static inline void ifft4(buf)
{
uint32_t tm4=(uint32_t)(buf);
S32LDI(xr1, tm4, 0);
S32LDI(xr2, tm4, 8);
S32LDI(xr3, tm4, 8);
S32LDI(xr4, tm4, 8);
D32ADD_AS(xr5, xr1, xr2, xr6);
D32ADD_AS(xr7, xr4, xr3, xr8);
D32ADD_AS(xr9, xr5, xr7, xr10);
S32SDI(xr10,tm4,-8);
S32SDI(xr9,tm4,-16);
S32LDI(xr1, tm4, 4);
S32LDI(xr2, tm4, 8);
S32LDI(xr3, tm4, 8);
S32LDI(xr4, tm4, 8);
D32ADD_AS(xr5, xr1, xr2, xr9);
D32ADD_AS(xr7, xr3, xr4, xr10);
D32ADD_AS(xr1,xr5,xr7,xr2);
D32ADD_AS(xr11,xr6,xr10,xr12);
D32ADD_AS(xr13,xr9,xr8,xr14);
S32SDI(xr14,tm4,0);
S32SDI(xr12,tm4,-4);
S32SDI(xr2, tm4,-4);
S32SDI(xr13,tm4,-8);
S32SDI(xr11,tm4,-4);
S32SDI(xr1, tm4,-4);
}
static inline void ifft2(buf)
{
S32LDD(xr1, buf, 0);
S32LDD(xr3, buf, 8);
S32LDD(xr2, buf, 4);
S32LDD(xr4, buf, 12);
D32ADD_AS(xr5, xr1, xr3, xr7);
D32ADD_AS(xr6, xr2, xr4, xr8);
S32STD(xr5, buf, 0);
S32STD(xr7, buf, 8);
S32STD(xr6, buf, 4);
S32STD(xr8, buf, 12);
}
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);
}
