static void
kf_bfly4 (kiss_fft_s16_cpx * Fout,
const size_t fstride, const kiss_fft_s16_cfg st, const size_t m)
{
kiss_fft_s16_cpx *tw1, *tw2, *tw3;
kiss_fft_s16_cpx scratch[6];
size_t k = m;
const size_t m2 = 2 * m;
const size_t m3 = 3 * m;
tw3 = tw2 = tw1 = st->twiddles;
do {
C_FIXDIV (*Fout, 4);
C_FIXDIV (Fout[m], 4);
C_FIXDIV (Fout[m2], 4);
C_FIXDIV (Fout[m3], 4);
C_MUL (scratch[0], Fout[m], *tw1);
C_MUL (scratch[1], Fout[m2], *tw2);
C_MUL (scratch[2], Fout[m3], *tw3);
C_SUB (scratch[5], *Fout, scratch[1]);
C_ADDTO (*Fout, scratch[1]);
C_ADD (scratch[3], scratch[0], scratch[2]);
C_SUB (scratch[4], scratch[0], scratch[2]);
C_SUB (Fout[m2], *Fout, scratch[3]);
tw1 += fstride;
tw2 += fstride * 2;
tw3 += fstride * 3;
C_ADDTO (*Fout, scratch[3]);
if (st->inverse) {
Fout[m].r = scratch[5].r - scratch[4].i;
Fout[m].i = scratch[5].i + scratch[4].r;
Fout[m3].r = scratch[5].r + scratch[4].i;
Fout[m3].i = scratch[5].i - scratch[4].r;
} else {
Fout[m].r = scratch[5].r + scratch[4].i;
Fout[m].i = scratch[5].i - scratch[4].r;
Fout[m3].r = scratch[5].r - scratch[4].i;
Fout[m3].i = scratch[5].i + scratch[4].r;
}
++Fout;
} while (--k);
}
static void ki_bfly4(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_state *st,
int m,
int N,
int mm
)
{
const kiss_twiddle_cpx *tw1,*tw2,*tw3;
kiss_fft_cpx scratch[6];
const size_t m2=2*m;
const size_t m3=3*m;
int i, j;
kiss_fft_cpx * Fout_beg = Fout;
for (i=0;i<N;i++)
{
Fout = Fout_beg + i*mm;
tw3 = tw2 = tw1 = st->twiddles;
for (j=0;j<m;j++)
{
C_MULC(scratch[0],Fout[m] , *tw1 );
C_MULC(scratch[1],Fout[m2] , *tw2 );
C_MULC(scratch[2],Fout[m3] , *tw3 );
C_SUB( scratch[5] , *Fout, scratch[1] );
C_ADDTO(*Fout, scratch[1]);
C_ADD( scratch[3] , scratch[0] , scratch[2] );
C_SUB( scratch[4] , scratch[0] , scratch[2] );
C_SUB( Fout[m2], *Fout, scratch[3] );
tw1 += fstride;
tw2 += fstride*2;
tw3 += fstride*3;
C_ADDTO( *Fout , scratch[3] );
Fout[m].r = scratch[5].r - scratch[4].i;
Fout[m].i = scratch[5].i + scratch[4].r;
Fout[m3].r = scratch[5].r + scratch[4].i;
Fout[m3].i = scratch[5].i - scratch[4].r;
++Fout;
}
}
}
static void kf_bfly2(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
int m,
int N,
int mm
)
{
kiss_fft_cpx * Fout2;
kiss_fft_cpx * tw1;
kiss_fft_cpx t;
if (!st->inverse) {
int i,j;
kiss_fft_cpx * Fout_beg = Fout;
for (i=0;i<N;i++)
{
Fout = Fout_beg + i*mm;
Fout2 = Fout + m;
tw1 = st->twiddles;
for(j=0;j<m;j++)
{
/* Almost the same as the code path below, except that we divide the input by two
(while keeping the best accuracy possible) */
ms_word32_t tr, ti;
tr = SHR32(SUB32(MULT16_16(Fout2->r , tw1->r),MULT16_16(Fout2->i , tw1->i)), 1);
ti = SHR32(ADD32(MULT16_16(Fout2->i , tw1->r),MULT16_16(Fout2->r , tw1->i)), 1);
tw1 += fstride;
Fout2->r = PSHR32(SUB32(SHL32(EXTEND32(Fout->r), 14), tr), 15);
Fout2->i = PSHR32(SUB32(SHL32(EXTEND32(Fout->i), 14), ti), 15);
Fout->r = PSHR32(ADD32(SHL32(EXTEND32(Fout->r), 14), tr), 15);
Fout->i = PSHR32(ADD32(SHL32(EXTEND32(Fout->i), 14), ti), 15);
++Fout2;
++Fout;
}
}
} else {
int i,j;
kiss_fft_cpx * Fout_beg = Fout;
for (i=0;i<N;i++)
{
Fout = Fout_beg + i*mm;
Fout2 = Fout + m;
tw1 = st->twiddles;
for(j=0;j<m;j++)
{
C_MUL (t, *Fout2 , *tw1);
tw1 += fstride;
C_SUB( *Fout2 , *Fout , t );
C_ADDTO( *Fout , t );
++Fout2;
++Fout;
}
}
}
}
static void kf_bfly3(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_state *st,
int m,
int N,
int mm
)
{
int i;
size_t k;
const size_t m2 = 2*m;
const kiss_twiddle_cpx *tw1,*tw2;
kiss_fft_cpx scratch[5];
kiss_twiddle_cpx epi3;
kiss_fft_cpx * Fout_beg = Fout;
#ifdef FIXED_POINT
/*epi3.r = -16384;*/ /* Unused */
epi3.i = -28378;
#else
epi3 = st->twiddles[fstride*m];
#endif
for (i=0;i<N;i++)
{
Fout = Fout_beg + i*mm;
tw1=tw2=st->twiddles;
/* For non-custom modes, m is guaranteed to be a multiple of 4. */
k=m;
do {
C_MUL(scratch[1],Fout[m] , *tw1);
C_MUL(scratch[2],Fout[m2] , *tw2);
C_ADD(scratch[3],scratch[1],scratch[2]);
C_SUB(scratch[0],scratch[1],scratch[2]);
tw1 += fstride;
tw2 += fstride*2;
Fout[m].r = SUB32_ovflw(Fout->r, HALF_OF(scratch[3].r));
Fout[m].i = SUB32_ovflw(Fout->i, HALF_OF(scratch[3].i));
C_MULBYSCALAR( scratch[0] , epi3.i );
C_ADDTO(*Fout,scratch[3]);
Fout[m2].r = ADD32_ovflw(Fout[m].r, scratch[0].i);
Fout[m2].i = SUB32_ovflw(Fout[m].i, scratch[0].r);
Fout[m].r = SUB32_ovflw(Fout[m].r, scratch[0].i);
Fout[m].i = ADD32_ovflw(Fout[m].i, scratch[0].r);
++Fout;
} while(--k);
}
}
static void kf_bfly3(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_state *st,
int m,
int N,
int mm
)
{
int i;
size_t k;
const size_t m2 = 2*m;
const kiss_twiddle_cpx *tw1,*tw2;
kiss_fft_cpx scratch[5];
kiss_twiddle_cpx epi3;
kiss_fft_cpx * Fout_beg = Fout;
epi3 = st->twiddles[fstride*m];
for (i=0;i<N;i++)
{
Fout = Fout_beg + i*mm;
tw1=tw2=st->twiddles;
k=m;
do {
C_FIXDIV(*Fout,3); C_FIXDIV(Fout[m],3); C_FIXDIV(Fout[m2],3);
C_MUL(scratch[1],Fout[m] , *tw1);
C_MUL(scratch[2],Fout[m2] , *tw2);
C_ADD(scratch[3],scratch[1],scratch[2]);
C_SUB(scratch[0],scratch[1],scratch[2]);
tw1 += fstride;
tw2 += fstride*2;
Fout[m].r = Fout->r - HALF_OF(scratch[3].r);
Fout[m].i = Fout->i - HALF_OF(scratch[3].i);
C_MULBYSCALAR( scratch[0] , epi3.i );
C_ADDTO(*Fout,scratch[3]);
Fout[m2].r = Fout[m].r + scratch[0].i;
Fout[m2].i = Fout[m].i - scratch[0].r;
Fout[m].r -= scratch[0].i;
Fout[m].i += scratch[0].r;
++Fout;
} while(--k);
}
}
static void kf_bfly3(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
size_t m
)
{
size_t k=m;
const size_t m2 = 2*m;
kiss_fft_cpx *tw1,*tw2;
kiss_fft_cpx scratch[5];
kiss_fft_cpx epi3;
epi3 = st->twiddles[fstride*m];
tw1=tw2=st->twiddles;
do {
if (!st->inverse) {
C_FIXDIV(*Fout,3);
C_FIXDIV(Fout[m],3);
C_FIXDIV(Fout[m2],3);
}
C_MUL(scratch[1],Fout[m] , *tw1);
C_MUL(scratch[2],Fout[m2] , *tw2);
C_ADD(scratch[3],scratch[1],scratch[2]);
C_SUB(scratch[0],scratch[1],scratch[2]);
tw1 += fstride;
tw2 += fstride*2;
Fout[m].r = Fout->r - HALF_OF(scratch[3].r);
Fout[m].i = Fout->i - HALF_OF(scratch[3].i);
C_MULBYSCALAR( scratch[0] , epi3.i );
C_ADDTO(*Fout,scratch[3]);
Fout[m2].r = Fout[m].r + scratch[0].i;
Fout[m2].i = Fout[m].i - scratch[0].r;
Fout[m].r -= scratch[0].i;
Fout[m].i += scratch[0].r;
++Fout;
} while(--k);
}
static void kf_bfly2(kiss_fft_cpx *Fout, const size_t fstride,
const kiss_fft_cfg st, int m) {
kiss_fft_cpx *Fout2;
kiss_fft_cpx *tw1 = st->twiddles;
kiss_fft_cpx t;
Fout2 = Fout + m;
do {
C_FIXDIV(*Fout, 2);
C_FIXDIV(*Fout2, 2);
C_MUL(t, *Fout2, *tw1);
tw1 += fstride;
C_SUB(*Fout2, *Fout, t);
C_ADDTO(*Fout, t);
++Fout2;
++Fout;
} while (--m);
}
/* perform the butterfly for one stage of a mixed radix FFT */
static void kf_bfly_generic(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
int m,
int p
)
{
int u,k,q1,q;
kiss_fft_cpx * twiddles = st->twiddles;
kiss_fft_cpx t;
kiss_fft_cpx scratchbuf[17];
int Norig = st->nfft;
/*CHECKBUF(scratchbuf,nscratchbuf,p);*/
if (p>17)
ms_fatal("KissFFT: max radix supported is 17");
for ( u=0; u<m; ++u ) {
k=u;
for ( q1=0 ; q1<p ; ++q1 ) {
scratchbuf[q1] = Fout[ k ];
if (!st->inverse) {
C_FIXDIV(scratchbuf[q1],p);
}
k += m;
}
k=u;
for ( q1=0 ; q1<p ; ++q1 ) {
int twidx=0;
Fout[ k ] = scratchbuf[0];
for (q=1;q<p;++q ) {
twidx += fstride * k;
if (twidx>=Norig) twidx-=Norig;
C_MUL(t,scratchbuf[q] , twiddles[twidx] );
C_ADDTO( Fout[ k ] ,t);
}
k += m;
}
}
}
/* perform the butterfly for one stage of a mixed radix FFT */
static void kf_bfly_generic(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
int m,
int p
)
{
int u,k,q1,q;
kiss_fft_cpx * twiddles = st->twiddles;
kiss_fft_cpx t;
int Norig = st->nfft;
kiss_fft_cpx * scratch = (kiss_fft_cpx*)KISS_FFT_TMP_ALLOC(sizeof(kiss_fft_cpx)*p);
// see http://sourceforge.net/p/kissfft/bugs/9/
#pragma omp critical
for ( u=0; u<m; ++u ) {
k=u;
for ( q1=0 ; q1<p ; ++q1 ) {
scratch[q1] = Fout[ k ];
C_FIXDIV(scratch[q1],p);
k += m;
}
k=u;
for ( q1=0 ; q1<p ; ++q1 ) {
int twidx=0;
Fout[ k ] = scratch[0];
for (q=1;q<p;++q ) {
twidx += fstride * k;
if (twidx>=Norig) twidx-=Norig;
C_MUL(t,scratch[q] , twiddles[twidx] );
C_ADDTO( Fout[ k ] ,t);
}
k += m;
}
}
KISS_FFT_TMP_FREE(scratch);
}
/* perform the butterfly for one stage of a mixed radix FFT */
static void kf_bfly_generic(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
int m,
int p
)
{
int u,k,q1,q;
kiss_fft_cpx * twiddles = st->twiddles;
kiss_fft_cpx t;
int Norig = st->nfft;
kiss_fft_cpx *scratchbuf=(kiss_fft_cpx *)malloc( sizeof(kiss_fft_cpx) * p );
for ( u=0; u<m; ++u ) {
k=u;
for ( q1=0 ; q1<p ; ++q1 ) {
scratchbuf[q1] = Fout[ k ];
C_FIXDIV(scratchbuf[q1],p);
k += m;
}
k=u;
for ( q1=0 ; q1<p ; ++q1 ) {
int twidx=0;
Fout[ k ] = scratchbuf[0];
for (q=1; q<p; ++q ) {
twidx += fstride * k;
if (twidx>=Norig) twidx-=Norig;
C_MUL(t,scratchbuf[q] , twiddles[twidx] );
C_ADDTO( Fout[ k ] ,t);
}
k += m;
}
}
free( scratchbuf );
}
static void kf_bfly4(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
const size_t m
)
{
kiss_fft_cpx *tw1,*tw2,*tw3;
kiss_fft_cpx scratch[6];
size_t k=m;
const size_t m2=2*m;
const size_t m3=3*m;
tw3 = tw2 = tw1 = st->twiddles;
if (!st->inverse) {
int i;
kiss_fft_cpx *x=Fout;
for (i=0; i<4*m; i++)
{
//C_FIXDIV(x[i],4);
x[i].r = PSHR16(x[i].r,2);
x[i].i = PSHR16(x[i].i,2);
}
}
if (st->inverse)
{
do {
C_MUL(scratch[0],Fout[m] , *tw1 );
C_MUL(scratch[1],Fout[m2] , *tw2 );
C_MUL(scratch[2],Fout[m3] , *tw3 );
C_SUB( scratch[5] , *Fout, scratch[1] );
C_ADDTO(*Fout, scratch[1]);
C_ADD( scratch[3] , scratch[0] , scratch[2] );
C_SUB( scratch[4] , scratch[0] , scratch[2] );
C_SUB( Fout[m2], *Fout, scratch[3] );
tw1 += fstride;
tw2 += fstride*2;
tw3 += fstride*3;
C_ADDTO( *Fout , scratch[3] );
Fout[m].r = scratch[5].r - scratch[4].i;
Fout[m].i = scratch[5].i + scratch[4].r;
Fout[m3].r = scratch[5].r + scratch[4].i;
Fout[m3].i = scratch[5].i - scratch[4].r;
++Fout;
} while(--k);
} else
{
do {
C_MUL(scratch[0],Fout[m] , *tw1 );
C_MUL(scratch[1],Fout[m2] , *tw2 );
C_MUL(scratch[2],Fout[m3] , *tw3 );
C_SUB( scratch[5] , *Fout, scratch[1] );
C_ADDTO(*Fout, scratch[1]);
C_ADD( scratch[3] , scratch[0] , scratch[2] );
C_SUB( scratch[4] , scratch[0] , scratch[2] );
C_SUB( Fout[m2], *Fout, scratch[3] );
tw1 += fstride;
tw2 += fstride*2;
tw3 += fstride*3;
C_ADDTO( *Fout , scratch[3] );
Fout[m].r = scratch[5].r + scratch[4].i;
Fout[m].i = scratch[5].i - scratch[4].r;
Fout[m3].r = scratch[5].r - scratch[4].i;
Fout[m3].i = scratch[5].i + scratch[4].r;
++Fout;
} while(--k);
}
}
static void kf_bfly4(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_state *st,
int m,
int N,
int mm
)
{
int i;
if (m==1)
{
/* Degenerate case where all the twiddles are 1. */
for (i=0;i<N;i++)
{
kiss_fft_cpx scratch0, scratch1;
C_SUB( scratch0 , *Fout, Fout[2] );
C_ADDTO(*Fout, Fout[2]);
C_ADD( scratch1 , Fout[1] , Fout[3] );
C_SUB( Fout[2], *Fout, scratch1 );
C_ADDTO( *Fout , scratch1 );
C_SUB( scratch1 , Fout[1] , Fout[3] );
Fout[1].r = ADD32_ovflw(scratch0.r, scratch1.i);
Fout[1].i = SUB32_ovflw(scratch0.i, scratch1.r);
Fout[3].r = SUB32_ovflw(scratch0.r, scratch1.i);
Fout[3].i = ADD32_ovflw(scratch0.i, scratch1.r);
Fout+=4;
}
} else {
int j;
kiss_fft_cpx scratch[6];
const kiss_twiddle_cpx *tw1,*tw2,*tw3;
const int m2=2*m;
const int m3=3*m;
kiss_fft_cpx * Fout_beg = Fout;
for (i=0;i<N;i++)
{
Fout = Fout_beg + i*mm;
tw3 = tw2 = tw1 = st->twiddles;
/* m is guaranteed to be a multiple of 4. */
for (j=0;j<m;j++)
{
C_MUL(scratch[0],Fout[m] , *tw1 );
C_MUL(scratch[1],Fout[m2] , *tw2 );
C_MUL(scratch[2],Fout[m3] , *tw3 );
C_SUB( scratch[5] , *Fout, scratch[1] );
C_ADDTO(*Fout, scratch[1]);
C_ADD( scratch[3] , scratch[0] , scratch[2] );
C_SUB( scratch[4] , scratch[0] , scratch[2] );
C_SUB( Fout[m2], *Fout, scratch[3] );
tw1 += fstride;
tw2 += fstride*2;
tw3 += fstride*3;
C_ADDTO( *Fout , scratch[3] );
Fout[m].r = ADD32_ovflw(scratch[5].r, scratch[4].i);
Fout[m].i = SUB32_ovflw(scratch[5].i, scratch[4].r);
Fout[m3].r = SUB32_ovflw(scratch[5].r, scratch[4].i);
Fout[m3].i = ADD32_ovflw(scratch[5].i, scratch[4].r);
++Fout;
}
}
}
}