static void kf_bfly2(
kiss_fft_cpx * Fout,
int m,
int N
)
{
kiss_fft_cpx * Fout2;
int i;
(void)m;
#ifdef CUSTOM_MODES
if (m==1)
{
celt_assert(m==1);
for (i=0;i<N;i++)
{
kiss_fft_cpx t;
Fout2 = Fout + 1;
t = *Fout2;
C_SUB( *Fout2 , *Fout , t );
C_ADDTO( *Fout , t );
Fout += 2;
}
} else
#endif
{
opus_val16 tw;
tw = QCONST16(0.7071067812f, 15);
/* We know that m==4 here because the radix-2 is just after a radix-4 */
celt_assert(m==4);
for (i=0;i<N;i++)
{
kiss_fft_cpx t;
Fout2 = Fout + 4;
t = Fout2[0];
C_SUB( Fout2[0] , Fout[0] , t );
C_ADDTO( Fout[0] , t );
t.r = S_MUL(ADD32_ovflw(Fout2[1].r, Fout2[1].i), tw);
t.i = S_MUL(SUB32_ovflw(Fout2[1].i, Fout2[1].r), tw);
C_SUB( Fout2[1] , Fout[1] , t );
C_ADDTO( Fout[1] , t );
t.r = Fout2[2].i;
t.i = -Fout2[2].r;
C_SUB( Fout2[2] , Fout[2] , t );
C_ADDTO( Fout[2] , t );
t.r = S_MUL(SUB32_ovflw(Fout2[3].i, Fout2[3].r), tw);
t.i = S_MUL(NEG32_ovflw(ADD32_ovflw(Fout2[3].i, Fout2[3].r)), tw);
C_SUB( Fout2[3] , Fout[3] , t );
C_ADDTO( Fout[3] , t );
Fout += 8;
}
}
}
static void kf_bfly5(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
int m
)
{
kiss_fft_cpx *Fout0,*Fout1,*Fout2,*Fout3,*Fout4;
int u;
kiss_fft_cpx scratch[13];
kiss_fft_cpx * twiddles = st->twiddles;
kiss_fft_cpx *tw;
kiss_fft_cpx ya,yb;
ya = twiddles[fstride*m];
yb = twiddles[fstride*2*m];
Fout0=Fout;
Fout1=Fout0+m;
Fout2=Fout0+2*m;
Fout3=Fout0+3*m;
Fout4=Fout0+4*m;
tw=st->twiddles;
for ( u=0; u<m; ++u ) {
if (!st->inverse) {
C_FIXDIV( *Fout0,5); C_FIXDIV( *Fout1,5); C_FIXDIV( *Fout2,5); C_FIXDIV( *Fout3,5); C_FIXDIV( *Fout4,5);
}
scratch[0] = *Fout0;
C_MUL(scratch[1] ,*Fout1, tw[u*fstride]);
C_MUL(scratch[2] ,*Fout2, tw[2*u*fstride]);
C_MUL(scratch[3] ,*Fout3, tw[3*u*fstride]);
C_MUL(scratch[4] ,*Fout4, tw[4*u*fstride]);
C_ADD( scratch[7],scratch[1],scratch[4]);
C_SUB( scratch[10],scratch[1],scratch[4]);
C_ADD( scratch[8],scratch[2],scratch[3]);
C_SUB( scratch[9],scratch[2],scratch[3]);
Fout0->r += scratch[7].r + scratch[8].r;
Fout0->i += scratch[7].i + scratch[8].i;
scratch[5].r = scratch[0].r + S_MUL(scratch[7].r,ya.r) + S_MUL(scratch[8].r,yb.r);
scratch[5].i = scratch[0].i + S_MUL(scratch[7].i,ya.r) + S_MUL(scratch[8].i,yb.r);
scratch[6].r = S_MUL(scratch[10].i,ya.i) + S_MUL(scratch[9].i,yb.i);
scratch[6].i = -S_MUL(scratch[10].r,ya.i) - S_MUL(scratch[9].r,yb.i);
C_SUB(*Fout1,scratch[5],scratch[6]);
C_ADD(*Fout4,scratch[5],scratch[6]);
scratch[11].r = scratch[0].r + S_MUL(scratch[7].r,yb.r) + S_MUL(scratch[8].r,ya.r);
scratch[11].i = scratch[0].i + S_MUL(scratch[7].i,yb.r) + S_MUL(scratch[8].i,ya.r);
scratch[12].r = - S_MUL(scratch[10].i,yb.i) + S_MUL(scratch[9].i,ya.i);
scratch[12].i = S_MUL(scratch[10].r,yb.i) - S_MUL(scratch[9].r,ya.i);
C_ADD(*Fout2,scratch[11],scratch[12]);
C_SUB(*Fout3,scratch[11],scratch[12]);
++Fout0;++Fout1;++Fout2;++Fout3;++Fout4;
}
}
static void kf_bfly5(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_state *st,
int m,
int N,
int mm
)
{
kiss_fft_cpx *Fout0,*Fout1,*Fout2,*Fout3,*Fout4;
int i, u;
kiss_fft_cpx scratch[13];
const kiss_twiddle_cpx *tw;
kiss_twiddle_cpx ya,yb;
kiss_fft_cpx * Fout_beg = Fout;
#ifdef FIXED_POINT
ya.r = 10126;
ya.i = -31164;
yb.r = -26510;
yb.i = -19261;
#else
ya = st->twiddles[fstride*m];
yb = st->twiddles[fstride*2*m];
#endif
tw=st->twiddles;
for (i=0;i<N;i++)
{
Fout = Fout_beg + i*mm;
Fout0=Fout;
Fout1=Fout0+m;
Fout2=Fout0+2*m;
Fout3=Fout0+3*m;
Fout4=Fout0+4*m;
/* For non-custom modes, m is guaranteed to be a multiple of 4. */
for ( u=0; u<m; ++u ) {
scratch[0] = *Fout0;
C_MUL(scratch[1] ,*Fout1, tw[u*fstride]);
C_MUL(scratch[2] ,*Fout2, tw[2*u*fstride]);
C_MUL(scratch[3] ,*Fout3, tw[3*u*fstride]);
C_MUL(scratch[4] ,*Fout4, tw[4*u*fstride]);
C_ADD( scratch[7],scratch[1],scratch[4]);
C_SUB( scratch[10],scratch[1],scratch[4]);
C_ADD( scratch[8],scratch[2],scratch[3]);
C_SUB( scratch[9],scratch[2],scratch[3]);
Fout0->r = ADD32_ovflw(Fout0->r, ADD32_ovflw(scratch[7].r, scratch[8].r));
Fout0->i = ADD32_ovflw(Fout0->i, ADD32_ovflw(scratch[7].i, scratch[8].i));
scratch[5].r = ADD32_ovflw(scratch[0].r, ADD32_ovflw(S_MUL(scratch[7].r,ya.r), S_MUL(scratch[8].r,yb.r)));
scratch[5].i = ADD32_ovflw(scratch[0].i, ADD32_ovflw(S_MUL(scratch[7].i,ya.r), S_MUL(scratch[8].i,yb.r)));
scratch[6].r = ADD32_ovflw(S_MUL(scratch[10].i,ya.i), S_MUL(scratch[9].i,yb.i));
scratch[6].i = NEG32_ovflw(ADD32_ovflw(S_MUL(scratch[10].r,ya.i), S_MUL(scratch[9].r,yb.i)));
C_SUB(*Fout1,scratch[5],scratch[6]);
C_ADD(*Fout4,scratch[5],scratch[6]);
scratch[11].r = ADD32_ovflw(scratch[0].r, ADD32_ovflw(S_MUL(scratch[7].r,yb.r), S_MUL(scratch[8].r,ya.r)));
scratch[11].i = ADD32_ovflw(scratch[0].i, ADD32_ovflw(S_MUL(scratch[7].i,yb.r), S_MUL(scratch[8].i,ya.r)));
scratch[12].r = SUB32_ovflw(S_MUL(scratch[9].i,ya.i), S_MUL(scratch[10].i,yb.i));
scratch[12].i = SUB32_ovflw(S_MUL(scratch[10].r,yb.i), S_MUL(scratch[9].r,ya.i));
C_ADD(*Fout2,scratch[11],scratch[12]);
C_SUB(*Fout3,scratch[11],scratch[12]);
++Fout0;++Fout1;++Fout2;++Fout3;++Fout4;
}
}
}
static void ki_bfly5(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_state *st,
int m,
int N,
int mm
)
{
kiss_fft_cpx *Fout0,*Fout1,*Fout2,*Fout3,*Fout4;
int i, u;
kiss_fft_cpx scratch[13];
const kiss_twiddle_cpx * twiddles = st->twiddles;
const kiss_twiddle_cpx *tw;
kiss_twiddle_cpx ya,yb;
kiss_fft_cpx * Fout_beg = Fout;
ya = twiddles[fstride*m];
yb = twiddles[fstride*2*m];
tw=st->twiddles;
for (i=0;i<N;i++)
{
Fout = Fout_beg + i*mm;
Fout0=Fout;
Fout1=Fout0+m;
Fout2=Fout0+2*m;
Fout3=Fout0+3*m;
Fout4=Fout0+4*m;
for ( u=0; u<m; ++u ) {
scratch[0] = *Fout0;
C_MULC(scratch[1] ,*Fout1, tw[u*fstride]);
C_MULC(scratch[2] ,*Fout2, tw[2*u*fstride]);
C_MULC(scratch[3] ,*Fout3, tw[3*u*fstride]);
C_MULC(scratch[4] ,*Fout4, tw[4*u*fstride]);
C_ADD( scratch[7],scratch[1],scratch[4]);
C_SUB( scratch[10],scratch[1],scratch[4]);
C_ADD( scratch[8],scratch[2],scratch[3]);
C_SUB( scratch[9],scratch[2],scratch[3]);
Fout0->r += scratch[7].r + scratch[8].r;
Fout0->i += scratch[7].i + scratch[8].i;
scratch[5].r = scratch[0].r + S_MUL(scratch[7].r,ya.r) + S_MUL(scratch[8].r,yb.r);
scratch[5].i = scratch[0].i + S_MUL(scratch[7].i,ya.r) + S_MUL(scratch[8].i,yb.r);
scratch[6].r = -S_MUL(scratch[10].i,ya.i) - S_MUL(scratch[9].i,yb.i);
scratch[6].i = S_MUL(scratch[10].r,ya.i) + S_MUL(scratch[9].r,yb.i);
C_SUB(*Fout1,scratch[5],scratch[6]);
C_ADD(*Fout4,scratch[5],scratch[6]);
scratch[11].r = scratch[0].r + S_MUL(scratch[7].r,yb.r) + S_MUL(scratch[8].r,ya.r);
scratch[11].i = scratch[0].i + S_MUL(scratch[7].i,yb.r) + S_MUL(scratch[8].i,ya.r);
scratch[12].r = S_MUL(scratch[10].i,yb.i) - S_MUL(scratch[9].i,ya.i);
scratch[12].i = -S_MUL(scratch[10].r,yb.i) + S_MUL(scratch[9].r,ya.i);
C_ADD(*Fout2,scratch[11],scratch[12]);
C_SUB(*Fout3,scratch[11],scratch[12]);
++Fout0;++Fout1;++Fout2;++Fout3;++Fout4;
}
}
}
