Skip to content

Commit

Permalink
Fix segfault during start
Browse files Browse the repository at this point in the history
  • Loading branch information
@AlwinEsch
AlwinEsch committed Oct 14, 2015
1 parent 942882a commit cfe0f9b
Show file tree
Hide file tree
Showing 3 changed files with 33 additions and 34 deletions.
4 changes: 1 addition & 3 deletions src/DSPProcessFreeSurround.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -30,8 +30,6 @@
#include "addon.h"
#include "FreeSurroundSettings.h"

#define FFT_LENGTH 1024

// our default internal block size, in floats
static const unsigned default_block_size = SURROUND_BUFSIZE;
unsigned int block_size = default_block_size;
Expand Down Expand Up @@ -176,7 +174,7 @@ unsigned int CDSPProcess_FreeSurround::StreamProcess(float **array_in, float **a
if (!m_Decoder)
return samples;

float **outputs = m_Decoder->getOutputBuffers();
float **outputs = m_Decoder->getOutputBuffers();

for (unsigned int pos = 0; pos < samples; ++pos)
{
Expand Down
44 changes: 23 additions & 21 deletions src/FreeSurroundDecoder.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -33,9 +33,8 @@

#include "FreeSurroundDecoder.h"

const float pi = 3.141592654f;
const float epsilon = 0.000001f;
using namespace std;
const double pi = 3.14159265358979323846;
const double epsilon = 0.000001;

#ifndef TARGET_WINDOWS
#undef min
Expand Down Expand Up @@ -69,9 +68,9 @@ CFreeSurroundDecoder::CFreeSurroundDecoder(channel_setup setup, unsigned blocksi
for (unsigned k=0; k < m_BufferSize; ++k)
m_Window[k] = sqrt(0.5 * (1 - cos(2 * pi * k / m_BufferSize)) / m_BufferSize);

m_Destination = (double*)calloc(m_BufferSize, sizeof(double));
m_TotalLeft = (double*)calloc(m_BufferSize, sizeof(double));
m_TotalRight = (double*)calloc(m_BufferSize, sizeof(double));
m_Destination = (float*)calloc(m_BufferSize, sizeof(float));
m_TotalLeft = (float*)calloc(m_BufferSize, sizeof(float));
m_TotalRight = (float*)calloc(m_BufferSize, sizeof(float));
m_FreqLeft = (cplx*)calloc(m_BufferSize / 2 + 1, sizeof(cplx));
m_FreqRight = (cplx*)calloc(m_BufferSize / 2 + 1, sizeof(cplx));

Expand Down Expand Up @@ -116,12 +115,15 @@ float **CFreeSurroundDecoder::getOutputBuffers()

void CFreeSurroundDecoder::Decode(float **inputs)
{
//! append incoming data to the end of the input buffer
memcpy(&m_InputBuffer[0][m_BufferSizeHalf], &inputs[0][0], m_BufferSize*sizeof(float));
memcpy(&m_InputBuffer[1][m_BufferSizeHalf], &inputs[1][0], m_BufferSize*sizeof(float));

//! process first and second half, overlapped
BufferedDecode(m_InputBuffer, 0);
BufferedDecode(m_InputBuffer, m_BufferSizeHalf);

//! shift last half of the input to the beginning (for overlapping with a future block)
memcpy(&m_InputBuffer[0][0], &m_InputBuffer[0][m_BufferSize], m_BufferSizeHalf*sizeof(float));
memcpy(&m_InputBuffer[1][0], &m_InputBuffer[1][m_BufferSize], m_BufferSizeHalf*sizeof(float));

Expand All @@ -136,7 +138,7 @@ void CFreeSurroundDecoder::Flush()
m_BufferEmpty = true;
}

inline float CFreeSurroundDecoder::sqr(double x)
inline double CFreeSurroundDecoder::sqr(double x)
{
return x * x;
}
Expand All @@ -156,29 +158,29 @@ inline cplx CFreeSurroundDecoder::polar(double a, double p)
return cplx(a * cos(p), a * sin(p));
}

inline float CFreeSurroundDecoder::Min(double a, double b)
inline double CFreeSurroundDecoder::Min(double a, double b)
{
return a < b ? a : b;
}

inline float CFreeSurroundDecoder::Max(double a, double b)
inline double CFreeSurroundDecoder::Max(double a, double b)
{
return a > b ? a : b;
}

inline float CFreeSurroundDecoder::clamp(double x)
inline double CFreeSurroundDecoder::clamp(double x)
{
return Max(-1, Min(1, x));
}

inline float CFreeSurroundDecoder::sign(double x)
inline double CFreeSurroundDecoder::sign(double x)
{
return x < 0 ? -1 : (x > 0 ? 1 : 0);
}

inline double CFreeSurroundDecoder::EdgeDistance(double a)
{
return min(sqrt(1 + sqr(tan(a))), sqrt(1 + sqr(1 / tan(a))));
return Min(sqrt(1 + sqr(tan(a))), sqrt(1 + sqr(1 / tan(a))));
}

void CFreeSurroundDecoder::BufferedDecode(float *input[2], int ptr)
Expand All @@ -191,11 +193,11 @@ void CFreeSurroundDecoder::BufferedDecode(float *input[2], int ptr)
}

//! map into spectral domain
kiss_fftr(m_FFTForward, &m_TotalLeft[0], (kiss_fft_cpx*)&m_FreqLeft[0]);
kiss_fftr(m_FFTForward, &m_TotalRight[0], (kiss_fft_cpx*)&m_FreqRight[0]);
kiss_fftr(m_FFTForward, m_TotalLeft, (kiss_fft_cpx*)m_FreqLeft);
kiss_fftr(m_FFTForward, m_TotalRight, (kiss_fft_cpx*)m_FreqRight);

// compute multichannel output signal in the spectral domain
for (unsigned f = 1; f < m_BufferSizeHalf; ++f)
//! compute multichannel output signal in the spectral domain
for (unsigned int f = 1; f < m_BufferSizeHalf; ++f)
{
//! get Lt/Rt amplitudes & phases
double ampL = amplitude(m_FreqLeft[f]);
Expand Down Expand Up @@ -232,7 +234,7 @@ void CFreeSurroundDecoder::BufferedDecode(float *input[2], int ptr)
for (unsigned c = 0; c < m_Channels-1; ++c)
{
//! look up channel map at respective position (with bilinear interpolation) and build the signal
vector<float*> &a = chn_alloc[m_ChannelSetup][c];
std::vector<float*> &a = chn_alloc[m_ChannelSetup][c];

m_Signal[c][f] = polar(amp_total*((1-x)*(1-y)*a[q][p] + x*(1-y)*a[q][p+1] + (1-x)*y*a[q+1][p] + x*y*a[q+1][p+1]), phase_of[1+(int)sign(chn_xsf[m_ChannelSetup][c])]);
}
Expand All @@ -241,7 +243,7 @@ void CFreeSurroundDecoder::BufferedDecode(float *input[2], int ptr)
if (m_UseLFE && f < m_HighCutoff)
{
//! level of LFE channel according to normalized frequency
double lfe_level = f < m_LowCutoff ? 1 : 0.5*(1+cos(pi*(f-m_LowCutoff)/(m_HighCutoff-m_LowCutoff)));
float lfe_level = f < m_LowCutoff ? 1 : 0.5*(1+cos(pi*(f-m_LowCutoff)/(m_HighCutoff-m_LowCutoff)));
//! assign LFE channel
m_Signal[m_Channels-1][f] = lfe_level * polar(amp_total,phase_of[1]);
//! subtract the signal from the other channels
Expand All @@ -253,11 +255,11 @@ void CFreeSurroundDecoder::BufferedDecode(float *input[2], int ptr)
for (unsigned channel = 0; channel < m_Channels; ++channel)
{
//! shift the last 2/3 to the first 2/3 of the output buffer
memcpy(&m_OutputBuffer[channel][0], &m_OutputBuffer[channel][m_BufferSizeHalf], m_BufferSize*sizeof(float));
memcpy(&m_OutputBuffer[channel][0], &m_OutputBuffer[channel][m_BufferSizeHalf], m_BufferSize*sizeof(double));
//! and clear the rest
memset(&m_OutputBuffer[channel][m_BufferSize], 0, m_BufferSizeHalf*sizeof(float));
memset(&m_OutputBuffer[channel][m_BufferSize], 0, m_BufferSizeHalf*sizeof(double));

kiss_fftri(m_FFTInverse,(kiss_fft_cpx*)&m_Signal[channel][0], &m_Destination[0]);
kiss_fftri(m_FFTInverse, (kiss_fft_cpx*)m_Signal[channel], &m_Destination[0]);

//! add the result to the last 2/3 of the output buffer, windowed (and remultiplex)
for (unsigned k = 0; k < m_BufferSize; ++k)
Expand Down
19 changes: 9 additions & 10 deletions src/FreeSurroundDecoder.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,6 @@

#include <complex>

#define kiss_fft_scalar double //!< we're using doubles here...
#include "kissfft/kiss_fftr.h"
#include "ChannelMaps.h"

Expand Down Expand Up @@ -66,7 +65,7 @@ typedef enum {
cs_legacy = 0 // same channels as cs_5point1 but different upmixing transform; does not support the focus control
} channel_setup;

typedef std::complex<double> cplx;
typedef std::complex<kiss_fft_scalar> cplx;

/**
* The FreeSurround decoder.
Expand Down Expand Up @@ -221,9 +220,9 @@ class CFreeSurroundDecoder
bool m_UseLFE; //!< whether to use the LFE channel

//! FFT data structures
double *m_Destination;
double *m_TotalLeft; //!< left total, (source arrays), time-domain destination buffer array
double *m_TotalRight; //!< right total, "
float *m_Destination;
float *m_TotalLeft; //!< left total, (source arrays), time-domain destination buffer array
float *m_TotalRight; //!< right total, "
cplx *m_FreqLeft; //!< left total in frequency domain
cplx *m_FreqRight; //!< right total in frequency domain
kiss_fftr_cfg m_FFTForward; //!< FFT buffers
Expand All @@ -237,14 +236,14 @@ class CFreeSurroundDecoder
double *m_Window; //!< the window function, precomputed

//! helper functions
static inline float sqr(double x);
static inline double sqr(double x);

This comment has been minimized.

Sign in to view

Copy link
@AchimTuran

AchimTuran Oct 15, 2015

Member

I'm not familiar enough with the complete sources. Is it really necessary to use doubles for audio processing?
ActiveAE only uses floats.
static inline double amplitude(const cplx &x);
static inline double phase(const cplx &x);
static inline cplx polar(double a, double p);
static inline float Min(double a, double b);
static inline float Max(double a, double b);
static inline float clamp(double x);
static inline float sign(double x);
static inline double Min(double a, double b);
static inline double Max(double a, double b);
static inline double clamp(double x);
static inline double sign(double x);

//!> get the distance of the soundfield edge, along a given angle
static inline double EdgeDistance(double a);
Expand Down

1 comment on commit cfe0f9b

@AlwinEsch
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not direct but during the Fourier process I like to use double, has seen in the past that the quality after change them back to normal is a bit better (for me audible on headset) , there are on some time value changes far a away from the comma (,) which was on normal then a bigger value.

The other for the change there is to prevent value replaces between double and float which can create a bigger machine code.

Please sign in to comment.