blob: 335d771dc3d12bd6165bb76aaec774bbe41cf324 [file] [log] [blame]
/*****************************************************************************
FFTReal.hpp
Copyright (c) 2005 Laurent de Soras
--- Legal stuff ---
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*Tab=3***********************************************************************/
#if defined (FFTReal_CURRENT_CODEHEADER)
#error Recursive inclusion of FFTReal code header.
#endif
#define FFTReal_CURRENT_CODEHEADER
#if ! defined (FFTReal_CODEHEADER_INCLUDED)
#define FFTReal_CODEHEADER_INCLUDED
/*\\\ INCLUDE FILES \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
#include <cassert>
#include <cmath>
static inline bool FFTReal_is_pow2 (long x)
{
assert (x > 0);
return ((x & -x) == x);
}
static inline int FFTReal_get_next_pow2 (long x)
{
--x;
int p = 0;
while ((x & ~0xFFFFL) != 0)
{
p += 16;
x >>= 16;
}
while ((x & ~0xFL) != 0)
{
p += 4;
x >>= 4;
}
while (x > 0)
{
++p;
x >>= 1;
}
return (p);
}
/*\\\ PUBLIC \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
/*
==============================================================================
Name: ctor
Input parameters:
- length: length of the array on which we want to do a FFT. Range: power of
2 only, > 0.
Throws: std::bad_alloc
==============================================================================
*/
template <class DT>
FFTReal <DT>::FFTReal (long length)
: _length (length)
, _nbr_bits (FFTReal_get_next_pow2 (length))
, _br_lut ()
, _trigo_lut ()
, _buffer (length)
, _trigo_osc ()
{
assert (FFTReal_is_pow2 (length));
assert (_nbr_bits <= MAX_BIT_DEPTH);
init_br_lut ();
init_trigo_lut ();
init_trigo_osc ();
}
/*
==============================================================================
Name: get_length
Description:
Returns the number of points processed by this FFT object.
Returns: The number of points, power of 2, > 0.
Throws: Nothing
==============================================================================
*/
template <class DT>
long FFTReal <DT>::get_length () const
{
return (_length);
}
/*
==============================================================================
Name: do_fft
Description:
Compute the FFT of the array.
Input parameters:
- x: pointer on the source array (time).
Output parameters:
- f: pointer on the destination array (frequencies).
f [0...length(x)/2] = real values,
f [length(x)/2+1...length(x)-1] = negative imaginary values of
coefficents 1...length(x)/2-1.
Throws: Nothing
==============================================================================
*/
template <class DT>
void FFTReal <DT>::do_fft (DataType f [], const DataType x []) const
{
assert (f != 0);
assert (f != use_buffer ());
assert (x != 0);
assert (x != use_buffer ());
assert (x != f);
// General case
if (_nbr_bits > 2)
{
compute_fft_general (f, x);
}
// 4-point FFT
else if (_nbr_bits == 2)
{
f [1] = x [0] - x [2];
f [3] = x [1] - x [3];
const DataType b_0 = x [0] + x [2];
const DataType b_2 = x [1] + x [3];
f [0] = b_0 + b_2;
f [2] = b_0 - b_2;
}
// 2-point FFT
else if (_nbr_bits == 1)
{
f [0] = x [0] + x [1];
f [1] = x [0] - x [1];
}
// 1-point FFT
else
{
f [0] = x [0];
}
}
/*
==============================================================================
Name: do_ifft
Description:
Compute the inverse FFT of the array. Note that data must be post-scaled:
IFFT (FFT (x)) = x * length (x).
Input parameters:
- f: pointer on the source array (frequencies).
f [0...length(x)/2] = real values
f [length(x)/2+1...length(x)-1] = negative imaginary values of
coefficents 1...length(x)/2-1.
Output parameters:
- x: pointer on the destination array (time).
Throws: Nothing
==============================================================================
*/
template <class DT>
void FFTReal <DT>::do_ifft (const DataType f [], DataType x []) const
{
assert (f != 0);
assert (f != use_buffer ());
assert (x != 0);
assert (x != use_buffer ());
assert (x != f);
// General case
if (_nbr_bits > 2)
{
compute_ifft_general (f, x);
}
// 4-point IFFT
else if (_nbr_bits == 2)
{
const DataType b_0 = f [0] + f [2];
const DataType b_2 = f [0] - f [2];
x [0] = b_0 + f [1] * 2;
x [2] = b_0 - f [1] * 2;
x [1] = b_2 + f [3] * 2;
x [3] = b_2 - f [3] * 2;
}
// 2-point IFFT
else if (_nbr_bits == 1)
{
x [0] = f [0] + f [1];
x [1] = f [0] - f [1];
}
// 1-point IFFT
else
{
x [0] = f [0];
}
}
/*
==============================================================================
Name: rescale
Description:
Scale an array by divide each element by its length. This function should
be called after FFT + IFFT.
Input parameters:
- x: pointer on array to rescale (time or frequency).
Throws: Nothing
==============================================================================
*/
template <class DT>
void FFTReal <DT>::rescale (DataType x []) const
{
const DataType mul = DataType (1.0 / _length);
if (_length < 4)
{
long i = _length - 1;
do
{
x [i] *= mul;
--i;
}
while (i >= 0);
}
else
{
assert ((_length & 3) == 0);
// Could be optimized with SIMD instruction sets (needs alignment check)
long i = _length - 4;
do
{
x [i + 0] *= mul;
x [i + 1] *= mul;
x [i + 2] *= mul;
x [i + 3] *= mul;
i -= 4;
}
while (i >= 0);
}
}
/*
==============================================================================
Name: use_buffer
Description:
Access the internal buffer, whose length is the FFT one.
Buffer content will be erased at each do_fft() / do_ifft() call!
This buffer cannot be used as:
- source for FFT or IFFT done with this object
- destination for FFT or IFFT done with this object
Returns:
Buffer start address
Throws: Nothing
==============================================================================
*/
template <class DT>
typename FFTReal <DT>::DataType * FFTReal <DT>::use_buffer () const
{
return (&_buffer [0]);
}
/*\\\ PROTECTED \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
/*\\\ PRIVATE \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
template <class DT>
void FFTReal <DT>::init_br_lut ()
{
const long length = 1L << _nbr_bits;
_br_lut.resize (length);
_br_lut [0] = 0;
long br_index = 0;
for (long cnt = 1; cnt < length; ++cnt)
{
// ++br_index (bit reversed)
long bit = length >> 1;
while (((br_index ^= bit) & bit) == 0)
{
bit >>= 1;
}
_br_lut [cnt] = br_index;
}
}
template <class DT>
void FFTReal <DT>::init_trigo_lut ()
{
using namespace std;
if (_nbr_bits > 3)
{
const long total_len = (1L << (_nbr_bits - 1)) - 4;
_trigo_lut.resize (total_len);
for (int level = 3; level < _nbr_bits; ++level)
{
const long level_len = 1L << (level - 1);
DataType * const level_ptr =
&_trigo_lut [get_trigo_level_index (level)];
const double mul = PI / (level_len << 1);
for (long i = 0; i < level_len; ++ i)
{
level_ptr [i] = static_cast <DataType> (cos (i * mul));
}
}
}
}
template <class DT>
void FFTReal <DT>::init_trigo_osc ()
{
const int nbr_osc = _nbr_bits - TRIGO_BD_LIMIT;
if (nbr_osc > 0)
{
_trigo_osc.resize (nbr_osc);
for (int osc_cnt = 0; osc_cnt < nbr_osc; ++osc_cnt)
{
OscType & osc = _trigo_osc [osc_cnt];
const long len = 1L << (TRIGO_BD_LIMIT + osc_cnt);
const double mul = (0.5 * PI) / len;
osc.set_step (mul);
}
}
}
template <class DT>
const long * FFTReal <DT>::get_br_ptr () const
{
return (&_br_lut [0]);
}
template <class DT>
const typename FFTReal <DT>::DataType * FFTReal <DT>::get_trigo_ptr (int level) const
{
assert (level >= 3);
return (&_trigo_lut [get_trigo_level_index (level)]);
}
template <class DT>
long FFTReal <DT>::get_trigo_level_index (int level) const
{
assert (level >= 3);
return ((1L << (level - 1)) - 4);
}
// Transform in several passes
template <class DT>
void FFTReal <DT>::compute_fft_general (DataType f [], const DataType x []) const
{
assert (f != 0);
assert (f != use_buffer ());
assert (x != 0);
assert (x != use_buffer ());
assert (x != f);
DataType * sf;
DataType * df;
if ((_nbr_bits & 1) != 0)
{
df = use_buffer ();
sf = f;
}
else
{
df = f;
sf = use_buffer ();
}
compute_direct_pass_1_2 (df, x);
compute_direct_pass_3 (sf, df);
for (int pass = 3; pass < _nbr_bits; ++ pass)
{
compute_direct_pass_n (df, sf, pass);
DataType * const temp_ptr = df;
df = sf;
sf = temp_ptr;
}
}
template <class DT>
void FFTReal <DT>::compute_direct_pass_1_2 (DataType df [], const DataType x []) const
{
assert (df != 0);
assert (x != 0);
assert (df != x);
const long * const bit_rev_lut_ptr = get_br_ptr ();
long coef_index = 0;
do
{
const long rev_index_0 = bit_rev_lut_ptr [coef_index];
const long rev_index_1 = bit_rev_lut_ptr [coef_index + 1];
const long rev_index_2 = bit_rev_lut_ptr [coef_index + 2];
const long rev_index_3 = bit_rev_lut_ptr [coef_index + 3];
DataType * const df2 = df + coef_index;
df2 [1] = x [rev_index_0] - x [rev_index_1];
df2 [3] = x [rev_index_2] - x [rev_index_3];
const DataType sf_0 = x [rev_index_0] + x [rev_index_1];
const DataType sf_2 = x [rev_index_2] + x [rev_index_3];
df2 [0] = sf_0 + sf_2;
df2 [2] = sf_0 - sf_2;
coef_index += 4;
}
while (coef_index < _length);
}
template <class DT>
void FFTReal <DT>::compute_direct_pass_3 (DataType df [], const DataType sf []) const
{
assert (df != 0);
assert (sf != 0);
assert (df != sf);
const DataType sqrt2_2 = DataType (SQRT2 * 0.5);
long coef_index = 0;
do
{
DataType v;
df [coef_index] = sf [coef_index] + sf [coef_index + 4];
df [coef_index + 4] = sf [coef_index] - sf [coef_index + 4];
df [coef_index + 2] = sf [coef_index + 2];
df [coef_index + 6] = sf [coef_index + 6];
v = (sf [coef_index + 5] - sf [coef_index + 7]) * sqrt2_2;
df [coef_index + 1] = sf [coef_index + 1] + v;
df [coef_index + 3] = sf [coef_index + 1] - v;
v = (sf [coef_index + 5] + sf [coef_index + 7]) * sqrt2_2;
df [coef_index + 5] = v + sf [coef_index + 3];
df [coef_index + 7] = v - sf [coef_index + 3];
coef_index += 8;
}
while (coef_index < _length);
}
template <class DT>
void FFTReal <DT>::compute_direct_pass_n (DataType df [], const DataType sf [], int pass) const
{
assert (df != 0);
assert (sf != 0);
assert (df != sf);
assert (pass >= 3);
assert (pass < _nbr_bits);
if (pass <= TRIGO_BD_LIMIT)
{
compute_direct_pass_n_lut (df, sf, pass);
}
else
{
compute_direct_pass_n_osc (df, sf, pass);
}
}
template <class DT>
void FFTReal <DT>::compute_direct_pass_n_lut (DataType df [], const DataType sf [], int pass) const
{
assert (df != 0);
assert (sf != 0);
assert (df != sf);
assert (pass >= 3);
assert (pass < _nbr_bits);
const long nbr_coef = 1 << pass;
const long h_nbr_coef = nbr_coef >> 1;
const long d_nbr_coef = nbr_coef << 1;
long coef_index = 0;
const DataType * const cos_ptr = get_trigo_ptr (pass);
do
{
const DataType * const sf1r = sf + coef_index;
const DataType * const sf2r = sf1r + nbr_coef;
DataType * const dfr = df + coef_index;
DataType * const dfi = dfr + nbr_coef;
// Extreme coefficients are always real
dfr [0] = sf1r [0] + sf2r [0];
dfi [0] = sf1r [0] - sf2r [0]; // dfr [nbr_coef] =
dfr [h_nbr_coef] = sf1r [h_nbr_coef];
dfi [h_nbr_coef] = sf2r [h_nbr_coef];
// Others are conjugate complex numbers
const DataType * const sf1i = sf1r + h_nbr_coef;
const DataType * const sf2i = sf1i + nbr_coef;
for (long i = 1; i < h_nbr_coef; ++ i)
{
const DataType c = cos_ptr [i]; // cos (i*PI/nbr_coef);
const DataType s = cos_ptr [h_nbr_coef - i]; // sin (i*PI/nbr_coef);
DataType v;
v = sf2r [i] * c - sf2i [i] * s;
dfr [i] = sf1r [i] + v;
dfi [-i] = sf1r [i] - v; // dfr [nbr_coef - i] =
v = sf2r [i] * s + sf2i [i] * c;
dfi [i] = v + sf1i [i];
dfi [nbr_coef - i] = v - sf1i [i];
}
coef_index += d_nbr_coef;
}
while (coef_index < _length);
}
template <class DT>
void FFTReal <DT>::compute_direct_pass_n_osc (DataType df [], const DataType sf [], int pass) const
{
assert (df != 0);
assert (sf != 0);
assert (df != sf);
assert (pass > TRIGO_BD_LIMIT);
assert (pass < _nbr_bits);
const long nbr_coef = 1 << pass;
const long h_nbr_coef = nbr_coef >> 1;
const long d_nbr_coef = nbr_coef << 1;
long coef_index = 0;
OscType & osc = _trigo_osc [pass - (TRIGO_BD_LIMIT + 1)];
do
{
const DataType * const sf1r = sf + coef_index;
const DataType * const sf2r = sf1r + nbr_coef;
DataType * const dfr = df + coef_index;
DataType * const dfi = dfr + nbr_coef;
osc.clear_buffers ();
// Extreme coefficients are always real
dfr [0] = sf1r [0] + sf2r [0];
dfi [0] = sf1r [0] - sf2r [0]; // dfr [nbr_coef] =
dfr [h_nbr_coef] = sf1r [h_nbr_coef];
dfi [h_nbr_coef] = sf2r [h_nbr_coef];
// Others are conjugate complex numbers
const DataType * const sf1i = sf1r + h_nbr_coef;
const DataType * const sf2i = sf1i + nbr_coef;
for (long i = 1; i < h_nbr_coef; ++ i)
{
osc.step ();
const DataType c = osc.get_cos ();
const DataType s = osc.get_sin ();
DataType v;
v = sf2r [i] * c - sf2i [i] * s;
dfr [i] = sf1r [i] + v;
dfi [-i] = sf1r [i] - v; // dfr [nbr_coef - i] =
v = sf2r [i] * s + sf2i [i] * c;
dfi [i] = v + sf1i [i];
dfi [nbr_coef - i] = v - sf1i [i];
}
coef_index += d_nbr_coef;
}
while (coef_index < _length);
}
// Transform in several pass
template <class DT>
void FFTReal <DT>::compute_ifft_general (const DataType f [], DataType x []) const
{
assert (f != 0);
assert (f != use_buffer ());
assert (x != 0);
assert (x != use_buffer ());
assert (x != f);
DataType * sf = const_cast <DataType *> (f);
DataType * df;
DataType * df_temp;
if (_nbr_bits & 1)
{
df = use_buffer ();
df_temp = x;
}
else
{
df = x;
df_temp = use_buffer ();
}
for (int pass = _nbr_bits - 1; pass >= 3; -- pass)
{
compute_inverse_pass_n (df, sf, pass);
if (pass < _nbr_bits - 1)
{
DataType * const temp_ptr = df;
df = sf;
sf = temp_ptr;
}
else
{
sf = df;
df = df_temp;
}
}
compute_inverse_pass_3 (df, sf);
compute_inverse_pass_1_2 (x, df);
}
template <class DT>
void FFTReal <DT>::compute_inverse_pass_n (DataType df [], const DataType sf [], int pass) const
{
assert (df != 0);
assert (sf != 0);
assert (df != sf);
assert (pass >= 3);
assert (pass < _nbr_bits);
if (pass <= TRIGO_BD_LIMIT)
{
compute_inverse_pass_n_lut (df, sf, pass);
}
else
{
compute_inverse_pass_n_osc (df, sf, pass);
}
}
template <class DT>
void FFTReal <DT>::compute_inverse_pass_n_lut (DataType df [], const DataType sf [], int pass) const
{
assert (df != 0);
assert (sf != 0);
assert (df != sf);
assert (pass >= 3);
assert (pass < _nbr_bits);
const long nbr_coef = 1 << pass;
const long h_nbr_coef = nbr_coef >> 1;
const long d_nbr_coef = nbr_coef << 1;
long coef_index = 0;
const DataType * const cos_ptr = get_trigo_ptr (pass);
do
{
const DataType * const sfr = sf + coef_index;
const DataType * const sfi = sfr + nbr_coef;
DataType * const df1r = df + coef_index;
DataType * const df2r = df1r + nbr_coef;
// Extreme coefficients are always real
df1r [0] = sfr [0] + sfi [0]; // + sfr [nbr_coef]
df2r [0] = sfr [0] - sfi [0]; // - sfr [nbr_coef]
df1r [h_nbr_coef] = sfr [h_nbr_coef] * 2;
df2r [h_nbr_coef] = sfi [h_nbr_coef] * 2;
// Others are conjugate complex numbers
DataType * const df1i = df1r + h_nbr_coef;
DataType * const df2i = df1i + nbr_coef;
for (long i = 1; i < h_nbr_coef; ++ i)
{
df1r [i] = sfr [i] + sfi [-i]; // + sfr [nbr_coef - i]
df1i [i] = sfi [i] - sfi [nbr_coef - i];
const DataType c = cos_ptr [i]; // cos (i*PI/nbr_coef);
const DataType s = cos_ptr [h_nbr_coef - i]; // sin (i*PI/nbr_coef);
const DataType vr = sfr [i] - sfi [-i]; // - sfr [nbr_coef - i]
const DataType vi = sfi [i] + sfi [nbr_coef - i];
df2r [i] = vr * c + vi * s;
df2i [i] = vi * c - vr * s;
}
coef_index += d_nbr_coef;
}
while (coef_index < _length);
}
template <class DT>
void FFTReal <DT>::compute_inverse_pass_n_osc (DataType df [], const DataType sf [], int pass) const
{
assert (df != 0);
assert (sf != 0);
assert (df != sf);
assert (pass > TRIGO_BD_LIMIT);
assert (pass < _nbr_bits);
const long nbr_coef = 1 << pass;
const long h_nbr_coef = nbr_coef >> 1;
const long d_nbr_coef = nbr_coef << 1;
long coef_index = 0;
OscType & osc = _trigo_osc [pass - (TRIGO_BD_LIMIT + 1)];
do
{
const DataType * const sfr = sf + coef_index;
const DataType * const sfi = sfr + nbr_coef;
DataType * const df1r = df + coef_index;
DataType * const df2r = df1r + nbr_coef;
osc.clear_buffers ();
// Extreme coefficients are always real
df1r [0] = sfr [0] + sfi [0]; // + sfr [nbr_coef]
df2r [0] = sfr [0] - sfi [0]; // - sfr [nbr_coef]
df1r [h_nbr_coef] = sfr [h_nbr_coef] * 2;
df2r [h_nbr_coef] = sfi [h_nbr_coef] * 2;
// Others are conjugate complex numbers
DataType * const df1i = df1r + h_nbr_coef;
DataType * const df2i = df1i + nbr_coef;
for (long i = 1; i < h_nbr_coef; ++ i)
{
df1r [i] = sfr [i] + sfi [-i]; // + sfr [nbr_coef - i]
df1i [i] = sfi [i] - sfi [nbr_coef - i];
osc.step ();
const DataType c = osc.get_cos ();
const DataType s = osc.get_sin ();
const DataType vr = sfr [i] - sfi [-i]; // - sfr [nbr_coef - i]
const DataType vi = sfi [i] + sfi [nbr_coef - i];
df2r [i] = vr * c + vi * s;
df2i [i] = vi * c - vr * s;
}
coef_index += d_nbr_coef;
}
while (coef_index < _length);
}
template <class DT>
void FFTReal <DT>::compute_inverse_pass_3 (DataType df [], const DataType sf []) const
{
assert (df != 0);
assert (sf != 0);
assert (df != sf);
const DataType sqrt2_2 = DataType (SQRT2 * 0.5);
long coef_index = 0;
do
{
df [coef_index] = sf [coef_index] + sf [coef_index + 4];
df [coef_index + 4] = sf [coef_index] - sf [coef_index + 4];
df [coef_index + 2] = sf [coef_index + 2] * 2;
df [coef_index + 6] = sf [coef_index + 6] * 2;
df [coef_index + 1] = sf [coef_index + 1] + sf [coef_index + 3];
df [coef_index + 3] = sf [coef_index + 5] - sf [coef_index + 7];
const DataType vr = sf [coef_index + 1] - sf [coef_index + 3];
const DataType vi = sf [coef_index + 5] + sf [coef_index + 7];
df [coef_index + 5] = (vr + vi) * sqrt2_2;
df [coef_index + 7] = (vi - vr) * sqrt2_2;
coef_index += 8;
}
while (coef_index < _length);
}
template <class DT>
void FFTReal <DT>::compute_inverse_pass_1_2 (DataType x [], const DataType sf []) const
{
assert (x != 0);
assert (sf != 0);
assert (x != sf);
const long * bit_rev_lut_ptr = get_br_ptr ();
const DataType * sf2 = sf;
long coef_index = 0;
do
{
{
const DataType b_0 = sf2 [0] + sf2 [2];
const DataType b_2 = sf2 [0] - sf2 [2];
const DataType b_1 = sf2 [1] * 2;
const DataType b_3 = sf2 [3] * 2;
x [bit_rev_lut_ptr [0]] = b_0 + b_1;
x [bit_rev_lut_ptr [1]] = b_0 - b_1;
x [bit_rev_lut_ptr [2]] = b_2 + b_3;
x [bit_rev_lut_ptr [3]] = b_2 - b_3;
}
{
const DataType b_0 = sf2 [4] + sf2 [6];
const DataType b_2 = sf2 [4] - sf2 [6];
const DataType b_1 = sf2 [5] * 2;
const DataType b_3 = sf2 [7] * 2;
x [bit_rev_lut_ptr [4]] = b_0 + b_1;
x [bit_rev_lut_ptr [5]] = b_0 - b_1;
x [bit_rev_lut_ptr [6]] = b_2 + b_3;
x [bit_rev_lut_ptr [7]] = b_2 - b_3;
}
sf2 += 8;
coef_index += 8;
bit_rev_lut_ptr += 8;
}
while (coef_index < _length);
}
#endif // FFTReal_CODEHEADER_INCLUDED
#undef FFTReal_CURRENT_CODEHEADER
/*\\\ EOF \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/