/**
* Compute inverse MDCT of size N = 2^nbits
* @param output N samples
* @param input N/2 samples
* @param tmp N/2 samples
*/
void imdct_calc_fix(MDCTContext_fix *s, FFTSample_fix *output,
const FFTSample_fix *input, FFTSample_fix *tmp)
{
int k, n8, n4, n2, n, j;
const unsigned short *revtab = s->fft.revtab;
const FFTSample_fix *tcos = s->tcos;
const FFTSample_fix *tsin = s->tsin;
const FFTSample_fix *in1, *in2;
FFTComplex_fix *z = (FFTComplex_fix *)tmp;
n = 1 << s->nbits;
n2 = n >> 1;
n4 = n >> 2;
n8 = n >> 3;
/* pre rotation */
in1 = input;
in2 = input + n2 - 1;
for(k = 0; k < n4; k++) {
j=revtab[k];
FFT_CMUL_fix(z[j].re, z[j].im, *in2, *in1, tcos[k], tsin[k]);
in1 += 2;
in2 -= 2;
}
s->fft.fft_calc(&s->fft, z);
/* post rotation + reordering */
/* XXX: optimize */
for(k = 0; k < n4; k++) {
FFT_CMUL_fix(z[k].re, z[k].im, z[k].re, z[k].im, tcos[k], tsin[k]);
}
for(k = 0; k < n8; k++) {
output[2*k] = -z[n8 + k].im;
output[n2-1-2*k] = z[n8 + k].im;
output[2*k+1] = z[n8-1-k].re;
output[n2-1-2*k-1] = -z[n8-1-k].re;
output[n2 + 2*k]=-z[k+n8].re;
output[n-1- 2*k]=-z[k+n8].re;
output[n2 + 2*k+1]=z[n8-k-1].im;
output[n-2 - 2 * k] = z[n8-k-1].im;
}
}
/**
* Compute MDCT of size N = 2^nbits
* @param input N samples
* @param out N/2 samples
* @param tmp temporary storage of N/2 samples
*/
void mdct_calc_fix(MDCTContext_fix *s, FFTSample_fix *out,
const FFTSample_fix *input, FFTSample_fix *tmp)
{
int i, j, n, n8, n4, n2, n3;
FFTSample_fix re, im, re1, im1;
const unsigned short *revtab = s->fft.revtab;
const FFTSample_fix *tcos = s->tcos;
const FFTSample_fix *tsin = s->tsin;
FFTComplex_fix *x = (FFTComplex_fix *)tmp;
n = 1 << s->nbits;
n2 = n >> 1;
n4 = n >> 2;
n8 = n >> 3;
n3 = 3 * n4;
/* pre rotation */
for(i=0; i<n8; i++) {
re = -input[2*i+3*n4] - input[n3-1-2*i];
im = -input[n4+2*i] + input[n4-1-2*i];
j = revtab[i];
FFT_CMUL_fix(x[j].re, x[j].im, re, im, -tcos[i], tsin[i]);
re = input[2*i] - input[n2-1-2*i];
im = -(input[n2+2*i] + input[n-1-2*i]);
j = revtab[n8 + i];
FFT_CMUL_fix(x[j].re, x[j].im, re, im, -tcos[n8 + i], tsin[n8 + i]);
}
s->fft.fft_calc(&s->fft, x);
/* post rotation */
for(i=0; i<n4; i++) {
re = x[i].re;
im = x[i].im;
FFT_CMUL_fix(re1, im1, re, im, -tsin[i], -tcos[i]);
out[2*i] = im1;
out[n2-1-2*i] = re1;
}
}
void imdct_half_fix_c(MDCTContext_fix *s, FFTSample_fix *output,
const FFTSample_fix *input)
{
//PMON_ON(qmf);
int k, n8, n4, n2, n, j,j1;
const FFTSample_fix *in1, *in2;
const unsigned short *revtab = s->fft.revtab;
const FFTSample_fix *tcos = s->tcos;
const FFTSample_fix *tsin = s->tsin;
FFTComplex_fix *z = (FFTComplex_fix *)output;
n = 1 << s->nbits;//64
n2 = n >> 1;//32
n4 = n >> 2;//16
n8 = n >> 3;//8
/* pre rotation */
in1 = input; //head
in2 = input + n2 - 1;//tail
for(k = 0; k < n8; k++) {
#if 0
j=revtab[k];
FFT_CMUL_fix(z[j].re, z[j].im, *in2, *in1, tcos[k], tsin[k]);
in1 += 2;
in2 -= 2;
#else
FFTSample_fix _are,_bre,_aim,_bim,are,aim;
_are = *in2;
_bre = tcos[k];
_aim = *in1;
_bim = tsin[k];
j=revtab[k];
n=n4-k-1;
j1=revtab[n];
S32MUL(xr1,xr2, _are, _bre);
S32MUL(xr3, xr4, _are, _bim);
in2--;
are = *in2;
S32MUL(xr7,xr8, are, _bre);
S32MUL(xr9, xr10, are, _bim);
S32MSUB(xr1, xr2, _aim, _bim);
S32MADD(xr3, xr4, _aim, _bre); ;
in1++;
aim = *in1;
D32SLL(xr5,xr1,xr3,xr6,1);
S32MSUB(xr7, xr8, aim, _bim);
S32MADD(xr9, xr10, aim, _bre);
z[j].re=S32M2I(xr5);
D32SLL(xr11,xr7,xr9,xr12,1);
z[j].im=S32M2I(xr6);
in1++;
in2--;
z[j1].re=S32M2I(xr11);
z[j1].im=S32M2I(xr12);
#endif
}
s->fft.fft_calc(&s->fft, z);
/* post rotation + reordering */
/* XXX: optimize */
for(k = 0; k < n8; k++) {
FFTSample_fix r0, i0, r1, i1;
FFT_CMUL_fix(r0, i1, z[n8-k-1].im, z[n8-k-1].re, tsin[n8-k-1], tcos[n8-k-1]);
FFT_CMUL_fix(r1, i0, z[n8+k ].im, z[n8+k ].re, tsin[n8+k ], tcos[n8+k ]);
z[n8-k-1].re = r0;
z[n8-k-1].im = i0;
z[n8+k ].re = r1;
z[n8+k ].im = i1;
}
//PMON_OFF(qmf);
}
void fft_calc_fix_inverse(FFTContext_fix *s, FFTComplex_fix *z)
{
int ln = s->nbits;
int j, np, np2;
int nblocks, nloops;
register FFTComplex_fix *p, *q;
FFTComplex_fix *exptab = s->exptab;
int l;
FFTSample_fix tmp_re, tmp_im;
np = 1 << ln;
/* function is :butterfly all 4 step ,N=16 */
/* pass 0 */
#if 0
p=&z[0];
j=(np >> 1);
do {
/*
X(k) = G(k)+H(k)*W (= e j*0)
*/
FFT_BF_fix(p[0].re, p[0].im, p[1].re, p[1].im,
p[0].re, p[0].im, p[1].re, p[1].im);
p+=2;
} while (--j);
#endif
/* pass 1 */
p=&z[0];
j=np >> 2;
do {
#if 1
S32LDD(xr1,p,0);
S32LDD(xr2,p,4);
S32LDD(xr3,p,8);
S32LDD(xr4,p,12);
S32LDD(xr5,p,16);
S32LDD(xr6,p,20);
S32LDD(xr7,p,24);
S32LDD(xr8,p,28);
D32ADD_AS(xr1,xr1,xr3,xr3);
D32ADD_AS(xr2,xr2,xr4,xr4);
D32ADD_AS(xr5,xr5,xr7,xr7);
D32ADD_AS(xr6,xr6,xr8,xr8);
D32ADD_AS(xr1,xr1,xr5,xr5);
D32ADD_AS(xr2,xr2,xr6,xr6);
D32ADD_SA(xr3,xr3,xr8,xr9);
D32ADD_AS(xr4,xr4,xr7,xr8);
S32STD(xr1,p,0);
S32STD(xr2,p,4);
S32STD(xr3,p,8);
S32STD(xr4,p,12);
S32STD(xr5,p,16);
S32STD(xr6,p,20);
S32STD(xr9,p,24);
S32STD(xr8,p,28);
#else
FFT_BF_fix(p[0].re, p[0].im, p[1].re, p[1].im,
p[0].re, p[0].im, p[1].re, p[1].im);
FFT_BF_fix(p[2].re, p[2].im, p[3].re, p[3].im,
p[2].re, p[2].im, p[3].re, p[3].im);
FFT_BF_fix(p[0].re, p[0].im, p[2].re, p[2].im,
p[0].re, p[0].im, p[2].re, p[2].im);
FFT_BF_fix(p[1].re, p[1].im, p[3].re, p[3].im,
p[1].re, p[1].im, -p[3].im, p[3].re);
#endif
p+=4;
} while (--j);
/* pass 2 .. ln-1 */
nblocks = np >> 3;
nloops = 1 << 2;
np2 = np >> 1;
do {
p = z;
q = z + nloops;
for (j = 0; j < nblocks; ++j) {
#if 1
S32LDD(xr1,p,0);
S32LDD(xr2,p,4);
S32LDD(xr3,q,0);
S32LDD(xr4,q,4);
D32ADD_AS(xr1,xr1,xr3,xr3);
D32ADD_AS(xr2,xr2,xr4,xr4);
S32STD(xr1,p,0);
S32STD(xr2,p,4);
S32STD(xr3,q,0);
S32STD(xr4,q,4);
#else
FFT_BF_fix(p->re, p->im, q->re, q->im,
p->re, p->im, q->re, q->im);
#endif
p++;
q++;
for(l = nblocks; l < np2; l += nblocks) {
/* FFT_CMUL_fix( ) fuction is :
(-j 2*PI/N *km)
H(i) * E
*/
#if 1
FFTSample_fix _are = exptab[l].re;
FFTSample_fix _bre = q->re;
FFTSample_fix _aim = exptab[l].im;
FFTSample_fix _bim = q->im;
S32MUL(xr1, xr2, _are, _bre);
S32MUL(xr5, xr6, _are, _bim);
S32LDD(xr7,p,0);
S32MSUB(xr1, xr2, _aim, _bim);
S32MADD(xr5, xr6, _aim, _bre);
S32LDD(xr8,p,4);
D32SLL(xr1, xr1, xr5, xr5, 1);
D32ADD_AS(xr7,xr7,xr1,xr1);
D32ADD_AS(xr8,xr8,xr5,xr5);
S32STD(xr7,p,0);
S32STD(xr8,p,4);
S32STD(xr1,q,0);
S32STD(xr5,q,4);
#else
FFT_CMUL_fix(tmp_re, tmp_im, exptab[l].re, exptab[l].im, q->re, q->im);
FFT_BF_fix(p->re, p->im, q->re, q->im,
p->re, p->im, tmp_re, tmp_im);
#endif
p++;
q++;
}
p += nloops;
q += nloops;
}
nblocks = nblocks >> 1;
nloops = nloops << 1;
} while (nblocks);
}
void fft_calc_fix(FFTContext_fix *s, FFTComplex_fix *z)
{
int ln = s->nbits;
int j, np, np2;
int nblocks, nloops;
register FFTComplex_fix *p, *q;
FFTComplex_fix *exptab = s->exptab;
int l;
FFTSample_fix tmp_re, tmp_im;
np = 1 << ln;
/* function is :butterfly all 4 step ,N=16 */
/* pass 0 */
p=&z[0];
j=(np >> 1);
do {
/*
X(k) = G(k)+H(k)*W (= e j*0)
*/
FFT_BF_fix(p[0].re, p[0].im, p[1].re, p[1].im,
p[0].re, p[0].im, p[1].re, p[1].im);
p+=2;
} while (--j);
/* pass 1 */
p=&z[0];
j=np >> 2;
do {
FFT_BF_fix(p[0].re, p[0].im, p[2].re, p[2].im,
p[0].re, p[0].im, p[2].re, p[2].im);
FFT_BF_fix(p[1].re, p[1].im, p[3].re, p[3].im,
p[1].re, p[1].im, p[3].im, -p[3].re);
p+=4;
} while (--j);
/* pass 2 .. ln-1 */
nblocks = np >> 3;
nloops = 1 << 2;
np2 = np >> 1;
do {
p = z;
q = z + nloops;
for (j = 0; j < nblocks; ++j) {
FFT_BF_fix(p->re, p->im, q->re, q->im,
p->re, p->im, q->re, q->im);
p++;
q++;
for(l = nblocks; l < np2; l += nblocks) {
/* FFT_CMUL_fix( ) fuction is :
(-j 2*PI/N *km)
H(i) * E
*/
FFT_CMUL_fix(tmp_re, tmp_im, exptab[l].re, exptab[l].im, q->re, q->im);
FFT_BF_fix(p->re, p->im, q->re, q->im,
p->re, p->im, tmp_re, tmp_im);
p++;
q++;
}
p += nloops;
q += nloops;
}
nblocks = nblocks >> 1;
nloops = nloops << 1;
} while (nblocks);
}
void fft_calc_fix(FFTContext_fix *s, FFTComplex_fix *z)
{
int ln = s->nbits;
int j, np, np2;
int nblocks, nloops;
register FFTComplex_fix *p, *q;
FFTComplex_fix *exptab = s->exptab;
int l;
FFTSample_fix tmp_re, tmp_im;
np = 1 << ln;
/* pass 0 */
p=&z[0];
j=(np >> 1);
do {
FFT_BF_fix(p[0].re, p[0].im, p[1].re, p[1].im,
p[0].re, p[0].im, p[1].re, p[1].im);
p+=2;
} while (--j);
/* pass 1 */
p=&z[0];
j=np >> 2;
if (s->inverse) {
do {
FFT_BF_fix(p[0].re, p[0].im, p[2].re, p[2].im,
p[0].re, p[0].im, p[2].re, p[2].im);
FFT_BF_fix(p[1].re, p[1].im, p[3].re, p[3].im,
p[1].re, p[1].im, -p[3].im, p[3].re);
p+=4;
} while (--j);
} else {
do {
FFT_BF_fix(p[0].re, p[0].im, p[2].re, p[2].im,
p[0].re, p[0].im, p[2].re, p[2].im);
FFT_BF_fix(p[1].re, p[1].im, p[3].re, p[3].im,
p[1].re, p[1].im, p[3].im, -p[3].re);
p+=4;
} while (--j);
}
/* pass 2 .. ln-1 */
nblocks = np >> 3;
nloops = 1 << 2;
np2 = np >> 1;
do {
p = z;
q = z + nloops;
for (j = 0; j < nblocks; ++j) {
FFT_BF_fix(p->re, p->im, q->re, q->im,
p->re, p->im, q->re, q->im);
p++;
q++;
for(l = nblocks; l < np2; l += nblocks) {
FFT_CMUL_fix(tmp_re, tmp_im, exptab[l].re, exptab[l].im, q->re, q->im);
FFT_BF_fix(p->re, p->im, q->re, q->im,
p->re, p->im, tmp_re, tmp_im);
p++;
q++;
}
p += nloops;
q += nloops;
}
nblocks = nblocks >> 1;
nloops = nloops << 1;
} while (nblocks);
}
