/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */ /* Rubber Band An audio time-stretching and pitch-shifting library. Copyright 2007 Chris Cannam. This program 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. See the file COPYING included with this distribution for more information. */ #include "FFT.h" #include "Thread.h" #include #include #include #include #include #include namespace RubberBand { class FFTImpl { public: virtual ~FFTImpl() { } virtual void initFloat() = 0; virtual void initDouble() = 0; virtual void forward(double *realIn, double *realOut, double *imagOut) = 0; virtual void forwardPolar(double *realIn, double *magOut, double *phaseOut) = 0; virtual void forwardMagnitude(double *realIn, double *magOut) = 0; virtual void forward(float *realIn, float *realOut, float *imagOut) = 0; virtual void forwardPolar(float *realIn, float *magOut, float *phaseOut) = 0; virtual void forwardMagnitude(float *realIn, float *magOut) = 0; virtual void inverse(double *realIn, double *imagIn, double *realOut) = 0; virtual void inversePolar(double *magIn, double *phaseIn, double *realOut) = 0; virtual void inverse(float *realIn, float *imagIn, float *realOut) = 0; virtual void inversePolar(float *magIn, float *phaseIn, float *realOut) = 0; virtual float *getFloatTimeBuffer() = 0; virtual double *getDoubleTimeBuffer() = 0; }; // Define FFTW_DOUBLE_ONLY to make all uses of FFTW functions be // double-precision (so "float" FFTs are calculated by casting to // doubles and using the double-precision FFTW function). // // Define FFTW_FLOAT_ONLY to make all uses of FFTW functions be // single-precision (so "double" FFTs are calculated by casting to // floats and using the single-precision FFTW function). // // Neither of these flags is terribly desirable -- FFTW_FLOAT_ONLY // obviously loses you precision, and neither is handled in the most // efficient way so any performance improvement will be small at best. // The only real reason to define either flag would be to avoid // linking against both fftw3 and fftw3f libraries. //#define FFTW_DOUBLE_ONLY 1 //#define FFTW_FLOAT_ONLY 1 #ifdef FFTW_DOUBLE_ONLY #ifdef FFTW_FLOAT_ONLY #error Building for FFTW-DOUBLE BOTH // Can't meaningfully define both #undef FFTW_DOUBLE_ONLY #undef FFTW_FLOAT_ONLY #else /* !FFTW_FLOAT_ONLY */ #define fftwf_complex fftw_complex #define fftwf_plan fftw_plan #define fftwf_plan_dft_r2c_1d fftw_plan_dft_r2c_1d #define fftwf_plan_dft_c2r_1d fftw_plan_dft_c2r_1d #define fftwf_destroy_plan fftw_destroy_plan #define fftwf_malloc fftw_malloc #define fftwf_free fftw_free #define fftwf_execute fftw_execute #define atan2f atan2 #define sqrtf sqrt #define cosf cos #define sinf sin #endif /* !FFTW_FLOAT_ONLY */ #endif #ifdef FFTW_FLOAT_ONLY #define fftw_complex fftwf_complex #define fftw_plan fftwf_plan #define fftw_plan_dft_r2c_1d fftwf_plan_dft_r2c_1d #define fftw_plan_dft_c2r_1d fftwf_plan_dft_c2r_1d #define fftw_destroy_plan fftwf_destroy_plan #define fftw_malloc fftwf_malloc #define fftw_free fftwf_free #define fftw_execute fftwf_execute #define atan2 atan2f #define sqrt sqrtf #define cos cosf #define sif sinf #endif /* FFTW_FLOAT_ONLY */ class D_FFTW : public FFTImpl { public: D_FFTW(unsigned int size) : m_fplanf(0) #ifdef FFTW_DOUBLE_ONLY , m_frb(0) #endif , m_dplanf(0) #ifdef FFTW_FLOAT_ONLY , m_drb(0) #endif , m_size(size) { } ~D_FFTW() { if (m_fplanf) { bool save = false; m_extantMutex.lock(); if (m_extantf > 0 && --m_extantf == 0) save = true; m_extantMutex.unlock(); if (save) saveWisdom('f'); fftwf_destroy_plan(m_fplanf); fftwf_destroy_plan(m_fplani); fftwf_free(m_fbuf); fftwf_free(m_fpacked); #ifdef FFTW_DOUBLE_ONLY if (m_frb) fftw_free(m_frb); #endif } if (m_dplanf) { bool save = false; m_extantMutex.lock(); if (m_extantd > 0 && --m_extantd == 0) save = true; m_extantMutex.unlock(); if (save) saveWisdom('d'); fftw_destroy_plan(m_dplanf); fftw_destroy_plan(m_dplani); fftw_free(m_dbuf); fftw_free(m_dpacked); #ifdef FFTW_FLOAT_ONLY if (m_drb) fftwf_free(m_drb); #endif } } void initFloat() { if (m_fplanf) return; bool load = false; m_extantMutex.lock(); if (m_extantf++ == 0) load = true; m_extantMutex.unlock(); #ifdef FFTW_DOUBLE_ONLY if (load) loadWisdom('d'); m_fbuf = (double *)fftw_malloc(m_size * sizeof(double)); #else if (load) loadWisdom('f'); m_fbuf = (float *)fftwf_malloc(m_size * sizeof(float)); #endif m_fpacked = (fftwf_complex *)fftw_malloc ((m_size/2 + 1) * sizeof(fftwf_complex)); m_fplanf = fftwf_plan_dft_r2c_1d (m_size, m_fbuf, m_fpacked, FFTW_MEASURE); m_fplani = fftwf_plan_dft_c2r_1d (m_size, m_fpacked, m_fbuf, FFTW_MEASURE); } void initDouble() { if (m_dplanf) return; bool load = false; m_extantMutex.lock(); if (m_extantd++ == 0) load = true; m_extantMutex.unlock(); #ifdef FFTW_FLOAT_ONLY if (load) loadWisdom('f'); m_dbuf = (float *)fftwf_malloc(m_size * sizeof(float)); #else if (load) loadWisdom('d'); m_dbuf = (double *)fftw_malloc(m_size * sizeof(double)); #endif m_dpacked = (fftw_complex *)fftw_malloc ((m_size/2 + 1) * sizeof(fftw_complex)); m_dplanf = fftw_plan_dft_r2c_1d (m_size, m_dbuf, m_dpacked, FFTW_MEASURE); m_dplani = fftw_plan_dft_c2r_1d (m_size, m_dpacked, m_dbuf, FFTW_MEASURE); } void loadWisdom(char type) { wisdom(false, type); } void saveWisdom(char type) { wisdom(true, type); } void wisdom(bool save, char type) { #ifdef FFTW_DOUBLE_ONLY if (type == 'f') return; #endif #ifdef FFTW_FLOAT_ONLY if (type == 'd') return; #endif const char *home = getenv("HOME"); if (!home) return; char fn[256]; snprintf(fn, 256, "%s/%s.%c", home, ".rubberband.wisdom", type); FILE *f = fopen(fn, save ? "wb" : "rb"); if (!f) return; if (save) { switch (type) { #ifdef FFTW_DOUBLE_ONLY case 'f': break; #else case 'f': fftwf_export_wisdom_to_file(f); break; #endif #ifdef FFTW_FLOAT_ONLY case 'd': break; #else case 'd': fftw_export_wisdom_to_file(f); break; #endif default: break; } } else { switch (type) { #ifdef FFTW_DOUBLE_ONLY case 'f': break; #else case 'f': fftwf_import_wisdom_from_file(f); break; #endif #ifdef FFTW_FLOAT_ONLY case 'd': break; #else case 'd': fftw_import_wisdom_from_file(f); break; #endif default: break; } } fclose(f); } void packFloat(float *re, float *im) { for (unsigned int i = 0; i <= m_size/2; ++i) { m_fpacked[i][0] = re[i]; m_fpacked[i][1] = im[i]; } } void packDouble(double *re, double *im) { for (unsigned int i = 0; i <= m_size/2; ++i) { m_dpacked[i][0] = re[i]; m_dpacked[i][1] = im[i]; } } void unpackFloat(float *re, float *im) { for (unsigned int i = 0; i <= m_size/2; ++i) { re[i] = m_fpacked[i][0]; im[i] = m_fpacked[i][1]; } } void unpackDouble(double *re, double *im) { for (unsigned int i = 0; i <= m_size/2; ++i) { re[i] = m_dpacked[i][0]; im[i] = m_dpacked[i][1]; } } void forward(double *realIn, double *realOut, double *imagOut) { if (!m_dplanf) initDouble(); #ifndef FFTW_FLOAT_ONLY if (realIn != m_dbuf) #endif for (unsigned int i = 0; i < m_size; ++i) { m_dbuf[i] = realIn[i]; } fftw_execute(m_dplanf); unpackDouble(realOut, imagOut); } void forwardPolar(double *realIn, double *magOut, double *phaseOut) { if (!m_dplanf) initDouble(); #ifndef FFTW_FLOAT_ONLY if (realIn != m_dbuf) #endif for (unsigned int i = 0; i < m_size; ++i) { m_dbuf[i] = realIn[i]; } fftw_execute(m_dplanf); for (unsigned int i = 0; i <= m_size/2; ++i) { magOut[i] = sqrt(m_dpacked[i][0] * m_dpacked[i][0] + m_dpacked[i][1] * m_dpacked[i][1]); } for (unsigned int i = 0; i <= m_size/2; ++i) { phaseOut[i] = atan2(m_dpacked[i][1], m_dpacked[i][0]); } } void forwardMagnitude(double *realIn, double *magOut) { if (!m_dplanf) initDouble(); #ifndef FFTW_FLOAT_ONLY if (realIn != m_dbuf) #endif for (unsigned int i = 0; i < m_size; ++i) { m_dbuf[i] = realIn[i]; } fftw_execute(m_dplanf); for (unsigned int i = 0; i <= m_size/2; ++i) { magOut[i] = sqrt(m_dpacked[i][0] * m_dpacked[i][0] + m_dpacked[i][1] * m_dpacked[i][1]); } } void forward(float *realIn, float *realOut, float *imagOut) { if (!m_fplanf) initFloat(); #ifndef FFTW_DOUBLE_ONLY if (realIn != m_fbuf) #endif for (unsigned int i = 0; i < m_size; ++i) { m_fbuf[i] = realIn[i]; } fftwf_execute(m_fplanf); unpackFloat(realOut, imagOut); } void forwardPolar(float *realIn, float *magOut, float *phaseOut) { if (!m_fplanf) initFloat(); #ifndef FFTW_DOUBLE_ONLY if (realIn != m_fbuf) #endif for (unsigned int i = 0; i < m_size; ++i) { m_fbuf[i] = realIn[i]; } fftwf_execute(m_fplanf); for (unsigned int i = 0; i <= m_size/2; ++i) { magOut[i] = sqrtf(m_fpacked[i][0] * m_fpacked[i][0] + m_fpacked[i][1] * m_fpacked[i][1]); } for (unsigned int i = 0; i <= m_size/2; ++i) { phaseOut[i] = atan2f(m_fpacked[i][1], m_fpacked[i][0]) ; } } void forwardMagnitude(float *realIn, float *magOut) { if (!m_fplanf) initFloat(); #ifndef FFTW_DOUBLE_ONLY if (realIn != m_fbuf) #endif for (unsigned int i = 0; i < m_size; ++i) { m_fbuf[i] = realIn[i]; } fftwf_execute(m_fplanf); for (unsigned int i = 0; i <= m_size/2; ++i) { magOut[i] = sqrtf(m_fpacked[i][0] * m_fpacked[i][0] + m_fpacked[i][1] * m_fpacked[i][1]); } } void inverse(double *realIn, double *imagIn, double *realOut) { if (!m_dplanf) initDouble(); packDouble(realIn, imagIn); fftw_execute(m_dplani); #ifndef FFTW_FLOAT_ONLY if (realOut != m_dbuf) #endif for (unsigned int i = 0; i < m_size; ++i) { realOut[i] = m_dbuf[i]; } } void inversePolar(double *magIn, double *phaseIn, double *realOut) { if (!m_dplanf) initDouble(); for (unsigned int i = 0; i <= m_size/2; ++i) { m_dpacked[i][0] = magIn[i] * cos(phaseIn[i]); m_dpacked[i][1] = magIn[i] * sin(phaseIn[i]); } fftw_execute(m_dplani); #ifndef FFTW_FLOAT_ONLY if (realOut != m_dbuf) #endif for (unsigned int i = 0; i < m_size; ++i) { realOut[i] = m_dbuf[i]; } } void inverse(float *realIn, float *imagIn, float *realOut) { if (!m_fplanf) initFloat(); packFloat(realIn, imagIn); fftwf_execute(m_fplani); #ifndef FFTW_DOUBLE_ONLY if (realOut != m_fbuf) #endif for (unsigned int i = 0; i < m_size; ++i) { realOut[i] = m_fbuf[i]; } } void inversePolar(float *magIn, float *phaseIn, float *realOut) { if (!m_fplanf) initFloat(); for (unsigned int i = 0; i <= m_size/2; ++i) { m_fpacked[i][0] = magIn[i] * cosf(phaseIn[i]); m_fpacked[i][1] = magIn[i] * sinf(phaseIn[i]); } fftwf_execute(m_fplani); #ifndef FFTW_DOUBLE_ONLY if (realOut != m_fbuf) #endif for (unsigned int i = 0; i < m_size; ++i) { realOut[i] = m_fbuf[i]; } } float *getFloatTimeBuffer() { initFloat(); #ifdef FFTW_DOUBLE_ONLY if (!m_frb) m_frb = (float *)fftw_malloc(m_size * sizeof(float)); return m_frb; #else return m_fbuf; #endif } double *getDoubleTimeBuffer() { initDouble(); #ifdef FFTW_FLOAT_ONLY if (!m_drb) m_drb = (double *)fftwf_malloc(m_size * sizeof(double)); return m_drb; #else return m_dbuf; #endif } private: fftwf_plan m_fplanf; fftwf_plan m_fplani; #ifdef FFTW_DOUBLE_ONLY float *m_frb; double *m_fbuf; #else float *m_fbuf; #endif fftwf_complex *m_fpacked; fftw_plan m_dplanf; fftw_plan m_dplani; #ifdef FFTW_FLOAT_ONLY float *m_dbuf; double *m_drb; #else double *m_dbuf; #endif fftw_complex *m_dpacked; unsigned int m_size; static unsigned int m_extantf; static unsigned int m_extantd; static Mutex m_extantMutex; }; unsigned int D_FFTW::m_extantf = 0; unsigned int D_FFTW::m_extantd = 0; Mutex D_FFTW::m_extantMutex; class D_Cross : public FFTImpl { public: D_Cross(unsigned int size) : m_size(size), m_table(0), m_frb(0), m_drb(0) { m_a = new double[size]; m_b = new double[size]; m_c = new double[size]; m_d = new double[size]; m_table = new int[m_size]; unsigned int bits; unsigned int i, j, k, m; for (i = 0; ; ++i) { if (m_size & (1 << i)) { bits = i; break; } } for (i = 0; i < m_size; ++i) { m = i; for (j = k = 0; j < bits; ++j) { k = (k << 1) | (m & 1); m >>= 1; } m_table[i] = k; } } ~D_Cross() { delete[] m_table; delete[] m_a; delete[] m_b; delete[] m_c; delete[] m_d; delete[] m_frb; delete[] m_drb; } void initFloat() { } void initDouble() { } void forward(double *realIn, double *realOut, double *imagOut) { basefft(false, realIn, 0, m_c, m_d); for (size_t i = 0; i <= m_size/2; ++i) realOut[i] = m_c[i]; for (size_t i = 0; i <= m_size/2; ++i) imagOut[i] = m_d[i]; } void forwardPolar(double *realIn, double *magOut, double *phaseOut) { basefft(false, realIn, 0, m_c, m_d); for (unsigned int i = 0; i <= m_size/2; ++i) { magOut[i] = sqrt(m_c[i] * m_c[i] + m_d[i] * m_d[i]); phaseOut[i] = atan2(m_d[i], m_c[i]) ; } } void forwardMagnitude(double *realIn, double *magOut) { basefft(false, realIn, 0, m_c, m_d); for (unsigned int i = 0; i <= m_size/2; ++i) { magOut[i] = sqrt(m_c[i] * m_c[i] + m_d[i] * m_d[i]); } } void forward(float *realIn, float *realOut, float *imagOut) { for (size_t i = 0; i < m_size; ++i) m_a[i] = realIn[i]; basefft(false, m_a, 0, m_c, m_d); for (size_t i = 0; i <= m_size/2; ++i) realOut[i] = m_c[i]; for (size_t i = 0; i <= m_size/2; ++i) imagOut[i] = m_d[i]; } void forwardPolar(float *realIn, float *magOut, float *phaseOut) { for (size_t i = 0; i < m_size; ++i) m_a[i] = realIn[i]; basefft(false, m_a, 0, m_c, m_d); for (unsigned int i = 0; i <= m_size/2; ++i) { magOut[i] = sqrt(m_c[i] * m_c[i] + m_d[i] * m_d[i]); phaseOut[i] = atan2(m_d[i], m_c[i]) ; } } void forwardMagnitude(float *realIn, float *magOut) { for (size_t i = 0; i < m_size; ++i) m_a[i] = realIn[i]; basefft(false, m_a, 0, m_c, m_d); for (unsigned int i = 0; i <= m_size/2; ++i) { magOut[i] = sqrt(m_c[i] * m_c[i] + m_d[i] * m_d[i]); } } void inverse(double *realIn, double *imagIn, double *realOut) { for (unsigned int i = 0; i <= m_size/2; ++i) { double real = realIn[i]; double imag = imagIn[i]; m_a[i] = real; m_b[i] = imag; if (i > 0) { m_a[m_size-i] = real; m_b[m_size-i] = -imag; } } basefft(true, m_a, m_b, realOut, m_d); } void inversePolar(double *magIn, double *phaseIn, double *realOut) { for (unsigned int i = 0; i <= m_size/2; ++i) { double real = magIn[i] * cos(phaseIn[i]); double imag = magIn[i] * sin(phaseIn[i]); m_a[i] = real; m_b[i] = imag; if (i > 0) { m_a[m_size-i] = real; m_b[m_size-i] = -imag; } } basefft(true, m_a, m_b, realOut, m_d); } void inverse(float *realIn, float *imagIn, float *realOut) { for (unsigned int i = 0; i <= m_size/2; ++i) { float real = realIn[i]; float imag = imagIn[i]; m_a[i] = real; m_b[i] = imag; if (i > 0) { m_a[m_size-i] = real; m_b[m_size-i] = -imag; } } basefft(true, m_a, m_b, m_c, m_d); for (unsigned int i = 0; i < m_size; ++i) realOut[i] = m_c[i]; } void inversePolar(float *magIn, float *phaseIn, float *realOut) { for (unsigned int i = 0; i <= m_size/2; ++i) { float real = magIn[i] * cosf(phaseIn[i]); float imag = magIn[i] * sinf(phaseIn[i]); m_a[i] = real; m_b[i] = imag; if (i > 0) { m_a[m_size-i] = real; m_b[m_size-i] = -imag; } } basefft(true, m_a, m_b, m_c, m_d); for (unsigned int i = 0; i < m_size; ++i) realOut[i] = m_c[i]; } float *getFloatTimeBuffer() { if (!m_frb) m_frb = new float[m_size]; return m_frb; } double *getDoubleTimeBuffer() { if (!m_drb) m_drb = new double[m_size]; return m_drb; } private: unsigned int m_size; int *m_table; float *m_frb; double *m_drb; double *m_a; double *m_b; double *m_c; double *m_d; void basefft(bool inverse, double *ri, double *ii, double *ro, double *io); }; void D_Cross::basefft(bool inverse, double *ri, double *ii, double *ro, double *io) { if (!ri || !ro || !io) return; unsigned int i, j, k, m; unsigned int blockSize, blockEnd; double tr, ti; double angle = 2.0 * M_PI; if (inverse) angle = -angle; const unsigned int n = m_size; if (ii) { for (i = 0; i < n; ++i) { ro[m_table[i]] = ri[i]; io[m_table[i]] = ii[i]; } } else { for (i = 0; i < n; ++i) { ro[m_table[i]] = ri[i]; io[m_table[i]] = 0.0; } } blockEnd = 1; for (blockSize = 2; blockSize <= n; blockSize <<= 1) { double delta = angle / (double)blockSize; double sm2 = -sin(-2 * delta); double sm1 = -sin(-delta); double cm2 = cos(-2 * delta); double cm1 = cos(-delta); double w = 2 * cm1; double ar[3], ai[3]; for (i = 0; i < n; i += blockSize) { ar[2] = cm2; ar[1] = cm1; ai[2] = sm2; ai[1] = sm1; for (j = i, m = 0; m < blockEnd; j++, m++) { ar[0] = w * ar[1] - ar[2]; ar[2] = ar[1]; ar[1] = ar[0]; ai[0] = w * ai[1] - ai[2]; ai[2] = ai[1]; ai[1] = ai[0]; k = j + blockEnd; tr = ar[0] * ro[k] - ai[0] * io[k]; ti = ar[0] * io[k] + ai[0] * ro[k]; ro[k] = ro[j] - tr; io[k] = io[j] - ti; ro[j] += tr; io[j] += ti; } } blockEnd = blockSize; } /* fftw doesn't rescale, so nor will we if (inverse) { double denom = (double)n; for (i = 0; i < n; i++) { ro[i] /= denom; io[i] /= denom; } } */ } int FFT::m_method = -1; FFT::FFT(unsigned int size) { if (size < 2) throw InvalidSize; if (size & (size-1)) throw InvalidSize; if (m_method == -1) { m_method = 1; } switch (m_method) { case 0: d = new D_Cross(size); break; case 1: // std::cerr << "FFT::FFT(" << size << "): using FFTW3 implementation" // << std::endl; d = new D_FFTW(size); break; default: std::cerr << "FFT::FFT(" << size << "): WARNING: using slow built-in implementation" << std::endl; d = new D_Cross(size); break; } } FFT::~FFT() { delete d; } void FFT::forward(double *realIn, double *realOut, double *imagOut) { d->forward(realIn, realOut, imagOut); } void FFT::forwardPolar(double *realIn, double *magOut, double *phaseOut) { d->forwardPolar(realIn, magOut, phaseOut); } void FFT::forwardMagnitude(double *realIn, double *magOut) { d->forwardMagnitude(realIn, magOut); } void FFT::forward(float *realIn, float *realOut, float *imagOut) { d->forward(realIn, realOut, imagOut); } void FFT::forwardPolar(float *realIn, float *magOut, float *phaseOut) { d->forwardPolar(realIn, magOut, phaseOut); } void FFT::forwardMagnitude(float *realIn, float *magOut) { d->forwardMagnitude(realIn, magOut); } void FFT::inverse(double *realIn, double *imagIn, double *realOut) { d->inverse(realIn, imagIn, realOut); } void FFT::inversePolar(double *magIn, double *phaseIn, double *realOut) { d->inversePolar(magIn, phaseIn, realOut); } void FFT::inverse(float *realIn, float *imagIn, float *realOut) { d->inverse(realIn, imagIn, realOut); } void FFT::inversePolar(float *magIn, float *phaseIn, float *realOut) { d->inversePolar(magIn, phaseIn, realOut); } void FFT::initFloat() { d->initFloat(); } void FFT::initDouble() { d->initDouble(); } float * FFT::getFloatTimeBuffer() { return d->getFloatTimeBuffer(); } double * FFT::getDoubleTimeBuffer() { return d->getDoubleTimeBuffer(); } void FFT::tune() { } }