mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-11-14 05:12:09 +00:00
cb90622b41
Original commit message from CVS: Remove all config.h includes from header files, add it to each source file and remove duplicate config.h includes from several source files
105 lines
2.5 KiB
C++
105 lines
2.5 KiB
C++
/*
|
|
* This source code is public domain.
|
|
*
|
|
* Authors: Olivier Lapicque <olivierl@jps.net>
|
|
*/
|
|
|
|
#ifdef HAVE_CONFIG_H
|
|
#include "config.h"
|
|
#endif
|
|
|
|
#include "stdafx.h"
|
|
#include "sndfile.h"
|
|
|
|
// AWE32: cutoff = reg[0-255] * 31.25 + 100 -> [100Hz-8060Hz]
|
|
// EMU10K1 docs: cutoff = reg[0-127]*62+100
|
|
#define FILTER_PRECISION 8192
|
|
|
|
#ifndef NO_FILTER
|
|
|
|
#ifdef MSC_VER
|
|
#define _ASM_MATH
|
|
#endif
|
|
|
|
#ifdef _ASM_MATH
|
|
|
|
// pow(a,b) returns a^^b -> 2^^(b.log2(a))
|
|
static float pow(float a, float b)
|
|
{
|
|
long tmpint;
|
|
float result;
|
|
_asm {
|
|
fld b // Load b
|
|
fld a // Load a
|
|
fyl2x // ST(0) = b.log2(a)
|
|
fist tmpint // Store integer exponent
|
|
fisub tmpint // ST(0) = -1 <= (b*log2(a)) <= 1
|
|
f2xm1 // ST(0) = 2^(x)-1
|
|
fild tmpint // load integer exponent
|
|
fld1 // Load 1
|
|
fscale // ST(0) = 2^ST(1)
|
|
fstp ST(1) // Remove the integer from the stack
|
|
fmul ST(1), ST(0) // multiply with fractional part
|
|
faddp ST(1), ST(0) // add integer_part
|
|
fstp result // Store the result
|
|
}
|
|
return result;
|
|
}
|
|
|
|
|
|
#else
|
|
|
|
#include <math.h>
|
|
|
|
#endif // _ASM_MATH
|
|
|
|
|
|
DWORD CSoundFile::CutOffToFrequency(UINT nCutOff, int flt_modifier) const
|
|
//-----------------------------------------------------------------------
|
|
{
|
|
float Fc;
|
|
|
|
if (m_dwSongFlags & SONG_EXFILTERRANGE)
|
|
Fc = 110.0f * pow(2.0f, 0.25f + ((float)(nCutOff*(flt_modifier+256)))/(21.0f*512.0f));
|
|
else
|
|
Fc = 110.0f * pow(2.0f, 0.25f + ((float)(nCutOff*(flt_modifier+256)))/(24.0f*512.0f));
|
|
LONG freq = (LONG)Fc;
|
|
if (freq < 120) return 120;
|
|
if (freq > 10000) return 10000;
|
|
if (freq*2 > (LONG)gdwMixingFreq) freq = gdwMixingFreq>>1;
|
|
return (DWORD)freq;
|
|
}
|
|
|
|
|
|
// Simple 2-poles resonant filter
|
|
void CSoundFile::SetupChannelFilter(MODCHANNEL *pChn, BOOL bReset, int flt_modifier) const
|
|
//----------------------------------------------------------------------------------------
|
|
{
|
|
float fc = (float)CutOffToFrequency(pChn->nCutOff, flt_modifier);
|
|
float fs = (float)gdwMixingFreq;
|
|
float fg, fb0, fb1;
|
|
|
|
fc *= (float)(2.0*3.14159265358/fs);
|
|
float dmpfac = pow(10.0f, -((24.0f / 128.0f)*(float)pChn->nResonance) / 20.0f);
|
|
float d = (1.0f-2.0f*dmpfac)* fc;
|
|
if (d>2.0) d = 2.0;
|
|
d = (2.0f*dmpfac - d)/fc;
|
|
float e = pow(1.0f/fc, 2);
|
|
|
|
fg=1/(1+d+e);
|
|
fb0=(d+e+e)/(1+d+e);
|
|
fb1=-e/(1+d+e);
|
|
|
|
pChn->nFilter_A0 = (int)(fg * FILTER_PRECISION);
|
|
pChn->nFilter_B0 = (int)(fb0 * FILTER_PRECISION);
|
|
pChn->nFilter_B1 = (int)(fb1 * FILTER_PRECISION);
|
|
|
|
if (bReset)
|
|
{
|
|
pChn->nFilter_Y1 = pChn->nFilter_Y2 = 0;
|
|
pChn->nFilter_Y3 = pChn->nFilter_Y4 = 0;
|
|
}
|
|
pChn->dwFlags |= CHN_FILTER;
|
|
}
|
|
|
|
#endif // NO_FILTER
|