameters
{
private:
void throwIllegalParamExp(const STRING& str) const;
void throwLicense() const;
void parseSwitchParam(const STRING& str);
void checkLimits();
float parseSwitchValue(const STRING& tr) const;
public:
STRING inFileName;
STRING outFileName;
float tempoDelta{ 0 };
float pitchDelta{ 0 };
float rateDelta{ 0 };
int quick{ 0 };
int noAntiAlias{ 0 };
float goalBPM{ 0 };
bool detectBPM{ false };
bool speech{ false };
RunParameters(int nParams, const CHARTYPE* paramStr[]);
};
}
#endif
////////////////////////////////////////////////////////////////////////////////
///
/// A class for parsing the 'soundstretch' application command line parameters
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef RUNPARAMETERS_H
#define RUNPARAMETERS_H
#include <string>
#include "STTypes.h"
#include "SS_CharTypes.h"
#include "WavFile.h"
namespace soundstretch
{
/// Parses command line parameters into program parameters
class RunParameters
{
private:
void throwIllegalParamExp(const STRING& str) const;
void throwLicense() const;
void parseSwitchParam(const STRING& str);
void checkLimits();
float parseSwitchValue(const STRING& tr) const;
public:
STRING inFileName;
STRING outFileName;
float tempoDelta{ 0 };
float pitchDelta{ 0 };
float rateDelta{ 0 };
int quick{ 0 };
int noAntiAlias{ 0 };
float goalBPM{ 0 };
bool detectBPM{ false };
bool speech{ false };
RunParameters(int nParams, const CHARTYPE* paramStr[]);
};
}
#endif

104
source/SoundStretch/SS_CharTypes.h
View File

@ -1,52 +1,52 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Char type for SoundStretch
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef SS_CHARTYPE_H
#define SS_CHARTYPE_H
#include <string>
namespace soundstretch
{
#if _WIN32
// wide-char types for supporting non-latin file paths in Windows
using CHARTYPE = wchar_t;
using STRING = std::wstring;
#define STRING_CONST(x) (L"" x)
#else
// gnu platform can natively support UTF-8 paths using "char*" set
using CHARTYPE = char;
using STRING = std::string;
#define STRING_CONST(x) (x)
#endif
}
#endif //SS_CHARTYPE_H
////////////////////////////////////////////////////////////////////////////////
///
/// Char type for SoundStretch
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef SS_CHARTYPE_H
#define SS_CHARTYPE_H
#include <string>
namespace soundstretch
{
#if _WIN32
// wide-char types for supporting non-latin file paths in Windows
using CHARTYPE = wchar_t;
using STRING = std::wstring;
#define STRING_CONST(x) (L"" x)
#else
// gnu platform can natively support UTF-8 paths using "char*" set
using CHARTYPE = char;
using STRING = std::string;
#define STRING_CONST(x) (x)
#endif
}
#endif //SS_CHARTYPE_H

1968
source/SoundStretch/WavFile.cpp
View File

File diff suppressed because it is too large Load Diff

562
source/SoundStretch/WavFile.h
View File

@ -1,281 +1,281 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Classes for easy reading & writing of WAV sound files.
///
/// For big-endian CPU, define BIG_ENDIAN during compile-time to correctly
/// parse the WAV files with such processors.
///
/// Admittingly, more complete WAV reader routines may exist in public domain, but
/// the reason for 'yet another' one is that those generic WAV reader libraries are
/// exhaustingly large and cumbersome! Wanted to have something simpler here, i.e.
/// something that's not already larger than rest of the SoundTouch/SoundStretch program...
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef WAVFILE_H
#define WAVFILE_H
#include <cstdio>
#include <string>
#include "SS_CharTypes.h"
namespace soundstretch
{
#ifndef uint
typedef unsigned int uint;
#endif
/// WAV audio file 'riff' section header
typedef struct
{
char riff_char[4];
uint package_len;
char wave[4];
} WavRiff;
/// WAV audio file 'format' section header
typedef struct
{
char fmt[4];
unsigned int format_len;
unsigned short fixed;
unsigned short channel_number;
unsigned int sample_rate;
unsigned int byte_rate;
unsigned short byte_per_sample;
unsigned short bits_per_sample;
} WavFormat;
/// WAV audio file 'fact' section header
typedef struct
{
char fact_field[4];
uint fact_len;
uint fact_sample_len;
} WavFact;
/// WAV audio file 'data' section header
typedef struct
{
char data_field[4];
uint data_len;
} WavData;
/// WAV audio file header
typedef struct
{
WavRiff riff;
WavFormat format;
WavFact fact;
WavData data;
} WavHeader;
/// Base class for processing WAV audio files.
class WavFileBase
{
private:
/// Conversion working buffer;
char *convBuff;
int convBuffSize;
protected:
WavFileBase();
virtual ~WavFileBase();
/// Get pointer to conversion buffer of at min. given size
void *getConvBuffer(int sizeByte);
};
/// Class for reading WAV audio files.
class WavInFile : protected WavFileBase
{
private:
/// File pointer.
FILE *fptr;
/// Counter of how many bytes of sample data have been read from the file.
long dataRead;
/// WAV header information
WavHeader header;
/// Init the WAV file stream
void init();
/// Read WAV file headers.
/// \return zero if all ok, nonzero if file format is invalid.
int readWavHeaders();
/// Checks WAV file header tags.
/// \return zero if all ok, nonzero if file format is invalid.
int checkCharTags() const;
/// Reads a single WAV file header block.
/// \return zero if all ok, nonzero if file format is invalid.
int readHeaderBlock();
/// Reads WAV file 'riff' block
int readRIFFBlock();
public:
/// Constructor: Opens the given WAV file. If the file can't be opened,
/// throws 'runtime_error' exception.
WavInFile(const STRING& filename);
WavInFile(FILE *file);
/// Destructor: Closes the file.
~WavInFile();
/// Rewind to beginning of the file
void rewind();
/// Get sample rate.
uint getSampleRate() const;
/// Get number of bits per sample, i.e. 8 or 16.
uint getNumBits() const;
/// Get sample data size in bytes. Ahem, this should return same information as
/// 'getBytesPerSample'...
uint getDataSizeInBytes() const;
/// Get total number of samples in file.
uint getNumSamples() const;
/// Get number of bytes per audio sample (e.g. 16bit stereo = 4 bytes/sample)
uint getBytesPerSample() const;
/// Get number of audio channels in the file (1=mono, 2=stereo)
uint getNumChannels() const;
/// Get the audio file length in milliseconds
uint getLengthMS() const;
/// Returns how many milliseconds of audio have so far been read from the file
///
/// \return elapsed duration in milliseconds
uint getElapsedMS() const;
/// Reads audio samples from the WAV file. This routine works only for 8 bit samples.
/// Reads given number of elements from the file or if end-of-file reached, as many
/// elements as are left in the file.
///
/// \return Number of 8-bit integers read from the file.
int read(unsigned char *buffer, int maxElems);
/// Reads audio samples from the WAV file to 16 bit integer format. Reads given number
/// of elements from the file or if end-of-file reached, as many elements as are
/// left in the file.
///
/// \return Number of 16-bit integers read from the file.
int read(short *buffer, ///< Pointer to buffer where to read data.
int maxElems ///< Size of 'buffer' array (number of array elements).
);
/// Reads audio samples from the WAV file to floating point format, converting
/// sample values to range [-1,1[. Reads given number of elements from the file
/// or if end-of-file reached, as many elements as are left in the file.
/// Notice that reading in float format supports 8/16/24/32bit sample formats.
///
/// \return Number of elements read from the file.
int read(float *buffer, ///< Pointer to buffer where to read data.
int maxElems ///< Size of 'buffer' array (number of array elements).
);
/// Check end-of-file.
///
/// \return Nonzero if end-of-file reached.
int eof() const;
};
/// Class for writing WAV audio files.
class WavOutFile : protected WavFileBase
{
private:
/// Pointer to the WAV file
FILE *fptr;
/// WAV file header data.
WavHeader header;
/// Counter of how many bytes have been written to the file so far.
int bytesWritten;
/// Fills in WAV file header information.
void fillInHeader(const uint sampleRate, const uint bits, const uint channels);
/// Finishes the WAV file header by supplementing information of amount of
/// data written to file etc
void finishHeader();
/// Writes the WAV file header.
void writeHeader();
public:
/// Constructor: Creates a new WAV file. Throws a 'runtime_error' exception
/// if file creation fails.
WavOutFile(const STRING& fileName, ///< Filename
int sampleRate, ///< Sample rate (e.g. 44100 etc)
int bits, ///< Bits per sample (8 or 16 bits)
int channels ///< Number of channels (1=mono, 2=stereo)
);
WavOutFile(FILE *file, int sampleRate, int bits, int channels);
/// Destructor: Finalizes & closes the WAV file.
~WavOutFile();
/// Write data to WAV file. This function works only with 8bit samples.
/// Throws a 'runtime_error' exception if writing to file fails.
void write(const unsigned char *buffer, ///< Pointer to sample data buffer.
int numElems ///< How many array items are to be written to file.
);
/// Write data to WAV file. Throws a 'runtime_error' exception if writing to
/// file fails.
void write(const short *buffer, ///< Pointer to sample data buffer.
int numElems ///< How many array items are to be written to file.
);
/// Write data to WAV file in floating point format, saturating sample values to range
/// [-1..+1[. Throws a 'runtime_error' exception if writing to file fails.
void write(const float *buffer, ///< Pointer to sample data buffer.
int numElems ///< How many array items are to be written to file.
);
};
}
#endif
////////////////////////////////////////////////////////////////////////////////
///
/// Classes for easy reading & writing of WAV sound files.
///
/// For big-endian CPU, define BIG_ENDIAN during compile-time to correctly
/// parse the WAV files with such processors.
///
/// Admittingly, more complete WAV reader routines may exist in public domain, but
/// the reason for 'yet another' one is that those generic WAV reader libraries are
/// exhaustingly large and cumbersome! Wanted to have something simpler here, i.e.
/// something that's not already larger than rest of the SoundTouch/SoundStretch program...
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef WAVFILE_H
#define WAVFILE_H
#include <cstdio>
#include <string>
#include "SS_CharTypes.h"
namespace soundstretch
{
#ifndef uint
typedef unsigned int uint;
#endif
/// WAV audio file 'riff' section header
typedef struct
{
char riff_char[4];
uint package_len;
char wave[4];
} WavRiff;
/// WAV audio file 'format' section header
typedef struct
{
char fmt[4];
unsigned int format_len;
unsigned short fixed;
unsigned short channel_number;
unsigned int sample_rate;
unsigned int byte_rate;
unsigned short byte_per_sample;
unsigned short bits_per_sample;
} WavFormat;
/// WAV audio file 'fact' section header
typedef struct
{
char fact_field[4];
uint fact_len;
uint fact_sample_len;
} WavFact;
/// WAV audio file 'data' section header
typedef struct
{
char data_field[4];
uint data_len;
} WavData;
/// WAV audio file header
typedef struct
{
WavRiff riff;
WavFormat format;
WavFact fact;
WavData data;
} WavHeader;
/// Base class for processing WAV audio files.
class WavFileBase
{
private:
/// Conversion working buffer;
char *convBuff;
int convBuffSize;
protected:
WavFileBase();
virtual ~WavFileBase();
/// Get pointer to conversion buffer of at min. given size
void *getConvBuffer(int sizeByte);
};
/// Class for reading WAV audio files.
class WavInFile : protected WavFileBase
{
private:
/// File pointer.
FILE *fptr;
/// Counter of how many bytes of sample data have been read from the file.
long dataRead;
/// WAV header information
WavHeader header;
/// Init the WAV file stream
void init();
/// Read WAV file headers.
/// \return zero if all ok, nonzero if file format is invalid.
int readWavHeaders();
/// Checks WAV file header tags.
/// \return zero if all ok, nonzero if file format is invalid.
int checkCharTags() const;
/// Reads a single WAV file header block.
/// \return zero if all ok, nonzero if file format is invalid.
int readHeaderBlock();
/// Reads WAV file 'riff' block
int readRIFFBlock();
public:
/// Constructor: Opens the given WAV file. If the file can't be opened,
/// throws 'runtime_error' exception.
WavInFile(const STRING& filename);
WavInFile(FILE *file);
/// Destructor: Closes the file.
~WavInFile();
/// Rewind to beginning of the file
void rewind();
/// Get sample rate.
uint getSampleRate() const;
/// Get number of bits per sample, i.e. 8 or 16.
uint getNumBits() const;
/// Get sample data size in bytes. Ahem, this should return same information as
/// 'getBytesPerSample'...
uint getDataSizeInBytes() const;
/// Get total number of samples in file.
uint getNumSamples() const;
/// Get number of bytes per audio sample (e.g. 16bit stereo = 4 bytes/sample)
uint getBytesPerSample() const;
/// Get number of audio channels in the file (1=mono, 2=stereo)
uint getNumChannels() const;
/// Get the audio file length in milliseconds
uint getLengthMS() const;
/// Returns how many milliseconds of audio have so far been read from the file
///
/// \return elapsed duration in milliseconds
uint getElapsedMS() const;
/// Reads audio samples from the WAV file. This routine works only for 8 bit samples.
/// Reads given number of elements from the file or if end-of-file reached, as many
/// elements as are left in the file.
///
/// \return Number of 8-bit integers read from the file.
int read(unsigned char *buffer, int maxElems);
/// Reads audio samples from the WAV file to 16 bit integer format. Reads given number
/// of elements from the file or if end-of-file reached, as many elements as are
/// left in the file.
///
/// \return Number of 16-bit integers read from the file.
int read(short *buffer, ///< Pointer to buffer where to read data.
int maxElems ///< Size of 'buffer' array (number of array elements).
);
/// Reads audio samples from the WAV file to floating point format, converting
/// sample values to range [-1,1[. Reads given number of elements from the file
/// or if end-of-file reached, as many elements as are left in the file.
/// Notice that reading in float format supports 8/16/24/32bit sample formats.
///
/// \return Number of elements read from the file.
int read(float *buffer, ///< Pointer to buffer where to read data.
int maxElems ///< Size of 'buffer' array (number of array elements).
);
/// Check end-of-file.
///
/// \return Nonzero if end-of-file reached.
int eof() const;
};
/// Class for writing WAV audio files.
class WavOutFile : protected WavFileBase
{
private:
/// Pointer to the WAV file
FILE *fptr;
/// WAV file header data.
WavHeader header;
/// Counter of how many bytes have been written to the file so far.
int bytesWritten;
/// Fills in WAV file header information.
void fillInHeader(const uint sampleRate, const uint bits, const uint channels);
/// Finishes the WAV file header by supplementing information of amount of
/// data written to file etc
void finishHeader();
/// Writes the WAV file header.
void writeHeader();
public:
/// Constructor: Creates a new WAV file. Throws a 'runtime_error' exception
/// if file creation fails.
WavOutFile(const STRING& fileName, ///< Filename
int sampleRate, ///< Sample rate (e.g. 44100 etc)
int bits, ///< Bits per sample (8 or 16 bits)
int channels ///< Number of channels (1=mono, 2=stereo)
);
WavOutFile(FILE *file, int sampleRate, int bits, int channels);
/// Destructor: Finalizes & closes the WAV file.
~WavOutFile();
/// Write data to WAV file. This function works only with 8bit samples.
/// Throws a 'runtime_error' exception if writing to file fails.
void write(const unsigned char *buffer, ///< Pointer to sample data buffer.
int numElems ///< How many array items are to be written to file.
);
/// Write data to WAV file. Throws a 'runtime_error' exception if writing to
/// file fails.
void write(const short *buffer, ///< Pointer to sample data buffer.
int numElems ///< How many array items are to be written to file.
);
/// Write data to WAV file in floating point format, saturating sample values to range
/// [-1..+1[. Throws a 'runtime_error' exception if writing to file fails.
void write(const float *buffer, ///< Pointer to sample data buffer.
int numElems ///< How many array items are to be written to file.
);
};
}
#endif

622
source/SoundStretch/main.cpp
View File

@ -1,311 +1,311 @@
////////////////////////////////////////////////////////////////////////////////
///
/// SoundStretch main routine.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <iostream>
#include <memory>
#include <stdexcept>
#include <string>
#include <cstdio>
#include <ctime>
#include "RunParameters.h"
#include "WavFile.h"
#include "SoundTouch.h"
#include "BPMDetect.h"
using namespace soundtouch;
using namespace std;
namespace soundstretch
{
// Processing chunk size (size chosen to be divisible by 2, 4, 6, 8, 10, 12, 14, 16 channels ...)
#define BUFF_SIZE 6720
#if _WIN32
#include <io.h>
#include <fcntl.h>
// Macro for Win32 standard input/output stream support: Sets a file stream into binary mode
#define SET_STREAM_TO_BIN_MODE(f) (_setmode(_fileno(f), _O_BINARY))
#else
// Not needed for GNU environment...
#define SET_STREAM_TO_BIN_MODE(f) {}
#endif
static const char _helloText[] =
"\n"
" SoundStretch v%s - Copyright (c) Olli Parviainen\n"
"=========================================================\n"
"author e-mail: <oparviai"
"@"
"iki.fi> - WWW: http://www.surina.net/soundtouch\n"
"\n"
"This program is subject to (L)GPL license. Run \"soundstretch -license\" for\n"
"more information.\n"
"\n";
static void openFiles(unique_ptr<WavInFile>& inFile, unique_ptr<WavOutFile>& outFile, const RunParameters& params)
{
if (params.inFileName == STRING_CONST("stdin"))
{
// used 'stdin' as input file
SET_STREAM_TO_BIN_MODE(stdin);
inFile = make_unique<WavInFile>(stdin);
}
else
{
// open input file...
inFile = make_unique<WavInFile>(params.inFileName.c_str());
}
// ... open output file with same sound parameters
const int bits = (int)inFile->getNumBits();
const int samplerate = (int)inFile->getSampleRate();
const int channels = (int)inFile->getNumChannels();
if (!params.outFileName.empty())
{
if (params.outFileName == STRING_CONST("stdout"))
{
SET_STREAM_TO_BIN_MODE(stdout);
outFile = make_unique<WavOutFile>(stdout, samplerate, bits, channels);
}
else
{
outFile = make_unique<WavOutFile>(params.outFileName.c_str(), samplerate, bits, channels);
}
}
}
// Sets the 'SoundTouch' object up according to input file sound format &
// command line parameters
static void setup(SoundTouch& soundTouch, const WavInFile& inFile, const RunParameters& params)
{
const int sampleRate = (int)inFile.getSampleRate();
const int channels = (int)inFile.getNumChannels();
soundTouch.setSampleRate(sampleRate);
soundTouch.setChannels(channels);
soundTouch.setTempoChange(params.tempoDelta);
soundTouch.setPitchSemiTones(params.pitchDelta);
soundTouch.setRateChange(params.rateDelta);
soundTouch.setSetting(SETTING_USE_QUICKSEEK, params.quick);
soundTouch.setSetting(SETTING_USE_AA_FILTER, !(params.noAntiAlias));
if (params.speech)
{
// use settings for speech processing
soundTouch.setSetting(SETTING_SEQUENCE_MS, 40);
soundTouch.setSetting(SETTING_SEEKWINDOW_MS, 15);
soundTouch.setSetting(SETTING_OVERLAP_MS, 8);
fprintf(stderr, "Tune processing parameters for speech processing.\n");
}
// print processing information
if (!params.outFileName.empty())
{
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
fprintf(stderr, "Uses 16bit integer sample type in processing.\n\n");
#else
#ifndef SOUNDTOUCH_FLOAT_SAMPLES
#error "Sampletype not defined"
#endif
fprintf(stderr, "Uses 32bit floating point sample type in processing.\n\n");
#endif
// print processing information only if outFileName given i.e. some processing will happen
fprintf(stderr, "Processing the file with the following changes:\n");
fprintf(stderr, " tempo change = %+g %%\n", params.tempoDelta);
fprintf(stderr, " pitch change = %+g semitones\n", params.pitchDelta);
fprintf(stderr, " rate change = %+g %%\n\n", params.rateDelta);
fprintf(stderr, "Working...");
}
else
{
// outFileName not given
fprintf(stderr, "Warning: output file name missing, won't output anything.\n\n");
}
fflush(stderr);
}
// Processes the sound
static void process(SoundTouch& soundTouch, WavInFile& inFile, WavOutFile& outFile)
{
SAMPLETYPE sampleBuffer[BUFF_SIZE];
int nSamples;
const int nChannels = (int)inFile.getNumChannels();
assert(nChannels > 0);
const int buffSizeSamples = BUFF_SIZE / nChannels;
// Process samples read from the input file
while (inFile.eof() == 0)
{
// Read a chunk of samples from the input file
const int num = inFile.read(sampleBuffer, BUFF_SIZE);
int nSamples = num / (int)inFile.getNumChannels();
// Feed the samples into SoundTouch processor
soundTouch.putSamples(sampleBuffer, nSamples);
// Read ready samples from SoundTouch processor & write them output file.
// NOTES:
// - 'receiveSamples' doesn't necessarily return any samples at all
// during some rounds!
// - On the other hand, during some round 'receiveSamples' may have more
// ready samples than would fit into 'sampleBuffer', and for this reason
// the 'receiveSamples' call is iterated for as many times as it
// outputs samples.
do
{
nSamples = soundTouch.receiveSamples(sampleBuffer, buffSizeSamples);
outFile.write(sampleBuffer, nSamples * nChannels);
} while (nSamples != 0);
}
// Now the input file is processed, yet 'flush' few last samples that are
// hiding in the SoundTouch's internal processing pipeline.
soundTouch.flush();
do
{
nSamples = soundTouch.receiveSamples(sampleBuffer, buffSizeSamples);
outFile.write(sampleBuffer, nSamples * nChannels);
} while (nSamples != 0);
}
// Detect BPM rate of inFile and adjust tempo setting accordingly if necessary
static void detectBPM(WavInFile& inFile, RunParameters& params)
{
BPMDetect bpm(inFile.getNumChannels(), inFile.getSampleRate());
SAMPLETYPE sampleBuffer[BUFF_SIZE];
// detect bpm rate
fprintf(stderr, "Detecting BPM rate...");
fflush(stderr);
const int nChannels = (int)inFile.getNumChannels();
int readSize = BUFF_SIZE - BUFF_SIZE % nChannels; // round read size down to multiple of num.channels
// Process the 'inFile' in small blocks, repeat until whole file has
// been processed
while (inFile.eof() == 0)
{
// Read sample data from input file
const int num = inFile.read(sampleBuffer, readSize);
// Enter the new samples to the bpm analyzer class
const int samples = num / nChannels;
bpm.inputSamples(sampleBuffer, samples);
}
// Now the whole song data has been analyzed. Read the resulting bpm.
const float bpmValue = bpm.getBpm();
fprintf(stderr, "Done!\n");
// rewind the file after bpm detection
inFile.rewind();
if (bpmValue > 0)
{
fprintf(stderr, "Detected BPM rate %.1f\n\n", bpmValue);
}
else
{
fprintf(stderr, "Couldn't detect BPM rate.\n\n");
return;
}
if (params.goalBPM > 0)
{
// adjust tempo to given bpm
params.tempoDelta = (params.goalBPM / bpmValue - 1.0f) * 100.0f;
fprintf(stderr, "The file will be converted to %.1f BPM\n\n", params.goalBPM);
}
}
void ss_main(RunParameters& params)
{
unique_ptr<WavInFile> inFile;
unique_ptr<WavOutFile> outFile;
SoundTouch soundTouch;
fprintf(stderr, _helloText, soundTouch.getVersionString());
// Open input & output files
openFiles(inFile, outFile, params);
if (params.detectBPM == true)
{
// detect sound BPM (and adjust processing parameters
// accordingly if necessary)
detectBPM(*inFile, params);
}
// Setup the 'SoundTouch' object for processing the sound
setup(soundTouch, *inFile, params);
// clock_t cs = clock(); // for benchmarking processing duration
// Process the sound
if (inFile && outFile)
{
process(soundTouch, *inFile, *outFile);
}
// clock_t ce = clock(); // for benchmarking processing duration
// printf("duration: %lf\n", (double)(ce-cs)/CLOCKS_PER_SEC);
fprintf(stderr, "Done!\n");
}
}
#if _WIN32
int wmain(int argc, const wchar_t* args[])
#else
int main(int argc, const char* args[])
#endif
{
try
{
soundstretch::RunParameters params(argc, args);
soundstretch::ss_main(params);
}
catch (const runtime_error& e)
{
fprintf(stderr, "%s\n", e.what());
return -1;
}
return 0;
}
////////////////////////////////////////////////////////////////////////////////
///
/// SoundStretch main routine.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <iostream>
#include <memory>
#include <stdexcept>
#include <string>
#include <cstdio>
#include <ctime>
#include "RunParameters.h"
#include "WavFile.h"
#include "SoundTouch.h"
#include "BPMDetect.h"
using namespace soundtouch;
using namespace std;
namespace soundstretch
{
// Processing chunk size (size chosen to be divisible by 2, 4, 6, 8, 10, 12, 14, 16 channels ...)
#define BUFF_SIZE 6720
#if _WIN32
#include <io.h>
#include <fcntl.h>
// Macro for Win32 standard input/output stream support: Sets a file stream into binary mode
#define SET_STREAM_TO_BIN_MODE(f) (_setmode(_fileno(f), _O_BINARY))
#else
// Not needed for GNU environment...
#define SET_STREAM_TO_BIN_MODE(f) {}
#endif
static const char _helloText[] =
"\n"
" SoundStretch v%s - Copyright (c) Olli Parviainen\n"
"=========================================================\n"
"author e-mail: <oparviai"
"@"
"iki.fi> - WWW: http://www.surina.net/soundtouch\n"
"\n"
"This program is subject to (L)GPL license. Run \"soundstretch -license\" for\n"
"more information.\n"
"\n";
static void openFiles(unique_ptr<WavInFile>& inFile, unique_ptr<WavOutFile>& outFile, const RunParameters& params)
{
if (params.inFileName == STRING_CONST("stdin"))
{
// used 'stdin' as input file
SET_STREAM_TO_BIN_MODE(stdin);
inFile = make_unique<WavInFile>(stdin);
}
else
{
// open input file...
inFile = make_unique<WavInFile>(params.inFileName.c_str());
}
// ... open output file with same sound parameters
const int bits = (int)inFile->getNumBits();
const int samplerate = (int)inFile->getSampleRate();
const int channels = (int)inFile->getNumChannels();
if (!params.outFileName.empty())
{
if (params.outFileName == STRING_CONST("stdout"))
{
SET_STREAM_TO_BIN_MODE(stdout);
outFile = make_unique<WavOutFile>(stdout, samplerate, bits, channels);
}
else
{
outFile = make_unique<WavOutFile>(params.outFileName.c_str(), samplerate, bits, channels);
}
}
}
// Sets the 'SoundTouch' object up according to input file sound format &
// command line parameters
static void setup(SoundTouch& soundTouch, const WavInFile& inFile, const RunParameters& params)
{
const int sampleRate = (int)inFile.getSampleRate();
const int channels = (int)inFile.getNumChannels();
soundTouch.setSampleRate(sampleRate);
soundTouch.setChannels(channels);
soundTouch.setTempoChange(params.tempoDelta);
soundTouch.setPitchSemiTones(params.pitchDelta);
soundTouch.setRateChange(params.rateDelta);
soundTouch.setSetting(SETTING_USE_QUICKSEEK, params.quick);
soundTouch.setSetting(SETTING_USE_AA_FILTER, !(params.noAntiAlias));
if (params.speech)
{
// use settings for speech processing
soundTouch.setSetting(SETTING_SEQUENCE_MS, 40);
soundTouch.setSetting(SETTING_SEEKWINDOW_MS, 15);
soundTouch.setSetting(SETTING_OVERLAP_MS, 8);
fprintf(stderr, "Tune processing parameters for speech processing.\n");
}
// print processing information
if (!params.outFileName.empty())
{
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
fprintf(stderr, "Uses 16bit integer sample type in processing.\n\n");
#else
#ifndef SOUNDTOUCH_FLOAT_SAMPLES
#error "Sampletype not defined"
#endif
fprintf(stderr, "Uses 32bit floating point sample type in processing.\n\n");
#endif
// print processing information only if outFileName given i.e. some processing will happen
fprintf(stderr, "Processing the file with the following changes:\n");
fprintf(stderr, " tempo change = %+g %%\n", params.tempoDelta);
fprintf(stderr, " pitch change = %+g semitones\n", params.pitchDelta);
fprintf(stderr, " rate change = %+g %%\n\n", params.rateDelta);
fprintf(stderr, "Working...");
}
else
{
// outFileName not given
fprintf(stderr, "Warning: output file name missing, won't output anything.\n\n");
}
fflush(stderr);
}
// Processes the sound
static void process(SoundTouch& soundTouch, WavInFile& inFile, WavOutFile& outFile)
{
SAMPLETYPE sampleBuffer[BUFF_SIZE];
int nSamples;
const int nChannels = (int)inFile.getNumChannels();
assert(nChannels > 0);
const int buffSizeSamples = BUFF_SIZE / nChannels;
// Process samples read from the input file
while (inFile.eof() == 0)
{
// Read a chunk of samples from the input file
const int num = inFile.read(sampleBuffer, BUFF_SIZE);
int nSamples = num / (int)inFile.getNumChannels();
// Feed the samples into SoundTouch processor
soundTouch.putSamples(sampleBuffer, nSamples);
// Read ready samples from SoundTouch processor & write them output file.
// NOTES:
// - 'receiveSamples' doesn't necessarily return any samples at all
// during some rounds!
// - On the other hand, during some round 'receiveSamples' may have more
// ready samples than would fit into 'sampleBuffer', and for this reason
// the 'receiveSamples' call is iterated for as many times as it
// outputs samples.
do
{
nSamples = soundTouch.receiveSamples(sampleBuffer, buffSizeSamples);
outFile.write(sampleBuffer, nSamples * nChannels);
} while (nSamples != 0);
}
// Now the input file is processed, yet 'flush' few last samples that are
// hiding in the SoundTouch's internal processing pipeline.
soundTouch.flush();
do
{
nSamples = soundTouch.receiveSamples(sampleBuffer, buffSizeSamples);
outFile.write(sampleBuffer, nSamples * nChannels);
} while (nSamples != 0);
}
// Detect BPM rate of inFile and adjust tempo setting accordingly if necessary
static void detectBPM(WavInFile& inFile, RunParameters& params)
{
BPMDetect bpm(inFile.getNumChannels(), inFile.getSampleRate());
SAMPLETYPE sampleBuffer[BUFF_SIZE];
// detect bpm rate
fprintf(stderr, "Detecting BPM rate...");
fflush(stderr);
const int nChannels = (int)inFile.getNumChannels();
int readSize = BUFF_SIZE - BUFF_SIZE % nChannels; // round read size down to multiple of num.channels
// Process the 'inFile' in small blocks, repeat until whole file has
// been processed
while (inFile.eof() == 0)
{
// Read sample data from input file
const int num = inFile.read(sampleBuffer, readSize);
// Enter the new samples to the bpm analyzer class
const int samples = num / nChannels;
bpm.inputSamples(sampleBuffer, samples);
}
// Now the whole song data has been analyzed. Read the resulting bpm.
const float bpmValue = bpm.getBpm();
fprintf(stderr, "Done!\n");
// rewind the file after bpm detection
inFile.rewind();
if (bpmValue > 0)
{
fprintf(stderr, "Detected BPM rate %.1f\n\n", bpmValue);
}
else
{
fprintf(stderr, "Couldn't detect BPM rate.\n\n");
return;
}
if (params.goalBPM > 0)
{
// adjust tempo to given bpm
params.tempoDelta = (params.goalBPM / bpmValue - 1.0f) * 100.0f;
fprintf(stderr, "The file will be converted to %.1f BPM\n\n", params.goalBPM);
}
}
void ss_main(RunParameters& params)
{
unique_ptr<WavInFile> inFile;
unique_ptr<WavOutFile> outFile;
SoundTouch soundTouch;
fprintf(stderr, _helloText, soundTouch.getVersionString());
// Open input & output files
openFiles(inFile, outFile, params);
if (params.detectBPM == true)
{
// detect sound BPM (and adjust processing parameters
// accordingly if necessary)
detectBPM(*inFile, params);
}
// Setup the 'SoundTouch' object for processing the sound
setup(soundTouch, *inFile, params);
// clock_t cs = clock(); // for benchmarking processing duration
// Process the sound
if (inFile && outFile)
{
process(soundTouch, *inFile, *outFile);
}
// clock_t ce = clock(); // for benchmarking processing duration
// printf("duration: %lf\n", (double)(ce-cs)/CLOCKS_PER_SEC);
fprintf(stderr, "Done!\n");
}
}
#if _WIN32
int wmain(int argc, const wchar_t* args[])
#else
int main(int argc, const char* args[])
#endif
{
try
{
soundstretch::RunParameters params(argc, args);
soundstretch::ss_main(params);
}
catch (const runtime_error& e)
{
fprintf(stderr, "%s\n", e.what());
return -1;
}
return 0;
}

444
source/SoundTouch/AAFilter.cpp
View File

@ -1,222 +1,222 @@
////////////////////////////////////////////////////////////////////////////////
///
/// FIR low-pass (anti-alias) filter with filter coefficient design routine and
/// MMX optimization.
///
/// Anti-alias filter is used to prevent folding of high frequencies when
/// transposing the sample rate with interpolation.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <memory.h>
#include <assert.h>
#include <math.h>
#include <stdlib.h>
#include "AAFilter.h"
#include "FIRFilter.h"
using namespace soundtouch;
#define PI 3.14159265358979323846
#define TWOPI (2 * PI)
// define this to save AA filter coefficients to a file
// #define _DEBUG_SAVE_AAFILTER_COEFFICIENTS 1
#ifdef _DEBUG_SAVE_AAFILTER_COEFFICIENTS
#include <stdio.h>
static void _DEBUG_SAVE_AAFIR_COEFFS(SAMPLETYPE *coeffs, int len)
{
FILE *fptr = fopen("aa_filter_coeffs.txt", "wt");
if (fptr == nullptr) return;
for (int i = 0; i < len; i ++)
{
double temp = coeffs[i];
fprintf(fptr, "%lf\n", temp);
}
fclose(fptr);
}
#else
#define _DEBUG_SAVE_AAFIR_COEFFS(x, y)
#endif
/*****************************************************************************
*
* Implementation of the class 'AAFilter'
*
*****************************************************************************/
AAFilter::AAFilter(uint len)
{
pFIR = FIRFilter::newInstance();
cutoffFreq = 0.5;
setLength(len);
}
AAFilter::~AAFilter()
{
delete pFIR;
}
// Sets new anti-alias filter cut-off edge frequency, scaled to
// sampling frequency (nyquist frequency = 0.5).
// The filter will cut frequencies higher than the given frequency.
void AAFilter::setCutoffFreq(double newCutoffFreq)
{
cutoffFreq = newCutoffFreq;
calculateCoeffs();
}
// Sets number of FIR filter taps
void AAFilter::setLength(uint newLength)
{
length = newLength;
calculateCoeffs();
}
// Calculates coefficients for a low-pass FIR filter using Hamming window
void AAFilter::calculateCoeffs()
{
uint i;
double cntTemp, temp, tempCoeff,h, w;
double wc;
double scaleCoeff, sum;
double *work;
SAMPLETYPE *coeffs;
assert(length >= 2);
assert(length % 4 == 0);
assert(cutoffFreq >= 0);
assert(cutoffFreq <= 0.5);
work = new double[length];
coeffs = new SAMPLETYPE[length];
wc = 2.0 * PI * cutoffFreq;
tempCoeff = TWOPI / (double)length;
sum = 0;
for (i = 0; i < length; i ++)
{
cntTemp = (double)i - (double)(length / 2);
temp = cntTemp * wc;
if (temp != 0)
{
h = sin(temp) / temp; // sinc function
}
else
{
h = 1.0;
}
w = 0.54 + 0.46 * cos(tempCoeff * cntTemp); // hamming window
temp = w * h;
work[i] = temp;
// calc net sum of coefficients
sum += temp;
}
// ensure the sum of coefficients is larger than zero
assert(sum > 0);
// ensure we've really designed a lowpass filter...
assert(work[length/2] > 0);
assert(work[length/2 + 1] > -1e-6);
assert(work[length/2 - 1] > -1e-6);
// Calculate a scaling coefficient in such a way that the result can be
// divided by 16384
scaleCoeff = 16384.0f / sum;
for (i = 0; i < length; i ++)
{
temp = work[i] * scaleCoeff;
// scale & round to nearest integer
temp += (temp >= 0) ? 0.5 : -0.5;
// ensure no overfloods
assert(temp >= -32768 && temp <= 32767);
coeffs[i] = (SAMPLETYPE)temp;
}
// Set coefficients. Use divide factor 14 => divide result by 2^14 = 16384
pFIR->setCoefficients(coeffs, length, 14);
_DEBUG_SAVE_AAFIR_COEFFS(coeffs, length);
delete[] work;
delete[] coeffs;
}
// Applies the filter to the given sequence of samples.
// Note : The amount of outputted samples is by value of 'filter length'
// smaller than the amount of input samples.
uint AAFilter::evaluate(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples, uint numChannels) const
{
return pFIR->evaluate(dest, src, numSamples, numChannels);
}
/// Applies the filter to the given src & dest pipes, so that processed amount of
/// samples get removed from src, and produced amount added to dest
/// Note : The amount of outputted samples is by value of 'filter length'
/// smaller than the amount of input samples.
uint AAFilter::evaluate(FIFOSampleBuffer &dest, FIFOSampleBuffer &src) const
{
SAMPLETYPE *pdest;
const SAMPLETYPE *psrc;
uint numSrcSamples;
uint result;
int numChannels = src.getChannels();
assert(numChannels == dest.getChannels());
numSrcSamples = src.numSamples();
psrc = src.ptrBegin();
pdest = dest.ptrEnd(numSrcSamples);
result = pFIR->evaluate(pdest, psrc, numSrcSamples, numChannels);
src.receiveSamples(result);
dest.putSamples(result);
return result;
}
uint AAFilter::getLength() const
{
return pFIR->getLength();
}
////////////////////////////////////////////////////////////////////////////////
///
/// FIR low-pass (anti-alias) filter with filter coefficient design routine and
/// MMX optimization.
///
/// Anti-alias filter is used to prevent folding of high frequencies when
/// transposing the sample rate with interpolation.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <memory.h>
#include <assert.h>
#include <math.h>
#include <stdlib.h>
#include "AAFilter.h"
#include "FIRFilter.h"
using namespace soundtouch;
#define PI 3.14159265358979323846
#define TWOPI (2 * PI)
// define this to save AA filter coefficients to a file
// #define _DEBUG_SAVE_AAFILTER_COEFFICIENTS 1
#ifdef _DEBUG_SAVE_AAFILTER_COEFFICIENTS
#include <stdio.h>
static void _DEBUG_SAVE_AAFIR_COEFFS(SAMPLETYPE *coeffs, int len)
{
FILE *fptr = fopen("aa_filter_coeffs.txt", "wt");
if (fptr == nullptr) return;
for (int i = 0; i < len; i ++)
{
double temp = coeffs[i];
fprintf(fptr, "%lf\n", temp);
}
fclose(fptr);
}
#else
#define _DEBUG_SAVE_AAFIR_COEFFS(x, y)
#endif
/*****************************************************************************
*
* Implementation of the class 'AAFilter'
*
*****************************************************************************/
AAFilter::AAFilter(uint len)
{
pFIR = FIRFilter::newInstance();
cutoffFreq = 0.5;
setLength(len);
}
AAFilter::~AAFilter()
{
delete pFIR;
}
// Sets new anti-alias filter cut-off edge frequency, scaled to
// sampling frequency (nyquist frequency = 0.5).
// The filter will cut frequencies higher than the given frequency.
void AAFilter::setCutoffFreq(double newCutoffFreq)
{
cutoffFreq = newCutoffFreq;
calculateCoeffs();
}
// Sets number of FIR filter taps
void AAFilter::setLength(uint newLength)
{
length = newLength;
calculateCoeffs();
}
// Calculates coefficients for a low-pass FIR filter using Hamming window
void AAFilter::calculateCoeffs()
{
uint i;
double cntTemp, temp, tempCoeff,h, w;
double wc;
double scaleCoeff, sum;
double *work;
SAMPLETYPE *coeffs;
assert(length >= 2);
assert(length % 4 == 0);
assert(cutoffFreq >= 0);
assert(cutoffFreq <= 0.5);
work = new double[length];
coeffs = new SAMPLETYPE[length];
wc = 2.0 * PI * cutoffFreq;
tempCoeff = TWOPI / (double)length;
sum = 0;
for (i = 0; i < length; i ++)
{
cntTemp = (double)i - (double)(length / 2);
temp = cntTemp * wc;
if (temp != 0)
{
h = sin(temp) / temp; // sinc function
}
else
{
h = 1.0;
}
w = 0.54 + 0.46 * cos(tempCoeff * cntTemp); // hamming window
temp = w * h;
work[i] = temp;
// calc net sum of coefficients
sum += temp;
}
// ensure the sum of coefficients is larger than zero
assert(sum > 0);
// ensure we've really designed a lowpass filter...
assert(work[length/2] > 0);
assert(work[length/2 + 1] > -1e-6);
assert(work[length/2 - 1] > -1e-6);
// Calculate a scaling coefficient in such a way that the result can be
// divided by 16384
scaleCoeff = 16384.0f / sum;
for (i = 0; i < length; i ++)
{
temp = work[i] * scaleCoeff;
// scale & round to nearest integer
temp += (temp >= 0) ? 0.5 : -0.5;
// ensure no overfloods
assert(temp >= -32768 && temp <= 32767);
coeffs[i] = (SAMPLETYPE)temp;
}
// Set coefficients. Use divide factor 14 => divide result by 2^14 = 16384
pFIR->setCoefficients(coeffs, length, 14);
_DEBUG_SAVE_AAFIR_COEFFS(coeffs, length);
delete[] work;
delete[] coeffs;
}
// Applies the filter to the given sequence of samples.
// Note : The amount of outputted samples is by value of 'filter length'
// smaller than the amount of input samples.
uint AAFilter::evaluate(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples, uint numChannels) const
{
return pFIR->evaluate(dest, src, numSamples, numChannels);
}
/// Applies the filter to the given src & dest pipes, so that processed amount of
/// samples get removed from src, and produced amount added to dest
/// Note : The amount of outputted samples is by value of 'filter length'
/// smaller than the amount of input samples.
uint AAFilter::evaluate(FIFOSampleBuffer &dest, FIFOSampleBuffer &src) const
{
SAMPLETYPE *pdest;
const SAMPLETYPE *psrc;
uint numSrcSamples;
uint result;
int numChannels = src.getChannels();
assert(numChannels == dest.getChannels());
numSrcSamples = src.numSamples();
psrc = src.ptrBegin();
pdest = dest.ptrEnd(numSrcSamples);
result = pFIR->evaluate(pdest, psrc, numSrcSamples, numChannels);
src.receiveSamples(result);
dest.putSamples(result);
return result;
}
uint AAFilter::getLength() const
{
return pFIR->getLength();
}

186
source/SoundTouch/AAFilter.h
View File

@ -1,93 +1,93 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Sampled sound tempo changer/time stretch algorithm. Changes the sound tempo
/// while maintaining the original pitch by using a time domain WSOLA-like method
/// with several performance-increasing tweaks.
///
/// Anti-alias filter is used to prevent folding of high frequencies when
/// transposing the sample rate with interpolation.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef AAFilter_H
#define AAFilter_H
#include "STTypes.h"
#include "FIFOSampleBuffer.h"
namespace soundtouch
{
class AAFilter
{
protected:
class FIRFilter *pFIR;
/// Low-pass filter cut-off frequency, negative = invalid
double cutoffFreq;
/// num of filter taps
uint length;
/// Calculate the FIR coefficients realizing the given cutoff-frequency
void calculateCoeffs();
public:
AAFilter(uint length);
~AAFilter();
/// Sets new anti-alias filter cut-off edge frequency, scaled to sampling
/// frequency (nyquist frequency = 0.5). The filter will cut off the
/// frequencies than that.
void setCutoffFreq(double newCutoffFreq);
/// Sets number of FIR filter taps, i.e. ~filter complexity
void setLength(uint newLength);
uint getLength() const;
/// Applies the filter to the given sequence of samples.
/// Note : The amount of outputted samples is by value of 'filter length'
/// smaller than the amount of input samples.
uint evaluate(SAMPLETYPE *dest,
const SAMPLETYPE *src,
uint numSamples,
uint numChannels) const;
/// Applies the filter to the given src & dest pipes, so that processed amount of
/// samples get removed from src, and produced amount added to dest
/// Note : The amount of outputted samples is by value of 'filter length'
/// smaller than the amount of input samples.
uint evaluate(FIFOSampleBuffer &dest,
FIFOSampleBuffer &src) const;
};
}
#endif
////////////////////////////////////////////////////////////////////////////////
///
/// Sampled sound tempo changer/time stretch algorithm. Changes the sound tempo
/// while maintaining the original pitch by using a time domain WSOLA-like method
/// with several performance-increasing tweaks.
///
/// Anti-alias filter is used to prevent folding of high frequencies when
/// transposing the sample rate with interpolation.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef AAFilter_H
#define AAFilter_H
#include "STTypes.h"
#include "FIFOSampleBuffer.h"
namespace soundtouch
{
class AAFilter
{
protected:
class FIRFilter *pFIR;
/// Low-pass filter cut-off frequency, negative = invalid
double cutoffFreq;
/// num of filter taps
uint length;
/// Calculate the FIR coefficients realizing the given cutoff-frequency
void calculateCoeffs();
public:
AAFilter(uint length);
~AAFilter();
/// Sets new anti-alias filter cut-off edge frequency, scaled to sampling
/// frequency (nyquist frequency = 0.5). The filter will cut off the
/// frequencies than that.
void setCutoffFreq(double newCutoffFreq);
/// Sets number of FIR filter taps, i.e. ~filter complexity
void setLength(uint newLength);
uint getLength() const;
/// Applies the filter to the given sequence of samples.
/// Note : The amount of outputted samples is by value of 'filter length'
/// smaller than the amount of input samples.
uint evaluate(SAMPLETYPE *dest,
const SAMPLETYPE *src,
uint numSamples,
uint numChannels) const;
/// Applies the filter to the given src & dest pipes, so that processed amount of
/// samples get removed from src, and produced amount added to dest
/// Note : The amount of outputted samples is by value of 'filter length'
/// smaller than the amount of input samples.
uint evaluate(FIFOSampleBuffer &dest,
FIFOSampleBuffer &src) const;
};
}
#endif

1142
source/SoundTouch/BPMDetect.cpp
View File

File diff suppressed because it is too large Load Diff

550
source/SoundTouch/FIFOSampleBuffer.cpp
View File

@ -1,275 +1,275 @@
////////////////////////////////////////////////////////////////////////////////
///
/// A buffer class for temporarily storaging sound samples, operates as a
/// first-in-first-out pipe.
///
/// Samples are added to the end of the sample buffer with the 'putSamples'
/// function, and are received from the beginning of the buffer by calling
/// the 'receiveSamples' function. The class automatically removes the
/// outputted samples from the buffer, as well as grows the buffer size
/// whenever necessary.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <stdlib.h>
#include <memory.h>
#include <string.h>
#include <assert.h>
#include "FIFOSampleBuffer.h"
using namespace soundtouch;
// Constructor
FIFOSampleBuffer::FIFOSampleBuffer(int numChannels)
{
assert(numChannels > 0);
sizeInBytes = 0; // reasonable initial value
buffer = nullptr;
bufferUnaligned = nullptr;
samplesInBuffer = 0;
bufferPos = 0;
channels = (uint)numChannels;
ensureCapacity(32); // allocate initial capacity
}
// destructor
FIFOSampleBuffer::~FIFOSampleBuffer()
{
delete[] bufferUnaligned;
bufferUnaligned = nullptr;
buffer = nullptr;
}
// Sets number of channels, 1 = mono, 2 = stereo
void FIFOSampleBuffer::setChannels(int numChannels)
{
uint usedBytes;
if (!verifyNumberOfChannels(numChannels)) return;
usedBytes = channels * samplesInBuffer;
channels = (uint)numChannels;
samplesInBuffer = usedBytes / channels;
}
// if output location pointer 'bufferPos' isn't zero, 'rewinds' the buffer and
// zeroes this pointer by copying samples from the 'bufferPos' pointer
// location on to the beginning of the buffer.
void FIFOSampleBuffer::rewind()
{
if (buffer && bufferPos)
{
memmove(buffer, ptrBegin(), sizeof(SAMPLETYPE) * channels * samplesInBuffer);
bufferPos = 0;
}
}
// Adds 'numSamples' pcs of samples from the 'samples' memory position to
// the sample buffer.
void FIFOSampleBuffer::putSamples(const SAMPLETYPE *samples, uint nSamples)
{
memcpy(ptrEnd(nSamples), samples, sizeof(SAMPLETYPE) * nSamples * channels);
samplesInBuffer += nSamples;
}
// Increases the number of samples in the buffer without copying any actual
// samples.
//
// This function is used to update the number of samples in the sample buffer
// when accessing the buffer directly with 'ptrEnd' function. Please be
// careful though!
void FIFOSampleBuffer::putSamples(uint nSamples)
{
uint req;
req = samplesInBuffer + nSamples;
ensureCapacity(req);
samplesInBuffer += nSamples;
}
// Returns a pointer to the end of the used part of the sample buffer (i.e.
// where the new samples are to be inserted). This function may be used for
// inserting new samples into the sample buffer directly. Please be careful!
//
// Parameter 'slackCapacity' tells the function how much free capacity (in
// terms of samples) there _at least_ should be, in order to the caller to
// successfully insert all the required samples to the buffer. When necessary,
// the function grows the buffer size to comply with this requirement.
//
// When using this function as means for inserting new samples, also remember
// to increase the sample count afterwards, by calling the
// 'putSamples(numSamples)' function.
SAMPLETYPE *FIFOSampleBuffer::ptrEnd(uint slackCapacity)
{
ensureCapacity(samplesInBuffer + slackCapacity);
return buffer + samplesInBuffer * channels;
}
// Returns a pointer to the beginning of the currently non-outputted samples.
// This function is provided for accessing the output samples directly.
// Please be careful!
//
// When using this function to output samples, also remember to 'remove' the
// outputted samples from the buffer by calling the
// 'receiveSamples(numSamples)' function
SAMPLETYPE *FIFOSampleBuffer::ptrBegin()
{
assert(buffer);
return buffer + bufferPos * channels;
}
// Ensures that the buffer has enough capacity, i.e. space for _at least_
// 'capacityRequirement' number of samples. The buffer is grown in steps of
// 4 kilobytes to eliminate the need for frequently growing up the buffer,
// as well as to round the buffer size up to the virtual memory page size.
void FIFOSampleBuffer::ensureCapacity(uint capacityRequirement)
{
SAMPLETYPE *tempUnaligned, *temp;
if (capacityRequirement > getCapacity())
{
// enlarge the buffer in 4kbyte steps (round up to next 4k boundary)
sizeInBytes = (capacityRequirement * channels * sizeof(SAMPLETYPE) + 4095) & (uint)-4096;
assert(sizeInBytes % 2 == 0);
tempUnaligned = new SAMPLETYPE[sizeInBytes / sizeof(SAMPLETYPE) + 16 / sizeof(SAMPLETYPE)];
if (tempUnaligned == nullptr)
{
ST_THROW_RT_ERROR("Couldn't allocate memory!\n");
}
// Align the buffer to begin at 16byte cache line boundary for optimal performance
temp = (SAMPLETYPE *)SOUNDTOUCH_ALIGN_POINTER_16(tempUnaligned);
if (samplesInBuffer)
{
memcpy(temp, ptrBegin(), samplesInBuffer * channels * sizeof(SAMPLETYPE));
}
delete[] bufferUnaligned;
buffer = temp;
bufferUnaligned = tempUnaligned;
bufferPos = 0;
}
else
{
// simply rewind the buffer (if necessary)
rewind();
}
}
// Returns the current buffer capacity in terms of samples
uint FIFOSampleBuffer::getCapacity() const
{
return sizeInBytes / (channels * sizeof(SAMPLETYPE));
}
// Returns the number of samples currently in the buffer
uint FIFOSampleBuffer::numSamples() const
{
return samplesInBuffer;
}
// Output samples from beginning of the sample buffer. Copies demanded number
// of samples to output and removes them from the sample buffer. If there
// are less than 'numsample' samples in the buffer, returns all available.
//
// Returns number of samples copied.
uint FIFOSampleBuffer::receiveSamples(SAMPLETYPE *output, uint maxSamples)
{
uint num;
num = (maxSamples > samplesInBuffer) ? samplesInBuffer : maxSamples;
memcpy(output, ptrBegin(), channels * sizeof(SAMPLETYPE) * num);
return receiveSamples(num);
}
// Removes samples from the beginning of the sample buffer without copying them
// anywhere. Used to reduce the number of samples in the buffer, when accessing
// the sample buffer with the 'ptrBegin' function.
uint FIFOSampleBuffer::receiveSamples(uint maxSamples)
{
if (maxSamples >= samplesInBuffer)
{
uint temp;
temp = samplesInBuffer;
samplesInBuffer = 0;
return temp;
}
samplesInBuffer -= maxSamples;
bufferPos += maxSamples;
return maxSamples;
}
// Returns nonzero if the sample buffer is empty
int FIFOSampleBuffer::isEmpty() const
{
return (samplesInBuffer == 0) ? 1 : 0;
}
// Clears the sample buffer
void FIFOSampleBuffer::clear()
{
samplesInBuffer = 0;
bufferPos = 0;
}
/// allow trimming (downwards) amount of samples in pipeline.
/// Returns adjusted amount of samples
uint FIFOSampleBuffer::adjustAmountOfSamples(uint numSamples)
{
if (numSamples < samplesInBuffer)
{
samplesInBuffer = numSamples;
}
return samplesInBuffer;
}
/// Add silence to end of buffer
void FIFOSampleBuffer::addSilent(uint nSamples)
{
memset(ptrEnd(nSamples), 0, sizeof(SAMPLETYPE) * nSamples * channels);
samplesInBuffer += nSamples;
}
////////////////////////////////////////////////////////////////////////////////
///
/// A buffer class for temporarily storaging sound samples, operates as a
/// first-in-first-out pipe.
///
/// Samples are added to the end of the sample buffer with the 'putSamples'
/// function, and are received from the beginning of the buffer by calling
/// the 'receiveSamples' function. The class automatically removes the
/// outputted samples from the buffer, as well as grows the buffer size
/// whenever necessary.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <stdlib.h>
#include <memory.h>
#include <string.h>
#include <assert.h>
#include "FIFOSampleBuffer.h"
using namespace soundtouch;
// Constructor
FIFOSampleBuffer::FIFOSampleBuffer(int numChannels)
{
assert(numChannels > 0);
sizeInBytes = 0; // reasonable initial value
buffer = nullptr;
bufferUnaligned = nullptr;
samplesInBuffer = 0;
bufferPos = 0;
channels = (uint)numChannels;
ensureCapacity(32); // allocate initial capacity
}
// destructor
FIFOSampleBuffer::~FIFOSampleBuffer()
{
delete[] bufferUnaligned;
bufferUnaligned = nullptr;
buffer = nullptr;
}
// Sets number of channels, 1 = mono, 2 = stereo
void FIFOSampleBuffer::setChannels(int numChannels)
{
uint usedBytes;
if (!verifyNumberOfChannels(numChannels)) return;
usedBytes = channels * samplesInBuffer;
channels = (uint)numChannels;
samplesInBuffer = usedBytes / channels;
}
// if output location pointer 'bufferPos' isn't zero, 'rewinds' the buffer and
// zeroes this pointer by copying samples from the 'bufferPos' pointer
// location on to the beginning of the buffer.
void FIFOSampleBuffer::rewind()
{
if (buffer && bufferPos)
{
memmove(buffer, ptrBegin(), sizeof(SAMPLETYPE) * channels * samplesInBuffer);
bufferPos = 0;
}
}
// Adds 'numSamples' pcs of samples from the 'samples' memory position to
// the sample buffer.
void FIFOSampleBuffer::putSamples(const SAMPLETYPE *samples, uint nSamples)
{
memcpy(ptrEnd(nSamples), samples, sizeof(SAMPLETYPE) * nSamples * channels);
samplesInBuffer += nSamples;
}
// Increases the number of samples in the buffer without copying any actual
// samples.
//
// This function is used to update the number of samples in the sample buffer
// when accessing the buffer directly with 'ptrEnd' function. Please be
// careful though!
void FIFOSampleBuffer::putSamples(uint nSamples)
{
uint req;
req = samplesInBuffer + nSamples;
ensureCapacity(req);
samplesInBuffer += nSamples;
}
// Returns a pointer to the end of the used part of the sample buffer (i.e.
// where the new samples are to be inserted). This function may be used for
// inserting new samples into the sample buffer directly. Please be careful!
//
// Parameter 'slackCapacity' tells the function how much free capacity (in
// terms of samples) there _at least_ should be, in order to the caller to
// successfully insert all the required samples to the buffer. When necessary,
// the function grows the buffer size to comply with this requirement.
//
// When using this function as means for inserting new samples, also remember
// to increase the sample count afterwards, by calling the
// 'putSamples(numSamples)' function.
SAMPLETYPE *FIFOSampleBuffer::ptrEnd(uint slackCapacity)
{
ensureCapacity(samplesInBuffer + slackCapacity);
return buffer + samplesInBuffer * channels;
}
// Returns a pointer to the beginning of the currently non-outputted samples.
// This function is provided for accessing the output samples directly.
// Please be careful!
//
// When using this function to output samples, also remember to 'remove' the
// outputted samples from the buffer by calling the
// 'receiveSamples(numSamples)' function
SAMPLETYPE *FIFOSampleBuffer::ptrBegin()
{
assert(buffer);
return buffer + bufferPos * channels;
}
// Ensures that the buffer has enough capacity, i.e. space for _at least_
// 'capacityRequirement' number of samples. The buffer is grown in steps of
// 4 kilobytes to eliminate the need for frequently growing up the buffer,
// as well as to round the buffer size up to the virtual memory page size.
void FIFOSampleBuffer::ensureCapacity(uint capacityRequirement)
{
SAMPLETYPE *tempUnaligned, *temp;
if (capacityRequirement > getCapacity())
{
// enlarge the buffer in 4kbyte steps (round up to next 4k boundary)
sizeInBytes = (capacityRequirement * channels * sizeof(SAMPLETYPE) + 4095) & (uint)-4096;
assert(sizeInBytes % 2 == 0);
tempUnaligned = new SAMPLETYPE[sizeInBytes / sizeof(SAMPLETYPE) + 16 / sizeof(SAMPLETYPE)];
if (tempUnaligned == nullptr)
{
ST_THROW_RT_ERROR("Couldn't allocate memory!\n");
}
// Align the buffer to begin at 16byte cache line boundary for optimal performance
temp = (SAMPLETYPE *)SOUNDTOUCH_ALIGN_POINTER_16(tempUnaligned);
if (samplesInBuffer)
{
memcpy(temp, ptrBegin(), samplesInBuffer * channels * sizeof(SAMPLETYPE));
}
delete[] bufferUnaligned;
buffer = temp;
bufferUnaligned = tempUnaligned;
bufferPos = 0;
}
else
{
// simply rewind the buffer (if necessary)
rewind();
}
}
// Returns the current buffer capacity in terms of samples
uint FIFOSampleBuffer::getCapacity() const
{
return sizeInBytes / (channels * sizeof(SAMPLETYPE));
}
// Returns the number of samples currently in the buffer
uint FIFOSampleBuffer::numSamples() const
{
return samplesInBuffer;
}
// Output samples from beginning of the sample buffer. Copies demanded number
// of samples to output and removes them from the sample buffer. If there
// are less than 'numsample' samples in the buffer, returns all available.
//
// Returns number of samples copied.
uint FIFOSampleBuffer::receiveSamples(SAMPLETYPE *output, uint maxSamples)
{
uint num;
num = (maxSamples > samplesInBuffer) ? samplesInBuffer : maxSamples;
memcpy(output, ptrBegin(), channels * sizeof(SAMPLETYPE) * num);
return receiveSamples(num);
}
// Removes samples from the beginning of the sample buffer without copying them
// anywhere. Used to reduce the number of samples in the buffer, when accessing
// the sample buffer with the 'ptrBegin' function.
uint FIFOSampleBuffer::receiveSamples(uint maxSamples)
{
if (maxSamples >= samplesInBuffer)
{
uint temp;
temp = samplesInBuffer;
samplesInBuffer = 0;
return temp;
}
samplesInBuffer -= maxSamples;
bufferPos += maxSamples;
return maxSamples;
}
// Returns nonzero if the sample buffer is empty
int FIFOSampleBuffer::isEmpty() const
{
return (samplesInBuffer == 0) ? 1 : 0;
}
// Clears the sample buffer
void FIFOSampleBuffer::clear()
{
samplesInBuffer = 0;
bufferPos = 0;
}
/// allow trimming (downwards) amount of samples in pipeline.
/// Returns adjusted amount of samples
uint FIFOSampleBuffer::adjustAmountOfSamples(uint numSamples)
{
if (numSamples < samplesInBuffer)
{
samplesInBuffer = numSamples;
}
return samplesInBuffer;
}
/// Add silence to end of buffer
void FIFOSampleBuffer::addSilent(uint nSamples)
{
memset(ptrEnd(nSamples), 0, sizeof(SAMPLETYPE) * nSamples * channels);
samplesInBuffer += nSamples;
}

630
source/SoundTouch/FIRFilter.cpp
View File

@ -1,315 +1,315 @@
////////////////////////////////////////////////////////////////////////////////
///
/// General FIR digital filter routines with MMX optimization.
///
/// Notes : MMX optimized functions reside in a separate, platform-specific file,
/// e.g. 'mmx_win.cpp' or 'mmx_gcc.cpp'
///
/// This source file contains OpenMP optimizations that allow speeding up the
/// corss-correlation algorithm by executing it in several threads / CPU cores
/// in parallel. See the following article link for more detailed discussion
/// about SoundTouch OpenMP optimizations:
/// http://www.softwarecoven.com/parallel-computing-in-embedded-mobile-devices
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <memory.h>
#include <assert.h>
#include <math.h>
#include <stdlib.h>
#include "FIRFilter.h"
#include "cpu_detect.h"
using namespace soundtouch;
/*****************************************************************************
*
* Implementation of the class 'FIRFilter'
*
*****************************************************************************/
FIRFilter::FIRFilter()
{
resultDivFactor = 0;
resultDivider = 0;
length = 0;
lengthDiv8 = 0;
filterCoeffs = nullptr;
filterCoeffsStereo = nullptr;
}
FIRFilter::~FIRFilter()
{
delete[] filterCoeffs;
delete[] filterCoeffsStereo;
}
// Usual C-version of the filter routine for stereo sound
uint FIRFilter::evaluateFilterStereo(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples) const
{
int j, end;
// hint compiler autovectorization that loop length is divisible by 8
uint ilength = length & -8;
assert((length != 0) && (length == ilength) && (src != nullptr) && (dest != nullptr) && (filterCoeffs != nullptr));
assert(numSamples > ilength);
end = 2 * (numSamples - ilength);
#pragma omp parallel for
for (j = 0; j < end; j += 2)
{
const SAMPLETYPE *ptr;
LONG_SAMPLETYPE suml, sumr;
suml = sumr = 0;
ptr = src + j;
for (uint i = 0; i < ilength; i ++)
{
suml += ptr[2 * i] * filterCoeffsStereo[2 * i];
sumr += ptr[2 * i + 1] * filterCoeffsStereo[2 * i + 1];
}
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
suml >>= resultDivFactor;
sumr >>= resultDivFactor;
// saturate to 16 bit integer limits
suml = (suml < -32768) ? -32768 : (suml > 32767) ? 32767 : suml;
// saturate to 16 bit integer limits
sumr = (sumr < -32768) ? -32768 : (sumr > 32767) ? 32767 : sumr;
#endif // SOUNDTOUCH_INTEGER_SAMPLES
dest[j] = (SAMPLETYPE)suml;
dest[j + 1] = (SAMPLETYPE)sumr;
}
return numSamples - ilength;
}
// Usual C-version of the filter routine for mono sound
uint FIRFilter::evaluateFilterMono(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples) const
{
int j, end;
// hint compiler autovectorization that loop length is divisible by 8
int ilength = length & -8;
assert(ilength != 0);
end = numSamples - ilength;
#pragma omp parallel for
for (j = 0; j < end; j ++)
{
const SAMPLETYPE *pSrc = src + j;
LONG_SAMPLETYPE sum;
int i;
sum = 0;
for (i = 0; i < ilength; i ++)
{
sum += pSrc[i] * filterCoeffs[i];
}
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
sum >>= resultDivFactor;
// saturate to 16 bit integer limits
sum = (sum < -32768) ? -32768 : (sum > 32767) ? 32767 : sum;
#endif // SOUNDTOUCH_INTEGER_SAMPLES
dest[j] = (SAMPLETYPE)sum;
}
return end;
}
uint FIRFilter::evaluateFilterMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples, uint numChannels)
{
int j, end;
assert(length != 0);
assert(src != nullptr);
assert(dest != nullptr);
assert(filterCoeffs != nullptr);
assert(numChannels < 16);
// hint compiler autovectorization that loop length is divisible by 8
int ilength = length & -8;
end = numChannels * (numSamples - ilength);
#pragma omp parallel for
for (j = 0; j < end; j += numChannels)
{
const SAMPLETYPE *ptr;
LONG_SAMPLETYPE sums[16];
uint c;
int i;
for (c = 0; c < numChannels; c ++)
{
sums[c] = 0;
}
ptr = src + j;
for (i = 0; i < ilength; i ++)
{
SAMPLETYPE coef=filterCoeffs[i];
for (c = 0; c < numChannels; c ++)
{
sums[c] += ptr[0] * coef;
ptr ++;
}
}
for (c = 0; c < numChannels; c ++)
{
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
sums[c] >>= resultDivFactor;
#endif // SOUNDTOUCH_INTEGER_SAMPLES
dest[j+c] = (SAMPLETYPE)sums[c];
}
}
return numSamples - ilength;
}
// Set filter coeffiecients and length.
//
// Throws an exception if filter length isn't divisible by 8
void FIRFilter::setCoefficients(const SAMPLETYPE *coeffs, uint newLength, uint uResultDivFactor)
{
assert(newLength > 0);
if (newLength % 8) ST_THROW_RT_ERROR("FIR filter length not divisible by 8");
#ifdef SOUNDTOUCH_FLOAT_SAMPLES
// scale coefficients already here if using floating samples
double scale = 1.0 / resultDivider;
#else
short scale = 1;
#endif
lengthDiv8 = newLength / 8;
length = lengthDiv8 * 8;
assert(length == newLength);
resultDivFactor = uResultDivFactor;
resultDivider = (SAMPLETYPE)::pow(2.0, (int)resultDivFactor);
delete[] filterCoeffs;
filterCoeffs = new SAMPLETYPE[length];
delete[] filterCoeffsStereo;
filterCoeffsStereo = new SAMPLETYPE[length*2];
for (uint i = 0; i < length; i ++)
{
filterCoeffs[i] = (SAMPLETYPE)(coeffs[i] * scale);
// create also stereo set of filter coefficients: this allows compiler
// to autovectorize filter evaluation much more efficiently
filterCoeffsStereo[2 * i] = (SAMPLETYPE)(coeffs[i] * scale);
filterCoeffsStereo[2 * i + 1] = (SAMPLETYPE)(coeffs[i] * scale);
}
}
uint FIRFilter::getLength() const
{
return length;
}
// Applies the filter to the given sequence of samples.
//
// Note : The amount of outputted samples is by value of 'filter_length'
// smaller than the amount of input samples.
uint FIRFilter::evaluate(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples, uint numChannels)
{
assert(length > 0);
assert(lengthDiv8 * 8 == length);
if (numSamples < length) return 0;
#ifndef USE_MULTICH_ALWAYS
if (numChannels == 1)
{
return evaluateFilterMono(dest, src, numSamples);
}
else if (numChannels == 2)
{
return evaluateFilterStereo(dest, src, numSamples);
}
else
#endif // USE_MULTICH_ALWAYS
{
assert(numChannels > 0);
return evaluateFilterMulti(dest, src, numSamples, numChannels);
}
}
// Operator 'new' is overloaded so that it automatically creates a suitable instance
// depending on if we've a MMX-capable CPU available or not.
void * FIRFilter::operator new(size_t)
{
// Notice! don't use "new FIRFilter" directly, use "newInstance" to create a new instance instead!
ST_THROW_RT_ERROR("Error in FIRFilter::new: Don't use 'new FIRFilter', use 'newInstance' member instead!");
return newInstance();
}
FIRFilter * FIRFilter::newInstance()
{
uint uExtensions;
uExtensions = detectCPUextensions();
(void)uExtensions;
// Check if MMX/SSE instruction set extensions supported by CPU
#ifdef SOUNDTOUCH_ALLOW_MMX
// MMX routines available only with integer sample types
if (uExtensions & SUPPORT_MMX)
{
return ::new FIRFilterMMX;
}
else
#endif // SOUNDTOUCH_ALLOW_MMX
#ifdef SOUNDTOUCH_ALLOW_SSE
if (uExtensions & SUPPORT_SSE)
{
// SSE support
return ::new FIRFilterSSE;
}
else
#endif // SOUNDTOUCH_ALLOW_SSE
{
// ISA optimizations not supported, use plain C version
return ::new FIRFilter;
}
}
////////////////////////////////////////////////////////////////////////////////
///
/// General FIR digital filter routines with MMX optimization.
///
/// Notes : MMX optimized functions reside in a separate, platform-specific file,
/// e.g. 'mmx_win.cpp' or 'mmx_gcc.cpp'
///
/// This source file contains OpenMP optimizations that allow speeding up the
/// corss-correlation algorithm by executing it in several threads / CPU cores
/// in parallel. See the following article link for more detailed discussion
/// about SoundTouch OpenMP optimizations:
/// http://www.softwarecoven.com/parallel-computing-in-embedded-mobile-devices
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <memory.h>
#include <assert.h>
#include <math.h>
#include <stdlib.h>
#include "FIRFilter.h"
#include "cpu_detect.h"
using namespace soundtouch;
/*****************************************************************************
*
* Implementation of the class 'FIRFilter'
*
*****************************************************************************/
FIRFilter::FIRFilter()
{
resultDivFactor = 0;
resultDivider = 0;
length = 0;
lengthDiv8 = 0;
filterCoeffs = nullptr;
filterCoeffsStereo = nullptr;
}
FIRFilter::~FIRFilter()
{
delete[] filterCoeffs;
delete[] filterCoeffsStereo;
}
// Usual C-version of the filter routine for stereo sound
uint FIRFilter::evaluateFilterStereo(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples) const
{
int j, end;
// hint compiler autovectorization that loop length is divisible by 8
uint ilength = length & -8;
assert((length != 0) && (length == ilength) && (src != nullptr) && (dest != nullptr) && (filterCoeffs != nullptr));
assert(numSamples > ilength);
end = 2 * (numSamples - ilength);
#pragma omp parallel for
for (j = 0; j < end; j += 2)
{
const SAMPLETYPE *ptr;
LONG_SAMPLETYPE suml, sumr;
suml = sumr = 0;
ptr = src + j;
for (uint i = 0; i < ilength; i ++)
{
suml += ptr[2 * i] * filterCoeffsStereo[2 * i];
sumr += ptr[2 * i + 1] * filterCoeffsStereo[2 * i + 1];
}
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
suml >>= resultDivFactor;
sumr >>= resultDivFactor;
// saturate to 16 bit integer limits
suml = (suml < -32768) ? -32768 : (suml > 32767) ? 32767 : suml;
// saturate to 16 bit integer limits
sumr = (sumr < -32768) ? -32768 : (sumr > 32767) ? 32767 : sumr;
#endif // SOUNDTOUCH_INTEGER_SAMPLES
dest[j] = (SAMPLETYPE)suml;
dest[j + 1] = (SAMPLETYPE)sumr;
}
return numSamples - ilength;
}
// Usual C-version of the filter routine for mono sound
uint FIRFilter::evaluateFilterMono(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples) const
{
int j, end;
// hint compiler autovectorization that loop length is divisible by 8
int ilength = length & -8;
assert(ilength != 0);
end = numSamples - ilength;
#pragma omp parallel for
for (j = 0; j < end; j ++)
{
const SAMPLETYPE *pSrc = src + j;
LONG_SAMPLETYPE sum;
int i;
sum = 0;
for (i = 0; i < ilength; i ++)
{
sum += pSrc[i] * filterCoeffs[i];
}
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
sum >>= resultDivFactor;
// saturate to 16 bit integer limits
sum = (sum < -32768) ? -32768 : (sum > 32767) ? 32767 : sum;
#endif // SOUNDTOUCH_INTEGER_SAMPLES
dest[j] = (SAMPLETYPE)sum;
}
return end;
}
uint FIRFilter::evaluateFilterMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples, uint numChannels)
{
int j, end;
assert(length != 0);
assert(src != nullptr);
assert(dest != nullptr);
assert(filterCoeffs != nullptr);
assert(numChannels < 16);
// hint compiler autovectorization that loop length is divisible by 8
int ilength = length & -8;
end = numChannels * (numSamples - ilength);
#pragma omp parallel for
for (j = 0; j < end; j += numChannels)
{
const SAMPLETYPE *ptr;
LONG_SAMPLETYPE sums[16];
uint c;
int i;
for (c = 0; c < numChannels; c ++)
{
sums[c] = 0;
}
ptr = src + j;
for (i = 0; i < ilength; i ++)
{
SAMPLETYPE coef=filterCoeffs[i];
for (c = 0; c < numChannels; c ++)
{
sums[c] += ptr[0] * coef;
ptr ++;
}
}
for (c = 0; c < numChannels; c ++)
{
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
sums[c] >>= resultDivFactor;
#endif // SOUNDTOUCH_INTEGER_SAMPLES
dest[j+c] = (SAMPLETYPE)sums[c];
}
}
return numSamples - ilength;
}
// Set filter coeffiecients and length.
//
// Throws an exception if filter length isn't divisible by 8
void FIRFilter::setCoefficients(const SAMPLETYPE *coeffs, uint newLength, uint uResultDivFactor)
{
assert(newLength > 0);
if (newLength % 8) ST_THROW_RT_ERROR("FIR filter length not divisible by 8");
#ifdef SOUNDTOUCH_FLOAT_SAMPLES
// scale coefficients already here if using floating samples
double scale = 1.0 / resultDivider;
#else
short scale = 1;
#endif
lengthDiv8 = newLength / 8;
length = lengthDiv8 * 8;
assert(length == newLength);
resultDivFactor = uResultDivFactor;
resultDivider = (SAMPLETYPE)::pow(2.0, (int)resultDivFactor);
delete[] filterCoeffs;
filterCoeffs = new SAMPLETYPE[length];
delete[] filterCoeffsStereo;
filterCoeffsStereo = new SAMPLETYPE[length*2];
for (uint i = 0; i < length; i ++)
{
filterCoeffs[i] = (SAMPLETYPE)(coeffs[i] * scale);
// create also stereo set of filter coefficients: this allows compiler
// to autovectorize filter evaluation much more efficiently
filterCoeffsStereo[2 * i] = (SAMPLETYPE)(coeffs[i] * scale);
filterCoeffsStereo[2 * i + 1] = (SAMPLETYPE)(coeffs[i] * scale);
}
}
uint FIRFilter::getLength() const
{
return length;
}
// Applies the filter to the given sequence of samples.
//
// Note : The amount of outputted samples is by value of 'filter_length'
// smaller than the amount of input samples.
uint FIRFilter::evaluate(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples, uint numChannels)
{
assert(length > 0);
assert(lengthDiv8 * 8 == length);
if (numSamples < length) return 0;
#ifndef USE_MULTICH_ALWAYS
if (numChannels == 1)
{
return evaluateFilterMono(dest, src, numSamples);
}
else if (numChannels == 2)
{
return evaluateFilterStereo(dest, src, numSamples);
}
else
#endif // USE_MULTICH_ALWAYS
{
assert(numChannels > 0);
return evaluateFilterMulti(dest, src, numSamples, numChannels);
}
}
// Operator 'new' is overloaded so that it automatically creates a suitable instance
// depending on if we've a MMX-capable CPU available or not.
void * FIRFilter::operator new(size_t)
{
// Notice! don't use "new FIRFilter" directly, use "newInstance" to create a new instance instead!
ST_THROW_RT_ERROR("Error in FIRFilter::new: Don't use 'new FIRFilter', use 'newInstance' member instead!");
return newInstance();
}
FIRFilter * FIRFilter::newInstance()
{
uint uExtensions;
uExtensions = detectCPUextensions();
(void)uExtensions;
// Check if MMX/SSE instruction set extensions supported by CPU
#ifdef SOUNDTOUCH_ALLOW_MMX
// MMX routines available only with integer sample types
if (uExtensions & SUPPORT_MMX)
{
return ::new FIRFilterMMX;
}
else
#endif // SOUNDTOUCH_ALLOW_MMX
#ifdef SOUNDTOUCH_ALLOW_SSE
if (uExtensions & SUPPORT_SSE)
{
// SSE support
return ::new FIRFilterSSE;
}
else
#endif // SOUNDTOUCH_ALLOW_SSE
{
// ISA optimizations not supported, use plain C version
return ::new FIRFilter;
}
}

280
source/SoundTouch/FIRFilter.h
View File

@ -1,140 +1,140 @@
////////////////////////////////////////////////////////////////////////////////
///
/// General FIR digital filter routines with MMX optimization.
///
/// Note : MMX optimized functions reside in a separate, platform-specific file,
/// e.g. 'mmx_win.cpp' or 'mmx_gcc.cpp'
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef FIRFilter_H
#define FIRFilter_H
#include <stddef.h>
#include "STTypes.h"
namespace soundtouch
{
class FIRFilter
{
protected:
// Number of FIR filter taps
uint length;
// Number of FIR filter taps divided by 8
uint lengthDiv8;
// Result divider factor in 2^k format
uint resultDivFactor;
// Result divider value.
SAMPLETYPE resultDivider;
// Memory for filter coefficients
SAMPLETYPE *filterCoeffs;
SAMPLETYPE *filterCoeffsStereo;
virtual uint evaluateFilterStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
uint numSamples) const;
virtual uint evaluateFilterMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
uint numSamples) const;
virtual uint evaluateFilterMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples, uint numChannels);
public:
FIRFilter();
virtual ~FIRFilter();
/// Operator 'new' is overloaded so that it automatically creates a suitable instance
/// depending on if we've a MMX-capable CPU available or not.
static void * operator new(size_t s);
static FIRFilter *newInstance();
/// Applies the filter to the given sequence of samples.
/// Note : The amount of outputted samples is by value of 'filter_length'
/// smaller than the amount of input samples.
///
/// \return Number of samples copied to 'dest'.
uint evaluate(SAMPLETYPE *dest,
const SAMPLETYPE *src,
uint numSamples,
uint numChannels);
uint getLength() const;
virtual void setCoefficients(const SAMPLETYPE *coeffs,
uint newLength,
uint uResultDivFactor);
};
// Optional subclasses that implement CPU-specific optimizations:
#ifdef SOUNDTOUCH_ALLOW_MMX
/// Class that implements MMX optimized functions exclusive for 16bit integer samples type.
class FIRFilterMMX : public FIRFilter
{
protected:
short *filterCoeffsUnalign;
short *filterCoeffsAlign;
virtual uint evaluateFilterStereo(short *dest, const short *src, uint numSamples) const override;
public:
FIRFilterMMX();
~FIRFilterMMX();
virtual void setCoefficients(const short *coeffs, uint newLength, uint uResultDivFactor) override;
};
#endif // SOUNDTOUCH_ALLOW_MMX
#ifdef SOUNDTOUCH_ALLOW_SSE
/// Class that implements SSE optimized functions exclusive for floating point samples type.
class FIRFilterSSE : public FIRFilter
{
protected:
float *filterCoeffsUnalign;
float *filterCoeffsAlign;
virtual uint evaluateFilterStereo(float *dest, const float *src, uint numSamples) const override;
public:
FIRFilterSSE();
~FIRFilterSSE();
virtual void setCoefficients(const float *coeffs, uint newLength, uint uResultDivFactor) override;
};
#endif // SOUNDTOUCH_ALLOW_SSE
}
#endif // FIRFilter_H
////////////////////////////////////////////////////////////////////////////////
///
/// General FIR digital filter routines with MMX optimization.
///
/// Note : MMX optimized functions reside in a separate, platform-specific file,
/// e.g. 'mmx_win.cpp' or 'mmx_gcc.cpp'
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef FIRFilter_H
#define FIRFilter_H
#include <stddef.h>
#include "STTypes.h"
namespace soundtouch
{
class FIRFilter
{
protected:
// Number of FIR filter taps
uint length;
// Number of FIR filter taps divided by 8
uint lengthDiv8;
// Result divider factor in 2^k format
uint resultDivFactor;
// Result divider value.
SAMPLETYPE resultDivider;
// Memory for filter coefficients
SAMPLETYPE *filterCoeffs;
SAMPLETYPE *filterCoeffsStereo;
virtual uint evaluateFilterStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
uint numSamples) const;
virtual uint evaluateFilterMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
uint numSamples) const;
virtual uint evaluateFilterMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples, uint numChannels);
public:
FIRFilter();
virtual ~FIRFilter();
/// Operator 'new' is overloaded so that it automatically creates a suitable instance
/// depending on if we've a MMX-capable CPU available or not.
static void * operator new(size_t s);
static FIRFilter *newInstance();
/// Applies the filter to the given sequence of samples.
/// Note : The amount of outputted samples is by value of 'filter_length'
/// smaller than the amount of input samples.
///
/// \return Number of samples copied to 'dest'.
uint evaluate(SAMPLETYPE *dest,
const SAMPLETYPE *src,
uint numSamples,
uint numChannels);
uint getLength() const;
virtual void setCoefficients(const SAMPLETYPE *coeffs,
uint newLength,
uint uResultDivFactor);
};
// Optional subclasses that implement CPU-specific optimizations:
#ifdef SOUNDTOUCH_ALLOW_MMX
/// Class that implements MMX optimized functions exclusive for 16bit integer samples type.
class FIRFilterMMX : public FIRFilter
{
protected:
short *filterCoeffsUnalign;
short *filterCoeffsAlign;
virtual uint evaluateFilterStereo(short *dest, const short *src, uint numSamples) const override;
public:
FIRFilterMMX();
~FIRFilterMMX();
virtual void setCoefficients(const short *coeffs, uint newLength, uint uResultDivFactor) override;
};
#endif // SOUNDTOUCH_ALLOW_MMX
#ifdef SOUNDTOUCH_ALLOW_SSE
/// Class that implements SSE optimized functions exclusive for floating point samples type.
class FIRFilterSSE : public FIRFilter
{
protected:
float *filterCoeffsUnalign;
float *filterCoeffsAlign;
virtual uint evaluateFilterStereo(float *dest, const float *src, uint numSamples) const override;
public:
FIRFilterSSE();
~FIRFilterSSE();
virtual void setCoefficients(const float *coeffs, uint newLength, uint uResultDivFactor) override;
};
#endif // SOUNDTOUCH_ALLOW_SSE
}
#endif // FIRFilter_H

392
source/SoundTouch/InterpolateCubic.cpp
View File

@ -1,196 +1,196 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Cubic interpolation routine.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <stddef.h>
#include <math.h>
#include "InterpolateCubic.h"
#include "STTypes.h"
using namespace soundtouch;
// cubic interpolation coefficients
static const float _coeffs[]=
{ -0.5f, 1.0f, -0.5f, 0.0f,
1.5f, -2.5f, 0.0f, 1.0f,
-1.5f, 2.0f, 0.5f, 0.0f,
0.5f, -0.5f, 0.0f, 0.0f};
InterpolateCubic::InterpolateCubic()
{
fract = 0;
}
void InterpolateCubic::resetRegisters()
{
fract = 0;
}
/// Transpose mono audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateCubic::transposeMono(SAMPLETYPE *pdest,
const SAMPLETYPE *psrc,
int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 4;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
float out;
const float x3 = 1.0f;
const float x2 = (float)fract; // x
const float x1 = x2*x2; // x^2
const float x0 = x1*x2; // x^3
float y0, y1, y2, y3;
assert(fract < 1.0);
y0 = _coeffs[0] * x0 + _coeffs[1] * x1 + _coeffs[2] * x2 + _coeffs[3] * x3;
y1 = _coeffs[4] * x0 + _coeffs[5] * x1 + _coeffs[6] * x2 + _coeffs[7] * x3;
y2 = _coeffs[8] * x0 + _coeffs[9] * x1 + _coeffs[10] * x2 + _coeffs[11] * x3;
y3 = _coeffs[12] * x0 + _coeffs[13] * x1 + _coeffs[14] * x2 + _coeffs[15] * x3;
out = y0 * psrc[0] + y1 * psrc[1] + y2 * psrc[2] + y3 * psrc[3];
pdest[i] = (SAMPLETYPE)out;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
psrc += whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
/// Transpose stereo audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateCubic::transposeStereo(SAMPLETYPE *pdest,
const SAMPLETYPE *psrc,
int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 4;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
const float x3 = 1.0f;
const float x2 = (float)fract; // x
const float x1 = x2*x2; // x^2
const float x0 = x1*x2; // x^3
float y0, y1, y2, y3;
float out0, out1;
assert(fract < 1.0);
y0 = _coeffs[0] * x0 + _coeffs[1] * x1 + _coeffs[2] * x2 + _coeffs[3] * x3;
y1 = _coeffs[4] * x0 + _coeffs[5] * x1 + _coeffs[6] * x2 + _coeffs[7] * x3;
y2 = _coeffs[8] * x0 + _coeffs[9] * x1 + _coeffs[10] * x2 + _coeffs[11] * x3;
y3 = _coeffs[12] * x0 + _coeffs[13] * x1 + _coeffs[14] * x2 + _coeffs[15] * x3;
out0 = y0 * psrc[0] + y1 * psrc[2] + y2 * psrc[4] + y3 * psrc[6];
out1 = y0 * psrc[1] + y1 * psrc[3] + y2 * psrc[5] + y3 * psrc[7];
pdest[2*i] = (SAMPLETYPE)out0;
pdest[2*i+1] = (SAMPLETYPE)out1;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
psrc += 2*whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
/// Transpose multi-channel audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateCubic::transposeMulti(SAMPLETYPE *pdest,
const SAMPLETYPE *psrc,
int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 4;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
const float x3 = 1.0f;
const float x2 = (float)fract; // x
const float x1 = x2*x2; // x^2
const float x0 = x1*x2; // x^3
float y0, y1, y2, y3;
assert(fract < 1.0);
y0 = _coeffs[0] * x0 + _coeffs[1] * x1 + _coeffs[2] * x2 + _coeffs[3] * x3;
y1 = _coeffs[4] * x0 + _coeffs[5] * x1 + _coeffs[6] * x2 + _coeffs[7] * x3;
y2 = _coeffs[8] * x0 + _coeffs[9] * x1 + _coeffs[10] * x2 + _coeffs[11] * x3;
y3 = _coeffs[12] * x0 + _coeffs[13] * x1 + _coeffs[14] * x2 + _coeffs[15] * x3;
for (int c = 0; c < numChannels; c ++)
{
float out;
out = y0 * psrc[c] + y1 * psrc[c + numChannels] + y2 * psrc[c + 2 * numChannels] + y3 * psrc[c + 3 * numChannels];
pdest[0] = (SAMPLETYPE)out;
pdest ++;
}
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
psrc += numChannels*whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
////////////////////////////////////////////////////////////////////////////////
///
/// Cubic interpolation routine.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <stddef.h>
#include <math.h>
#include "InterpolateCubic.h"
#include "STTypes.h"
using namespace soundtouch;
// cubic interpolation coefficients
static const float _coeffs[]=
{ -0.5f, 1.0f, -0.5f, 0.0f,
1.5f, -2.5f, 0.0f, 1.0f,
-1.5f, 2.0f, 0.5f, 0.0f,
0.5f, -0.5f, 0.0f, 0.0f};
InterpolateCubic::InterpolateCubic()
{
fract = 0;
}
void InterpolateCubic::resetRegisters()
{
fract = 0;
}
/// Transpose mono audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateCubic::transposeMono(SAMPLETYPE *pdest,
const SAMPLETYPE *psrc,
int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 4;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
float out;
const float x3 = 1.0f;
const float x2 = (float)fract; // x
const float x1 = x2*x2; // x^2
const float x0 = x1*x2; // x^3
float y0, y1, y2, y3;
assert(fract < 1.0);
y0 = _coeffs[0] * x0 + _coeffs[1] * x1 + _coeffs[2] * x2 + _coeffs[3] * x3;
y1 = _coeffs[4] * x0 + _coeffs[5] * x1 + _coeffs[6] * x2 + _coeffs[7] * x3;
y2 = _coeffs[8] * x0 + _coeffs[9] * x1 + _coeffs[10] * x2 + _coeffs[11] * x3;
y3 = _coeffs[12] * x0 + _coeffs[13] * x1 + _coeffs[14] * x2 + _coeffs[15] * x3;
out = y0 * psrc[0] + y1 * psrc[1] + y2 * psrc[2] + y3 * psrc[3];
pdest[i] = (SAMPLETYPE)out;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
psrc += whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
/// Transpose stereo audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateCubic::transposeStereo(SAMPLETYPE *pdest,
const SAMPLETYPE *psrc,
int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 4;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
const float x3 = 1.0f;
const float x2 = (float)fract; // x
const float x1 = x2*x2; // x^2
const float x0 = x1*x2; // x^3
float y0, y1, y2, y3;
float out0, out1;
assert(fract < 1.0);
y0 = _coeffs[0] * x0 + _coeffs[1] * x1 + _coeffs[2] * x2 + _coeffs[3] * x3;
y1 = _coeffs[4] * x0 + _coeffs[5] * x1 + _coeffs[6] * x2 + _coeffs[7] * x3;
y2 = _coeffs[8] * x0 + _coeffs[9] * x1 + _coeffs[10] * x2 + _coeffs[11] * x3;
y3 = _coeffs[12] * x0 + _coeffs[13] * x1 + _coeffs[14] * x2 + _coeffs[15] * x3;
out0 = y0 * psrc[0] + y1 * psrc[2] + y2 * psrc[4] + y3 * psrc[6];
out1 = y0 * psrc[1] + y1 * psrc[3] + y2 * psrc[5] + y3 * psrc[7];
pdest[2*i] = (SAMPLETYPE)out0;
pdest[2*i+1] = (SAMPLETYPE)out1;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
psrc += 2*whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
/// Transpose multi-channel audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateCubic::transposeMulti(SAMPLETYPE *pdest,
const SAMPLETYPE *psrc,
int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 4;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
const float x3 = 1.0f;
const float x2 = (float)fract; // x
const float x1 = x2*x2; // x^2
const float x0 = x1*x2; // x^3
float y0, y1, y2, y3;
assert(fract < 1.0);
y0 = _coeffs[0] * x0 + _coeffs[1] * x1 + _coeffs[2] * x2 + _coeffs[3] * x3;
y1 = _coeffs[4] * x0 + _coeffs[5] * x1 + _coeffs[6] * x2 + _coeffs[7] * x3;
y2 = _coeffs[8] * x0 + _coeffs[9] * x1 + _coeffs[10] * x2 + _coeffs[11] * x3;
y3 = _coeffs[12] * x0 + _coeffs[13] * x1 + _coeffs[14] * x2 + _coeffs[15] * x3;
for (int c = 0; c < numChannels; c ++)
{
float out;
out = y0 * psrc[c] + y1 * psrc[c + numChannels] + y2 * psrc[c + 2 * numChannels] + y3 * psrc[c + 3 * numChannels];
pdest[0] = (SAMPLETYPE)out;
pdest ++;
}
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
psrc += numChannels*whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}

138
source/SoundTouch/InterpolateCubic.h
View File

@ -1,69 +1,69 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Cubic interpolation routine.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _InterpolateCubic_H_
#define _InterpolateCubic_H_
#include "RateTransposer.h"
#include "STTypes.h"
namespace soundtouch
{
class InterpolateCubic : public TransposerBase
{
protected:
virtual int transposeMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
virtual int transposeStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
virtual int transposeMulti(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
double fract;
public:
InterpolateCubic();
virtual void resetRegisters() override;
virtual int getLatency() const override
{
return 1;
}
};
}
#endif
////////////////////////////////////////////////////////////////////////////////
///
/// Cubic interpolation routine.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _InterpolateCubic_H_
#define _InterpolateCubic_H_
#include "RateTransposer.h"
#include "STTypes.h"
namespace soundtouch
{
class InterpolateCubic : public TransposerBase
{
protected:
virtual int transposeMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
virtual int transposeStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
virtual int transposeMulti(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
double fract;
public:
InterpolateCubic();
virtual void resetRegisters() override;
virtual int getLatency() const override
{
return 1;
}
};
}
#endif

592
source/SoundTouch/InterpolateLinear.cpp
View File

@ -1,296 +1,296 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Linear interpolation algorithm.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <assert.h>
#include <stdlib.h>
#include "InterpolateLinear.h"
using namespace soundtouch;
//////////////////////////////////////////////////////////////////////////////
//
// InterpolateLinearInteger - integer arithmetic implementation
//
/// fixed-point interpolation routine precision
#define SCALE 65536
// Constructor
InterpolateLinearInteger::InterpolateLinearInteger() : TransposerBase()
{
// Notice: use local function calling syntax for sake of clarity,
// to indicate the fact that C++ constructor can't call virtual functions.
resetRegisters();
setRate(1.0f);
}
void InterpolateLinearInteger::resetRegisters()
{
iFract = 0;
}
// Transposes the sample rate of the given samples using linear interpolation.
// 'Mono' version of the routine. Returns the number of samples returned in
// the "dest" buffer
int InterpolateLinearInteger::transposeMono(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
LONG_SAMPLETYPE temp;
assert(iFract < SCALE);
temp = (SCALE - iFract) * src[0] + iFract * src[1];
dest[i] = (SAMPLETYPE)(temp / SCALE);
i++;
iFract += iRate;
int iWhole = iFract / SCALE;
iFract -= iWhole * SCALE;
srcCount += iWhole;
src += iWhole;
}
srcSamples = srcCount;
return i;
}
// Transposes the sample rate of the given samples using linear interpolation.
// 'Stereo' version of the routine. Returns the number of samples returned in
// the "dest" buffer
int InterpolateLinearInteger::transposeStereo(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
LONG_SAMPLETYPE temp0;
LONG_SAMPLETYPE temp1;
assert(iFract < SCALE);
temp0 = (SCALE - iFract) * src[0] + iFract * src[2];
temp1 = (SCALE - iFract) * src[1] + iFract * src[3];
dest[0] = (SAMPLETYPE)(temp0 / SCALE);
dest[1] = (SAMPLETYPE)(temp1 / SCALE);
dest += 2;
i++;
iFract += iRate;
int iWhole = iFract / SCALE;
iFract -= iWhole * SCALE;
srcCount += iWhole;
src += 2*iWhole;
}
srcSamples = srcCount;
return i;
}
int InterpolateLinearInteger::transposeMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
LONG_SAMPLETYPE temp, vol1;
assert(iFract < SCALE);
vol1 = (LONG_SAMPLETYPE)(SCALE - iFract);
for (int c = 0; c < numChannels; c ++)
{
temp = vol1 * src[c] + iFract * src[c + numChannels];
dest[0] = (SAMPLETYPE)(temp / SCALE);
dest ++;
}
i++;
iFract += iRate;
int iWhole = iFract / SCALE;
iFract -= iWhole * SCALE;
srcCount += iWhole;
src += iWhole * numChannels;
}
srcSamples = srcCount;
return i;
}
// Sets new target iRate. Normal iRate = 1.0, smaller values represent slower
// iRate, larger faster iRates.
void InterpolateLinearInteger::setRate(double newRate)
{
iRate = (int)(newRate * SCALE + 0.5);
TransposerBase::setRate(newRate);
}
//////////////////////////////////////////////////////////////////////////////
//
// InterpolateLinearFloat - floating point arithmetic implementation
//
//////////////////////////////////////////////////////////////////////////////
// Constructor
InterpolateLinearFloat::InterpolateLinearFloat() : TransposerBase()
{
// Notice: use local function calling syntax for sake of clarity,
// to indicate the fact that C++ constructor can't call virtual functions.
resetRegisters();
setRate(1.0);
}
void InterpolateLinearFloat::resetRegisters()
{
fract = 0;
}
// Transposes the sample rate of the given samples using linear interpolation.
// 'Mono' version of the routine. Returns the number of samples returned in
// the "dest" buffer
int InterpolateLinearFloat::transposeMono(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
double out;
assert(fract < 1.0);
out = (1.0 - fract) * src[0] + fract * src[1];
dest[i] = (SAMPLETYPE)out;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
src += whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
// Transposes the sample rate of the given samples using linear interpolation.
// 'Mono' version of the routine. Returns the number of samples returned in
// the "dest" buffer
int InterpolateLinearFloat::transposeStereo(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
double out0, out1;
assert(fract < 1.0);
out0 = (1.0 - fract) * src[0] + fract * src[2];
out1 = (1.0 - fract) * src[1] + fract * src[3];
dest[2*i] = (SAMPLETYPE)out0;
dest[2*i+1] = (SAMPLETYPE)out1;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
src += 2*whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
int InterpolateLinearFloat::transposeMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
float temp, vol1, fract_float;
vol1 = (float)(1.0 - fract);
fract_float = (float)fract;
for (int c = 0; c < numChannels; c ++)
{
temp = vol1 * src[c] + fract_float * src[c + numChannels];
*dest = (SAMPLETYPE)temp;
dest ++;
}
i++;
fract += rate;
int iWhole = (int)fract;
fract -= iWhole;
srcCount += iWhole;
src += iWhole * numChannels;
}
srcSamples = srcCount;
return i;
}
////////////////////////////////////////////////////////////////////////////////
///
/// Linear interpolation algorithm.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <assert.h>
#include <stdlib.h>
#include "InterpolateLinear.h"
using namespace soundtouch;
//////////////////////////////////////////////////////////////////////////////
//
// InterpolateLinearInteger - integer arithmetic implementation
//
/// fixed-point interpolation routine precision
#define SCALE 65536
// Constructor
InterpolateLinearInteger::InterpolateLinearInteger() : TransposerBase()
{
// Notice: use local function calling syntax for sake of clarity,
// to indicate the fact that C++ constructor can't call virtual functions.
resetRegisters();
setRate(1.0f);
}
void InterpolateLinearInteger::resetRegisters()
{
iFract = 0;
}
// Transposes the sample rate of the given samples using linear interpolation.
// 'Mono' version of the routine. Returns the number of samples returned in
// the "dest" buffer
int InterpolateLinearInteger::transposeMono(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
LONG_SAMPLETYPE temp;
assert(iFract < SCALE);
temp = (SCALE - iFract) * src[0] + iFract * src[1];
dest[i] = (SAMPLETYPE)(temp / SCALE);
i++;
iFract += iRate;
int iWhole = iFract / SCALE;
iFract -= iWhole * SCALE;
srcCount += iWhole;
src += iWhole;
}
srcSamples = srcCount;
return i;
}
// Transposes the sample rate of the given samples using linear interpolation.
// 'Stereo' version of the routine. Returns the number of samples returned in
// the "dest" buffer
int InterpolateLinearInteger::transposeStereo(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
LONG_SAMPLETYPE temp0;
LONG_SAMPLETYPE temp1;
assert(iFract < SCALE);
temp0 = (SCALE - iFract) * src[0] + iFract * src[2];
temp1 = (SCALE - iFract) * src[1] + iFract * src[3];
dest[0] = (SAMPLETYPE)(temp0 / SCALE);
dest[1] = (SAMPLETYPE)(temp1 / SCALE);
dest += 2;
i++;
iFract += iRate;
int iWhole = iFract / SCALE;
iFract -= iWhole * SCALE;
srcCount += iWhole;
src += 2*iWhole;
}
srcSamples = srcCount;
return i;
}
int InterpolateLinearInteger::transposeMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
LONG_SAMPLETYPE temp, vol1;
assert(iFract < SCALE);
vol1 = (LONG_SAMPLETYPE)(SCALE - iFract);
for (int c = 0; c < numChannels; c ++)
{
temp = vol1 * src[c] + iFract * src[c + numChannels];
dest[0] = (SAMPLETYPE)(temp / SCALE);
dest ++;
}
i++;
iFract += iRate;
int iWhole = iFract / SCALE;
iFract -= iWhole * SCALE;
srcCount += iWhole;
src += iWhole * numChannels;
}
srcSamples = srcCount;
return i;
}
// Sets new target iRate. Normal iRate = 1.0, smaller values represent slower
// iRate, larger faster iRates.
void InterpolateLinearInteger::setRate(double newRate)
{
iRate = (int)(newRate * SCALE + 0.5);
TransposerBase::setRate(newRate);
}
//////////////////////////////////////////////////////////////////////////////
//
// InterpolateLinearFloat - floating point arithmetic implementation
//
//////////////////////////////////////////////////////////////////////////////
// Constructor
InterpolateLinearFloat::InterpolateLinearFloat() : TransposerBase()
{
// Notice: use local function calling syntax for sake of clarity,
// to indicate the fact that C++ constructor can't call virtual functions.
resetRegisters();
setRate(1.0);
}
void InterpolateLinearFloat::resetRegisters()
{
fract = 0;
}
// Transposes the sample rate of the given samples using linear interpolation.
// 'Mono' version of the routine. Returns the number of samples returned in
// the "dest" buffer
int InterpolateLinearFloat::transposeMono(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
double out;
assert(fract < 1.0);
out = (1.0 - fract) * src[0] + fract * src[1];
dest[i] = (SAMPLETYPE)out;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
src += whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
// Transposes the sample rate of the given samples using linear interpolation.
// 'Mono' version of the routine. Returns the number of samples returned in
// the "dest" buffer
int InterpolateLinearFloat::transposeStereo(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
double out0, out1;
assert(fract < 1.0);
out0 = (1.0 - fract) * src[0] + fract * src[2];
out1 = (1.0 - fract) * src[1] + fract * src[3];
dest[2*i] = (SAMPLETYPE)out0;
dest[2*i+1] = (SAMPLETYPE)out1;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
src += 2*whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
int InterpolateLinearFloat::transposeMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 1;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
float temp, vol1, fract_float;
vol1 = (float)(1.0 - fract);
fract_float = (float)fract;
for (int c = 0; c < numChannels; c ++)
{
temp = vol1 * src[c] + fract_float * src[c + numChannels];
*dest = (SAMPLETYPE)temp;
dest ++;
}
i++;
fract += rate;
int iWhole = (int)fract;
fract -= iWhole;
srcCount += iWhole;
src += iWhole * numChannels;
}
srcSamples = srcCount;
return i;
}

196
source/SoundTouch/InterpolateLinear.h
View File

@ -1,98 +1,98 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Linear interpolation routine.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _InterpolateLinear_H_
#define _InterpolateLinear_H_
#include "RateTransposer.h"
#include "STTypes.h"
namespace soundtouch
{
/// Linear transposer class that uses integer arithmetic
class InterpolateLinearInteger : public TransposerBase
{
protected:
int iFract;
int iRate;
virtual int transposeMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
virtual int transposeStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
virtual int transposeMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples) override;
public:
InterpolateLinearInteger();
/// Sets new target rate. Normal rate = 1.0, smaller values represent slower
/// rate, larger faster rates.
virtual void setRate(double newRate) override;
virtual void resetRegisters() override;
virtual int getLatency() const override
{
return 0;
}
};
/// Linear transposer class that uses floating point arithmetic
class InterpolateLinearFloat : public TransposerBase
{
protected:
double fract;
virtual int transposeMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples);
virtual int transposeStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples);
virtual int transposeMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples);
public:
InterpolateLinearFloat();
virtual void resetRegisters();
int getLatency() const
{
return 0;
}
};
}
#endif
////////////////////////////////////////////////////////////////////////////////
///
/// Linear interpolation routine.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _InterpolateLinear_H_
#define _InterpolateLinear_H_
#include "RateTransposer.h"
#include "STTypes.h"
namespace soundtouch
{
/// Linear transposer class that uses integer arithmetic
class InterpolateLinearInteger : public TransposerBase
{
protected:
int iFract;
int iRate;
virtual int transposeMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
virtual int transposeStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
virtual int transposeMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples) override;
public:
InterpolateLinearInteger();
/// Sets new target rate. Normal rate = 1.0, smaller values represent slower
/// rate, larger faster rates.
virtual void setRate(double newRate) override;
virtual void resetRegisters() override;
virtual int getLatency() const override
{
return 0;
}
};
/// Linear transposer class that uses floating point arithmetic
class InterpolateLinearFloat : public TransposerBase
{
protected:
double fract;
virtual int transposeMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples);
virtual int transposeStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples);
virtual int transposeMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, int &srcSamples);
public:
InterpolateLinearFloat();
virtual void resetRegisters();
int getLatency() const
{
return 0;
}
};
}
#endif

362
source/SoundTouch/InterpolateShannon.cpp
View File

@ -1,181 +1,181 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Sample interpolation routine using 8-tap band-limited Shannon interpolation
/// with kaiser window.
///
/// Notice. This algorithm is remarkably much heavier than linear or cubic
/// interpolation, and not remarkably better than cubic algorithm. Thus mostly
/// for experimental purposes
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <math.h>
#include "InterpolateShannon.h"
#include "STTypes.h"
using namespace soundtouch;
/// Kaiser window with beta = 2.0
/// Values scaled down by 5% to avoid overflows
static const double _kaiser8[8] =
{
0.41778693317814,
0.64888025049173,
0.83508562409944,
0.93887857733412,
0.93887857733412,
0.83508562409944,
0.64888025049173,
0.41778693317814
};
InterpolateShannon::InterpolateShannon()
{
fract = 0;
}
void InterpolateShannon::resetRegisters()
{
fract = 0;
}
#define PI 3.1415926536
#define sinc(x) (sin(PI * (x)) / (PI * (x)))
/// Transpose mono audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateShannon::transposeMono(SAMPLETYPE *pdest,
const SAMPLETYPE *psrc,
int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 8;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
double out;
assert(fract < 1.0);
out = psrc[0] * sinc(-3.0 - fract) * _kaiser8[0];
out += psrc[1] * sinc(-2.0 - fract) * _kaiser8[1];
out += psrc[2] * sinc(-1.0 - fract) * _kaiser8[2];
if (fract < 1e-6)
{
out += psrc[3] * _kaiser8[3]; // sinc(0) = 1
}
else
{
out += psrc[3] * sinc(- fract) * _kaiser8[3];
}
out += psrc[4] * sinc( 1.0 - fract) * _kaiser8[4];
out += psrc[5] * sinc( 2.0 - fract) * _kaiser8[5];
out += psrc[6] * sinc( 3.0 - fract) * _kaiser8[6];
out += psrc[7] * sinc( 4.0 - fract) * _kaiser8[7];
pdest[i] = (SAMPLETYPE)out;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
psrc += whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
/// Transpose stereo audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateShannon::transposeStereo(SAMPLETYPE *pdest,
const SAMPLETYPE *psrc,
int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 8;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
double out0, out1, w;
assert(fract < 1.0);
w = sinc(-3.0 - fract) * _kaiser8[0];
out0 = psrc[0] * w; out1 = psrc[1] * w;
w = sinc(-2.0 - fract) * _kaiser8[1];
out0 += psrc[2] * w; out1 += psrc[3] * w;
w = sinc(-1.0 - fract) * _kaiser8[2];
out0 += psrc[4] * w; out1 += psrc[5] * w;
w = _kaiser8[3] * ((fract < 1e-5) ? 1.0 : sinc(- fract)); // sinc(0) = 1
out0 += psrc[6] * w; out1 += psrc[7] * w;
w = sinc( 1.0 - fract) * _kaiser8[4];
out0 += psrc[8] * w; out1 += psrc[9] * w;
w = sinc( 2.0 - fract) * _kaiser8[5];
out0 += psrc[10] * w; out1 += psrc[11] * w;
w = sinc( 3.0 - fract) * _kaiser8[6];
out0 += psrc[12] * w; out1 += psrc[13] * w;
w = sinc( 4.0 - fract) * _kaiser8[7];
out0 += psrc[14] * w; out1 += psrc[15] * w;
pdest[2*i] = (SAMPLETYPE)out0;
pdest[2*i+1] = (SAMPLETYPE)out1;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
psrc += 2*whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
/// Transpose stereo audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateShannon::transposeMulti(SAMPLETYPE *,
const SAMPLETYPE *,
int &)
{
// not implemented
assert(false);
return 0;
}
////////////////////////////////////////////////////////////////////////////////
///
/// Sample interpolation routine using 8-tap band-limited Shannon interpolation
/// with kaiser window.
///
/// Notice. This algorithm is remarkably much heavier than linear or cubic
/// interpolation, and not remarkably better than cubic algorithm. Thus mostly
/// for experimental purposes
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <math.h>
#include "InterpolateShannon.h"
#include "STTypes.h"
using namespace soundtouch;
/// Kaiser window with beta = 2.0
/// Values scaled down by 5% to avoid overflows
static const double _kaiser8[8] =
{
0.41778693317814,
0.64888025049173,
0.83508562409944,
0.93887857733412,
0.93887857733412,
0.83508562409944,
0.64888025049173,
0.41778693317814
};
InterpolateShannon::InterpolateShannon()
{
fract = 0;
}
void InterpolateShannon::resetRegisters()
{
fract = 0;
}
#define PI 3.1415926536
#define sinc(x) (sin(PI * (x)) / (PI * (x)))
/// Transpose mono audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateShannon::transposeMono(SAMPLETYPE *pdest,
const SAMPLETYPE *psrc,
int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 8;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
double out;
assert(fract < 1.0);
out = psrc[0] * sinc(-3.0 - fract) * _kaiser8[0];
out += psrc[1] * sinc(-2.0 - fract) * _kaiser8[1];
out += psrc[2] * sinc(-1.0 - fract) * _kaiser8[2];
if (fract < 1e-6)
{
out += psrc[3] * _kaiser8[3]; // sinc(0) = 1
}
else
{
out += psrc[3] * sinc(- fract) * _kaiser8[3];
}
out += psrc[4] * sinc( 1.0 - fract) * _kaiser8[4];
out += psrc[5] * sinc( 2.0 - fract) * _kaiser8[5];
out += psrc[6] * sinc( 3.0 - fract) * _kaiser8[6];
out += psrc[7] * sinc( 4.0 - fract) * _kaiser8[7];
pdest[i] = (SAMPLETYPE)out;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
psrc += whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
/// Transpose stereo audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateShannon::transposeStereo(SAMPLETYPE *pdest,
const SAMPLETYPE *psrc,
int &srcSamples)
{
int i;
int srcSampleEnd = srcSamples - 8;
int srcCount = 0;
i = 0;
while (srcCount < srcSampleEnd)
{
double out0, out1, w;
assert(fract < 1.0);
w = sinc(-3.0 - fract) * _kaiser8[0];
out0 = psrc[0] * w; out1 = psrc[1] * w;
w = sinc(-2.0 - fract) * _kaiser8[1];
out0 += psrc[2] * w; out1 += psrc[3] * w;
w = sinc(-1.0 - fract) * _kaiser8[2];
out0 += psrc[4] * w; out1 += psrc[5] * w;
w = _kaiser8[3] * ((fract < 1e-5) ? 1.0 : sinc(- fract)); // sinc(0) = 1
out0 += psrc[6] * w; out1 += psrc[7] * w;
w = sinc( 1.0 - fract) * _kaiser8[4];
out0 += psrc[8] * w; out1 += psrc[9] * w;
w = sinc( 2.0 - fract) * _kaiser8[5];
out0 += psrc[10] * w; out1 += psrc[11] * w;
w = sinc( 3.0 - fract) * _kaiser8[6];
out0 += psrc[12] * w; out1 += psrc[13] * w;
w = sinc( 4.0 - fract) * _kaiser8[7];
out0 += psrc[14] * w; out1 += psrc[15] * w;
pdest[2*i] = (SAMPLETYPE)out0;
pdest[2*i+1] = (SAMPLETYPE)out1;
i ++;
// update position fraction
fract += rate;
// update whole positions
int whole = (int)fract;
fract -= whole;
psrc += 2*whole;
srcCount += whole;
}
srcSamples = srcCount;
return i;
}
/// Transpose stereo audio. Returns number of produced output samples, and
/// updates "srcSamples" to amount of consumed source samples
int InterpolateShannon::transposeMulti(SAMPLETYPE *,
const SAMPLETYPE *,
int &)
{
// not implemented
assert(false);
return 0;
}

148
source/SoundTouch/InterpolateShannon.h
View File

@ -1,74 +1,74 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Sample interpolation routine using 8-tap band-limited Shannon interpolation
/// with kaiser window.
///
/// Notice. This algorithm is remarkably much heavier than linear or cubic
/// interpolation, and not remarkably better than cubic algorithm. Thus mostly
/// for experimental purposes
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _InterpolateShannon_H_
#define _InterpolateShannon_H_
#include "RateTransposer.h"
#include "STTypes.h"
namespace soundtouch
{
class InterpolateShannon : public TransposerBase
{
protected:
int transposeMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
int transposeStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
int transposeMulti(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
double fract;
public:
InterpolateShannon();
void resetRegisters() override;
virtual int getLatency() const override
{
return 3;
}
};
}
#endif
////////////////////////////////////////////////////////////////////////////////
///
/// Sample interpolation routine using 8-tap band-limited Shannon interpolation
/// with kaiser window.
///
/// Notice. This algorithm is remarkably much heavier than linear or cubic
/// interpolation, and not remarkably better than cubic algorithm. Thus mostly
/// for experimental purposes
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _InterpolateShannon_H_
#define _InterpolateShannon_H_
#include "RateTransposer.h"
#include "STTypes.h"
namespace soundtouch
{
class InterpolateShannon : public TransposerBase
{
protected:
int transposeMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
int transposeStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
int transposeMulti(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) override;
double fract;
public:
InterpolateShannon();
void resetRegisters() override;
virtual int getLatency() const override
{
return 3;
}
};
}
#endif

148
source/SoundTouch/Makefile.am
View File

@ -1,74 +1,74 @@
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA
include $(top_srcdir)/config/am_include.mk
# set to something if you want other stuff to be included in the distribution tarball
EXTRA_DIST=SoundTouch.sln SoundTouch.vcxproj
noinst_HEADERS=AAFilter.h cpu_detect.h cpu_detect_x86.cpp FIRFilter.h RateTransposer.h TDStretch.h PeakFinder.h \
InterpolateCubic.h InterpolateLinear.h InterpolateShannon.h
lib_LTLIBRARIES=libSoundTouch.la
#
libSoundTouch_la_SOURCES=AAFilter.cpp FIRFilter.cpp FIFOSampleBuffer.cpp \
RateTransposer.cpp SoundTouch.cpp TDStretch.cpp cpu_detect_x86.cpp \
BPMDetect.cpp PeakFinder.cpp InterpolateLinear.cpp InterpolateCubic.cpp \
InterpolateShannon.cpp
# Compiler flags
#AM_CXXFLAGS+=
# Compile the files that need MMX and SSE individually.
libSoundTouch_la_LIBADD=libSoundTouchMMX.la libSoundTouchSSE.la
noinst_LTLIBRARIES=libSoundTouchMMX.la libSoundTouchSSE.la
libSoundTouchMMX_la_SOURCES=mmx_optimized.cpp
libSoundTouchSSE_la_SOURCES=sse_optimized.cpp
# We enable optimizations by default.
# If MMX is supported compile with -mmmx.
# Do not assume -msse is also supported.
if HAVE_MMX
libSoundTouchMMX_la_CXXFLAGS = -mmmx $(AM_CXXFLAGS)
else
libSoundTouchMMX_la_CXXFLAGS = $(AM_CXXFLAGS)
endif
# We enable optimizations by default.
# If SSE is supported compile with -msse.
if HAVE_SSE
libSoundTouchSSE_la_CXXFLAGS = -msse $(AM_CXXFLAGS)
else
libSoundTouchSSE_la_CXXFLAGS = $(AM_CXXFLAGS)
endif
# Let the user disable optimizations if he wishes to.
if !X86_OPTIMIZATIONS
libSoundTouchMMX_la_CXXFLAGS = $(AM_CXXFLAGS)
libSoundTouchSSE_la_CXXFLAGS = $(AM_CXXFLAGS)
endif
# Modify the default 0.0.0 to LIB_SONAME.0.0
libSoundTouch_la_LDFLAGS=-version-info @LIB_SONAME@
# other linking flags to add
# noinst_LTLIBRARIES = libSoundTouchOpt.la
# libSoundTouch_la_LIBADD = libSoundTouchOpt.la
# libSoundTouchOpt_la_SOURCES = mmx_optimized.cpp sse_optimized.cpp
# libSoundTouchOpt_la_CXXFLAGS = -O3 -msse -fcheck-new -I../../include
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA
include $(top_srcdir)/config/am_include.mk
# set to something if you want other stuff to be included in the distribution tarball
EXTRA_DIST=SoundTouch.sln SoundTouch.vcxproj
noinst_HEADERS=AAFilter.h cpu_detect.h cpu_detect_x86.cpp FIRFilter.h RateTransposer.h TDStretch.h PeakFinder.h \
InterpolateCubic.h InterpolateLinear.h InterpolateShannon.h
lib_LTLIBRARIES=libSoundTouch.la
#
libSoundTouch_la_SOURCES=AAFilter.cpp FIRFilter.cpp FIFOSampleBuffer.cpp \
RateTransposer.cpp SoundTouch.cpp TDStretch.cpp cpu_detect_x86.cpp \
BPMDetect.cpp PeakFinder.cpp InterpolateLinear.cpp InterpolateCubic.cpp \
InterpolateShannon.cpp
# Compiler flags
#AM_CXXFLAGS+=
# Compile the files that need MMX and SSE individually.
libSoundTouch_la_LIBADD=libSoundTouchMMX.la libSoundTouchSSE.la
noinst_LTLIBRARIES=libSoundTouchMMX.la libSoundTouchSSE.la
libSoundTouchMMX_la_SOURCES=mmx_optimized.cpp
libSoundTouchSSE_la_SOURCES=sse_optimized.cpp
# We enable optimizations by default.
# If MMX is supported compile with -mmmx.
# Do not assume -msse is also supported.
if HAVE_MMX
libSoundTouchMMX_la_CXXFLAGS = -mmmx $(AM_CXXFLAGS)
else
libSoundTouchMMX_la_CXXFLAGS = $(AM_CXXFLAGS)
endif
# We enable optimizations by default.
# If SSE is supported compile with -msse.
if HAVE_SSE
libSoundTouchSSE_la_CXXFLAGS = -msse $(AM_CXXFLAGS)
else
libSoundTouchSSE_la_CXXFLAGS = $(AM_CXXFLAGS)
endif
# Let the user disable optimizations if he wishes to.
if !X86_OPTIMIZATIONS
libSoundTouchMMX_la_CXXFLAGS = $(AM_CXXFLAGS)
libSoundTouchSSE_la_CXXFLAGS = $(AM_CXXFLAGS)
endif
# Modify the default 0.0.0 to LIB_SONAME.0.0
libSoundTouch_la_LDFLAGS=-version-info @LIB_SONAME@
# other linking flags to add
# noinst_LTLIBRARIES = libSoundTouchOpt.la
# libSoundTouch_la_LIBADD = libSoundTouchOpt.la
# libSoundTouchOpt_la_SOURCES = mmx_optimized.cpp sse_optimized.cpp
# libSoundTouchOpt_la_CXXFLAGS = -O3 -msse -fcheck-new -I../../include

554
source/SoundTouch/PeakFinder.cpp
View File

@ -1,277 +1,277 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Peak detection routine.
///
/// The routine detects highest value on an array of values and calculates the
/// precise peak location as a mass-center of the 'hump' around the peak value.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <math.h>
#include <assert.h>
#include "PeakFinder.h"
using namespace soundtouch;
#define max(x, y) (((x) > (y)) ? (x) : (y))
PeakFinder::PeakFinder()
{
minPos = maxPos = 0;
}
// Finds real 'top' of a peak hump from neighnourhood of the given 'peakpos'.
int PeakFinder::findTop(const float *data, int peakpos) const
{
int i;
int start, end;
float refvalue;
refvalue = data[peakpos];
// seek within ±10 points
start = peakpos - 10;
if (start < minPos) start = minPos;
end = peakpos + 10;
if (end > maxPos) end = maxPos;
for (i = start; i <= end; i ++)
{
if (data[i] > refvalue)
{
peakpos = i;
refvalue = data[i];
}
}
// failure if max value is at edges of seek range => it's not peak, it's at slope.
if ((peakpos == start) || (peakpos == end)) return 0;
return peakpos;
}
// Finds 'ground level' of a peak hump by starting from 'peakpos' and proceeding
// to direction defined by 'direction' until next 'hump' after minimum value will
// begin
int PeakFinder::findGround(const float *data, int peakpos, int direction) const
{
int lowpos;
int pos;
int climb_count;
float refvalue;
float delta;
climb_count = 0;
refvalue = data[peakpos];
lowpos = peakpos;
pos = peakpos;
while ((pos > minPos+1) && (pos < maxPos-1))
{
int prevpos;
prevpos = pos;
pos += direction;
// calculate derivate
delta = data[pos] - data[prevpos];
if (delta <= 0)
{
// going downhill, ok
if (climb_count)
{
climb_count --; // decrease climb count
}
// check if new minimum found
if (data[pos] < refvalue)
{
// new minimum found
lowpos = pos;
refvalue = data[pos];
}
}
else
{
// going uphill, increase climbing counter
climb_count ++;
if (climb_count > 5) break; // we've been climbing too long => it's next uphill => quit
}
}
return lowpos;
}
// Find offset where the value crosses the given level, when starting from 'peakpos' and
// proceeds to direction defined in 'direction'
int PeakFinder::findCrossingLevel(const float *data, float level, int peakpos, int direction) const
{
float peaklevel;
int pos;
peaklevel = data[peakpos];
assert(peaklevel >= level);
pos = peakpos;
while ((pos >= minPos) && (pos + direction < maxPos))
{
if (data[pos + direction] < level) return pos; // crossing found
pos += direction;
}
return -1; // not found
}
// Calculates the center of mass location of 'data' array items between 'firstPos' and 'lastPos'
double PeakFinder::calcMassCenter(const float *data, int firstPos, int lastPos) const
{
int i;
float sum;
float wsum;
sum = 0;
wsum = 0;
for (i = firstPos; i <= lastPos; i ++)
{
sum += (float)i * data[i];
wsum += data[i];
}
if (wsum < 1e-6) return 0;
return sum / wsum;
}
/// get exact center of peak near given position by calculating local mass of center
double PeakFinder::getPeakCenter(const float *data, int peakpos) const
{
float peakLevel; // peak level
int crosspos1, crosspos2; // position where the peak 'hump' crosses cutting level
float cutLevel; // cutting value
float groundLevel; // ground level of the peak
int gp1, gp2; // bottom positions of the peak 'hump'
// find ground positions.
gp1 = findGround(data, peakpos, -1);
gp2 = findGround(data, peakpos, 1);
peakLevel = data[peakpos];
if (gp1 == gp2)
{
// avoid rounding errors when all are equal
assert(gp1 == peakpos);
cutLevel = groundLevel = peakLevel;
} else {
// get average of the ground levels
groundLevel = 0.5f * (data[gp1] + data[gp2]);
// calculate 70%-level of the peak
cutLevel = 0.70f * peakLevel + 0.30f * groundLevel;
}
// find mid-level crossings
crosspos1 = findCrossingLevel(data, cutLevel, peakpos, -1);
crosspos2 = findCrossingLevel(data, cutLevel, peakpos, 1);
if ((crosspos1 < 0) || (crosspos2 < 0)) return 0; // no crossing, no peak..
// calculate mass center of the peak surroundings
return calcMassCenter(data, crosspos1, crosspos2);
}
double PeakFinder::detectPeak(const float *data, int aminPos, int amaxPos)
{
int i;
int peakpos; // position of peak level
double highPeak, peak;
this->minPos = aminPos;
this->maxPos = amaxPos;
// find absolute peak
peakpos = minPos;
peak = data[minPos];
for (i = minPos + 1; i < maxPos; i ++)
{
if (data[i] > peak)
{
peak = data[i];
peakpos = i;
}
}
// Calculate exact location of the highest peak mass center
highPeak = getPeakCenter(data, peakpos);
peak = highPeak;
// Now check if the highest peak were in fact harmonic of the true base beat peak
// - sometimes the highest peak can be Nth harmonic of the true base peak yet
// just a slightly higher than the true base
for (i = 1; i < 3; i ++)
{
double peaktmp, harmonic;
int i1,i2;
harmonic = (double)pow(2.0, i);
peakpos = (int)(highPeak / harmonic + 0.5f);
if (peakpos < minPos) break;
peakpos = findTop(data, peakpos); // seek true local maximum index
if (peakpos == 0) continue; // no local max here
// calculate mass-center of possible harmonic peak
peaktmp = getPeakCenter(data, peakpos);
// accept harmonic peak if
// (a) it is found
// (b) is within ±4% of the expected harmonic interval
// (c) has at least half x-corr value of the max. peak
double diff = harmonic * peaktmp / highPeak;
if ((diff < 0.96) || (diff > 1.04)) continue; // peak too afar from expected
// now compare to highest detected peak
i1 = (int)(highPeak + 0.5);
i2 = (int)(peaktmp + 0.5);
if (data[i2] >= 0.4*data[i1])
{
// The harmonic is at least half as high primary peak,
// thus use the harmonic peak instead
peak = peaktmp;
}
}
return peak;
}
////////////////////////////////////////////////////////////////////////////////
///
/// Peak detection routine.
///
/// The routine detects highest value on an array of values and calculates the
/// precise peak location as a mass-center of the 'hump' around the peak value.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <math.h>
#include <assert.h>
#include "PeakFinder.h"
using namespace soundtouch;
#define max(x, y) (((x) > (y)) ? (x) : (y))
PeakFinder::PeakFinder()
{
minPos = maxPos = 0;
}
// Finds real 'top' of a peak hump from neighnourhood of the given 'peakpos'.
int PeakFinder::findTop(const float *data, int peakpos) const
{
int i;
int start, end;
float refvalue;
refvalue = data[peakpos];
// seek within ±10 points
start = peakpos - 10;
if (start < minPos) start = minPos;
end = peakpos + 10;
if (end > maxPos) end = maxPos;
for (i = start; i <= end; i ++)
{
if (data[i] > refvalue)
{
peakpos = i;
refvalue = data[i];
}
}
// failure if max value is at edges of seek range => it's not peak, it's at slope.
if ((peakpos == start) || (peakpos == end)) return 0;
return peakpos;
}
// Finds 'ground level' of a peak hump by starting from 'peakpos' and proceeding
// to direction defined by 'direction' until next 'hump' after minimum value will
// begin
int PeakFinder::findGround(const float *data, int peakpos, int direction) const
{
int lowpos;
int pos;
int climb_count;
float refvalue;
float delta;
climb_count = 0;
refvalue = data[peakpos];
lowpos = peakpos;
pos = peakpos;
while ((pos > minPos+1) && (pos < maxPos-1))
{
int prevpos;
prevpos = pos;
pos += direction;
// calculate derivate
delta = data[pos] - data[prevpos];
if (delta <= 0)
{
// going downhill, ok
if (climb_count)
{
climb_count --; // decrease climb count
}
// check if new minimum found
if (data[pos] < refvalue)
{
// new minimum found
lowpos = pos;
refvalue = data[pos];
}
}
else
{
// going uphill, increase climbing counter
climb_count ++;
if (climb_count > 5) break; // we've been climbing too long => it's next uphill => quit
}
}
return lowpos;
}
// Find offset where the value crosses the given level, when starting from 'peakpos' and
// proceeds to direction defined in 'direction'
int PeakFinder::findCrossingLevel(const float *data, float level, int peakpos, int direction) const
{
float peaklevel;
int pos;
peaklevel = data[peakpos];
assert(peaklevel >= level);
pos = peakpos;
while ((pos >= minPos) && (pos + direction < maxPos))
{
if (data[pos + direction] < level) return pos; // crossing found
pos += direction;
}
return -1; // not found
}
// Calculates the center of mass location of 'data' array items between 'firstPos' and 'lastPos'
double PeakFinder::calcMassCenter(const float *data, int firstPos, int lastPos) const
{
int i;
float sum;
float wsum;
sum = 0;
wsum = 0;
for (i = firstPos; i <= lastPos; i ++)
{
sum += (float)i * data[i];
wsum += data[i];
}
if (wsum < 1e-6) return 0;
return sum / wsum;
}
/// get exact center of peak near given position by calculating local mass of center
double PeakFinder::getPeakCenter(const float *data, int peakpos) const
{
float peakLevel; // peak level
int crosspos1, crosspos2; // position where the peak 'hump' crosses cutting level
float cutLevel; // cutting value
float groundLevel; // ground level of the peak
int gp1, gp2; // bottom positions of the peak 'hump'
// find ground positions.
gp1 = findGround(data, peakpos, -1);
gp2 = findGround(data, peakpos, 1);
peakLevel = data[peakpos];
if (gp1 == gp2)
{
// avoid rounding errors when all are equal
assert(gp1 == peakpos);
cutLevel = groundLevel = peakLevel;
} else {
// get average of the ground levels
groundLevel = 0.5f * (data[gp1] + data[gp2]);
// calculate 70%-level of the peak
cutLevel = 0.70f * peakLevel + 0.30f * groundLevel;
}
// find mid-level crossings
crosspos1 = findCrossingLevel(data, cutLevel, peakpos, -1);
crosspos2 = findCrossingLevel(data, cutLevel, peakpos, 1);
if ((crosspos1 < 0) || (crosspos2 < 0)) return 0; // no crossing, no peak..
// calculate mass center of the peak surroundings
return calcMassCenter(data, crosspos1, crosspos2);
}
double PeakFinder::detectPeak(const float *data, int aminPos, int amaxPos)
{
int i;
int peakpos; // position of peak level
double highPeak, peak;
this->minPos = aminPos;
this->maxPos = amaxPos;
// find absolute peak
peakpos = minPos;
peak = data[minPos];
for (i = minPos + 1; i < maxPos; i ++)
{
if (data[i] > peak)
{
peak = data[i];
peakpos = i;
}
}
// Calculate exact location of the highest peak mass center
highPeak = getPeakCenter(data, peakpos);
peak = highPeak;
// Now check if the highest peak were in fact harmonic of the true base beat peak
// - sometimes the highest peak can be Nth harmonic of the true base peak yet
// just a slightly higher than the true base
for (i = 1; i < 3; i ++)
{
double peaktmp, harmonic;
int i1,i2;
harmonic = (double)pow(2.0, i);
peakpos = (int)(highPeak / harmonic + 0.5f);
if (peakpos < minPos) break;
peakpos = findTop(data, peakpos); // seek true local maximum index
if (peakpos == 0) continue; // no local max here
// calculate mass-center of possible harmonic peak
peaktmp = getPeakCenter(data, peakpos);
// accept harmonic peak if
// (a) it is found
// (b) is within ±4% of the expected harmonic interval
// (c) has at least half x-corr value of the max. peak
double diff = harmonic * peaktmp / highPeak;
if ((diff < 0.96) || (diff > 1.04)) continue; // peak too afar from expected
// now compare to highest detected peak
i1 = (int)(highPeak + 0.5);
i2 = (int)(peaktmp + 0.5);
if (data[i2] >= 0.4*data[i1])
{
// The harmonic is at least half as high primary peak,
// thus use the harmonic peak instead
peak = peaktmp;
}
}
return peak;
}

180
source/SoundTouch/PeakFinder.h
View File

@ -1,90 +1,90 @@
////////////////////////////////////////////////////////////////////////////////
///
/// The routine detects highest value on an array of values and calculates the
/// precise peak location as a mass-center of the 'hump' around the peak value.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _PeakFinder_H_
#define _PeakFinder_H_
namespace soundtouch
{
class PeakFinder
{
protected:
/// Min, max allowed peak positions within the data vector
int minPos, maxPos;
/// Calculates the mass center between given vector items.
double calcMassCenter(const float *data, ///< Data vector.
int firstPos, ///< Index of first vector item belonging to the peak.
int lastPos ///< Index of last vector item belonging to the peak.
) const;
/// Finds the data vector index where the monotoniously decreasing signal crosses the
/// given level.
int findCrossingLevel(const float *data, ///< Data vector.
float level, ///< Goal crossing level.
int peakpos, ///< Peak position index within the data vector.
int direction /// Direction where to proceed from the peak: 1 = right, -1 = left.
) const;
// Finds real 'top' of a peak hump from neighnourhood of the given 'peakpos'.
int findTop(const float *data, int peakpos) const;
/// Finds the 'ground' level, i.e. smallest level between two neighbouring peaks, to right-
/// or left-hand side of the given peak position.
int findGround(const float *data, /// Data vector.
int peakpos, /// Peak position index within the data vector.
int direction /// Direction where to proceed from the peak: 1 = right, -1 = left.
) const;
/// get exact center of peak near given position by calculating local mass of center
double getPeakCenter(const float *data, int peakpos) const;
public:
/// Constructor.
PeakFinder();
/// Detect exact peak position of the data vector by finding the largest peak 'hump'
/// and calculating the mass-center location of the peak hump.
///
/// \return The location of the largest base harmonic peak hump.
double detectPeak(const float *data, /// Data vector to be analyzed. The data vector has
/// to be at least 'maxPos' items long.
int minPos, ///< Min allowed peak location within the vector data.
int maxPos ///< Max allowed peak location within the vector data.
);
};
}
#endif // _PeakFinder_H_
////////////////////////////////////////////////////////////////////////////////
///
/// The routine detects highest value on an array of values and calculates the
/// precise peak location as a mass-center of the 'hump' around the peak value.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _PeakFinder_H_
#define _PeakFinder_H_
namespace soundtouch
{
class PeakFinder
{
protected:
/// Min, max allowed peak positions within the data vector
int minPos, maxPos;
/// Calculates the mass center between given vector items.
double calcMassCenter(const float *data, ///< Data vector.
int firstPos, ///< Index of first vector item belonging to the peak.
int lastPos ///< Index of last vector item belonging to the peak.
) const;
/// Finds the data vector index where the monotoniously decreasing signal crosses the
/// given level.
int findCrossingLevel(const float *data, ///< Data vector.
float level, ///< Goal crossing level.
int peakpos, ///< Peak position index within the data vector.
int direction /// Direction where to proceed from the peak: 1 = right, -1 = left.
) const;
// Finds real 'top' of a peak hump from neighnourhood of the given 'peakpos'.
int findTop(const float *data, int peakpos) const;
/// Finds the 'ground' level, i.e. smallest level between two neighbouring peaks, to right-
/// or left-hand side of the given peak position.
int findGround(const float *data, /// Data vector.
int peakpos, /// Peak position index within the data vector.
int direction /// Direction where to proceed from the peak: 1 = right, -1 = left.
) const;
/// get exact center of peak near given position by calculating local mass of center
double getPeakCenter(const float *data, int peakpos) const;
public:
/// Constructor.
PeakFinder();
/// Detect exact peak position of the data vector by finding the largest peak 'hump'
/// and calculating the mass-center location of the peak hump.
///
/// \return The location of the largest base harmonic peak hump.
double detectPeak(const float *data, /// Data vector to be analyzed. The data vector has
/// to be at least 'maxPos' items long.
int minPos, ///< Min allowed peak location within the vector data.
int maxPos ///< Max allowed peak location within the vector data.
);
};
}
#endif // _PeakFinder_H_

626
source/SoundTouch/RateTransposer.cpp
View File

@ -1,313 +1,313 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Sample rate transposer. Changes sample rate by using linear interpolation
/// together with anti-alias filtering (first order interpolation with anti-
/// alias filtering should be quite adequate for this application)
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <memory.h>
#include <assert.h>
#include <stdlib.h>
#include <stdio.h>
#include "RateTransposer.h"
#include "InterpolateLinear.h"
#include "InterpolateCubic.h"
#include "InterpolateShannon.h"
#include "AAFilter.h"
using namespace soundtouch;
// Define default interpolation algorithm here
TransposerBase::ALGORITHM TransposerBase::algorithm = TransposerBase::CUBIC;
// Constructor
RateTransposer::RateTransposer() : FIFOProcessor(&outputBuffer)
{
bUseAAFilter =
#ifndef SOUNDTOUCH_PREVENT_CLICK_AT_RATE_CROSSOVER
true;
#else
// Disable Anti-alias filter if desirable to avoid click at rate change zero value crossover
false;
#endif
// Instantiates the anti-alias filter
pAAFilter = new AAFilter(64);
pTransposer = TransposerBase::newInstance();
clear();
}
RateTransposer::~RateTransposer()
{
delete pAAFilter;
delete pTransposer;
}
/// Enables/disables the anti-alias filter. Zero to disable, nonzero to enable
void RateTransposer::enableAAFilter(bool newMode)
{
#ifndef SOUNDTOUCH_PREVENT_CLICK_AT_RATE_CROSSOVER
// Disable Anti-alias filter if desirable to avoid click at rate change zero value crossover
bUseAAFilter = newMode;
clear();
#endif
}
/// Returns nonzero if anti-alias filter is enabled.
bool RateTransposer::isAAFilterEnabled() const
{
return bUseAAFilter;
}
AAFilter *RateTransposer::getAAFilter()
{
return pAAFilter;
}
// Sets new target iRate. Normal iRate = 1.0, smaller values represent slower
// iRate, larger faster iRates.
void RateTransposer::setRate(double newRate)
{
double fCutoff;
pTransposer->setRate(newRate);
// design a new anti-alias filter
if (newRate > 1.0)
{
fCutoff = 0.5 / newRate;
}
else
{
fCutoff = 0.5 * newRate;
}
pAAFilter->setCutoffFreq(fCutoff);
}
// Adds 'nSamples' pcs of samples from the 'samples' memory position into
// the input of the object.
void RateTransposer::putSamples(const SAMPLETYPE *samples, uint nSamples)
{
processSamples(samples, nSamples);
}
// Transposes sample rate by applying anti-alias filter to prevent folding.
// Returns amount of samples returned in the "dest" buffer.
// The maximum amount of samples that can be returned at a time is set by
// the 'set_returnBuffer_size' function.
void RateTransposer::processSamples(const SAMPLETYPE *src, uint nSamples)
{
if (nSamples == 0) return;
// Store samples to input buffer
inputBuffer.putSamples(src, nSamples);
// If anti-alias filter is turned off, simply transpose without applying
// the filter
if (bUseAAFilter == false)
{
(void)pTransposer->transpose(outputBuffer, inputBuffer);
return;
}
assert(pAAFilter);
// Transpose with anti-alias filter
if (pTransposer->rate < 1.0f)
{
// If the parameter 'Rate' value is smaller than 1, first transpose
// the samples and then apply the anti-alias filter to remove aliasing.
// Transpose the samples, store the result to end of "midBuffer"
pTransposer->transpose(midBuffer, inputBuffer);
// Apply the anti-alias filter for transposed samples in midBuffer
pAAFilter->evaluate(outputBuffer, midBuffer);
}
else
{
// If the parameter 'Rate' value is larger than 1, first apply the
// anti-alias filter to remove high frequencies (prevent them from folding
// over the lover frequencies), then transpose.
// Apply the anti-alias filter for samples in inputBuffer
pAAFilter->evaluate(midBuffer, inputBuffer);
// Transpose the AA-filtered samples in "midBuffer"
pTransposer->transpose(outputBuffer, midBuffer);
}
}
// Sets the number of channels, 1 = mono, 2 = stereo
void RateTransposer::setChannels(int nChannels)
{
if (!verifyNumberOfChannels(nChannels) ||
(pTransposer->numChannels == nChannels)) return;
pTransposer->setChannels(nChannels);
inputBuffer.setChannels(nChannels);
midBuffer.setChannels(nChannels);
outputBuffer.setChannels(nChannels);
}
// Clears all the samples in the object
void RateTransposer::clear()
{
outputBuffer.clear();
midBuffer.clear();
inputBuffer.clear();
pTransposer->resetRegisters();
// prefill buffer to avoid losing first samples at beginning of stream
int prefill = getLatency();
inputBuffer.addSilent(prefill);
}
// Returns nonzero if there aren't any samples available for outputting.
int RateTransposer::isEmpty() const
{
int res;
res = FIFOProcessor::isEmpty();
if (res == 0) return 0;
return inputBuffer.isEmpty();
}
/// Return approximate initial input-output latency
int RateTransposer::getLatency() const
{
return pTransposer->getLatency() +
((bUseAAFilter) ? (pAAFilter->getLength() / 2) : 0);
}
//////////////////////////////////////////////////////////////////////////////
//
// TransposerBase - Base class for interpolation
//
// static function to set interpolation algorithm
void TransposerBase::setAlgorithm(TransposerBase::ALGORITHM a)
{
TransposerBase::algorithm = a;
}
// Transposes the sample rate of the given samples using linear interpolation.
// Returns the number of samples returned in the "dest" buffer
int TransposerBase::transpose(FIFOSampleBuffer &dest, FIFOSampleBuffer &src)
{
int numSrcSamples = src.numSamples();
int sizeDemand = (int)((double)numSrcSamples / rate) + 8;
int numOutput;
SAMPLETYPE *psrc = src.ptrBegin();
SAMPLETYPE *pdest = dest.ptrEnd(sizeDemand);
#ifndef USE_MULTICH_ALWAYS
if (numChannels == 1)
{
numOutput = transposeMono(pdest, psrc, numSrcSamples);
}
else if (numChannels == 2)
{
numOutput = transposeStereo(pdest, psrc, numSrcSamples);
}
else
#endif // USE_MULTICH_ALWAYS
{
assert(numChannels > 0);
numOutput = transposeMulti(pdest, psrc, numSrcSamples);
}
dest.putSamples(numOutput);
src.receiveSamples(numSrcSamples);
return numOutput;
}
TransposerBase::TransposerBase()
{
numChannels = 0;
rate = 1.0f;
}
TransposerBase::~TransposerBase()
{
}
void TransposerBase::setChannels(int channels)
{
numChannels = channels;
resetRegisters();
}
void TransposerBase::setRate(double newRate)
{
rate = newRate;
}
// static factory function
TransposerBase *TransposerBase::newInstance()
{
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
// Notice: For integer arithmetic support only linear algorithm (due to simplest calculus)
return ::new InterpolateLinearInteger;
#else
switch (algorithm)
{
case LINEAR:
return new InterpolateLinearFloat;
case CUBIC:
return new InterpolateCubic;
case SHANNON:
return new InterpolateShannon;
default:
assert(false);
return nullptr;
}
#endif
}
////////////////////////////////////////////////////////////////////////////////
///
/// Sample rate transposer. Changes sample rate by using linear interpolation
/// together with anti-alias filtering (first order interpolation with anti-
/// alias filtering should be quite adequate for this application)
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include <memory.h>
#include <assert.h>
#include <stdlib.h>
#include <stdio.h>
#include "RateTransposer.h"
#include "InterpolateLinear.h"
#include "InterpolateCubic.h"
#include "InterpolateShannon.h"
#include "AAFilter.h"
using namespace soundtouch;
// Define default interpolation algorithm here
TransposerBase::ALGORITHM TransposerBase::algorithm = TransposerBase::CUBIC;
// Constructor
RateTransposer::RateTransposer() : FIFOProcessor(&outputBuffer)
{
bUseAAFilter =
#ifndef SOUNDTOUCH_PREVENT_CLICK_AT_RATE_CROSSOVER
true;
#else
// Disable Anti-alias filter if desirable to avoid click at rate change zero value crossover
false;
#endif
// Instantiates the anti-alias filter
pAAFilter = new AAFilter(64);
pTransposer = TransposerBase::newInstance();
clear();
}
RateTransposer::~RateTransposer()
{
delete pAAFilter;
delete pTransposer;
}
/// Enables/disables the anti-alias filter. Zero to disable, nonzero to enable
void RateTransposer::enableAAFilter(bool newMode)
{
#ifndef SOUNDTOUCH_PREVENT_CLICK_AT_RATE_CROSSOVER
// Disable Anti-alias filter if desirable to avoid click at rate change zero value crossover
bUseAAFilter = newMode;
clear();
#endif
}
/// Returns nonzero if anti-alias filter is enabled.
bool RateTransposer::isAAFilterEnabled() const
{
return bUseAAFilter;
}
AAFilter *RateTransposer::getAAFilter()
{
return pAAFilter;
}
// Sets new target iRate. Normal iRate = 1.0, smaller values represent slower
// iRate, larger faster iRates.
void RateTransposer::setRate(double newRate)
{
double fCutoff;
pTransposer->setRate(newRate);
// design a new anti-alias filter
if (newRate > 1.0)
{
fCutoff = 0.5 / newRate;
}
else
{
fCutoff = 0.5 * newRate;
}
pAAFilter->setCutoffFreq(fCutoff);
}
// Adds 'nSamples' pcs of samples from the 'samples' memory position into
// the input of the object.
void RateTransposer::putSamples(const SAMPLETYPE *samples, uint nSamples)
{
processSamples(samples, nSamples);
}
// Transposes sample rate by applying anti-alias filter to prevent folding.
// Returns amount of samples returned in the "dest" buffer.
// The maximum amount of samples that can be returned at a time is set by
// the 'set_returnBuffer_size' function.
void RateTransposer::processSamples(const SAMPLETYPE *src, uint nSamples)
{
if (nSamples == 0) return;
// Store samples to input buffer
inputBuffer.putSamples(src, nSamples);
// If anti-alias filter is turned off, simply transpose without applying
// the filter
if (bUseAAFilter == false)
{
(void)pTransposer->transpose(outputBuffer, inputBuffer);
return;
}
assert(pAAFilter);
// Transpose with anti-alias filter
if (pTransposer->rate < 1.0f)
{
// If the parameter 'Rate' value is smaller than 1, first transpose
// the samples and then apply the anti-alias filter to remove aliasing.
// Transpose the samples, store the result to end of "midBuffer"
pTransposer->transpose(midBuffer, inputBuffer);
// Apply the anti-alias filter for transposed samples in midBuffer
pAAFilter->evaluate(outputBuffer, midBuffer);
}
else
{
// If the parameter 'Rate' value is larger than 1, first apply the
// anti-alias filter to remove high frequencies (prevent them from folding
// over the lover frequencies), then transpose.
// Apply the anti-alias filter for samples in inputBuffer
pAAFilter->evaluate(midBuffer, inputBuffer);
// Transpose the AA-filtered samples in "midBuffer"
pTransposer->transpose(outputBuffer, midBuffer);
}
}
// Sets the number of channels, 1 = mono, 2 = stereo
void RateTransposer::setChannels(int nChannels)
{
if (!verifyNumberOfChannels(nChannels) ||
(pTransposer->numChannels == nChannels)) return;
pTransposer->setChannels(nChannels);
inputBuffer.setChannels(nChannels);
midBuffer.setChannels(nChannels);
outputBuffer.setChannels(nChannels);
}
// Clears all the samples in the object
void RateTransposer::clear()
{
outputBuffer.clear();
midBuffer.clear();
inputBuffer.clear();
pTransposer->resetRegisters();
// prefill buffer to avoid losing first samples at beginning of stream
int prefill = getLatency();
inputBuffer.addSilent(prefill);
}
// Returns nonzero if there aren't any samples available for outputting.
int RateTransposer::isEmpty() const
{
int res;
res = FIFOProcessor::isEmpty();
if (res == 0) return 0;
return inputBuffer.isEmpty();
}
/// Return approximate initial input-output latency
int RateTransposer::getLatency() const
{
return pTransposer->getLatency() +
((bUseAAFilter) ? (pAAFilter->getLength() / 2) : 0);
}
//////////////////////////////////////////////////////////////////////////////
//
// TransposerBase - Base class for interpolation
//
// static function to set interpolation algorithm
void TransposerBase::setAlgorithm(TransposerBase::ALGORITHM a)
{
TransposerBase::algorithm = a;
}
// Transposes the sample rate of the given samples using linear interpolation.
// Returns the number of samples returned in the "dest" buffer
int TransposerBase::transpose(FIFOSampleBuffer &dest, FIFOSampleBuffer &src)
{
int numSrcSamples = src.numSamples();
int sizeDemand = (int)((double)numSrcSamples / rate) + 8;
int numOutput;
SAMPLETYPE *psrc = src.ptrBegin();
SAMPLETYPE *pdest = dest.ptrEnd(sizeDemand);
#ifndef USE_MULTICH_ALWAYS
if (numChannels == 1)
{
numOutput = transposeMono(pdest, psrc, numSrcSamples);
}
else if (numChannels == 2)
{
numOutput = transposeStereo(pdest, psrc, numSrcSamples);
}
else
#endif // USE_MULTICH_ALWAYS
{
assert(numChannels > 0);
numOutput = transposeMulti(pdest, psrc, numSrcSamples);
}
dest.putSamples(numOutput);
src.receiveSamples(numSrcSamples);
return numOutput;
}
TransposerBase::TransposerBase()
{
numChannels = 0;
rate = 1.0f;
}
TransposerBase::~TransposerBase()
{
}
void TransposerBase::setChannels(int channels)
{
numChannels = channels;
resetRegisters();
}
void TransposerBase::setRate(double newRate)
{
rate = newRate;
}
// static factory function
TransposerBase *TransposerBase::newInstance()
{
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
// Notice: For integer arithmetic support only linear algorithm (due to simplest calculus)
return ::new InterpolateLinearInteger;
#else
switch (algorithm)
{
case LINEAR:
return new InterpolateLinearFloat;
case CUBIC:
return new InterpolateCubic;
case SHANNON:
return new InterpolateShannon;
default:
assert(false);
return nullptr;
}
#endif
}

328
source/SoundTouch/RateTransposer.h
View File

@ -1,164 +1,164 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Sample rate transposer. Changes sample rate by using linear interpolation
/// together with anti-alias filtering (first order interpolation with anti-
/// alias filtering should be quite adequate for this application).
///
/// Use either of the derived classes of 'RateTransposerInteger' or
/// 'RateTransposerFloat' for corresponding integer/floating point tranposing
/// algorithm implementation.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef RateTransposer_H
#define RateTransposer_H
#include <stddef.h>
#include "AAFilter.h"
#include "FIFOSamplePipe.h"
#include "FIFOSampleBuffer.h"
#include "STTypes.h"
namespace soundtouch
{
/// Abstract base class for transposer implementations (linear, advanced vs integer, float etc)
class TransposerBase
{
public:
enum ALGORITHM {
LINEAR = 0,
CUBIC,
SHANNON
};
protected:
virtual int transposeMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) = 0;
virtual int transposeStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) = 0;
virtual int transposeMulti(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) = 0;
static ALGORITHM algorithm;
public:
double rate;
int numChannels;
TransposerBase();
virtual ~TransposerBase();
virtual int transpose(FIFOSampleBuffer &dest, FIFOSampleBuffer &src);
virtual void setRate(double newRate);
virtual void setChannels(int channels);
virtual int getLatency() const = 0;
virtual void resetRegisters() = 0;
// static factory function
static TransposerBase *newInstance();
// static function to set interpolation algorithm
static void setAlgorithm(ALGORITHM a);
};
/// A common linear samplerate transposer class.
///
class RateTransposer : public FIFOProcessor
{
protected:
/// Anti-alias filter object
AAFilter *pAAFilter;
TransposerBase *pTransposer;
/// Buffer for collecting samples to feed the anti-alias filter between
/// two batches
FIFOSampleBuffer inputBuffer;
/// Buffer for keeping samples between transposing & anti-alias filter
FIFOSampleBuffer midBuffer;
/// Output sample buffer
FIFOSampleBuffer outputBuffer;
bool bUseAAFilter;
/// Transposes sample rate by applying anti-alias filter to prevent folding.
/// Returns amount of samples returned in the "dest" buffer.
/// The maximum amount of samples that can be returned at a time is set by
/// the 'set_returnBuffer_size' function.
void processSamples(const SAMPLETYPE *src,
uint numSamples);
public:
RateTransposer();
virtual ~RateTransposer() override;
/// Returns the output buffer object
FIFOSamplePipe *getOutput() { return &outputBuffer; };
/// Return anti-alias filter object
AAFilter *getAAFilter();
/// Enables/disables the anti-alias filter. Zero to disable, nonzero to enable
void enableAAFilter(bool newMode);
/// Returns nonzero if anti-alias filter is enabled.
bool isAAFilterEnabled() const;
/// Sets new target rate. Normal rate = 1.0, smaller values represent slower
/// rate, larger faster rates.
virtual void setRate(double newRate);
/// Sets the number of channels, 1 = mono, 2 = stereo
void setChannels(int channels);
/// Adds 'numSamples' pcs of samples from the 'samples' memory position into
/// the input of the object.
void putSamples(const SAMPLETYPE *samples, uint numSamples) override;
/// Clears all the samples in the object
void clear() override;
/// Returns nonzero if there aren't any samples available for outputting.
int isEmpty() const override;
/// Return approximate initial input-output latency
int getLatency() const;
};
}
#endif
////////////////////////////////////////////////////////////////////////////////
///
/// Sample rate transposer. Changes sample rate by using linear interpolation
/// together with anti-alias filtering (first order interpolation with anti-
/// alias filtering should be quite adequate for this application).
///
/// Use either of the derived classes of 'RateTransposerInteger' or
/// 'RateTransposerFloat' for corresponding integer/floating point tranposing
/// algorithm implementation.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef RateTransposer_H
#define RateTransposer_H
#include <stddef.h>
#include "AAFilter.h"
#include "FIFOSamplePipe.h"
#include "FIFOSampleBuffer.h"
#include "STTypes.h"
namespace soundtouch
{
/// Abstract base class for transposer implementations (linear, advanced vs integer, float etc)
class TransposerBase
{
public:
enum ALGORITHM {
LINEAR = 0,
CUBIC,
SHANNON
};
protected:
virtual int transposeMono(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) = 0;
virtual int transposeStereo(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) = 0;
virtual int transposeMulti(SAMPLETYPE *dest,
const SAMPLETYPE *src,
int &srcSamples) = 0;
static ALGORITHM algorithm;
public:
double rate;
int numChannels;
TransposerBase();
virtual ~TransposerBase();
virtual int transpose(FIFOSampleBuffer &dest, FIFOSampleBuffer &src);
virtual void setRate(double newRate);
virtual void setChannels(int channels);
virtual int getLatency() const = 0;
virtual void resetRegisters() = 0;
// static factory function
static TransposerBase *newInstance();
// static function to set interpolation algorithm
static void setAlgorithm(ALGORITHM a);
};
/// A common linear samplerate transposer class.
///
class RateTransposer : public FIFOProcessor
{
protected:
/// Anti-alias filter object
AAFilter *pAAFilter;
TransposerBase *pTransposer;
/// Buffer for collecting samples to feed the anti-alias filter between
/// two batches
FIFOSampleBuffer inputBuffer;
/// Buffer for keeping samples between transposing & anti-alias filter
FIFOSampleBuffer midBuffer;
/// Output sample buffer
FIFOSampleBuffer outputBuffer;
bool bUseAAFilter;
/// Transposes sample rate by applying anti-alias filter to prevent folding.
/// Returns amount of samples returned in the "dest" buffer.
/// The maximum amount of samples that can be returned at a time is set by
/// the 'set_returnBuffer_size' function.
void processSamples(const SAMPLETYPE *src,
uint numSamples);
public:
RateTransposer();
virtual ~RateTransposer() override;
/// Returns the output buffer object
FIFOSamplePipe *getOutput() { return &outputBuffer; };
/// Return anti-alias filter object
AAFilter *getAAFilter();
/// Enables/disables the anti-alias filter. Zero to disable, nonzero to enable
void enableAAFilter(bool newMode);
/// Returns nonzero if anti-alias filter is enabled.
bool isAAFilterEnabled() const;
/// Sets new target rate. Normal rate = 1.0, smaller values represent slower
/// rate, larger faster rates.
virtual void setRate(double newRate);
/// Sets the number of channels, 1 = mono, 2 = stereo
void setChannels(int channels);
/// Adds 'numSamples' pcs of samples from the 'samples' memory position into
/// the input of the object.
void putSamples(const SAMPLETYPE *samples, uint numSamples) override;
/// Clears all the samples in the object
void clear() override;
/// Returns nonzero if there aren't any samples available for outputting.
int isEmpty() const override;
/// Return approximate initial input-output latency
int getLatency() const;
};
}
#endif

1076
source/SoundTouch/SoundTouch.cpp
View File

File diff suppressed because it is too large Load Diff

2176
source/SoundTouch/TDStretch.cpp
View File

File diff suppressed because it is too large Load Diff

558
source/SoundTouch/TDStretch.h
View File

@ -1,279 +1,279 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Sampled sound tempo changer/time stretch algorithm. Changes the sound tempo
/// while maintaining the original pitch by using a time domain WSOLA-like method
/// with several performance-increasing tweaks.
///
/// Note : MMX/SSE optimized functions reside in separate, platform-specific files
/// 'mmx_optimized.cpp' and 'sse_optimized.cpp'
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef TDStretch_H
#define TDStretch_H
#include <stddef.h>
#include "STTypes.h"
#include "RateTransposer.h"
#include "FIFOSamplePipe.h"
namespace soundtouch
{
/// Default values for sound processing parameters:
/// Notice that the default parameters are tuned for contemporary popular music
/// processing. For speech processing applications these parameters suit better:
/// #define DEFAULT_SEQUENCE_MS 40
/// #define DEFAULT_SEEKWINDOW_MS 15
/// #define DEFAULT_OVERLAP_MS 8
///
/// Default length of a single processing sequence, in milliseconds. This determines to how
/// long sequences the original sound is chopped in the time-stretch algorithm.
///
/// The larger this value is, the lesser sequences are used in processing. In principle
/// a bigger value sounds better when slowing down tempo, but worse when increasing tempo
/// and vice versa.
///
/// Increasing this value reduces computational burden & vice versa.
//#define DEFAULT_SEQUENCE_MS 40
#define DEFAULT_SEQUENCE_MS USE_AUTO_SEQUENCE_LEN
/// Giving this value for the sequence length sets automatic parameter value
/// according to tempo setting (recommended)
#define USE_AUTO_SEQUENCE_LEN 0
/// Seeking window default length in milliseconds for algorithm that finds the best possible
/// overlapping location. This determines from how wide window the algorithm may look for an
/// optimal joining location when mixing the sound sequences back together.
///
/// The bigger this window setting is, the higher the possibility to find a better mixing
/// position will become, but at the same time large values may cause a "drifting" artifact
/// because consequent sequences will be taken at more uneven intervals.
///
/// If there's a disturbing artifact that sounds as if a constant frequency was drifting
/// around, try reducing this setting.
///
/// Increasing this value increases computational burden & vice versa.
//#define DEFAULT_SEEKWINDOW_MS 15
#define DEFAULT_SEEKWINDOW_MS USE_AUTO_SEEKWINDOW_LEN
/// Giving this value for the seek window length sets automatic parameter value
/// according to tempo setting (recommended)
#define USE_AUTO_SEEKWINDOW_LEN 0
/// Overlap length in milliseconds. When the chopped sound sequences are mixed back together,
/// to form a continuous sound stream, this parameter defines over how long period the two
/// consecutive sequences are let to overlap each other.
///
/// This shouldn't be that critical parameter. If you reduce the DEFAULT_SEQUENCE_MS setting
/// by a large amount, you might wish to try a smaller value on this.
///
/// Increasing this value increases computational burden & vice versa.
#define DEFAULT_OVERLAP_MS 8
/// Class that does the time-stretch (tempo change) effect for the processed
/// sound.
class TDStretch : public FIFOProcessor
{
protected:
int channels;
int sampleReq;
int overlapLength;
int seekLength;
int seekWindowLength;
int overlapDividerBitsNorm;
int overlapDividerBitsPure;
int slopingDivider;
int sampleRate;
int sequenceMs;
int seekWindowMs;
int overlapMs;
unsigned long maxnorm;
float maxnormf;
double tempo;
double nominalSkip;
double skipFract;
bool bQuickSeek;
bool bAutoSeqSetting;
bool bAutoSeekSetting;
bool isBeginning;
SAMPLETYPE *pMidBuffer;
SAMPLETYPE *pMidBufferUnaligned;
FIFOSampleBuffer outputBuffer;
FIFOSampleBuffer inputBuffer;
void acceptNewOverlapLength(int newOverlapLength);
virtual void clearCrossCorrState();
void calculateOverlapLength(int overlapMs);
virtual double calcCrossCorr(const SAMPLETYPE *mixingPos, const SAMPLETYPE *compare, double &norm);
virtual double calcCrossCorrAccumulate(const SAMPLETYPE *mixingPos, const SAMPLETYPE *compare, double &norm);
virtual int seekBestOverlapPositionFull(const SAMPLETYPE *refPos);
virtual int seekBestOverlapPositionQuick(const SAMPLETYPE *refPos);
virtual int seekBestOverlapPosition(const SAMPLETYPE *refPos);
virtual void overlapStereo(SAMPLETYPE *output, const SAMPLETYPE *input) const;
virtual void overlapMono(SAMPLETYPE *output, const SAMPLETYPE *input) const;
virtual void overlapMulti(SAMPLETYPE *output, const SAMPLETYPE *input) const;
void clearMidBuffer();
void overlap(SAMPLETYPE *output, const SAMPLETYPE *input, uint ovlPos) const;
void calcSeqParameters();
void adaptNormalizer();
/// Changes the tempo of the given sound samples.
/// Returns amount of samples returned in the "output" buffer.
/// The maximum amount of samples that can be returned at a time is set by
/// the 'set_returnBuffer_size' function.
void processSamples();
public:
TDStretch();
virtual ~TDStretch() override;
/// Operator 'new' is overloaded so that it automatically creates a suitable instance
/// depending on if we've a MMX/SSE/etc-capable CPU available or not.
static void *operator new(size_t s);
/// Use this function instead of "new" operator to create a new instance of this class.
/// This function automatically chooses a correct feature set depending on if the CPU
/// supports MMX/SSE/etc extensions.
static TDStretch *newInstance();
/// Returns the output buffer object
FIFOSamplePipe *getOutput() { return &outputBuffer; };
/// Returns the input buffer object
FIFOSamplePipe *getInput() { return &inputBuffer; };
/// Sets new target tempo. Normal tempo = 'SCALE', smaller values represent slower
/// tempo, larger faster tempo.
void setTempo(double newTempo);
/// Returns nonzero if there aren't any samples available for outputting.
virtual void clear() override;
/// Clears the input buffer
void clearInput();
/// Sets the number of channels, 1 = mono, 2 = stereo
void setChannels(int numChannels);
/// Enables/disables the quick position seeking algorithm. Zero to disable,
/// nonzero to enable
void enableQuickSeek(bool enable);
/// Returns nonzero if the quick seeking algorithm is enabled.
bool isQuickSeekEnabled() const;
/// Sets routine control parameters. These control are certain time constants
/// defining how the sound is stretched to the desired duration.
//
/// 'sampleRate' = sample rate of the sound
/// 'sequenceMS' = one processing sequence length in milliseconds
/// 'seekwindowMS' = seeking window length for scanning the best overlapping
/// position
/// 'overlapMS' = overlapping length
void setParameters(int sampleRate, ///< Samplerate of sound being processed (Hz)
int sequenceMS = -1, ///< Single processing sequence length (ms)
int seekwindowMS = -1, ///< Offset seeking window length (ms)
int overlapMS = -1 ///< Sequence overlapping length (ms)
);
/// Get routine control parameters, see setParameters() function.
/// Any of the parameters to this function can be nullptr, in such case corresponding parameter
/// value isn't returned.
void getParameters(int *pSampleRate, int *pSequenceMs, int *pSeekWindowMs, int *pOverlapMs) const;
/// Adds 'numsamples' pcs of samples from the 'samples' memory position into
/// the input of the object.
virtual void putSamples(
const SAMPLETYPE *samples, ///< Input sample data
uint numSamples ///< Number of samples in 'samples' so that one sample
///< contains both channels if stereo
) override;
/// return nominal input sample requirement for triggering a processing batch
int getInputSampleReq() const
{
return (int)(nominalSkip + 0.5);
}
/// return nominal output sample amount when running a processing batch
int getOutputBatchSize() const
{
return seekWindowLength - overlapLength;
}
/// return approximate initial input-output latency
int getLatency() const
{
return sampleReq;
}
};
// Implementation-specific class declarations:
#ifdef SOUNDTOUCH_ALLOW_MMX
/// Class that implements MMX optimized routines for 16bit integer samples type.
class TDStretchMMX : public TDStretch
{
protected:
double calcCrossCorr(const short *mixingPos, const short *compare, double &norm) override;
double calcCrossCorrAccumulate(const short *mixingPos, const short *compare, double &norm) override;
virtual void overlapStereo(short *output, const short *input) const override;
virtual void clearCrossCorrState() override;
};
#endif /// SOUNDTOUCH_ALLOW_MMX
#ifdef SOUNDTOUCH_ALLOW_SSE
/// Class that implements SSE optimized routines for floating point samples type.
class TDStretchSSE : public TDStretch
{
protected:
double calcCrossCorr(const float *mixingPos, const float *compare, double &norm) override;
double calcCrossCorrAccumulate(const float *mixingPos, const float *compare, double &norm) override;
};
#endif /// SOUNDTOUCH_ALLOW_SSE
}
#endif /// TDStretch_H
////////////////////////////////////////////////////////////////////////////////
///
/// Sampled sound tempo changer/time stretch algorithm. Changes the sound tempo
/// while maintaining the original pitch by using a time domain WSOLA-like method
/// with several performance-increasing tweaks.
///
/// Note : MMX/SSE optimized functions reside in separate, platform-specific files
/// 'mmx_optimized.cpp' and 'sse_optimized.cpp'
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef TDStretch_H
#define TDStretch_H
#include <stddef.h>
#include "STTypes.h"
#include "RateTransposer.h"
#include "FIFOSamplePipe.h"
namespace soundtouch
{
/// Default values for sound processing parameters:
/// Notice that the default parameters are tuned for contemporary popular music
/// processing. For speech processing applications these parameters suit better:
/// #define DEFAULT_SEQUENCE_MS 40
/// #define DEFAULT_SEEKWINDOW_MS 15
/// #define DEFAULT_OVERLAP_MS 8
///
/// Default length of a single processing sequence, in milliseconds. This determines to how
/// long sequences the original sound is chopped in the time-stretch algorithm.
///
/// The larger this value is, the lesser sequences are used in processing. In principle
/// a bigger value sounds better when slowing down tempo, but worse when increasing tempo
/// and vice versa.
///
/// Increasing this value reduces computational burden & vice versa.
//#define DEFAULT_SEQUENCE_MS 40
#define DEFAULT_SEQUENCE_MS USE_AUTO_SEQUENCE_LEN
/// Giving this value for the sequence length sets automatic parameter value
/// according to tempo setting (recommended)
#define USE_AUTO_SEQUENCE_LEN 0
/// Seeking window default length in milliseconds for algorithm that finds the best possible
/// overlapping location. This determines from how wide window the algorithm may look for an
/// optimal joining location when mixing the sound sequences back together.
///
/// The bigger this window setting is, the higher the possibility to find a better mixing
/// position will become, but at the same time large values may cause a "drifting" artifact
/// because consequent sequences will be taken at more uneven intervals.
///
/// If there's a disturbing artifact that sounds as if a constant frequency was drifting
/// around, try reducing this setting.
///
/// Increasing this value increases computational burden & vice versa.
//#define DEFAULT_SEEKWINDOW_MS 15
#define DEFAULT_SEEKWINDOW_MS USE_AUTO_SEEKWINDOW_LEN
/// Giving this value for the seek window length sets automatic parameter value
/// according to tempo setting (recommended)
#define USE_AUTO_SEEKWINDOW_LEN 0
/// Overlap length in milliseconds. When the chopped sound sequences are mixed back together,
/// to form a continuous sound stream, this parameter defines over how long period the two
/// consecutive sequences are let to overlap each other.
///
/// This shouldn't be that critical parameter. If you reduce the DEFAULT_SEQUENCE_MS setting
/// by a large amount, you might wish to try a smaller value on this.
///
/// Increasing this value increases computational burden & vice versa.
#define DEFAULT_OVERLAP_MS 8
/// Class that does the time-stretch (tempo change) effect for the processed
/// sound.
class TDStretch : public FIFOProcessor
{
protected:
int channels;
int sampleReq;
int overlapLength;
int seekLength;
int seekWindowLength;
int overlapDividerBitsNorm;
int overlapDividerBitsPure;
int slopingDivider;
int sampleRate;
int sequenceMs;
int seekWindowMs;
int overlapMs;
unsigned long maxnorm;
float maxnormf;
double tempo;
double nominalSkip;
double skipFract;
bool bQuickSeek;
bool bAutoSeqSetting;
bool bAutoSeekSetting;
bool isBeginning;
SAMPLETYPE *pMidBuffer;
SAMPLETYPE *pMidBufferUnaligned;
FIFOSampleBuffer outputBuffer;
FIFOSampleBuffer inputBuffer;
void acceptNewOverlapLength(int newOverlapLength);
virtual void clearCrossCorrState();
void calculateOverlapLength(int overlapMs);
virtual double calcCrossCorr(const SAMPLETYPE *mixingPos, const SAMPLETYPE *compare, double &norm);
virtual double calcCrossCorrAccumulate(const SAMPLETYPE *mixingPos, const SAMPLETYPE *compare, double &norm);
virtual int seekBestOverlapPositionFull(const SAMPLETYPE *refPos);
virtual int seekBestOverlapPositionQuick(const SAMPLETYPE *refPos);
virtual int seekBestOverlapPosition(const SAMPLETYPE *refPos);
virtual void overlapStereo(SAMPLETYPE *output, const SAMPLETYPE *input) const;
virtual void overlapMono(SAMPLETYPE *output, const SAMPLETYPE *input) const;
virtual void overlapMulti(SAMPLETYPE *output, const SAMPLETYPE *input) const;
void clearMidBuffer();
void overlap(SAMPLETYPE *output, const SAMPLETYPE *input, uint ovlPos) const;
void calcSeqParameters();
void adaptNormalizer();
/// Changes the tempo of the given sound samples.
/// Returns amount of samples returned in the "output" buffer.
/// The maximum amount of samples that can be returned at a time is set by
/// the 'set_returnBuffer_size' function.
void processSamples();
public:
TDStretch();
virtual ~TDStretch() override;
/// Operator 'new' is overloaded so that it automatically creates a suitable instance
/// depending on if we've a MMX/SSE/etc-capable CPU available or not.
static void *operator new(size_t s);
/// Use this function instead of "new" operator to create a new instance of this class.
/// This function automatically chooses a correct feature set depending on if the CPU
/// supports MMX/SSE/etc extensions.
static TDStretch *newInstance();
/// Returns the output buffer object
FIFOSamplePipe *getOutput() { return &outputBuffer; };
/// Returns the input buffer object
FIFOSamplePipe *getInput() { return &inputBuffer; };
/// Sets new target tempo. Normal tempo = 'SCALE', smaller values represent slower
/// tempo, larger faster tempo.
void setTempo(double newTempo);
/// Returns nonzero if there aren't any samples available for outputting.
virtual void clear() override;
/// Clears the input buffer
void clearInput();
/// Sets the number of channels, 1 = mono, 2 = stereo
void setChannels(int numChannels);
/// Enables/disables the quick position seeking algorithm. Zero to disable,
/// nonzero to enable
void enableQuickSeek(bool enable);
/// Returns nonzero if the quick seeking algorithm is enabled.
bool isQuickSeekEnabled() const;
/// Sets routine control parameters. These control are certain time constants
/// defining how the sound is stretched to the desired duration.
//
/// 'sampleRate' = sample rate of the sound
/// 'sequenceMS' = one processing sequence length in milliseconds
/// 'seekwindowMS' = seeking window length for scanning the best overlapping
/// position
/// 'overlapMS' = overlapping length
void setParameters(int sampleRate, ///< Samplerate of sound being processed (Hz)
int sequenceMS = -1, ///< Single processing sequence length (ms)
int seekwindowMS = -1, ///< Offset seeking window length (ms)
int overlapMS = -1 ///< Sequence overlapping length (ms)
);
/// Get routine control parameters, see setParameters() function.
/// Any of the parameters to this function can be nullptr, in such case corresponding parameter
/// value isn't returned.
void getParameters(int *pSampleRate, int *pSequenceMs, int *pSeekWindowMs, int *pOverlapMs) const;
/// Adds 'numsamples' pcs of samples from the 'samples' memory position into
/// the input of the object.
virtual void putSamples(
const SAMPLETYPE *samples, ///< Input sample data
uint numSamples ///< Number of samples in 'samples' so that one sample
///< contains both channels if stereo
) override;
/// return nominal input sample requirement for triggering a processing batch
int getInputSampleReq() const
{
return (int)(nominalSkip + 0.5);
}
/// return nominal output sample amount when running a processing batch
int getOutputBatchSize() const
{
return seekWindowLength - overlapLength;
}
/// return approximate initial input-output latency
int getLatency() const
{
return sampleReq;
}
};
// Implementation-specific class declarations:
#ifdef SOUNDTOUCH_ALLOW_MMX
/// Class that implements MMX optimized routines for 16bit integer samples type.
class TDStretchMMX : public TDStretch
{
protected:
double calcCrossCorr(const short *mixingPos, const short *compare, double &norm) override;
double calcCrossCorrAccumulate(const short *mixingPos, const short *compare, double &norm) override;
virtual void overlapStereo(short *output, const short *input) const override;
virtual void clearCrossCorrState() override;
};
#endif /// SOUNDTOUCH_ALLOW_MMX
#ifdef SOUNDTOUCH_ALLOW_SSE
/// Class that implements SSE optimized routines for floating point samples type.
class TDStretchSSE : public TDStretch
{
protected:
double calcCrossCorr(const float *mixingPos, const float *compare, double &norm) override;
double calcCrossCorrAccumulate(const float *mixingPos, const float *compare, double &norm) override;
};
#endif /// SOUNDTOUCH_ALLOW_SSE
}
#endif /// TDStretch_H

110
source/SoundTouch/cpu_detect.h
View File

@ -1,55 +1,55 @@
////////////////////////////////////////////////////////////////////////////////
///
/// A header file for detecting the Intel MMX instructions set extension.
///
/// Please see 'mmx_win.cpp', 'mmx_cpp.cpp' and 'mmx_non_x86.cpp' for the
/// routine implementations for x86 Windows, x86 gnu version and non-x86
/// platforms, respectively.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _CPU_DETECT_H_
#define _CPU_DETECT_H_
#include "STTypes.h"
#define SUPPORT_MMX 0x0001
#define SUPPORT_3DNOW 0x0002
#define SUPPORT_ALTIVEC 0x0004
#define SUPPORT_SSE 0x0008
#define SUPPORT_SSE2 0x0010
/// Checks which instruction set extensions are supported by the CPU.
///
/// \return A bitmask of supported extensions, see SUPPORT_... defines.
uint detectCPUextensions(void);
/// Disables given set of instruction extensions. See SUPPORT_... defines.
void disableExtensions(uint wDisableMask);
#endif // _CPU_DETECT_H_
////////////////////////////////////////////////////////////////////////////////
///
/// A header file for detecting the Intel MMX instructions set extension.
///
/// Please see 'mmx_win.cpp', 'mmx_cpp.cpp' and 'mmx_non_x86.cpp' for the
/// routine implementations for x86 Windows, x86 gnu version and non-x86
/// platforms, respectively.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _CPU_DETECT_H_
#define _CPU_DETECT_H_
#include "STTypes.h"
#define SUPPORT_MMX 0x0001
#define SUPPORT_3DNOW 0x0002
#define SUPPORT_ALTIVEC 0x0004
#define SUPPORT_SSE 0x0008
#define SUPPORT_SSE2 0x0010
/// Checks which instruction set extensions are supported by the CPU.
///
/// \return A bitmask of supported extensions, see SUPPORT_... defines.
uint detectCPUextensions(void);
/// Disables given set of instruction extensions. See SUPPORT_... defines.
void disableExtensions(uint wDisableMask);
#endif // _CPU_DETECT_H_

260
source/SoundTouch/cpu_detect_x86.cpp
View File

@ -1,130 +1,130 @@
////////////////////////////////////////////////////////////////////////////////
///
/// Generic version of the x86 CPU extension detection routine.
///
/// This file is for GNU & other non-Windows compilers, see 'cpu_detect_x86_win.cpp'
/// for the Microsoft compiler version.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include "cpu_detect.h"
#include "STTypes.h"
#if defined(SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS)
#if defined(__GNUC__) && defined(__i386__)
// gcc
#include "cpuid.h"
#elif defined(_M_IX86)
// windows non-gcc
#include <intrin.h>
#endif
#define bit_MMX (1 << 23)
#define bit_SSE (1 << 25)
#define bit_SSE2 (1 << 26)
#endif
//////////////////////////////////////////////////////////////////////////////
//
// processor instructions extension detection routines
//
//////////////////////////////////////////////////////////////////////////////
// Flag variable indicating whick ISA extensions are disabled (for debugging)
static uint _dwDisabledISA = 0x00; // 0xffffffff; //<- use this to disable all extensions
// Disables given set of instruction extensions. See SUPPORT_... defines.
void disableExtensions(uint dwDisableMask)
{
_dwDisabledISA = dwDisableMask;
}
/// Checks which instruction set extensions are supported by the CPU.
uint detectCPUextensions(void)
{
/// If building for a 64bit system (no Itanium) and the user wants optimizations.
/// Return the OR of SUPPORT_{MMX,SSE,SSE2}. 11001 or 0x19.
/// Keep the _dwDisabledISA test (2 more operations, could be eliminated).
#if ((defined(__GNUC__) && defined(__x86_64__)) \
|| defined(_M_X64)) \
&& defined(SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS)
return 0x19 & ~_dwDisabledISA;
/// If building for a 32bit system and the user wants optimizations.
/// Keep the _dwDisabledISA test (2 more operations, could be eliminated).
#elif ((defined(__GNUC__) && defined(__i386__)) \
|| defined(_M_IX86)) \
&& defined(SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS)
if (_dwDisabledISA == 0xffffffff) return 0;
uint res = 0;
#if defined(__GNUC__)
// GCC version of cpuid. Requires GCC 4.3.0 or later for __cpuid intrinsic support.
uint eax, ebx, ecx, edx; // unsigned int is the standard type. uint is defined by the compiler and not guaranteed to be portable.
// Check if no cpuid support.
if (!__get_cpuid (1, &eax, &ebx, &ecx, &edx)) return 0; // always disable extensions.
if (edx & bit_MMX) res = res | SUPPORT_MMX;
if (edx & bit_SSE) res = res | SUPPORT_SSE;
if (edx & bit_SSE2) res = res | SUPPORT_SSE2;
#else
// Window / VS version of cpuid. Notice that Visual Studio 2005 or later required
// for __cpuid intrinsic support.
int reg[4] = {-1};
// Check if no cpuid support.
__cpuid(reg,0);
if ((unsigned int)reg[0] == 0) return 0; // always disable extensions.
__cpuid(reg,1);
if ((unsigned int)reg[3] & bit_MMX) res = res | SUPPORT_MMX;
if ((unsigned int)reg[3] & bit_SSE) res = res | SUPPORT_SSE;
if ((unsigned int)reg[3] & bit_SSE2) res = res | SUPPORT_SSE2;
#endif
return res & ~_dwDisabledISA;
#else
/// One of these is true:
/// 1) We don't want optimizations.
/// 2) Using an unsupported compiler.
/// 3) Running on a non-x86 platform.
return 0;
#endif
}
////////////////////////////////////////////////////////////////////////////////
///
/// Generic version of the x86 CPU extension detection routine.
///
/// This file is for GNU & other non-Windows compilers, see 'cpu_detect_x86_win.cpp'
/// for the Microsoft compiler version.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include "cpu_detect.h"
#include "STTypes.h"
#if defined(SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS)
#if defined(__GNUC__) && defined(__i386__)
// gcc
#include "cpuid.h"
#elif defined(_M_IX86)
// windows non-gcc
#include <intrin.h>
#endif
#define bit_MMX (1 << 23)
#define bit_SSE (1 << 25)
#define bit_SSE2 (1 << 26)
#endif
//////////////////////////////////////////////////////////////////////////////
//
// processor instructions extension detection routines
//
//////////////////////////////////////////////////////////////////////////////
// Flag variable indicating whick ISA extensions are disabled (for debugging)
static uint _dwDisabledISA = 0x00; // 0xffffffff; //<- use this to disable all extensions
// Disables given set of instruction extensions. See SUPPORT_... defines.
void disableExtensions(uint dwDisableMask)
{
_dwDisabledISA = dwDisableMask;
}
/// Checks which instruction set extensions are supported by the CPU.
uint detectCPUextensions(void)
{
/// If building for a 64bit system (no Itanium) and the user wants optimizations.
/// Return the OR of SUPPORT_{MMX,SSE,SSE2}. 11001 or 0x19.
/// Keep the _dwDisabledISA test (2 more operations, could be eliminated).
#if ((defined(__GNUC__) && defined(__x86_64__)) \
|| defined(_M_X64)) \
&& defined(SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS)
return 0x19 & ~_dwDisabledISA;
/// If building for a 32bit system and the user wants optimizations.
/// Keep the _dwDisabledISA test (2 more operations, could be eliminated).
#elif ((defined(__GNUC__) && defined(__i386__)) \
|| defined(_M_IX86)) \
&& defined(SOUNDTOUCH_ALLOW_X86_OPTIMIZATIONS)
if (_dwDisabledISA == 0xffffffff) return 0;
uint res = 0;
#if defined(__GNUC__)
// GCC version of cpuid. Requires GCC 4.3.0 or later for __cpuid intrinsic support.
uint eax, ebx, ecx, edx; // unsigned int is the standard type. uint is defined by the compiler and not guaranteed to be portable.
// Check if no cpuid support.
if (!__get_cpuid (1, &eax, &ebx, &ecx, &edx)) return 0; // always disable extensions.
if (edx & bit_MMX) res = res | SUPPORT_MMX;
if (edx & bit_SSE) res = res | SUPPORT_SSE;
if (edx & bit_SSE2) res = res | SUPPORT_SSE2;
#else
// Window / VS version of cpuid. Notice that Visual Studio 2005 or later required
// for __cpuid intrinsic support.
int reg[4] = {-1};
// Check if no cpuid support.
__cpuid(reg,0);
if ((unsigned int)reg[0] == 0) return 0; // always disable extensions.
__cpuid(reg,1);
if ((unsigned int)reg[3] & bit_MMX) res = res | SUPPORT_MMX;
if ((unsigned int)reg[3] & bit_SSE) res = res | SUPPORT_SSE;
if ((unsigned int)reg[3] & bit_SSE2) res = res | SUPPORT_SSE2;
#endif
return res & ~_dwDisabledISA;
#else
/// One of these is true:
/// 1) We don't want optimizations.
/// 2) Using an unsupported compiler.
/// 3) Running on a non-x86 platform.
return 0;
#endif
}

792
source/SoundTouch/mmx_optimized.cpp
View File

@ -1,396 +1,396 @@
////////////////////////////////////////////////////////////////////////////////
///
/// MMX optimized routines. All MMX optimized functions have been gathered into
/// this single source code file, regardless to their class or original source
/// code file, in order to ease porting the library to other compiler and
/// processor platforms.
///
/// The MMX-optimizations are programmed using MMX compiler intrinsics that
/// are supported both by Microsoft Visual C++ and GCC compilers, so this file
/// should compile with both toolsets.
///
/// NOTICE: If using Visual Studio 6.0, you'll need to install the "Visual C++
/// 6.0 processor pack" update to support compiler intrinsic syntax. The update
/// is available for download at Microsoft Developers Network, see here:
/// http://msdn.microsoft.com/en-us/vstudio/aa718349.aspx
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include "STTypes.h"
#ifdef SOUNDTOUCH_ALLOW_MMX
// MMX routines available only with integer sample type
using namespace soundtouch;
//////////////////////////////////////////////////////////////////////////////
//
// implementation of MMX optimized functions of class 'TDStretchMMX'
//
//////////////////////////////////////////////////////////////////////////////
#include "TDStretch.h"
#include <mmintrin.h>
#include <limits.h>
#include <math.h>
// Calculates cross correlation of two buffers
double TDStretchMMX::calcCrossCorr(const short *pV1, const short *pV2, double &dnorm)
{
const __m64 *pVec1, *pVec2;
__m64 shifter;
__m64 accu, normaccu;
long corr, norm;
int i;
pVec1 = (__m64*)pV1;
pVec2 = (__m64*)pV2;
shifter = _m_from_int(overlapDividerBitsNorm);
normaccu = accu = _mm_setzero_si64();
// Process 4 parallel sets of 2 * stereo samples or 4 * mono samples
// during each round for improved CPU-level parallellization.
for (i = 0; i < channels * overlapLength / 16; i ++)
{
__m64 temp, temp2;
// dictionary of instructions:
// _m_pmaddwd : 4*16bit multiply-add, resulting two 32bits = [a0*b0+a1*b1 ; a2*b2+a3*b3]
// _mm_add_pi32 : 2*32bit add
// _m_psrad : 32bit right-shift
temp = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[0], pVec2[0]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[1], pVec2[1]), shifter));
temp2 = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[0], pVec1[0]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[1], pVec1[1]), shifter));
accu = _mm_add_pi32(accu, temp);
normaccu = _mm_add_pi32(normaccu, temp2);
temp = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[2], pVec2[2]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[3], pVec2[3]), shifter));
temp2 = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[2], pVec1[2]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[3], pVec1[3]), shifter));
accu = _mm_add_pi32(accu, temp);
normaccu = _mm_add_pi32(normaccu, temp2);
pVec1 += 4;
pVec2 += 4;
}
// copy hi-dword of mm0 to lo-dword of mm1, then sum mmo+mm1
// and finally store the result into the variable "corr"
accu = _mm_add_pi32(accu, _mm_srli_si64(accu, 32));
corr = _m_to_int(accu);
normaccu = _mm_add_pi32(normaccu, _mm_srli_si64(normaccu, 32));
norm = _m_to_int(normaccu);
// Clear MMS state
_m_empty();
if (norm > (long)maxnorm)
{
// modify 'maxnorm' inside critical section to avoid multi-access conflict if in OpenMP mode
#pragma omp critical
if (norm > (long)maxnorm)
{
maxnorm = norm;
}
}
// Normalize result by dividing by sqrt(norm) - this step is easiest
// done using floating point operation
dnorm = (double)norm;
return (double)corr / sqrt(dnorm < 1e-9 ? 1.0 : dnorm);
// Note: Warning about the missing EMMS instruction is harmless
// as it'll be called elsewhere.
}
/// Update cross-correlation by accumulating "norm" coefficient by previously calculated value
double TDStretchMMX::calcCrossCorrAccumulate(const short *pV1, const short *pV2, double &dnorm)
{
const __m64 *pVec1, *pVec2;
__m64 shifter;
__m64 accu;
long corr, lnorm;
int i;
// cancel first normalizer tap from previous round
lnorm = 0;
for (i = 1; i <= channels; i ++)
{
lnorm -= (pV1[-i] * pV1[-i]) >> overlapDividerBitsNorm;
}
pVec1 = (__m64*)pV1;
pVec2 = (__m64*)pV2;
shifter = _m_from_int(overlapDividerBitsNorm);
accu = _mm_setzero_si64();
// Process 4 parallel sets of 2 * stereo samples or 4 * mono samples
// during each round for improved CPU-level parallellization.
for (i = 0; i < channels * overlapLength / 16; i ++)
{
__m64 temp;
// dictionary of instructions:
// _m_pmaddwd : 4*16bit multiply-add, resulting two 32bits = [a0*b0+a1*b1 ; a2*b2+a3*b3]
// _mm_add_pi32 : 2*32bit add
// _m_psrad : 32bit right-shift
temp = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[0], pVec2[0]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[1], pVec2[1]), shifter));
accu = _mm_add_pi32(accu, temp);
temp = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[2], pVec2[2]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[3], pVec2[3]), shifter));
accu = _mm_add_pi32(accu, temp);
pVec1 += 4;
pVec2 += 4;
}
// copy hi-dword of mm0 to lo-dword of mm1, then sum mmo+mm1
// and finally store the result into the variable "corr"
accu = _mm_add_pi32(accu, _mm_srli_si64(accu, 32));
corr = _m_to_int(accu);
// Clear MMS state
_m_empty();
// update normalizer with last samples of this round
pV1 = (short *)pVec1;
for (int j = 1; j <= channels; j ++)
{
lnorm += (pV1[-j] * pV1[-j]) >> overlapDividerBitsNorm;
}
dnorm += (double)lnorm;
if (lnorm > (long)maxnorm)
{
maxnorm = lnorm;
}
// Normalize result by dividing by sqrt(norm) - this step is easiest
// done using floating point operation
return (double)corr / sqrt((dnorm < 1e-9) ? 1.0 : dnorm);
}
void TDStretchMMX::clearCrossCorrState()
{
// Clear MMS state
_m_empty();
//_asm EMMS;
}
// MMX-optimized version of the function overlapStereo
void TDStretchMMX::overlapStereo(short *output, const short *input) const
{
const __m64 *pVinput, *pVMidBuf;
__m64 *pVdest;
__m64 mix1, mix2, adder, shifter;
int i;
pVinput = (const __m64*)input;
pVMidBuf = (const __m64*)pMidBuffer;
pVdest = (__m64*)output;
// mix1 = mixer values for 1st stereo sample
// mix1 = mixer values for 2nd stereo sample
// adder = adder for updating mixer values after each round
mix1 = _mm_set_pi16(0, overlapLength, 0, overlapLength);
adder = _mm_set_pi16(1, -1, 1, -1);
mix2 = _mm_add_pi16(mix1, adder);
adder = _mm_add_pi16(adder, adder);
// Overlaplength-division by shifter. "+1" is to account for "-1" deduced in
// overlapDividerBits calculation earlier.
shifter = _m_from_int(overlapDividerBitsPure + 1);
for (i = 0; i < overlapLength / 4; i ++)
{
__m64 temp1, temp2;
// load & shuffle data so that input & mixbuffer data samples are paired
temp1 = _mm_unpacklo_pi16(pVMidBuf[0], pVinput[0]); // = i0l m0l i0r m0r
temp2 = _mm_unpackhi_pi16(pVMidBuf[0], pVinput[0]); // = i1l m1l i1r m1r
// temp = (temp .* mix) >> shifter
temp1 = _mm_sra_pi32(_mm_madd_pi16(temp1, mix1), shifter);
temp2 = _mm_sra_pi32(_mm_madd_pi16(temp2, mix2), shifter);
pVdest[0] = _mm_packs_pi32(temp1, temp2); // pack 2*2*32bit => 4*16bit
// update mix += adder
mix1 = _mm_add_pi16(mix1, adder);
mix2 = _mm_add_pi16(mix2, adder);
// --- second round begins here ---
// load & shuffle data so that input & mixbuffer data samples are paired
temp1 = _mm_unpacklo_pi16(pVMidBuf[1], pVinput[1]); // = i2l m2l i2r m2r
temp2 = _mm_unpackhi_pi16(pVMidBuf[1], pVinput[1]); // = i3l m3l i3r m3r
// temp = (temp .* mix) >> shifter
temp1 = _mm_sra_pi32(_mm_madd_pi16(temp1, mix1), shifter);
temp2 = _mm_sra_pi32(_mm_madd_pi16(temp2, mix2), shifter);
pVdest[1] = _mm_packs_pi32(temp1, temp2); // pack 2*2*32bit => 4*16bit
// update mix += adder
mix1 = _mm_add_pi16(mix1, adder);
mix2 = _mm_add_pi16(mix2, adder);
pVinput += 2;
pVMidBuf += 2;
pVdest += 2;
}
_m_empty(); // clear MMS state
}
//////////////////////////////////////////////////////////////////////////////
//
// implementation of MMX optimized functions of class 'FIRFilter'
//
//////////////////////////////////////////////////////////////////////////////
#include "FIRFilter.h"
FIRFilterMMX::FIRFilterMMX() : FIRFilter()
{
filterCoeffsAlign = nullptr;
filterCoeffsUnalign = nullptr;
}
FIRFilterMMX::~FIRFilterMMX()
{
delete[] filterCoeffsUnalign;
}
// (overloaded) Calculates filter coefficients for MMX routine
void FIRFilterMMX::setCoefficients(const short *coeffs, uint newLength, uint uResultDivFactor)
{
uint i;
FIRFilter::setCoefficients(coeffs, newLength, uResultDivFactor);
// Ensure that filter coeffs array is aligned to 16-byte boundary
delete[] filterCoeffsUnalign;
filterCoeffsUnalign = new short[2 * newLength + 8];
filterCoeffsAlign = (short *)SOUNDTOUCH_ALIGN_POINTER_16(filterCoeffsUnalign);
// rearrange the filter coefficients for mmx routines
for (i = 0;i < length; i += 4)
{
filterCoeffsAlign[2 * i + 0] = coeffs[i + 0];
filterCoeffsAlign[2 * i + 1] = coeffs[i + 2];
filterCoeffsAlign[2 * i + 2] = coeffs[i + 0];
filterCoeffsAlign[2 * i + 3] = coeffs[i + 2];
filterCoeffsAlign[2 * i + 4] = coeffs[i + 1];
filterCoeffsAlign[2 * i + 5] = coeffs[i + 3];
filterCoeffsAlign[2 * i + 6] = coeffs[i + 1];
filterCoeffsAlign[2 * i + 7] = coeffs[i + 3];
}
}
// mmx-optimized version of the filter routine for stereo sound
uint FIRFilterMMX::evaluateFilterStereo(short *dest, const short *src, uint numSamples) const
{
// Create stack copies of the needed member variables for asm routines :
uint i, j;
__m64 *pVdest = (__m64*)dest;
if (length < 2) return 0;
for (i = 0; i < (numSamples - length) / 2; i ++)
{
__m64 accu1;
__m64 accu2;
const __m64 *pVsrc = (const __m64*)src;
const __m64 *pVfilter = (const __m64*)filterCoeffsAlign;
accu1 = accu2 = _mm_setzero_si64();
for (j = 0; j < lengthDiv8 * 2; j ++)
{
__m64 temp1, temp2;
temp1 = _mm_unpacklo_pi16(pVsrc[0], pVsrc[1]); // = l2 l0 r2 r0
temp2 = _mm_unpackhi_pi16(pVsrc[0], pVsrc[1]); // = l3 l1 r3 r1
accu1 = _mm_add_pi32(accu1, _mm_madd_pi16(temp1, pVfilter[0])); // += l2*f2+l0*f0 r2*f2+r0*f0
accu1 = _mm_add_pi32(accu1, _mm_madd_pi16(temp2, pVfilter[1])); // += l3*f3+l1*f1 r3*f3+r1*f1
temp1 = _mm_unpacklo_pi16(pVsrc[1], pVsrc[2]); // = l4 l2 r4 r2
accu2 = _mm_add_pi32(accu2, _mm_madd_pi16(temp2, pVfilter[0])); // += l3*f2+l1*f0 r3*f2+r1*f0
accu2 = _mm_add_pi32(accu2, _mm_madd_pi16(temp1, pVfilter[1])); // += l4*f3+l2*f1 r4*f3+r2*f1
// accu1 += l2*f2+l0*f0 r2*f2+r0*f0
// += l3*f3+l1*f1 r3*f3+r1*f1
// accu2 += l3*f2+l1*f0 r3*f2+r1*f0
// l4*f3+l2*f1 r4*f3+r2*f1
pVfilter += 2;
pVsrc += 2;
}
// accu >>= resultDivFactor
accu1 = _mm_srai_pi32(accu1, resultDivFactor);
accu2 = _mm_srai_pi32(accu2, resultDivFactor);
// pack 2*2*32bits => 4*16 bits
pVdest[0] = _mm_packs_pi32(accu1, accu2);
src += 4;
pVdest ++;
}
_m_empty(); // clear emms state
return (numSamples & 0xfffffffe) - length;
}
#else
// workaround to not complain about empty module
bool _dontcomplain_mmx_empty;
#endif // SOUNDTOUCH_ALLOW_MMX
////////////////////////////////////////////////////////////////////////////////
///
/// MMX optimized routines. All MMX optimized functions have been gathered into
/// this single source code file, regardless to their class or original source
/// code file, in order to ease porting the library to other compiler and
/// processor platforms.
///
/// The MMX-optimizations are programmed using MMX compiler intrinsics that
/// are supported both by Microsoft Visual C++ and GCC compilers, so this file
/// should compile with both toolsets.
///
/// NOTICE: If using Visual Studio 6.0, you'll need to install the "Visual C++
/// 6.0 processor pack" update to support compiler intrinsic syntax. The update
/// is available for download at Microsoft Developers Network, see here:
/// http://msdn.microsoft.com/en-us/vstudio/aa718349.aspx
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include "STTypes.h"
#ifdef SOUNDTOUCH_ALLOW_MMX
// MMX routines available only with integer sample type
using namespace soundtouch;
//////////////////////////////////////////////////////////////////////////////
//
// implementation of MMX optimized functions of class 'TDStretchMMX'
//
//////////////////////////////////////////////////////////////////////////////
#include "TDStretch.h"
#include <mmintrin.h>
#include <limits.h>
#include <math.h>
// Calculates cross correlation of two buffers
double TDStretchMMX::calcCrossCorr(const short *pV1, const short *pV2, double &dnorm)
{
const __m64 *pVec1, *pVec2;
__m64 shifter;
__m64 accu, normaccu;
long corr, norm;
int i;
pVec1 = (__m64*)pV1;
pVec2 = (__m64*)pV2;
shifter = _m_from_int(overlapDividerBitsNorm);
normaccu = accu = _mm_setzero_si64();
// Process 4 parallel sets of 2 * stereo samples or 4 * mono samples
// during each round for improved CPU-level parallellization.
for (i = 0; i < channels * overlapLength / 16; i ++)
{
__m64 temp, temp2;
// dictionary of instructions:
// _m_pmaddwd : 4*16bit multiply-add, resulting two 32bits = [a0*b0+a1*b1 ; a2*b2+a3*b3]
// _mm_add_pi32 : 2*32bit add
// _m_psrad : 32bit right-shift
temp = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[0], pVec2[0]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[1], pVec2[1]), shifter));
temp2 = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[0], pVec1[0]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[1], pVec1[1]), shifter));
accu = _mm_add_pi32(accu, temp);
normaccu = _mm_add_pi32(normaccu, temp2);
temp = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[2], pVec2[2]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[3], pVec2[3]), shifter));
temp2 = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[2], pVec1[2]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[3], pVec1[3]), shifter));
accu = _mm_add_pi32(accu, temp);
normaccu = _mm_add_pi32(normaccu, temp2);
pVec1 += 4;
pVec2 += 4;
}
// copy hi-dword of mm0 to lo-dword of mm1, then sum mmo+mm1
// and finally store the result into the variable "corr"
accu = _mm_add_pi32(accu, _mm_srli_si64(accu, 32));
corr = _m_to_int(accu);
normaccu = _mm_add_pi32(normaccu, _mm_srli_si64(normaccu, 32));
norm = _m_to_int(normaccu);
// Clear MMS state
_m_empty();
if (norm > (long)maxnorm)
{
// modify 'maxnorm' inside critical section to avoid multi-access conflict if in OpenMP mode
#pragma omp critical
if (norm > (long)maxnorm)
{
maxnorm = norm;
}
}
// Normalize result by dividing by sqrt(norm) - this step is easiest
// done using floating point operation
dnorm = (double)norm;
return (double)corr / sqrt(dnorm < 1e-9 ? 1.0 : dnorm);
// Note: Warning about the missing EMMS instruction is harmless
// as it'll be called elsewhere.
}
/// Update cross-correlation by accumulating "norm" coefficient by previously calculated value
double TDStretchMMX::calcCrossCorrAccumulate(const short *pV1, const short *pV2, double &dnorm)
{
const __m64 *pVec1, *pVec2;
__m64 shifter;
__m64 accu;
long corr, lnorm;
int i;
// cancel first normalizer tap from previous round
lnorm = 0;
for (i = 1; i <= channels; i ++)
{
lnorm -= (pV1[-i] * pV1[-i]) >> overlapDividerBitsNorm;
}
pVec1 = (__m64*)pV1;
pVec2 = (__m64*)pV2;
shifter = _m_from_int(overlapDividerBitsNorm);
accu = _mm_setzero_si64();
// Process 4 parallel sets of 2 * stereo samples or 4 * mono samples
// during each round for improved CPU-level parallellization.
for (i = 0; i < channels * overlapLength / 16; i ++)
{
__m64 temp;
// dictionary of instructions:
// _m_pmaddwd : 4*16bit multiply-add, resulting two 32bits = [a0*b0+a1*b1 ; a2*b2+a3*b3]
// _mm_add_pi32 : 2*32bit add
// _m_psrad : 32bit right-shift
temp = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[0], pVec2[0]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[1], pVec2[1]), shifter));
accu = _mm_add_pi32(accu, temp);
temp = _mm_add_pi32(_mm_sra_pi32(_mm_madd_pi16(pVec1[2], pVec2[2]), shifter),
_mm_sra_pi32(_mm_madd_pi16(pVec1[3], pVec2[3]), shifter));
accu = _mm_add_pi32(accu, temp);
pVec1 += 4;
pVec2 += 4;
}
// copy hi-dword of mm0 to lo-dword of mm1, then sum mmo+mm1
// and finally store the result into the variable "corr"
accu = _mm_add_pi32(accu, _mm_srli_si64(accu, 32));
corr = _m_to_int(accu);
// Clear MMS state
_m_empty();
// update normalizer with last samples of this round
pV1 = (short *)pVec1;
for (int j = 1; j <= channels; j ++)
{
lnorm += (pV1[-j] * pV1[-j]) >> overlapDividerBitsNorm;
}
dnorm += (double)lnorm;
if (lnorm > (long)maxnorm)
{
maxnorm = lnorm;
}
// Normalize result by dividing by sqrt(norm) - this step is easiest
// done using floating point operation
return (double)corr / sqrt((dnorm < 1e-9) ? 1.0 : dnorm);
}
void TDStretchMMX::clearCrossCorrState()
{
// Clear MMS state
_m_empty();
//_asm EMMS;
}
// MMX-optimized version of the function overlapStereo
void TDStretchMMX::overlapStereo(short *output, const short *input) const
{
const __m64 *pVinput, *pVMidBuf;
__m64 *pVdest;
__m64 mix1, mix2, adder, shifter;
int i;
pVinput = (const __m64*)input;
pVMidBuf = (const __m64*)pMidBuffer;
pVdest = (__m64*)output;
// mix1 = mixer values for 1st stereo sample
// mix1 = mixer values for 2nd stereo sample
// adder = adder for updating mixer values after each round
mix1 = _mm_set_pi16(0, overlapLength, 0, overlapLength);
adder = _mm_set_pi16(1, -1, 1, -1);
mix2 = _mm_add_pi16(mix1, adder);
adder = _mm_add_pi16(adder, adder);
// Overlaplength-division by shifter. "+1" is to account for "-1" deduced in
// overlapDividerBits calculation earlier.
shifter = _m_from_int(overlapDividerBitsPure + 1);
for (i = 0; i < overlapLength / 4; i ++)
{
__m64 temp1, temp2;
// load & shuffle data so that input & mixbuffer data samples are paired
temp1 = _mm_unpacklo_pi16(pVMidBuf[0], pVinput[0]); // = i0l m0l i0r m0r
temp2 = _mm_unpackhi_pi16(pVMidBuf[0], pVinput[0]); // = i1l m1l i1r m1r
// temp = (temp .* mix) >> shifter
temp1 = _mm_sra_pi32(_mm_madd_pi16(temp1, mix1), shifter);
temp2 = _mm_sra_pi32(_mm_madd_pi16(temp2, mix2), shifter);
pVdest[0] = _mm_packs_pi32(temp1, temp2); // pack 2*2*32bit => 4*16bit
// update mix += adder
mix1 = _mm_add_pi16(mix1, adder);
mix2 = _mm_add_pi16(mix2, adder);
// --- second round begins here ---
// load & shuffle data so that input & mixbuffer data samples are paired
temp1 = _mm_unpacklo_pi16(pVMidBuf[1], pVinput[1]); // = i2l m2l i2r m2r
temp2 = _mm_unpackhi_pi16(pVMidBuf[1], pVinput[1]); // = i3l m3l i3r m3r
// temp = (temp .* mix) >> shifter
temp1 = _mm_sra_pi32(_mm_madd_pi16(temp1, mix1), shifter);
temp2 = _mm_sra_pi32(_mm_madd_pi16(temp2, mix2), shifter);
pVdest[1] = _mm_packs_pi32(temp1, temp2); // pack 2*2*32bit => 4*16bit
// update mix += adder
mix1 = _mm_add_pi16(mix1, adder);
mix2 = _mm_add_pi16(mix2, adder);
pVinput += 2;
pVMidBuf += 2;
pVdest += 2;
}
_m_empty(); // clear MMS state
}
//////////////////////////////////////////////////////////////////////////////
//
// implementation of MMX optimized functions of class 'FIRFilter'
//
//////////////////////////////////////////////////////////////////////////////
#include "FIRFilter.h"
FIRFilterMMX::FIRFilterMMX() : FIRFilter()
{
filterCoeffsAlign = nullptr;
filterCoeffsUnalign = nullptr;
}
FIRFilterMMX::~FIRFilterMMX()
{
delete[] filterCoeffsUnalign;
}
// (overloaded) Calculates filter coefficients for MMX routine
void FIRFilterMMX::setCoefficients(const short *coeffs, uint newLength, uint uResultDivFactor)
{
uint i;
FIRFilter::setCoefficients(coeffs, newLength, uResultDivFactor);
// Ensure that filter coeffs array is aligned to 16-byte boundary
delete[] filterCoeffsUnalign;
filterCoeffsUnalign = new short[2 * newLength + 8];
filterCoeffsAlign = (short *)SOUNDTOUCH_ALIGN_POINTER_16(filterCoeffsUnalign);
// rearrange the filter coefficients for mmx routines
for (i = 0;i < length; i += 4)
{
filterCoeffsAlign[2 * i + 0] = coeffs[i + 0];
filterCoeffsAlign[2 * i + 1] = coeffs[i + 2];
filterCoeffsAlign[2 * i + 2] = coeffs[i + 0];
filterCoeffsAlign[2 * i + 3] = coeffs[i + 2];
filterCoeffsAlign[2 * i + 4] = coeffs[i + 1];
filterCoeffsAlign[2 * i + 5] = coeffs[i + 3];
filterCoeffsAlign[2 * i + 6] = coeffs[i + 1];
filterCoeffsAlign[2 * i + 7] = coeffs[i + 3];
}
}
// mmx-optimized version of the filter routine for stereo sound
uint FIRFilterMMX::evaluateFilterStereo(short *dest, const short *src, uint numSamples) const
{
// Create stack copies of the needed member variables for asm routines :
uint i, j;
__m64 *pVdest = (__m64*)dest;
if (length < 2) return 0;
for (i = 0; i < (numSamples - length) / 2; i ++)
{
__m64 accu1;
__m64 accu2;
const __m64 *pVsrc = (const __m64*)src;
const __m64 *pVfilter = (const __m64*)filterCoeffsAlign;
accu1 = accu2 = _mm_setzero_si64();
for (j = 0; j < lengthDiv8 * 2; j ++)
{
__m64 temp1, temp2;
temp1 = _mm_unpacklo_pi16(pVsrc[0], pVsrc[1]); // = l2 l0 r2 r0
temp2 = _mm_unpackhi_pi16(pVsrc[0], pVsrc[1]); // = l3 l1 r3 r1
accu1 = _mm_add_pi32(accu1, _mm_madd_pi16(temp1, pVfilter[0])); // += l2*f2+l0*f0 r2*f2+r0*f0
accu1 = _mm_add_pi32(accu1, _mm_madd_pi16(temp2, pVfilter[1])); // += l3*f3+l1*f1 r3*f3+r1*f1
temp1 = _mm_unpacklo_pi16(pVsrc[1], pVsrc[2]); // = l4 l2 r4 r2
accu2 = _mm_add_pi32(accu2, _mm_madd_pi16(temp2, pVfilter[0])); // += l3*f2+l1*f0 r3*f2+r1*f0
accu2 = _mm_add_pi32(accu2, _mm_madd_pi16(temp1, pVfilter[1])); // += l4*f3+l2*f1 r4*f3+r2*f1
// accu1 += l2*f2+l0*f0 r2*f2+r0*f0
// += l3*f3+l1*f1 r3*f3+r1*f1
// accu2 += l3*f2+l1*f0 r3*f2+r1*f0
// l4*f3+l2*f1 r4*f3+r2*f1
pVfilter += 2;
pVsrc += 2;
}
// accu >>= resultDivFactor
accu1 = _mm_srai_pi32(accu1, resultDivFactor);
accu2 = _mm_srai_pi32(accu2, resultDivFactor);
// pack 2*2*32bits => 4*16 bits
pVdest[0] = _mm_packs_pi32(accu1, accu2);
src += 4;
pVdest ++;
}
_m_empty(); // clear emms state
return (numSamples & 0xfffffffe) - length;
}
#else
// workaround to not complain about empty module
bool _dontcomplain_mmx_empty;
#endif // SOUNDTOUCH_ALLOW_MMX

730
source/SoundTouch/sse_optimized.cpp
View File

@ -1,365 +1,365 @@
////////////////////////////////////////////////////////////////////////////////
///
/// SSE optimized routines for Pentium-III, Athlon-XP and later CPUs. All SSE
/// optimized functions have been gathered into this single source
/// code file, regardless to their class or original source code file, in order
/// to ease porting the library to other compiler and processor platforms.
///
/// The SSE-optimizations are programmed using SSE compiler intrinsics that
/// are supported both by Microsoft Visual C++ and GCC compilers, so this file
/// should compile with both toolsets.
///
/// NOTICE: If using Visual Studio 6.0, you'll need to install the "Visual C++
/// 6.0 processor pack" update to support SSE instruction set. The update is
/// available for download at Microsoft Developers Network, see here:
/// http://msdn.microsoft.com/en-us/vstudio/aa718349.aspx
///
/// If the above URL is expired or removed, go to "http://msdn.microsoft.com" and
/// perform a search with keywords "processor pack".
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include "cpu_detect.h"
#include "STTypes.h"
using namespace soundtouch;
#ifdef SOUNDTOUCH_ALLOW_SSE
// SSE routines available only with float sample type
//////////////////////////////////////////////////////////////////////////////
//
// implementation of SSE optimized functions of class 'TDStretchSSE'
//
//////////////////////////////////////////////////////////////////////////////
#include "TDStretch.h"
#include <xmmintrin.h>
#include <math.h>
// Calculates cross correlation of two buffers
double TDStretchSSE::calcCrossCorr(const float *pV1, const float *pV2, double &anorm)
{
int i;
const float *pVec1;
const __m128 *pVec2;
__m128 vSum, vNorm;
// Note. It means a major slow-down if the routine needs to tolerate
// unaligned __m128 memory accesses. It's way faster if we can skip
// unaligned slots and use _mm_load_ps instruction instead of _mm_loadu_ps.
// This can mean up to ~ 10-fold difference (incl. part of which is
// due to skipping every second round for stereo sound though).
//
// Compile-time define SOUNDTOUCH_ALLOW_NONEXACT_SIMD_OPTIMIZATION is provided
// for choosing if this little cheating is allowed.
#ifdef ST_SIMD_AVOID_UNALIGNED
// Little cheating allowed, return valid correlation only for
// aligned locations, meaning every second round for stereo sound.
#define _MM_LOAD _mm_load_ps
if (((ulongptr)pV1) & 15) return -1e50; // skip unaligned locations
#else
// No cheating allowed, use unaligned load & take the resulting
// performance hit.
#define _MM_LOAD _mm_loadu_ps
#endif
// ensure overlapLength is divisible by 8
assert((overlapLength % 8) == 0);
// Calculates the cross-correlation value between 'pV1' and 'pV2' vectors
// Note: pV2 _must_ be aligned to 16-bit boundary, pV1 need not.
pVec1 = (const float*)pV1;
pVec2 = (const __m128*)pV2;
vSum = vNorm = _mm_setzero_ps();
// Unroll the loop by factor of 4 * 4 operations. Use same routine for
// stereo & mono, for mono it just means twice the amount of unrolling.
for (i = 0; i < channels * overlapLength / 16; i ++)
{
__m128 vTemp;
// vSum += pV1[0..3] * pV2[0..3]
vTemp = _MM_LOAD(pVec1);
vSum = _mm_add_ps(vSum, _mm_mul_ps(vTemp ,pVec2[0]));
vNorm = _mm_add_ps(vNorm, _mm_mul_ps(vTemp ,vTemp));
// vSum += pV1[4..7] * pV2[4..7]
vTemp = _MM_LOAD(pVec1 + 4);
vSum = _mm_add_ps(vSum, _mm_mul_ps(vTemp, pVec2[1]));
vNorm = _mm_add_ps(vNorm, _mm_mul_ps(vTemp ,vTemp));
// vSum += pV1[8..11] * pV2[8..11]
vTemp = _MM_LOAD(pVec1 + 8);
vSum = _mm_add_ps(vSum, _mm_mul_ps(vTemp, pVec2[2]));
vNorm = _mm_add_ps(vNorm, _mm_mul_ps(vTemp ,vTemp));
// vSum += pV1[12..15] * pV2[12..15]
vTemp = _MM_LOAD(pVec1 + 12);
vSum = _mm_add_ps(vSum, _mm_mul_ps(vTemp, pVec2[3]));
vNorm = _mm_add_ps(vNorm, _mm_mul_ps(vTemp ,vTemp));
pVec1 += 16;
pVec2 += 4;
}
// return value = vSum[0] + vSum[1] + vSum[2] + vSum[3]
float *pvNorm = (float*)&vNorm;
float norm = (pvNorm[0] + pvNorm[1] + pvNorm[2] + pvNorm[3]);
anorm = norm;
float *pvSum = (float*)&vSum;
return (double)(pvSum[0] + pvSum[1] + pvSum[2] + pvSum[3]) / sqrt(norm < 1e-9 ? 1.0 : norm);
/* This is approximately corresponding routine in C-language yet without normalization:
double corr, norm;
uint i;
// Calculates the cross-correlation value between 'pV1' and 'pV2' vectors
corr = norm = 0.0;
for (i = 0; i < channels * overlapLength / 16; i ++)
{
corr += pV1[0] * pV2[0] +
pV1[1] * pV2[1] +
pV1[2] * pV2[2] +
pV1[3] * pV2[3] +
pV1[4] * pV2[4] +
pV1[5] * pV2[5] +
pV1[6] * pV2[6] +
pV1[7] * pV2[7] +
pV1[8] * pV2[8] +
pV1[9] * pV2[9] +
pV1[10] * pV2[10] +
pV1[11] * pV2[11] +
pV1[12] * pV2[12] +
pV1[13] * pV2[13] +
pV1[14] * pV2[14] +
pV1[15] * pV2[15];
for (j = 0; j < 15; j ++) norm += pV1[j] * pV1[j];
pV1 += 16;
pV2 += 16;
}
return corr / sqrt(norm);
*/
}
double TDStretchSSE::calcCrossCorrAccumulate(const float *pV1, const float *pV2, double &norm)
{
// call usual calcCrossCorr function because SSE does not show big benefit of
// accumulating "norm" value, and also the "norm" rolling algorithm would get
// complicated due to SSE-specific alignment-vs-nonexact correlation rules.
return calcCrossCorr(pV1, pV2, norm);
}
//////////////////////////////////////////////////////////////////////////////
//
// implementation of SSE optimized functions of class 'FIRFilter'
//
//////////////////////////////////////////////////////////////////////////////
#include "FIRFilter.h"
FIRFilterSSE::FIRFilterSSE() : FIRFilter()
{
filterCoeffsAlign = nullptr;
filterCoeffsUnalign = nullptr;
}
FIRFilterSSE::~FIRFilterSSE()
{
delete[] filterCoeffsUnalign;
filterCoeffsAlign = nullptr;
filterCoeffsUnalign = nullptr;
}
// (overloaded) Calculates filter coefficients for SSE routine
void FIRFilterSSE::setCoefficients(const float *coeffs, uint newLength, uint uResultDivFactor)
{
uint i;
float fDivider;
FIRFilter::setCoefficients(coeffs, newLength, uResultDivFactor);
// Scale the filter coefficients so that it won't be necessary to scale the filtering result
// also rearrange coefficients suitably for SSE
// Ensure that filter coeffs array is aligned to 16-byte boundary
delete[] filterCoeffsUnalign;
filterCoeffsUnalign = new float[2 * newLength + 4];
filterCoeffsAlign = (float *)SOUNDTOUCH_ALIGN_POINTER_16(filterCoeffsUnalign);
fDivider = (float)resultDivider;
// rearrange the filter coefficients for mmx routines
for (i = 0; i < newLength; i ++)
{
filterCoeffsAlign[2 * i + 0] =
filterCoeffsAlign[2 * i + 1] = coeffs[i + 0] / fDivider;
}
}
// SSE-optimized version of the filter routine for stereo sound
uint FIRFilterSSE::evaluateFilterStereo(float *dest, const float *source, uint numSamples) const
{
int count = (int)((numSamples - length) & (uint)-2);
int j;
assert(count % 2 == 0);
if (count < 2) return 0;
assert(source != nullptr);
assert(dest != nullptr);
assert((length % 8) == 0);
assert(filterCoeffsAlign != nullptr);
assert(((ulongptr)filterCoeffsAlign) % 16 == 0);
// filter is evaluated for two stereo samples with each iteration, thus use of 'j += 2'
#pragma omp parallel for
for (j = 0; j < count; j += 2)
{
const float *pSrc;
float *pDest;
const __m128 *pFil;
__m128 sum1, sum2;
uint i;
pSrc = (const float*)source + j * 2; // source audio data
pDest = dest + j * 2; // destination audio data
pFil = (const __m128*)filterCoeffsAlign; // filter coefficients. NOTE: Assumes coefficients
// are aligned to 16-byte boundary
sum1 = sum2 = _mm_setzero_ps();
for (i = 0; i < length / 8; i ++)
{
// Unroll loop for efficiency & calculate filter for 2*2 stereo samples
// at each pass
// sum1 is accu for 2*2 filtered stereo sound data at the primary sound data offset
// sum2 is accu for 2*2 filtered stereo sound data for the next sound sample offset.
sum1 = _mm_add_ps(sum1, _mm_mul_ps(_mm_loadu_ps(pSrc) , pFil[0]));
sum2 = _mm_add_ps(sum2, _mm_mul_ps(_mm_loadu_ps(pSrc + 2), pFil[0]));
sum1 = _mm_add_ps(sum1, _mm_mul_ps(_mm_loadu_ps(pSrc + 4), pFil[1]));
sum2 = _mm_add_ps(sum2, _mm_mul_ps(_mm_loadu_ps(pSrc + 6), pFil[1]));
sum1 = _mm_add_ps(sum1, _mm_mul_ps(_mm_loadu_ps(pSrc + 8) , pFil[2]));
sum2 = _mm_add_ps(sum2, _mm_mul_ps(_mm_loadu_ps(pSrc + 10), pFil[2]));
sum1 = _mm_add_ps(sum1, _mm_mul_ps(_mm_loadu_ps(pSrc + 12), pFil[3]));
sum2 = _mm_add_ps(sum2, _mm_mul_ps(_mm_loadu_ps(pSrc + 14), pFil[3]));
pSrc += 16;
pFil += 4;
}
// Now sum1 and sum2 both have a filtered 2-channel sample each, but we still need
// to sum the two hi- and lo-floats of these registers together.
// post-shuffle & add the filtered values and store to dest.
_mm_storeu_ps(pDest, _mm_add_ps(
_mm_shuffle_ps(sum1, sum2, _MM_SHUFFLE(1,0,3,2)), // s2_1 s2_0 s1_3 s1_2
_mm_shuffle_ps(sum1, sum2, _MM_SHUFFLE(3,2,1,0)) // s2_3 s2_2 s1_1 s1_0
));
}
// Ideas for further improvement:
// 1. If it could be guaranteed that 'source' were always aligned to 16-byte
// boundary, a faster aligned '_mm_load_ps' instruction could be used.
// 2. If it could be guaranteed that 'dest' were always aligned to 16-byte
// boundary, a faster '_mm_store_ps' instruction could be used.
return (uint)count;
/* original routine in C-language. please notice the C-version has differently
organized coefficients though.
double suml1, suml2;
double sumr1, sumr2;
uint i, j;
for (j = 0; j < count; j += 2)
{
const float *ptr;
const float *pFil;
suml1 = sumr1 = 0.0;
suml2 = sumr2 = 0.0;
ptr = src;
pFil = filterCoeffs;
for (i = 0; i < lengthLocal; i ++)
{
// unroll loop for efficiency.
suml1 += ptr[0] * pFil[0] +
ptr[2] * pFil[2] +
ptr[4] * pFil[4] +
ptr[6] * pFil[6];
sumr1 += ptr[1] * pFil[1] +
ptr[3] * pFil[3] +
ptr[5] * pFil[5] +
ptr[7] * pFil[7];
suml2 += ptr[8] * pFil[0] +
ptr[10] * pFil[2] +
ptr[12] * pFil[4] +
ptr[14] * pFil[6];
sumr2 += ptr[9] * pFil[1] +
ptr[11] * pFil[3] +
ptr[13] * pFil[5] +
ptr[15] * pFil[7];
ptr += 16;
pFil += 8;
}
dest[0] = (float)suml1;
dest[1] = (float)sumr1;
dest[2] = (float)suml2;
dest[3] = (float)sumr2;
src += 4;
dest += 4;
}
*/
}
#endif // SOUNDTOUCH_ALLOW_SSE
////////////////////////////////////////////////////////////////////////////////
///
/// SSE optimized routines for Pentium-III, Athlon-XP and later CPUs. All SSE
/// optimized functions have been gathered into this single source
/// code file, regardless to their class or original source code file, in order
/// to ease porting the library to other compiler and processor platforms.
///
/// The SSE-optimizations are programmed using SSE compiler intrinsics that
/// are supported both by Microsoft Visual C++ and GCC compilers, so this file
/// should compile with both toolsets.
///
/// NOTICE: If using Visual Studio 6.0, you'll need to install the "Visual C++
/// 6.0 processor pack" update to support SSE instruction set. The update is
/// available for download at Microsoft Developers Network, see here:
/// http://msdn.microsoft.com/en-us/vstudio/aa718349.aspx
///
/// If the above URL is expired or removed, go to "http://msdn.microsoft.com" and
/// perform a search with keywords "processor pack".
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#include "cpu_detect.h"
#include "STTypes.h"
using namespace soundtouch;
#ifdef SOUNDTOUCH_ALLOW_SSE
// SSE routines available only with float sample type
//////////////////////////////////////////////////////////////////////////////
//
// implementation of SSE optimized functions of class 'TDStretchSSE'
//
//////////////////////////////////////////////////////////////////////////////
#include "TDStretch.h"
#include <xmmintrin.h>
#include <math.h>
// Calculates cross correlation of two buffers
double TDStretchSSE::calcCrossCorr(const float *pV1, const float *pV2, double &anorm)
{
int i;
const float *pVec1;
const __m128 *pVec2;
__m128 vSum, vNorm;
// Note. It means a major slow-down if the routine needs to tolerate
// unaligned __m128 memory accesses. It's way faster if we can skip
// unaligned slots and use _mm_load_ps instruction instead of _mm_loadu_ps.
// This can mean up to ~ 10-fold difference (incl. part of which is
// due to skipping every second round for stereo sound though).
//
// Compile-time define SOUNDTOUCH_ALLOW_NONEXACT_SIMD_OPTIMIZATION is provided
// for choosing if this little cheating is allowed.
#ifdef ST_SIMD_AVOID_UNALIGNED
// Little cheating allowed, return valid correlation only for
// aligned locations, meaning every second round for stereo sound.
#define _MM_LOAD _mm_load_ps
if (((ulongptr)pV1) & 15) return -1e50; // skip unaligned locations
#else
// No cheating allowed, use unaligned load & take the resulting
// performance hit.
#define _MM_LOAD _mm_loadu_ps
#endif
// ensure overlapLength is divisible by 8
assert((overlapLength % 8) == 0);
// Calculates the cross-correlation value between 'pV1' and 'pV2' vectors
// Note: pV2 _must_ be aligned to 16-bit boundary, pV1 need not.
pVec1 = (const float*)pV1;
pVec2 = (const __m128*)pV2;
vSum = vNorm = _mm_setzero_ps();
// Unroll the loop by factor of 4 * 4 operations. Use same routine for
// stereo & mono, for mono it just means twice the amount of unrolling.
for (i = 0; i < channels * overlapLength / 16; i ++)
{
__m128 vTemp;
// vSum += pV1[0..3] * pV2[0..3]
vTemp = _MM_LOAD(pVec1);
vSum = _mm_add_ps(vSum, _mm_mul_ps(vTemp ,pVec2[0]));
vNorm = _mm_add_ps(vNorm, _mm_mul_ps(vTemp ,vTemp));
// vSum += pV1[4..7] * pV2[4..7]
vTemp = _MM_LOAD(pVec1 + 4);
vSum = _mm_add_ps(vSum, _mm_mul_ps(vTemp, pVec2[1]));
vNorm = _mm_add_ps(vNorm, _mm_mul_ps(vTemp ,vTemp));
// vSum += pV1[8..11] * pV2[8..11]
vTemp = _MM_LOAD(pVec1 + 8);
vSum = _mm_add_ps(vSum, _mm_mul_ps(vTemp, pVec2[2]));
vNorm = _mm_add_ps(vNorm, _mm_mul_ps(vTemp ,vTemp));
// vSum += pV1[12..15] * pV2[12..15]
vTemp = _MM_LOAD(pVec1 + 12);
vSum = _mm_add_ps(vSum, _mm_mul_ps(vTemp, pVec2[3]));
vNorm = _mm_add_ps(vNorm, _mm_mul_ps(vTemp ,vTemp));
pVec1 += 16;
pVec2 += 4;
}
// return value = vSum[0] + vSum[1] + vSum[2] + vSum[3]
float *pvNorm = (float*)&vNorm;
float norm = (pvNorm[0] + pvNorm[1] + pvNorm[2] + pvNorm[3]);
anorm = norm;
float *pvSum = (float*)&vSum;
return (double)(pvSum[0] + pvSum[1] + pvSum[2] + pvSum[3]) / sqrt(norm < 1e-9 ? 1.0 : norm);
/* This is approximately corresponding routine in C-language yet without normalization:
double corr, norm;
uint i;
// Calculates the cross-correlation value between 'pV1' and 'pV2' vectors
corr = norm = 0.0;
for (i = 0; i < channels * overlapLength / 16; i ++)
{
corr += pV1[0] * pV2[0] +
pV1[1] * pV2[1] +
pV1[2] * pV2[2] +
pV1[3] * pV2[3] +
pV1[4] * pV2[4] +
pV1[5] * pV2[5] +
pV1[6] * pV2[6] +
pV1[7] * pV2[7] +
pV1[8] * pV2[8] +
pV1[9] * pV2[9] +
pV1[10] * pV2[10] +
pV1[11] * pV2[11] +
pV1[12] * pV2[12] +
pV1[13] * pV2[13] +
pV1[14] * pV2[14] +
pV1[15] * pV2[15];
for (j = 0; j < 15; j ++) norm += pV1[j] * pV1[j];
pV1 += 16;
pV2 += 16;
}
return corr / sqrt(norm);
*/
}
double TDStretchSSE::calcCrossCorrAccumulate(const float *pV1, const float *pV2, double &norm)
{
// call usual calcCrossCorr function because SSE does not show big benefit of
// accumulating "norm" value, and also the "norm" rolling algorithm would get
// complicated due to SSE-specific alignment-vs-nonexact correlation rules.
return calcCrossCorr(pV1, pV2, norm);
}
//////////////////////////////////////////////////////////////////////////////
//
// implementation of SSE optimized functions of class 'FIRFilter'
//
//////////////////////////////////////////////////////////////////////////////
#include "FIRFilter.h"
FIRFilterSSE::FIRFilterSSE() : FIRFilter()
{
filterCoeffsAlign = nullptr;
filterCoeffsUnalign = nullptr;
}
FIRFilterSSE::~FIRFilterSSE()
{
delete[] filterCoeffsUnalign;
filterCoeffsAlign = nullptr;
filterCoeffsUnalign = nullptr;
}
// (overloaded) Calculates filter coefficients for SSE routine
void FIRFilterSSE::setCoefficients(const float *coeffs, uint newLength, uint uResultDivFactor)
{
uint i;
float fDivider;
FIRFilter::setCoefficients(coeffs, newLength, uResultDivFactor);
// Scale the filter coefficients so that it won't be necessary to scale the filtering result
// also rearrange coefficients suitably for SSE
// Ensure that filter coeffs array is aligned to 16-byte boundary
delete[] filterCoeffsUnalign;
filterCoeffsUnalign = new float[2 * newLength + 4];
filterCoeffsAlign = (float *)SOUNDTOUCH_ALIGN_POINTER_16(filterCoeffsUnalign);
fDivider = (float)resultDivider;
// rearrange the filter coefficients for mmx routines
for (i = 0; i < newLength; i ++)
{
filterCoeffsAlign[2 * i + 0] =
filterCoeffsAlign[2 * i + 1] = coeffs[i + 0] / fDivider;
}
}
// SSE-optimized version of the filter routine for stereo sound
uint FIRFilterSSE::evaluateFilterStereo(float *dest, const float *source, uint numSamples) const
{
int count = (int)((numSamples - length) & (uint)-2);
int j;
assert(count % 2 == 0);
if (count < 2) return 0;
assert(source != nullptr);
assert(dest != nullptr);
assert((length % 8) == 0);
assert(filterCoeffsAlign != nullptr);
assert(((ulongptr)filterCoeffsAlign) % 16 == 0);
// filter is evaluated for two stereo samples with each iteration, thus use of 'j += 2'
#pragma omp parallel for
for (j = 0; j < count; j += 2)
{
const float *pSrc;
float *pDest;
const __m128 *pFil;
__m128 sum1, sum2;
uint i;
pSrc = (const float*)source + j * 2; // source audio data
pDest = dest + j * 2; // destination audio data
pFil = (const __m128*)filterCoeffsAlign; // filter coefficients. NOTE: Assumes coefficients
// are aligned to 16-byte boundary
sum1 = sum2 = _mm_setzero_ps();
for (i = 0; i < length / 8; i ++)
{
// Unroll loop for efficiency & calculate filter for 2*2 stereo samples
// at each pass
// sum1 is accu for 2*2 filtered stereo sound data at the primary sound data offset
// sum2 is accu for 2*2 filtered stereo sound data for the next sound sample offset.
sum1 = _mm_add_ps(sum1, _mm_mul_ps(_mm_loadu_ps(pSrc) , pFil[0]));
sum2 = _mm_add_ps(sum2, _mm_mul_ps(_mm_loadu_ps(pSrc + 2), pFil[0]));
sum1 = _mm_add_ps(sum1, _mm_mul_ps(_mm_loadu_ps(pSrc + 4), pFil[1]));
sum2 = _mm_add_ps(sum2, _mm_mul_ps(_mm_loadu_ps(pSrc + 6), pFil[1]));
sum1 = _mm_add_ps(sum1, _mm_mul_ps(_mm_loadu_ps(pSrc + 8) , pFil[2]));
sum2 = _mm_add_ps(sum2, _mm_mul_ps(_mm_loadu_ps(pSrc + 10), pFil[2]));
sum1 = _mm_add_ps(sum1, _mm_mul_ps(_mm_loadu_ps(pSrc + 12), pFil[3]));
sum2 = _mm_add_ps(sum2, _mm_mul_ps(_mm_loadu_ps(pSrc + 14), pFil[3]));
pSrc += 16;
pFil += 4;
}
// Now sum1 and sum2 both have a filtered 2-channel sample each, but we still need
// to sum the two hi- and lo-floats of these registers together.
// post-shuffle & add the filtered values and store to dest.
_mm_storeu_ps(pDest, _mm_add_ps(
_mm_shuffle_ps(sum1, sum2, _MM_SHUFFLE(1,0,3,2)), // s2_1 s2_0 s1_3 s1_2
_mm_shuffle_ps(sum1, sum2, _MM_SHUFFLE(3,2,1,0)) // s2_3 s2_2 s1_1 s1_0
));
}
// Ideas for further improvement:
// 1. If it could be guaranteed that 'source' were always aligned to 16-byte
// boundary, a faster aligned '_mm_load_ps' instruction could be used.
// 2. If it could be guaranteed that 'dest' were always aligned to 16-byte
// boundary, a faster '_mm_store_ps' instruction could be used.
return (uint)count;
/* original routine in C-language. please notice the C-version has differently
organized coefficients though.
double suml1, suml2;
double sumr1, sumr2;
uint i, j;
for (j = 0; j < count; j += 2)
{
const float *ptr;
const float *pFil;
suml1 = sumr1 = 0.0;
suml2 = sumr2 = 0.0;
ptr = src;
pFil = filterCoeffs;
for (i = 0; i < lengthLocal; i ++)
{
// unroll loop for efficiency.
suml1 += ptr[0] * pFil[0] +
ptr[2] * pFil[2] +
ptr[4] * pFil[4] +
ptr[6] * pFil[6];
sumr1 += ptr[1] * pFil[1] +
ptr[3] * pFil[3] +
ptr[5] * pFil[5] +
ptr[7] * pFil[7];
suml2 += ptr[8] * pFil[0] +
ptr[10] * pFil[2] +
ptr[12] * pFil[4] +
ptr[14] * pFil[6];
sumr2 += ptr[9] * pFil[1] +
ptr[11] * pFil[3] +
ptr[13] * pFil[5] +
ptr[15] * pFil[7];
ptr += 16;
pFil += 8;
}
dest[0] = (float)suml1;
dest[1] = (float)sumr1;
dest[2] = (float)suml2;
dest[3] = (float)sumr2;
src += 4;
dest += 4;
}
*/
}
#endif // SOUNDTOUCH_ALLOW_SSE

230
source/SoundTouchDLL/DllTest/DllTest.cpp
View File

@ -1,115 +1,115 @@
////////////////////////////////////////////////////////////////////////////////
///
/// DllTest.cpp : This is small app main routine used for testing sound processing
/// with SoundTouch.dll API
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
#include <string>
#include <iostream>
#include <fstream>
#include "../SoundTouchDLL.h"
#include "../../SoundStretch/WavFile.h"
using namespace std;
using namespace soundstretch;
// DllTest main
int wmain(int argc, const wchar_t *argv[])
{
// Check program arguments
if (argc < 4)
{
cout << "Too few arguments. Usage: DllTest [infile.wav] [outfile.wav] [sampletype]" << endl;
return -1;
}
wstring inFileName = argv[1];
wstring outFileName = argv[2];
wstring str_sampleType = argv[3];
bool floatSample;
if (str_sampleType == L"float")
{
floatSample = true;
}
else if (str_sampleType == L"short")
{
floatSample = false;
}
else
{
cerr << "Missing or invalid sampletype. Expected either short or float" << endl;
return -1;
}
try
{
// Open input & output WAV files
WavInFile inFile(inFileName);
int numChannels = inFile.getNumChannels();
int sampleRate = inFile.getSampleRate();
WavOutFile outFile(outFileName, sampleRate, inFile.getNumBits(), numChannels);
// Create SoundTouch DLL instance
HANDLE st = soundtouch_createInstance();
soundtouch_setChannels(st, numChannels);
soundtouch_setSampleRate(st, sampleRate);
soundtouch_setPitchSemiTones(st, 2);
cout << "processing with soundtouch.dll routines";
if (floatSample)
{
// Process file with SoundTouch.DLL float sample (default) API
float fbuffer[2048];
int nmax = 2048 / numChannels;
cout << " using float api ..." << endl;
while (inFile.eof() == false)
{
int n = inFile.read(fbuffer, nmax * numChannels) / numChannels;
soundtouch_putSamples(st, fbuffer, n);
do
{
n = soundtouch_receiveSamples(st, fbuffer, nmax);
outFile.write(fbuffer, n * numChannels);
} while (n > 0);
}
}
else
{
// Process file with SoundTouch.DLL int16 (short) sample API.
// Notice that SoundTouch.dll does internally processing using floating
// point routines so the int16 API is not any faster, but provided for
// convenience.
short i16buffer[2048];
int nmax = 2048 / numChannels;
cout << " using i16 api ..." << endl;
while (inFile.eof() == false)
{
int n = inFile.read(i16buffer, nmax * numChannels) / numChannels;
soundtouch_putSamples_i16(st, i16buffer, n);
do
{
n = soundtouch_receiveSamples_i16(st, i16buffer, nmax);
outFile.write(i16buffer, n * numChannels);
} while (n > 0);
}
}
soundtouch_destroyInstance(st);
cout << "done." << endl;
}
catch (const runtime_error &e)
{
cerr << e.what() << endl;
}
return 0;
}
////////////////////////////////////////////////////////////////////////////////
///
/// DllTest.cpp : This is small app main routine used for testing sound processing
/// with SoundTouch.dll API
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
#include <string>
#include <iostream>
#include <fstream>
#include "../SoundTouchDLL.h"
#include "../../SoundStretch/WavFile.h"
using namespace std;
using namespace soundstretch;
// DllTest main
int wmain(int argc, const wchar_t *argv[])
{
// Check program arguments
if (argc < 4)
{
cout << "Too few arguments. Usage: DllTest [infile.wav] [outfile.wav] [sampletype]" << endl;
return -1;
}
wstring inFileName = argv[1];
wstring outFileName = argv[2];
wstring str_sampleType = argv[3];
bool floatSample;
if (str_sampleType == L"float")
{
floatSample = true;
}
else if (str_sampleType == L"short")
{
floatSample = false;
}
else
{
cerr << "Missing or invalid sampletype. Expected either short or float" << endl;
return -1;
}
try
{
// Open input & output WAV files
WavInFile inFile(inFileName);
int numChannels = inFile.getNumChannels();
int sampleRate = inFile.getSampleRate();
WavOutFile outFile(outFileName, sampleRate, inFile.getNumBits(), numChannels);
// Create SoundTouch DLL instance
HANDLE st = soundtouch_createInstance();
soundtouch_setChannels(st, numChannels);
soundtouch_setSampleRate(st, sampleRate);
soundtouch_setPitchSemiTones(st, 2);
cout << "processing with soundtouch.dll routines";
if (floatSample)
{
// Process file with SoundTouch.DLL float sample (default) API
float fbuffer[2048];
int nmax = 2048 / numChannels;
cout << " using float api ..." << endl;
while (inFile.eof() == false)
{
int n = inFile.read(fbuffer, nmax * numChannels) / numChannels;
soundtouch_putSamples(st, fbuffer, n);
do
{
n = soundtouch_receiveSamples(st, fbuffer, nmax);
outFile.write(fbuffer, n * numChannels);
} while (n > 0);
}
}
else
{
// Process file with SoundTouch.DLL int16 (short) sample API.
// Notice that SoundTouch.dll does internally processing using floating
// point routines so the int16 API is not any faster, but provided for
// convenience.
short i16buffer[2048];
int nmax = 2048 / numChannels;
cout << " using i16 api ..." << endl;
while (inFile.eof() == false)
{
int n = inFile.read(i16buffer, nmax * numChannels) / numChannels;
soundtouch_putSamples_i16(st, i16buffer, n);
do
{
n = soundtouch_receiveSamples_i16(st, i16buffer, nmax);
outFile.write(i16buffer, n * numChannels);
} while (n > 0);
}
}
soundtouch_destroyInstance(st);
cout << "done." << endl;
}
catch (const runtime_error &e)
{
cerr << e.what() << endl;
}
return 0;
}

94
source/SoundTouchDLL/Makefile.am
View File

@ -1,47 +1,47 @@
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA
include $(top_srcdir)/config/am_include.mk
noinst_HEADERS=../SoundTouch/AAFilter.h ../SoundTouch/cpu_detect.h ../SoundTouch/cpu_detect_x86.cpp ../SoundTouch/FIRFilter.h \
../SoundTouch/RateTransposer.h ../SoundTouch/TDStretch.h ../SoundTouch/PeakFinder.h ../SoundTouch/InterpolateCubic.h \
../SoundTouch/InterpolateLinear.h ../SoundTouch/InterpolateShannon.h
include_HEADERS=SoundTouchDLL.h
lib_LTLIBRARIES=libSoundTouchDll.la
#
libSoundTouchDll_la_SOURCES=../SoundTouch/AAFilter.cpp ../SoundTouch/FIRFilter.cpp \
../SoundTouch/FIFOSampleBuffer.cpp ../SoundTouch/RateTransposer.cpp ../SoundTouch/SoundTouch.cpp \
../SoundTouch/TDStretch.cpp ../SoundTouch/sse_optimized.cpp ../SoundTouch/cpu_detect_x86.cpp \
../SoundTouch/BPMDetect.cpp ../SoundTouch/PeakFinder.cpp ../SoundTouch/InterpolateLinear.cpp \
../SoundTouch/InterpolateCubic.cpp ../SoundTouch/InterpolateShannon.cpp SoundTouchDLL.cpp
# Compiler flags
# Modify the default 0.0.0 to LIB_SONAME.0.0
AM_LDFLAGS=$(LDFLAGS) -version-info @LIB_SONAME@
if X86
CXXFLAGS1=-mstackrealign -msse
endif
if X86_64
CXXFLAGS2=-fPIC
endif
AM_CXXFLAGS=$(CXXFLAGS) $(CXXFLAGS1) $(CXXFLAGS2) -shared -DDLL_EXPORTS -fvisibility=hidden
## Process this file with automake to create Makefile.in
##
## This file is part of SoundTouch, an audio processing library for pitch/time adjustments
##
## SoundTouch is free software; you can redistribute it and/or modify it under the
## terms of the GNU General Public License as published by the Free Software
## Foundation; either version 2 of the License, or (at your option) any later
## version.
##
## SoundTouch is distributed in the hope that it will be useful, but WITHOUT ANY
## WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
## A PARTICULAR PURPOSE. See the GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License along with
## this program; if not, write to the Free Software Foundation, Inc., 59 Temple
## Place - Suite 330, Boston, MA 02111-1307, USA
include $(top_srcdir)/config/am_include.mk
noinst_HEADERS=../SoundTouch/AAFilter.h ../SoundTouch/cpu_detect.h ../SoundTouch/cpu_detect_x86.cpp ../SoundTouch/FIRFilter.h \
../SoundTouch/RateTransposer.h ../SoundTouch/TDStretch.h ../SoundTouch/PeakFinder.h ../SoundTouch/InterpolateCubic.h \
../SoundTouch/InterpolateLinear.h ../SoundTouch/InterpolateShannon.h
include_HEADERS=SoundTouchDLL.h
lib_LTLIBRARIES=libSoundTouchDll.la
#
libSoundTouchDll_la_SOURCES=../SoundTouch/AAFilter.cpp ../SoundTouch/FIRFilter.cpp \
../SoundTouch/FIFOSampleBuffer.cpp ../SoundTouch/RateTransposer.cpp ../SoundTouch/SoundTouch.cpp \
../SoundTouch/TDStretch.cpp ../SoundTouch/sse_optimized.cpp ../SoundTouch/cpu_detect_x86.cpp \
../SoundTouch/BPMDetect.cpp ../SoundTouch/PeakFinder.cpp ../SoundTouch/InterpolateLinear.cpp \
../SoundTouch/InterpolateCubic.cpp ../SoundTouch/InterpolateShannon.cpp SoundTouchDLL.cpp
# Compiler flags
# Modify the default 0.0.0 to LIB_SONAME.0.0
AM_LDFLAGS=$(LDFLAGS) -version-info @LIB_SONAME@
if X86
CXXFLAGS1=-mstackrealign -msse
endif
if X86_64
CXXFLAGS2=-fPIC
endif
AM_CXXFLAGS=$(CXXFLAGS) $(CXXFLAGS1) $(CXXFLAGS2) -shared -DDLL_EXPORTS -fvisibility=hidden

1164
source/SoundTouchDLL/SoundTouchDLL.cpp
View File

File diff suppressed because it is too large Load Diff

480
source/SoundTouchDLL/SoundTouchDLL.h
View File

@ -1,240 +1,240 @@
//////////////////////////////////////////////////////////////////////////////
///
/// SoundTouch DLL wrapper - wraps SoundTouch routines into a Dynamic Load
/// Library interface.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _SoundTouchDLL_h_
#define _SoundTouchDLL_h_
#if defined(_WIN32) || defined(WIN32)
// Windows
#ifndef __cplusplus
#error "Expected g++"
#endif
#ifdef DLL_EXPORTS
#define SOUNDTOUCHDLL_API extern "C" __declspec(dllexport)
#else
#define SOUNDTOUCHDLL_API extern "C" __declspec(dllimport)
#endif
#else
// GNU version
#if defined(DLL_EXPORTS) || defined(SoundTouchDLL_EXPORTS)
// GCC declaration for exporting functions
#define SOUNDTOUCHDLL_API extern "C" __attribute__((__visibility__("default")))
#else
// import function
#define SOUNDTOUCHDLL_API extern "C"
#endif
// Linux-replacements for Windows declarations:
#define __cdecl
typedef unsigned int DWORD;
#define FALSE 0
#define TRUE 1
#endif
typedef void * HANDLE;
/// Create a new instance of SoundTouch processor.
SOUNDTOUCHDLL_API HANDLE __cdecl soundtouch_createInstance();
/// Destroys a SoundTouch processor instance.
SOUNDTOUCHDLL_API void __cdecl soundtouch_destroyInstance(HANDLE h);
/// Get SoundTouch library version string
SOUNDTOUCHDLL_API const char *__cdecl soundtouch_getVersionString();
/// Get SoundTouch library version string - alternative function for
/// environments that can't properly handle character string as return value
SOUNDTOUCHDLL_API void __cdecl soundtouch_getVersionString2(char* versionString, int bufferSize);
/// Get SoundTouch library version Id
SOUNDTOUCHDLL_API unsigned int __cdecl soundtouch_getVersionId();
/// Sets new rate control value. Normal rate = 1.0, smaller values
/// represent slower rate, larger faster rates.
SOUNDTOUCHDLL_API void __cdecl soundtouch_setRate(HANDLE h, float newRate);
/// Sets new tempo control value. Normal tempo = 1.0, smaller values
/// represent slower tempo, larger faster tempo.
SOUNDTOUCHDLL_API void __cdecl soundtouch_setTempo(HANDLE h, float newTempo);
/// Sets new rate control value as a difference in percents compared
/// to the original rate (-50 .. +100 %);
SOUNDTOUCHDLL_API void __cdecl soundtouch_setRateChange(HANDLE h, float newRate);
/// Sets new tempo control value as a difference in percents compared
/// to the original tempo (-50 .. +100 %);
SOUNDTOUCHDLL_API void __cdecl soundtouch_setTempoChange(HANDLE h, float newTempo);
/// Sets new pitch control value. Original pitch = 1.0, smaller values
/// represent lower pitches, larger values higher pitch.
SOUNDTOUCHDLL_API void __cdecl soundtouch_setPitch(HANDLE h, float newPitch);
/// Sets pitch change in octaves compared to the original pitch
/// (-1.00 .. +1.00);
SOUNDTOUCHDLL_API void __cdecl soundtouch_setPitchOctaves(HANDLE h, float newPitch);
/// Sets pitch change in semi-tones compared to the original pitch
/// (-12 .. +12);
SOUNDTOUCHDLL_API void __cdecl soundtouch_setPitchSemiTones(HANDLE h, float newPitch);
/// Sets the number of channels, 1 = mono, 2 = stereo, n = multichannel
SOUNDTOUCHDLL_API int __cdecl soundtouch_setChannels(HANDLE h, unsigned int numChannels);
/// Sets sample rate.
SOUNDTOUCHDLL_API int __cdecl soundtouch_setSampleRate(HANDLE h, unsigned int srate);
/// Flushes the last samples from the processing pipeline to the output.
/// Clears also the internal processing buffers.
//
/// Note: This function is meant for extracting the last samples of a sound
/// stream. This function may introduce additional blank samples in the end
/// of the sound stream, and thus it's not recommended to call this function
/// in the middle of a sound stream.
SOUNDTOUCHDLL_API int __cdecl soundtouch_flush(HANDLE h);
/// Adds 'numSamples' pcs of samples from the 'samples' memory position into
/// the input of the object. Notice that sample rate _has_to_ be set before
/// calling this function, otherwise throws a runtime_error exception.
SOUNDTOUCHDLL_API int __cdecl soundtouch_putSamples(HANDLE h,
const float *samples, ///< Pointer to sample buffer.
unsigned int numSamples ///< Number of sample frames in buffer. Notice
///< that in case of multi-channel sound a single
///< sample frame contains data for all channels.
);
/// int16 version of soundtouch_putSamples(): This accept int16 (short) sample data
/// and internally converts it to float format before processing
SOUNDTOUCHDLL_API void __cdecl soundtouch_putSamples_i16(HANDLE h,
const short *samples, ///< Pointer to sample buffer.
unsigned int numSamples ///< Number of sample frames in buffer. Notice
///< that in case of multi-channel sound a single
///< sample frame contains data for all channels.
);
/// Clears all the samples in the object's output and internal processing
/// buffers.
SOUNDTOUCHDLL_API void __cdecl soundtouch_clear(HANDLE h);
/// Changes a setting controlling the processing system behaviour. See the
/// 'SETTING_...' defines for available setting ID's.
///
/// \return 'nonzero' if the setting was successfully changed, otherwise zero
SOUNDTOUCHDLL_API int __cdecl soundtouch_setSetting(HANDLE h,
int settingId, ///< Setting ID number. see SETTING_... defines.
int value ///< New setting value.
);
/// Reads a setting controlling the processing system behaviour. See the
/// 'SETTING_...' defines for available setting ID's.
///
/// \return the setting value.
SOUNDTOUCHDLL_API int __cdecl soundtouch_getSetting(HANDLE h,
int settingId ///< Setting ID number, see SETTING_... defines.
);
/// Returns number of samples currently unprocessed.
SOUNDTOUCHDLL_API unsigned int __cdecl soundtouch_numUnprocessedSamples(HANDLE h);
/// Adjusts book-keeping so that given number of samples are removed from beginning of the
/// sample buffer without copying them anywhere.
///
/// Used to reduce the number of samples in the buffer when accessing the sample buffer directly
/// with 'ptrBegin' function.
SOUNDTOUCHDLL_API unsigned int __cdecl soundtouch_receiveSamples(HANDLE h,
float *outBuffer, ///< Buffer where to copy output samples.
unsigned int maxSamples ///< How many samples to receive at max.
);
/// int16 version of soundtouch_receiveSamples(): This converts internal float samples
/// into int16 (short) return data type
SOUNDTOUCHDLL_API unsigned int __cdecl soundtouch_receiveSamples_i16(HANDLE h,
short *outBuffer, ///< Buffer where to copy output samples.
unsigned int maxSamples ///< How many samples to receive at max.
);
/// Returns number of samples currently available.
SOUNDTOUCHDLL_API unsigned int __cdecl soundtouch_numSamples(HANDLE h);
/// Returns nonzero if there aren't any samples available for outputting.
SOUNDTOUCHDLL_API int __cdecl soundtouch_isEmpty(HANDLE h);
/// Create a new instance of BPM detector
SOUNDTOUCHDLL_API HANDLE __cdecl bpm_createInstance(int numChannels, int sampleRate);
/// Destroys a BPM detector instance.
SOUNDTOUCHDLL_API void __cdecl bpm_destroyInstance(HANDLE h);
/// Feed 'numSamples' sample frames from 'samples' into the BPM detector.
SOUNDTOUCHDLL_API void __cdecl bpm_putSamples(HANDLE h,
const float *samples, ///< Pointer to sample buffer.
unsigned int numSamples ///< Number of samples in buffer. Notice
///< that in case of stereo-sound a single sample
///< contains data for both channels.
);
/// Feed 'numSamples' sample frames from 'samples' into the BPM detector.
/// 16bit int sample format version.
SOUNDTOUCHDLL_API void __cdecl bpm_putSamples_i16(HANDLE h,
const short *samples, ///< Pointer to sample buffer.
unsigned int numSamples ///< Number of samples in buffer. Notice
///< that in case of stereo-sound a single sample
///< contains data for both channels.
);
/// Analyzes the results and returns the BPM rate. Use this function to read result
/// after whole song data has been input to the class by consecutive calls of
/// 'inputSamples' function.
///
/// \return Beats-per-minute rate, or zero if detection failed.
SOUNDTOUCHDLL_API float __cdecl bpm_getBpm(HANDLE h);
/// Get beat position arrays. Note: The array includes also really low beat detection values
/// in absence of clear strong beats. Consumer may wish to filter low values away.
/// - "pos" receive array of beat positions
/// - "values" receive array of beat detection strengths
/// - max_num indicates max.size of "pos" and "values" array.
///
/// You can query a suitable array sized by calling this with nullptr in "pos" & "values".
///
/// \return number of beats in the arrays.
SOUNDTOUCHDLL_API int __cdecl bpm_getBeats(HANDLE h, float *pos, float *strength, int count);
#endif // _SoundTouchDLL_h_
//////////////////////////////////////////////////////////////////////////////
///
/// SoundTouch DLL wrapper - wraps SoundTouch routines into a Dynamic Load
/// Library interface.
///
/// Author : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
// SoundTouch audio processing library
// Copyright (c) Olli Parviainen
//
// This library is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 2.1 of the License, or (at your option) any later version.
//
// This library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
// Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
////////////////////////////////////////////////////////////////////////////////
#ifndef _SoundTouchDLL_h_
#define _SoundTouchDLL_h_
#if defined(_WIN32) || defined(WIN32)
// Windows
#ifndef __cplusplus
#error "Expected g++"
#endif
#ifdef DLL_EXPORTS
#define SOUNDTOUCHDLL_API extern "C" __declspec(dllexport)
#else
#define SOUNDTOUCHDLL_API extern "C" __declspec(dllimport)
#endif
#else
// GNU version
#if defined(DLL_EXPORTS) || defined(SoundTouchDLL_EXPORTS)
// GCC declaration for exporting functions
#define SOUNDTOUCHDLL_API extern "C" __attribute__((__visibility__("default")))
#else
// import function
#define SOUNDTOUCHDLL_API extern "C"
#endif
// Linux-replacements for Windows declarations:
#define __cdecl
typedef unsigned int DWORD;
#define FALSE 0
#define TRUE 1
#endif
typedef void * HANDLE;
/// Create a new instance of SoundTouch processor.
SOUNDTOUCHDLL_API HANDLE __cdecl soundtouch_createInstance();
/// Destroys a SoundTouch processor instance.
SOUNDTOUCHDLL_API void __cdecl soundtouch_destroyInstance(HANDLE h);
/// Get SoundTouch library version string
SOUNDTOUCHDLL_API const char *__cdecl soundtouch_getVersionString();
/// Get SoundTouch library version string - alternative function for
/// environments that can't properly handle character string as return value
SOUNDTOUCHDLL_API void __cdecl soundtouch_getVersionString2(char* versionString, int bufferSize);
/// Get SoundTouch library version Id
SOUNDTOUCHDLL_API unsigned int __cdecl soundtouch_getVersionId();
/// Sets new rate control value. Normal rate = 1.0, smaller values
/// represent slower rate, larger faster rates.
SOUNDTOUCHDLL_API void __cdecl soundtouch_setRate(HANDLE h, float newRate);
/// Sets new tempo control value. Normal tempo = 1.0, smaller values
/// represent slower tempo, larger faster tempo.
SOUNDTOUCHDLL_API void __cdecl soundtouch_setTempo(HANDLE h, float newTempo);
/// Sets new rate control value as a difference in percents compared
/// to the original rate (-50 .. +100 %);
SOUNDTOUCHDLL_API void __cdecl soundtouch_setRateChange(HANDLE h, float newRate);
/// Sets new tempo control value as a difference in percents compared
/// to the original tempo (-50 .. +100 %);
SOUNDTOUCHDLL_API void __cdecl soundtouch_setTempoChange(HANDLE h, float newTempo);
/// Sets new pitch control value. Original pitch = 1.0, smaller values
/// represent lower pitches, larger values higher pitch.
SOUNDTOUCHDLL_API void __cdecl soundtouch_setPitch(HANDLE h, float newPitch);
/// Sets pitch change in octaves compared to the original pitch
/// (-1.00 .. +1.00);
SOUNDTOUCHDLL_API void __cdecl soundtouch_setPitchOctaves(HANDLE h, float newPitch);
/// Sets pitch change in semi-tones compared to the original pitch
/// (-12 .. +12);
SOUNDTOUCHDLL_API void __cdecl soundtouch_setPitchSemiTones(HANDLE h, float newPitch);
/// Sets the number of channels, 1 = mono, 2 = stereo, n = multichannel
SOUNDTOUCHDLL_API int __cdecl soundtouch_setChannels(HANDLE h, unsigned int numChannels);
/// Sets sample rate.
SOUNDTOUCHDLL_API int __cdecl soundtouch_setSampleRate(HANDLE h, unsigned int srate);
/// Flushes the last samples from the processing pipeline to the output.
/// Clears also the internal processing buffers.
//
/// Note: This function is meant for extracting the last samples of a sound
/// stream. This function may introduce additional blank samples in the end
/// of the sound stream, and thus it's not recommended to call this function
/// in the middle of a sound stream.
SOUNDTOUCHDLL_API int __cdecl soundtouch_flush(HANDLE h);
/// Adds 'numSamples' pcs of samples from the 'samples' memory position into
/// the input of the object. Notice that sample rate _has_to_ be set before
/// calling this function, otherwise throws a runtime_error exception.
SOUNDTOUCHDLL_API int __cdecl soundtouch_putSamples(HANDLE h,
const float *samples, ///< Pointer to sample buffer.
unsigned int numSamples ///< Number of sample frames in buffer. Notice
///< that in case of multi-channel sound a single
///< sample frame contains data for all channels.
);
/// int16 version of soundtouch_putSamples(): This accept int16 (short) sample data
/// and internally converts it to float format before processing
SOUNDTOUCHDLL_API void __cdecl soundtouch_putSamples_i16(HANDLE h,
const short *samples, ///< Pointer to sample buffer.
unsigned int numSamples ///< Number of sample frames in buffer. Notice
///< that in case of multi-channel sound a single
///< sample frame contains data for all channels.
);
/// Clears all the samples in the object's output and internal processing
/// buffers.
SOUNDTOUCHDLL_API void __cdecl soundtouch_clear(HANDLE h);
/// Changes a setting controlling the processing system behaviour. See the
/// 'SETTING_...' defines for available setting ID's.
///
/// \return 'nonzero' if the setting was successfully changed, otherwise zero
SOUNDTOUCHDLL_API int __cdecl soundtouch_setSetting(HANDLE h,
int settingId, ///< Setting ID number. see SETTING_... defines.
int value ///< New setting value.
);
/// Reads a setting controlling the processing system behaviour. See the
/// 'SETTING_...' defines for available setting ID's.
///
/// \return the setting value.
SOUNDTOUCHDLL_API int __cdecl soundtouch_getSetting(HANDLE h,
int settingId ///< Setting ID number, see SETTING_... defines.
);
/// Returns number of samples currently unprocessed.
SOUNDTOUCHDLL_API unsigned int __cdecl soundtouch_numUnprocessedSamples(HANDLE h);
/// Adjusts book-keeping so that given number of samples are removed from beginning of the
/// sample buffer without copying them anywhere.
///
/// Used to reduce the number of samples in the buffer when accessing the sample buffer directly
/// with 'ptrBegin' function.
SOUNDTOUCHDLL_API unsigned int __cdecl soundtouch_receiveSamples(HANDLE h,
float *outBuffer, ///< Buffer where to copy output samples.
unsigned int maxSamples ///< How many samples to receive at max.
);
/// int16 version of soundtouch_receiveSamples(): This converts internal float samples
/// into int16 (short) return data type
SOUNDTOUCHDLL_API unsigned int __cdecl soundtouch_receiveSamples_i16(HANDLE h,
short *outBuffer, ///< Buffer where to copy output samples.
unsigned int maxSamples ///< How many samples to receive at max.
);
/// Returns number of samples currently available.
SOUNDTOUCHDLL_API unsigned int __cdecl soundtouch_numSamples(HANDLE h);
/// Returns nonzero if there aren't any samples available for outputting.
SOUNDTOUCHDLL_API int __cdecl soundtouch_isEmpty(HANDLE h);
/// Create a new instance of BPM detector
SOUNDTOUCHDLL_API HANDLE __cdecl bpm_createInstance(int numChannels, int sampleRate);
/// Destroys a BPM detector instance.
SOUNDTOUCHDLL_API void __cdecl bpm_destroyInstance(HANDLE h);
/// Feed 'numSamples' sample frames from 'samples' into the BPM detector.
SOUNDTOUCHDLL_API void __cdecl bpm_putSamples(HANDLE h,
const float *samples, ///< Pointer to sample buffer.
unsigned int numSamples ///< Number of samples in buffer. Notice
///< that in case of stereo-sound a single sample
///< contains data for both channels.
);
/// Feed 'numSamples' sample frames from 'samples' into the BPM detector.
/// 16bit int sample format version.
SOUNDTOUCHDLL_API void __cdecl bpm_putSamples_i16(HANDLE h,
const short *samples, ///< Pointer to sample buffer.
unsigned int numSamples ///< Number of samples in buffer. Notice
///< that in case of stereo-sound a single sample
///< contains data for both channels.
);
/// Analyzes the results and returns the BPM rate. Use this function to read result
/// after whole song data has been input to the class by consecutive calls of
/// 'inputSamples' function.
///
/// \return Beats-per-minute rate, or zero if detection failed.
SOUNDTOUCHDLL_API float __cdecl bpm_getBpm(HANDLE h);
/// Get beat position arrays. Note: The array includes also really low beat detection values
/// in absence of clear strong beats. Consumer may wish to filter low values away.
/// - "pos" receive array of beat positions
/// - "values" receive array of beat detection strengths
/// - max_num indicates max.size of "pos" and "values" array.
///
/// You can query a suitable array sized by calling this with nullptr in "pos" & "values".
///
/// \return number of beats in the arrays.
SOUNDTOUCHDLL_API int __cdecl bpm_getBeats(HANDLE h, float *pos, float *strength, int count);
#endif // _SoundTouchDLL_h_

6
source/SoundTouchDLL/SoundTouchDLL.rc
View File

@ -25,18 +25,18 @@ LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
// TEXTINCLUDE
//
1 TEXTINCLUDE
1 TEXTINCLUDE
BEGIN
"resource.h\0"
END
2 TEXTINCLUDE
2 TEXTINCLUDE
BEGIN
"#include ""afxres.h""\r\n"
"\0"
END
3 TEXTINCLUDE
3 TEXTINCLUDE
BEGIN
"\r\n"
"\0"

30
source/SoundTouchDLL/resource.h
View File

@ -1,15 +1,15 @@
//{{NO_DEPENDENCIES}}
// Microsoft Visual C++ generated include file.
// Used by SoundTouchDLL.rc
//
// Next default values for new objects
//
#ifdef APSTUDIO_INVOKED
#ifndef APSTUDIO_READONLY_SYMBOLS
#define _APS_NEXT_RESOURCE_VALUE 101
#define _APS_NEXT_COMMAND_VALUE 40001
#define _APS_NEXT_CONTROL_VALUE 1000
#define _APS_NEXT_SYMED_VALUE 101
#endif
#endif
//{{NO_DEPENDENCIES}}
// Microsoft Visual C++ generated include file.
// Used by SoundTouchDLL.rc
//
// Next default values for new objects
//
#ifdef APSTUDIO_INVOKED
#ifndef APSTUDIO_READONLY_SYMBOLS
#define _APS_NEXT_RESOURCE_VALUE 101
#define _APS_NEXT_COMMAND_VALUE 40001
#define _APS_NEXT_CONTROL_VALUE 1000
#define _APS_NEXT_SYMED_VALUE 101
#endif
#endif