void faad_mdct(mdct_info *mdct, real_t *X_in, real_t *X_out)
{
uint16_t k;
complex_t x;
ALIGN complex_t Z1[512];
complex_t *sincos = mdct->sincos;
uint16_t N = mdct->N;
uint16_t N2 = N >> 1;
uint16_t N4 = N >> 2;
uint16_t N8 = N >> 3;
#ifndef FIXED_POINT
real_t scale = REAL_CONST(N);
#else
real_t scale = REAL_CONST(4.0/N);
#endif
/* pre-FFT complex multiplication */
for (k = 0; k < N8; k++)
{
uint16_t n = k << 1;
RE(x) = X_in[N - N4 - 1 - n] + X_in[N - N4 + n];
IM(x) = X_in[ N4 + n] - X_in[ N4 - 1 - n];
ComplexMult(&RE(Z1[k]), &IM(Z1[k]),
RE(x), IM(x), RE(sincos[k]), IM(sincos[k]));
RE(Z1[k]) = MUL_R(RE(Z1[k]), scale);
IM(Z1[k]) = MUL_R(IM(Z1[k]), scale);
RE(x) = X_in[N2 - 1 - n] - X_in[ n];
IM(x) = X_in[N2 + n] + X_in[N - 1 - n];
ComplexMult(&RE(Z1[k + N8]), &IM(Z1[k + N8]),
RE(x), IM(x), RE(sincos[k + N8]), IM(sincos[k + N8]));
RE(Z1[k + N8]) = MUL_R(RE(Z1[k + N8]), scale);
IM(Z1[k + N8]) = MUL_R(IM(Z1[k + N8]), scale);
}
/* complex FFT, any non-scaling FFT can be used here */
cfftf(mdct->cfft, Z1);
/* post-FFT complex multiplication */
for (k = 0; k < N4; k++)
{
uint16_t n = k << 1;
ComplexMult(&RE(x), &IM(x),
RE(Z1[k]), IM(Z1[k]), RE(sincos[k]), IM(sincos[k]));
X_out[ n] = -RE(x);
X_out[N2 - 1 - n] = IM(x);
X_out[N2 + n] = -IM(x);
X_out[N - 1 - n] = RE(x);
}
}
static void AAC_Pass2_N(real_t* Dst, const real_t* Src, const real_t* Table, int n,int m)
{
int i,j,o = n*m*2;
for (i=0;i<n;++i)
{
for (j=0;j<m;++j)
{
real_t t2r, t2i;
Dst[0] = Src[0+0*m] + Src[0+2*m];
Dst[1] = Src[1+0*m] + Src[1+2*m];
t2r = Src[0+0*m] - Src[0+2*m];
t2i = Src[1+0*m] - Src[1+2*m];
ComplexMult(Dst+1+o, Dst+0+o, t2i, t2r, Table[0], Table[1]);
Table += 2;
Dst += 2;
Src += 2;
}
Src += 2*m;
Table -= 2*m;
}
}
static void passf2neg(const uint16_t ido, const uint16_t l1, const complex_t *cc,
complex_t *ch, const complex_t *wa)
{
uint16_t i, k, ah, ac;
if (ido == 1)
{
for (k = 0; k < l1; k++)
{
ah = 2*k;
ac = 4*k;
RE(ch[ah]) = RE(cc[ac]) + RE(cc[ac+1]);
RE(ch[ah+l1]) = RE(cc[ac]) - RE(cc[ac+1]);
IM(ch[ah]) = IM(cc[ac]) + IM(cc[ac+1]);
IM(ch[ah+l1]) = IM(cc[ac]) - IM(cc[ac+1]);
}
} else {
for (k = 0; k < l1; k++)
{
ah = k*ido;
ac = 2*k*ido;
for (i = 0; i < ido; i++)
{
complex_t t2;
RE(ch[ah+i]) = RE(cc[ac+i]) + RE(cc[ac+i+ido]);
RE(t2) = RE(cc[ac+i]) - RE(cc[ac+i+ido]);
IM(ch[ah+i]) = IM(cc[ac+i]) + IM(cc[ac+i+ido]);
IM(t2) = IM(cc[ac+i]) - IM(cc[ac+i+ido]);
#if 1
ComplexMult(&RE(ch[ah+i+l1*ido]), &IM(ch[ah+i+l1*ido]),
RE(t2), IM(t2), RE(wa[i]), IM(wa[i]));
#else
ComplexMult(&IM(ch[ah+i+l1*ido]), &RE(ch[ah+i+l1*ido]),
IM(t2), RE(t2), RE(wa[i]), IM(wa[i]));
#endif
}
}
}
}
static void AAC_Pre(real_t* Dst,const real_t* Src,const real_t* Table,int n)
{
const real_t* Src2 = Src + n*2 - 1;
do
{
ComplexMult(Dst+1,Dst+0,Src[0],Src2[0],Table[0],Table[1]);
Dst += 2; Src += 2; Src2 -= 2; Table += 2;
--n;
} while (n>0);
}
static void AAC_Pass4_N(real_t* Dst, const real_t* Src, const real_t* Table, int n,int m)
{
int i,j,o = n*m*2;
for (i=0;i<n;++i)
{
for (j=0;j<m;++j)
{
real_t t1r, t1i, t2r, t2i, t3r, t3i, t4r, t4i;
t2r = Src[0+0*m] + Src[0+4*m];
t1r = Src[0+0*m] - Src[0+4*m];
t2i = Src[1+0*m] + Src[1+4*m];
t1i = Src[1+0*m] - Src[1+4*m];
t3r = Src[0+2*m] + Src[0+6*m];
t4i = Src[0+2*m] - Src[0+6*m];
t3i = Src[1+6*m] + Src[1+2*m];
t4r = Src[1+6*m] - Src[1+2*m];
Dst[0] = t2r + t3r;
Dst[1] = t2i + t3i;
ComplexMult(Dst+1+o, Dst+0+o,
t1i + t4i, t1r + t4r,
Table[0], Table[1]);
ComplexMult(Dst+1+2*o, Dst+0+2*o,
t2i - t3i, t2r - t3r,
Table[m*2], Table[1+m*2]);
ComplexMult(Dst+1+3*o, Dst+0+3*o,
t1i - t4i, t1r - t4r,
Table[m*4], Table[1+m*4]);
Dst += 2;
Src += 2;
Table += 2;
}
Src += 6*m;
Table -= 2*m;
}
}
static void AAC_Post(real_t* Dst,const real_t* Src,const real_t* Table,int n)
{
int i;
real_t* Dst2;
Dst += n;
Dst2 = Dst+n;
n >>= 1;
for (i=0;i<n;++i)
{
ComplexMult(Dst2-1,Dst-1,Src[1],Src[0],-Table[0],-Table[1]);
// ComplexMult(Dst2-1,Dst-1,Src[1],Src[0],Table[0],Table[1]); Dst2[-1] = -Dst2[-1]; Dst[-1] = -Dst[-1];
Dst -= 2; Dst2 -= 2; Src += 2; Table += 2;
}
for (i=0;i<n;++i)
{
ComplexMult(Dst,Dst2,Src[1],Src[0],Table[0],Table[1]);
Dst += 2; Dst2 += 2; Src += 2; Table += 2;
}
}
static void passf3(const uint16_t ido, const uint16_t l1, const complex_t *cc,
complex_t *ch, const complex_t *wa1, const complex_t *wa2,
const int8_t isign)
{
static real_t taur = FRAC_CONST(-0.5);
static real_t taui = FRAC_CONST(0.866025403784439);
uint16_t i, k, ac, ah;
complex_t c2, c3, d2, d3, t2;
if (ido == 1)
{
if (isign == 1)
{
for (k = 0; k < l1; k++)
{
ac = 3*k+1;
ah = k;
RE(t2) = RE(cc[ac]) + RE(cc[ac+1]);
IM(t2) = IM(cc[ac]) + IM(cc[ac+1]);
RE(c2) = RE(cc[ac-1]) + MUL_F(RE(t2),taur);
IM(c2) = IM(cc[ac-1]) + MUL_F(IM(t2),taur);
RE(ch[ah]) = RE(cc[ac-1]) + RE(t2);
IM(ch[ah]) = IM(cc[ac-1]) + IM(t2);
RE(c3) = MUL_F((RE(cc[ac]) - RE(cc[ac+1])), taui);
IM(c3) = MUL_F((IM(cc[ac]) - IM(cc[ac+1])), taui);
RE(ch[ah+l1]) = RE(c2) - IM(c3);
IM(ch[ah+l1]) = IM(c2) + RE(c3);
RE(ch[ah+2*l1]) = RE(c2) + IM(c3);
IM(ch[ah+2*l1]) = IM(c2) - RE(c3);
}
} else {
for (k = 0; k < l1; k++)
{
ac = 3*k+1;
ah = k;
RE(t2) = RE(cc[ac]) + RE(cc[ac+1]);
IM(t2) = IM(cc[ac]) + IM(cc[ac+1]);
RE(c2) = RE(cc[ac-1]) + MUL_F(RE(t2),taur);
IM(c2) = IM(cc[ac-1]) + MUL_F(IM(t2),taur);
RE(ch[ah]) = RE(cc[ac-1]) + RE(t2);
IM(ch[ah]) = IM(cc[ac-1]) + IM(t2);
RE(c3) = MUL_F((RE(cc[ac]) - RE(cc[ac+1])), taui);
IM(c3) = MUL_F((IM(cc[ac]) - IM(cc[ac+1])), taui);
RE(ch[ah+l1]) = RE(c2) + IM(c3);
IM(ch[ah+l1]) = IM(c2) - RE(c3);
RE(ch[ah+2*l1]) = RE(c2) - IM(c3);
IM(ch[ah+2*l1]) = IM(c2) + RE(c3);
}
}
} else {
if (isign == 1)
{
for (k = 0; k < l1; k++)
{
for (i = 0; i < ido; i++)
{
ac = i + (3*k+1)*ido;
ah = i + k * ido;
RE(t2) = RE(cc[ac]) + RE(cc[ac+ido]);
RE(c2) = RE(cc[ac-ido]) + MUL_F(RE(t2),taur);
IM(t2) = IM(cc[ac]) + IM(cc[ac+ido]);
IM(c2) = IM(cc[ac-ido]) + MUL_F(IM(t2),taur);
RE(ch[ah]) = RE(cc[ac-ido]) + RE(t2);
IM(ch[ah]) = IM(cc[ac-ido]) + IM(t2);
RE(c3) = MUL_F((RE(cc[ac]) - RE(cc[ac+ido])), taui);
IM(c3) = MUL_F((IM(cc[ac]) - IM(cc[ac+ido])), taui);
RE(d2) = RE(c2) - IM(c3);
IM(d3) = IM(c2) - RE(c3);
RE(d3) = RE(c2) + IM(c3);
IM(d2) = IM(c2) + RE(c3);
#if 1
ComplexMult(&IM(ch[ah+l1*ido]), &RE(ch[ah+l1*ido]),
IM(d2), RE(d2), RE(wa1[i]), IM(wa1[i]));
ComplexMult(&IM(ch[ah+2*l1*ido]), &RE(ch[ah+2*l1*ido]),
IM(d3), RE(d3), RE(wa2[i]), IM(wa2[i]));
#else
ComplexMult(&RE(ch[ah+l1*ido]), &IM(ch[ah+l1*ido]),
RE(d2), IM(d2), RE(wa1[i]), IM(wa1[i]));
ComplexMult(&RE(ch[ah+2*l1*ido]), &IM(ch[ah+2*l1*ido]),
RE(d3), IM(d3), RE(wa2[i]), IM(wa2[i]));
#endif
}
}
} else {
for (k = 0; k < l1; k++)
{
for (i = 0; i < ido; i++)
{
ac = i + (3*k+1)*ido;
ah = i + k * ido;
RE(t2) = RE(cc[ac]) + RE(cc[ac+ido]);
RE(c2) = RE(cc[ac-ido]) + MUL_F(RE(t2),taur);
IM(t2) = IM(cc[ac]) + IM(cc[ac+ido]);
IM(c2) = IM(cc[ac-ido]) + MUL_F(IM(t2),taur);
RE(ch[ah]) = RE(cc[ac-ido]) + RE(t2);
IM(ch[ah]) = IM(cc[ac-ido]) + IM(t2);
RE(c3) = MUL_F((RE(cc[ac]) - RE(cc[ac+ido])), taui);
IM(c3) = MUL_F((IM(cc[ac]) - IM(cc[ac+ido])), taui);
RE(d2) = RE(c2) + IM(c3);
IM(d3) = IM(c2) + RE(c3);
RE(d3) = RE(c2) - IM(c3);
IM(d2) = IM(c2) - RE(c3);
#if 1
ComplexMult(&RE(ch[ah+l1*ido]), &IM(ch[ah+l1*ido]),
RE(d2), IM(d2), RE(wa1[i]), IM(wa1[i]));
ComplexMult(&RE(ch[ah+2*l1*ido]), &IM(ch[ah+2*l1*ido]),
RE(d3), IM(d3), RE(wa2[i]), IM(wa2[i]));
#else
ComplexMult(&IM(ch[ah+l1*ido]), &RE(ch[ah+l1*ido]),
IM(d2), RE(d2), RE(wa1[i]), IM(wa1[i]));
ComplexMult(&IM(ch[ah+2*l1*ido]), &RE(ch[ah+2*l1*ido]),
IM(d3), RE(d3), RE(wa2[i]), IM(wa2[i]));
#endif
}
}
}
}
}
void faad_mdct(mdct_info *mdct, real_t *X_in, real_t *X_out)
{
uint16_t k;
complex_t x;
ALIGN complex_t Z1[512];
complex_t *sincos = mdct->sincos;
uint16_t N = mdct->N;
uint16_t N2 = N >> 1;
uint16_t N4 = N >> 2;
uint16_t N8 = N >> 3;
#ifndef FIXED_POINT
real_t scale = REAL_CONST(N);
#else
real_t scale = REAL_CONST(4.0/N);
#endif
#ifdef ALLOW_SMALL_FRAMELENGTH
#ifdef FIXED_POINT
/* detect non-power of 2 */
if (N & (N-1))
{
/* adjust scale for non-power of 2 MDCT */
/* *= sqrt(2048/1920) */
scale = MUL_C(scale, COEF_CONST(1.0327955589886444));
}
#endif
#endif
/* pre-FFT complex multiplication */
for (k = 0; k < N8; k++)
{
uint16_t n = k << 1;
RE(x) = X_in[N - N4 - 1 - n] + X_in[N - N4 + n];
IM(x) = X_in[ N4 + n] - X_in[ N4 - 1 - n];
ComplexMult(&RE(Z1[k]), &IM(Z1[k]),
RE(x), IM(x), RE(sincos[k]), IM(sincos[k]));
RE(Z1[k]) = MUL_R(RE(Z1[k]), scale);
IM(Z1[k]) = MUL_R(IM(Z1[k]), scale);
RE(x) = X_in[N2 - 1 - n] - X_in[ n];
IM(x) = X_in[N2 + n] + X_in[N - 1 - n];
ComplexMult(&RE(Z1[k + N8]), &IM(Z1[k + N8]),
RE(x), IM(x), RE(sincos[k + N8]), IM(sincos[k + N8]));
RE(Z1[k + N8]) = MUL_R(RE(Z1[k + N8]), scale);
IM(Z1[k + N8]) = MUL_R(IM(Z1[k + N8]), scale);
}
/* complex FFT, any non-scaling FFT can be used here */
cfftf(mdct->cfft, Z1);
/* post-FFT complex multiplication */
for (k = 0; k < N4; k++)
{
uint16_t n = k << 1;
ComplexMult(&RE(x), &IM(x),
RE(Z1[k]), IM(Z1[k]), RE(sincos[k]), IM(sincos[k]));
X_out[ n] = -RE(x);
X_out[N2 - 1 - n] = IM(x);
X_out[N2 + n] = -IM(x);
X_out[N - 1 - n] = RE(x);
}
}
void faad_imdct(mdct_info *mdct, real_t *X_in, real_t *X_out)
{
uint16_t k;
complex_t x;
#ifdef ALLOW_SMALL_FRAMELENGTH
#ifdef FIXED_POINT
real_t scale, b_scale = 0;
#endif
#endif
ALIGN complex_t Z1[512];
complex_t *sincos = mdct->sincos;
uint16_t N = mdct->N;
uint16_t N2 = N >> 1;
uint16_t N4 = N >> 2;
uint16_t N8 = N >> 3;
#ifdef PROFILE
int64_t count1, count2 = faad_get_ts();
#endif
#ifdef ALLOW_SMALL_FRAMELENGTH
#ifdef FIXED_POINT
/* detect non-power of 2 */
if (N & (N-1))
{
/* adjust scale for non-power of 2 MDCT */
/* 2048/1920 */
b_scale = 1;
scale = COEF_CONST(1.0666666666666667);
}
#endif
#endif
/* pre-IFFT complex multiplication */
for (k = 0; k < N4; k++)
{
ComplexMult(&IM(Z1[k]), &RE(Z1[k]),
X_in[2*k], X_in[N2 - 1 - 2*k], RE(sincos[k]), IM(sincos[k]));
}
#ifdef PROFILE
count1 = faad_get_ts();
#endif
/* complex IFFT, any non-scaling FFT can be used here */
cfftb(mdct->cfft, Z1);
#ifdef PROFILE
count1 = faad_get_ts() - count1;
#endif
/* post-IFFT complex multiplication */
for (k = 0; k < N4; k++)
{
RE(x) = RE(Z1[k]);
IM(x) = IM(Z1[k]);
ComplexMult(&IM(Z1[k]), &RE(Z1[k]),
IM(x), RE(x), RE(sincos[k]), IM(sincos[k]));
#ifdef ALLOW_SMALL_FRAMELENGTH
#ifdef FIXED_POINT
/* non-power of 2 MDCT scaling */
if (b_scale)
{
RE(Z1[k]) = MUL_C(RE(Z1[k]), scale);
IM(Z1[k]) = MUL_C(IM(Z1[k]), scale);
}
#endif
#endif
}
/* reordering */
for (k = 0; k < N8; k+=2)
{
X_out[ 2*k] = IM(Z1[N8 + k]);
X_out[ 2 + 2*k] = IM(Z1[N8 + 1 + k]);
X_out[ 1 + 2*k] = -RE(Z1[N8 - 1 - k]);
X_out[ 3 + 2*k] = -RE(Z1[N8 - 2 - k]);
X_out[N4 + 2*k] = RE(Z1[ k]);
X_out[N4 + + 2 + 2*k] = RE(Z1[ 1 + k]);
X_out[N4 + 1 + 2*k] = -IM(Z1[N4 - 1 - k]);
X_out[N4 + 3 + 2*k] = -IM(Z1[N4 - 2 - k]);
X_out[N2 + 2*k] = RE(Z1[N8 + k]);
X_out[N2 + + 2 + 2*k] = RE(Z1[N8 + 1 + k]);
X_out[N2 + 1 + 2*k] = -IM(Z1[N8 - 1 - k]);
X_out[N2 + 3 + 2*k] = -IM(Z1[N8 - 2 - k]);
X_out[N2 + N4 + 2*k] = -IM(Z1[ k]);
X_out[N2 + N4 + 2 + 2*k] = -IM(Z1[ 1 + k]);
X_out[N2 + N4 + 1 + 2*k] = RE(Z1[N4 - 1 - k]);
X_out[N2 + N4 + 3 + 2*k] = RE(Z1[N4 - 2 - k]);
}
#ifdef PROFILE
count2 = faad_get_ts() - count2;
mdct->fft_cycles += count1;
mdct->cycles += (count2 - count1);
#endif
}
void faad_imdct(mdct_info *mdct, real_t *X_in, real_t *X_out)
{
uint16_t k;
complex_t x;
ALIGN complex_t Z1[512];
complex_t *sincos = mdct->sincos;
uint16_t N = mdct->N;
uint16_t N2 = N >> 1;
uint16_t N4 = N >> 2;
uint16_t N8 = N >> 3;
#ifdef PROFILE
int64_t count1, count2 = faad_get_ts();
#endif
/* pre-IFFT complex multiplication */
for (k = 0; k < N4; k++)
{
ComplexMult(&IM(Z1[k]), &RE(Z1[k]),
X_in[2*k], X_in[N2 - 1 - 2*k], RE(sincos[k]), IM(sincos[k]));
}
#ifdef PROFILE
count1 = faad_get_ts();
#endif
/* complex IFFT, any non-scaling FFT can be used here */
cfftb(mdct->cfft, Z1);
#ifdef PROFILE
count1 = faad_get_ts() - count1;
#endif
/* post-IFFT complex multiplication */
for (k = 0; k < N4; k++)
{
RE(x) = RE(Z1[k]);
IM(x) = IM(Z1[k]);
ComplexMult(&IM(Z1[k]), &RE(Z1[k]),
IM(x), RE(x), RE(sincos[k]), IM(sincos[k]));
}
/* reordering */
for (k = 0; k < N8; k+=2)
{
X_out[ 2*k] = IM(Z1[N8 + k]);
X_out[ 2 + 2*k] = IM(Z1[N8 + 1 + k]);
X_out[ 1 + 2*k] = -RE(Z1[N8 - 1 - k]);
X_out[ 3 + 2*k] = -RE(Z1[N8 - 2 - k]);
X_out[N4 + 2*k] = RE(Z1[ k]);
X_out[N4 + + 2 + 2*k] = RE(Z1[ 1 + k]);
X_out[N4 + 1 + 2*k] = -IM(Z1[N4 - 1 - k]);
X_out[N4 + 3 + 2*k] = -IM(Z1[N4 - 2 - k]);
X_out[N2 + 2*k] = RE(Z1[N8 + k]);
X_out[N2 + + 2 + 2*k] = RE(Z1[N8 + 1 + k]);
X_out[N2 + 1 + 2*k] = -IM(Z1[N8 - 1 - k]);
X_out[N2 + 3 + 2*k] = -IM(Z1[N8 - 2 - k]);
X_out[N2 + N4 + 2*k] = -IM(Z1[ k]);
X_out[N2 + N4 + 2 + 2*k] = -IM(Z1[ 1 + k]);
X_out[N2 + N4 + 1 + 2*k] = RE(Z1[N4 - 1 - k]);
X_out[N2 + N4 + 3 + 2*k] = RE(Z1[N4 - 2 - k]);
}
#ifdef PROFILE
count2 = faad_get_ts() - count2;
mdct->fft_cycles += count1;
mdct->cycles += (count2 - count1);
#endif
}
