diff options
| -rw-r--r-- | common/dct.cpp | 35 | ||||
| -rw-r--r-- | common/fft.cpp | 163 | 
2 files changed, 95 insertions, 103 deletions
diff --git a/common/dct.cpp b/common/dct.cpp index d98f2f4cc3..6525c727ae 100644 --- a/common/dct.cpp +++ b/common/dct.cpp @@ -51,21 +51,18 @@ DCT::~DCT() {  void DCT::calc(float *data) {  	switch (_trans) { -		case DCT_I: -			calcDCTI(data); -			break; - -		case DCT_II: -			calcDCTII(data); -			break; - -		case DCT_III: -			calcDCTIII(data); -			break; - -		case DST_I: -			calcDSTI(data); -			break; +	case DCT_I: +		calcDCTI(data); +		break; +	case DCT_II: +		calcDCTII(data); +		break; +	case DCT_III: +		calcDCTIII(data); +		break; +	case DST_I: +		calcDSTI(data); +		break;  	}  } @@ -119,8 +116,8 @@ void DCT::calcDCTII(float *data) {  		tmp1 = (tmp1 + tmp2) * 0.5; -		data[i    ] = tmp1 + s; -		data[n-i-1] = tmp1 - s; +		data[i        ] = tmp1 + s; +		data[n - i - 1] = tmp1 - s;  	}  	_rdft->calc(data); @@ -136,8 +133,8 @@ void DCT::calcDCTII(float *data) {  		float c = COS(n, i);  		float s = SIN(n, i); -		data[i  ] = c * inr + s * ini; -		data[i+1] = next; +		data[i    ] = c * inr + s * ini; +		data[i + 1] = next;  		next += s * inr - c * ini;  	} diff --git a/common/fft.cpp b/common/fft.cpp index 5dc378d886..5852698b61 100644 --- a/common/fft.cpp +++ b/common/fft.cpp @@ -57,7 +57,7 @@ void FFT::permute(Complex *z) {  	int np = 1 << _bits;  	if (_tmpBuf) { -		for(int j = 0; j < np; j++) +		for (int j = 0; j < np; j++)  			_tmpBuf[_revTab[j]] = z[j];  		memcpy(z, _tmpBuf, np * sizeof(Complex)); @@ -66,7 +66,7 @@ void FFT::permute(Complex *z) {  	}  	// Reverse -	for(int j = 0; j < np; j++) { +	for (int j = 0; j < np; j++) {  		int k = _revTab[j];  		if (k < j) @@ -80,12 +80,12 @@ int FFT::splitRadixPermutation(int i, int n, int inverse) {  	int m = n >> 1; -	if(!(i & m)) +	if (!(i & m))  		return splitRadixPermutation(i, m, inverse) * 2;  	m >>= 1; -	if(inverse == !(i & m)) +	if (inverse == !(i & m))  		return splitRadixPermutation(i, m, inverse) * 4 + 1;  	return splitRadixPermutation(i, m, inverse) * 4 - 1; @@ -93,68 +93,67 @@ int FFT::splitRadixPermutation(int i, int n, int inverse) {  #define sqrthalf (float)M_SQRT1_2 -#define BF(x,y,a,b) {\ -	x = a - b;\ -	y = a + b;\ +#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);\ +#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 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(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)\ +#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]);\ +#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);\  } @@ -163,17 +162,15 @@ PASS(pass)  #define BUTTERFLIES BUTTERFLIES_BIG  PASS(pass_big) -#define DECL_FFT(t,n,n2,n4)\ -static void fft##n(Complex *z)\ -{\ -	fft##n2(z);\ -	fft##n4(z+n4*2);\ -	fft##n4(z+n4*3);\ -	pass(z,getCosineTable(t),n4/2);\ +#define DECL_FFT(t, n, n2, n4) \ +static void fft##n(Complex *z) { \ +	fft##n2(z); \ +	fft##n4(z + n4 * 2); \ +	fft##n4(z + n4 * 3); \ +	pass(z, getCosineTable(t), n4 / 2);\  } -static void fft4(Complex *z) -{ +static void fft4(Complex *z) {  	float t1, t2, t3, t4, t5, t6, t7, t8;  	BF(t3, t1, z[0].re, z[1].re); @@ -186,8 +183,7 @@ static void fft4(Complex *z)  	BF(z[2].im, z[0].im, t2, t5);  } -static void fft8(Complex *z) -{ +static void fft8(Complex *z) {  	float t1, t2, t3, t4, t5, t6, t7, t8;  	fft4(z); @@ -203,38 +199,37 @@ static void fft8(Complex *z)  	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); +	TRANSFORM(z[1], z[3], z[5], z[7], sqrthalf, sqrthalf);  } -static void fft16(Complex *z) -{ +static void fft16(Complex *z) {  	float t1, t2, t3, t4, t5, t6;  	fft8(z); -	fft4(z+8); -	fft4(z+12); +	fft4(z + 8); +	fft4(z + 12);  	const float * const cosTable = getCosineTable(4); -	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],cosTable[1],cosTable[3]); -	TRANSFORM(z[3],z[7],z[11],z[15],cosTable[3],cosTable[1]); +	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], cosTable[1],cosTable[3]); +	TRANSFORM(z[3], z[7], z[11], z[15], cosTable[3], cosTable[1]);  } -DECL_FFT(5, 32,16,8) -DECL_FFT(6, 64,32,16) -DECL_FFT(7, 128,64,32) -DECL_FFT(8, 256,128,64) -DECL_FFT(9, 512,256,128) +DECL_FFT(5, 32, 16, 8) +DECL_FFT(6, 64, 32, 16) +DECL_FFT(7, 128, 64, 32) +DECL_FFT(8, 256, 128, 64) +DECL_FFT(9, 512, 256, 128)  #define pass pass_big -DECL_FFT(10, 1024,512,256) -DECL_FFT(11, 2048,1024,512) -DECL_FFT(12, 4096,2048,1024) -DECL_FFT(13, 8192,4096,2048) -DECL_FFT(14, 16384,8192,4096) -DECL_FFT(15, 32768,16384,8192) -DECL_FFT(16, 65536,32768,16384) +DECL_FFT(10, 1024, 512, 256) +DECL_FFT(11, 2048, 1024, 512) +DECL_FFT(12, 4096, 2048, 1024) +DECL_FFT(13, 8192, 4096, 2048) +DECL_FFT(14, 16384, 8192, 4096) +DECL_FFT(15, 32768, 16384, 8192) +DECL_FFT(16, 65536, 32768, 16384)  static void (* const fft_dispatch[])(Complex*) = {  	fft4, fft8, fft16, fft32, fft64, fft128, fft256, fft512, fft1024,  | 
