diff options
| author | Matthew Hoops | 2011-07-02 15:23:49 -0400 | 
|---|---|---|
| committer | Matthew Hoops | 2011-07-02 15:23:49 -0400 | 
| commit | 3aefa913faaf2a98512f02b0581ca8a077876b84 (patch) | |
| tree | 9bc5436e68e1dd83bf5013f181ef864c5c35e2d2 | |
| parent | 0aa58b9033ee83b3008a7f4ac48d10baa6c7335d (diff) | |
| download | scummvm-rg350-3aefa913faaf2a98512f02b0581ca8a077876b84.tar.gz scummvm-rg350-3aefa913faaf2a98512f02b0581ca8a077876b84.tar.bz2 scummvm-rg350-3aefa913faaf2a98512f02b0581ca8a077876b84.zip  | |
COMMON: Add an FFT math handler
Based on eos' code which is based on FFmpeg's code
| -rw-r--r-- | common/fft.cpp | 245 | ||||
| -rw-r--r-- | common/fft.h | 73 | ||||
| -rw-r--r-- | common/module.mk | 1 | 
3 files changed, 319 insertions, 0 deletions
diff --git a/common/fft.cpp b/common/fft.cpp new file mode 100644 index 0000000000..9398034c82 --- /dev/null +++ b/common/fft.cpp @@ -0,0 +1,245 @@ +/* ScummVM - Graphic Adventure Engine + * + * ScummVM is the legal property of its developers, whose names + * are too numerous to list here. Please refer to the COPYRIGHT + * file distributed with this source distribution. + * + * 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. + + * This program 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. + * + */ + +// Based on eos' (I)FFT code which is in turn +// Based upon the (I)FFT code in FFmpeg +// Copyright (c) 2008 Loren Merritt +// Copyright (c) 2002 Fabrice Bellard +// Partly based on libdjbfft by D. J. Bernstein + +#include "common/fft.h" +#include "common/util.h" + +namespace Common { + +FFT::FFT(int bits, int inverse) : _bits(bits), _inverse(inverse) { +	assert((_bits >= 2) && (_bits <= 16)); + +	int n = 1 << bits; + +	_tmpBuf = new Complex[n]; +	_expTab = new Complex[n / 2]; +	_revTab = new uint16[n]; + +	_splitRadix = 1; + +	for (int i = 0; i < n; i++) +		_revTab[-splitRadixPermutation(i, n, _inverse) & (n - 1)] = i; +} + +FFT::~FFT() { +	delete[] _revTab; +	delete[] _expTab; +	delete[] _tmpBuf; +} + +void FFT::permute(Complex *z) { +	int np = 1 << _bits; + +	if (_tmpBuf) { +		for(int j = 0; j < np; j++) +			_tmpBuf[_revTab[j]] = z[j]; + +		memcpy(z, _tmpBuf, np * sizeof(Complex)); + +		return; +	} + +	// Reverse +	for(int j = 0; j < np; j++) { +		int k = _revTab[j]; + +		if (k < j) +			SWAP(z[k], z[j]); +	} +} + +int FFT::splitRadixPermutation(int i, int n, int inverse) { +	if (n <= 2) +		return i & 1; + +	int m = n >> 1; + +	if(!(i & m)) +		return splitRadixPermutation(i, m, inverse) * 2; + +	m >>= 1; + +	if(inverse == !(i & m)) +		return splitRadixPermutation(i, m, inverse) * 4 + 1; + +	return splitRadixPermutation(i, m, inverse) * 4 - 1; +} + +#define sqrthalf (float)M_SQRT1_2 + +#define BF(x,y,a,b) {\ +	x = a - b;\ +	y = a + b;\ +} + +#define BUTTERFLIES(a0,a1,a2,a3) {\ +	BF(t3, t5, t5, t1);\ +	BF(a2.re, a0.re, a0.re, t5);\ +	BF(a3.im, a1.im, a1.im, t3);\ +	BF(t4, t6, t2, t6);\ +	BF(a3.re, a1.re, a1.re, t4);\ +	BF(a2.im, a0.im, a0.im, t6);\ +} + +// force loading all the inputs before storing any. +// this is slightly slower for small data, but avoids store->load aliasing +// for addresses separated by large powers of 2. +#define BUTTERFLIES_BIG(a0,a1,a2,a3) {\ +	float r0=a0.re, i0=a0.im, r1=a1.re, i1=a1.im;\ +	BF(t3, t5, t5, t1);\ +	BF(a2.re, a0.re, r0, t5);\ +	BF(a3.im, a1.im, i1, t3);\ +	BF(t4, t6, t2, t6);\ +	BF(a3.re, a1.re, r1, t4);\ +	BF(a2.im, a0.im, i0, t6);\ +} + +#define TRANSFORM(a0,a1,a2,a3,wre,wim) {\ +	t1 = a2.re * wre + a2.im * wim;\ +	t2 = a2.im * wre - a2.re * wim;\ +	t5 = a3.re * wre - a3.im * wim;\ +	t6 = a3.im * wre + a3.re * wim;\ +	BUTTERFLIES(a0,a1,a2,a3)\ +} + +#define TRANSFORM_ZERO(a0,a1,a2,a3) {\ +	t1 = a2.re;\ +	t2 = a2.im;\ +	t5 = a3.re;\ +	t6 = a3.im;\ +	BUTTERFLIES(a0,a1,a2,a3)\ +} + +/* z[0...8n-1], w[1...2n-1] */ +#define PASS(name)\ +static void name(Complex *z, const float *wre, unsigned int n)\ +{\ +	float t1, t2, t3, t4, t5, t6;\ +	int o1 = 2*n;\ +	int o2 = 4*n;\ +	int o3 = 6*n;\ +	const float *wim = wre+o1;\ +	n--;\ +\ +	TRANSFORM_ZERO(z[0],z[o1],z[o2],z[o3]);\ +	TRANSFORM(z[1],z[o1+1],z[o2+1],z[o3+1],wre[1],wim[-1]);\ +	do {\ +		z += 2;\ +		wre += 2;\ +		wim -= 2;\ +		TRANSFORM(z[0],z[o1],z[o2],z[o3],wre[0],wim[0]);\ +		TRANSFORM(z[1],z[o1+1],z[o2+1],z[o3+1],wre[1],wim[-1]);\ +	} while(--n);\ +} + +PASS(pass) +#undef BUTTERFLIES +#define BUTTERFLIES BUTTERFLIES_BIG +PASS(pass_big) + +#define DECL_FFT(n,n2,n4)\ +static void fft##n(Complex *z)\ +{\ +	fft##n2(z);\ +	fft##n4(z+n4*2);\ +	fft##n4(z+n4*3);\ +	pass(z,cosTable##n,n4/2);\ +} + +static void fft4(Complex *z) +{ +	float t1, t2, t3, t4, t5, t6, t7, t8; + +	BF(t3, t1, z[0].re, z[1].re); +	BF(t8, t6, z[3].re, z[2].re); +	BF(z[2].re, z[0].re, t1, t6); +	BF(t4, t2, z[0].im, z[1].im); +	BF(t7, t5, z[2].im, z[3].im); +	BF(z[3].im, z[1].im, t4, t8); +	BF(z[3].re, z[1].re, t3, t7); +	BF(z[2].im, z[0].im, t2, t5); +} + +static void fft8(Complex *z) +{ +	float t1, t2, t3, t4, t5, t6, t7, t8; + +	fft4(z); + +	BF(t1, z[5].re, z[4].re, -z[5].re); +	BF(t2, z[5].im, z[4].im, -z[5].im); +	BF(t3, z[7].re, z[6].re, -z[7].re); +	BF(t4, z[7].im, z[6].im, -z[7].im); +	BF(t8, t1, t3, t1); +	BF(t7, t2, t2, t4); +	BF(z[4].re, z[0].re, z[0].re, t1); +	BF(z[4].im, z[0].im, z[0].im, t2); +	BF(z[6].re, z[2].re, z[2].re, t7); +	BF(z[6].im, z[2].im, z[2].im, t8); + +	TRANSFORM(z[1],z[3],z[5],z[7],sqrthalf,sqrthalf); +} + +static void fft16(Complex *z) +{ +	float t1, t2, t3, t4, t5, t6; + +	fft8(z); +	fft4(z+8); +	fft4(z+12); + +	TRANSFORM_ZERO(z[0],z[4],z[8],z[12]); +	TRANSFORM(z[2],z[6],z[10],z[14],sqrthalf,sqrthalf); +	TRANSFORM(z[1],z[5],z[9],z[13],cosTable16[1],cosTable16[3]); +	TRANSFORM(z[3],z[7],z[11],z[15],cosTable16[3],cosTable16[1]); +} + +DECL_FFT(32,16,8) +DECL_FFT(64,32,16) +DECL_FFT(128,64,32) +DECL_FFT(256,128,64) +DECL_FFT(512,256,128) +#define pass pass_big +DECL_FFT(1024,512,256) +DECL_FFT(2048,1024,512) +DECL_FFT(4096,2048,1024) +DECL_FFT(8192,4096,2048) +DECL_FFT(16384,8192,4096) +DECL_FFT(32768,16384,8192) +DECL_FFT(65536,32768,16384) + +static void (* const fft_dispatch[])(Complex*) = { +	fft4, fft8, fft16, fft32, fft64, fft128, fft256, fft512, fft1024, +	fft2048, fft4096, fft8192, fft16384, fft32768, fft65536, +}; + +void FFT::calc(Complex *z) { +	fft_dispatch[_bits - 2](z); +} + +} // End of namespace Common diff --git a/common/fft.h b/common/fft.h new file mode 100644 index 0000000000..17d1ef0055 --- /dev/null +++ b/common/fft.h @@ -0,0 +1,73 @@ +/* ScummVM - Graphic Adventure Engine + * + * ScummVM is the legal property of its developers, whose names + * are too numerous to list here. Please refer to the COPYRIGHT + * file distributed with this source distribution. + * + * 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. + + * This program 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. + * + */ + +// Based on eos' (I)FFT code which is in turn +// based upon the (I)FFT code in FFmpeg +// Copyright (c) 2008 Loren Merritt +// Copyright (c) 2002 Fabrice Bellard +// Partly based on libdjbfft by D. J. Bernstein + +#ifndef COMMON_FFT_H +#define COMMON_FFT_H + +#include "common/scummsys.h" +#include "common/math.h" + +namespace Common { + +/** (Inverse) Fast Fourier Transform. */ +class FFT { +public: +	FFT(int bits, int inverse); +	~FFT(); + +	/** Do the permutation needed BEFORE calling calc(). */ +	void permute(Complex *z); + +	/** Do a complex FFT. +	 * +	 *  The input data must be permuted before. +	 *  No 1.0/sqrt(n) normalization is done. +	 */ +	void calc(Complex *z); + +private: +	int _bits; +	int _inverse; + +	uint16 *_revTab; + +	Complex *_expTab; +	Complex *_tmpBuf; + +	const float *_tSin; +	const float *_tCos; + +	int _splitRadix; +	int _permutation; + +	static int splitRadixPermutation(int i, int n, int inverse); +}; + +} // End of namespace Common + +#endif // COMMON_FFT_H diff --git a/common/module.mk b/common/module.mk index 7716f74d34..2ab4c113fa 100644 --- a/common/module.mk +++ b/common/module.mk @@ -10,6 +10,7 @@ MODULE_OBJS := \  	error.o \  	EventDispatcher.o \  	EventRecorder.o \ +	fft.o \  	file.o \  	fs.o \  	hashmap.o \  | 
