/* (C) 2008 Jean-Marc Valin, CSIRO
*/
/*
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of the Xiph.org Foundation nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* This is a simple MDCT implementation that uses a N/4 complex FFT
to do most of the work. It should be relatively straightforward to
plug in pretty much and FFT here.
This replaces the Vorbis FFT (and uses the exact same API), which
was a bit too messy and that was ending up duplicating code
(might as well use the same FFT everywhere).
The algorithm is similar to (and inspired from) Fabrice Bellard's
MDCT implementation in FFMPEG, but has differences in signs, ordering
and scaling in many places.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include "mdct.h"
#include "kfft_double.h"
#include
#include "os_support.h"
#include "mathops.h"
#include "stack_alloc.h"
#ifndef M_PI
#define M_PI 3.141592653
#endif
void mdct_init(mdct_lookup *l,int N)
{
int i;
int N2;
l->n = N;
N2 = N>>1;
l->kfft = cpx32_fft_alloc(N>>2);
l->trig = (kiss_twiddle_scalar*)celt_alloc(N2*sizeof(kiss_twiddle_scalar));
/* We have enough points that sine isn't necessary */
#if defined(FIXED_POINT)
#if defined(DOUBLE_PRECISION) & !defined(MIXED_PRECISION)
for (i=0;itrig[i] = SAMP_MAX*cos(2*M_PI*(i+1./8.)/N);
#else
for (i=0;itrig[i] = TRIG_UPSCALE*celt_cos_norm(DIV32(ADD32(SHL32(EXTEND32(i),17),16386),N));
#endif
#else
for (i=0;itrig[i] = cos(2*M_PI*(i+1./8.)/N);
#endif
}
void mdct_clear(mdct_lookup *l)
{
cpx32_fft_free(l->kfft);
celt_free(l->trig);
}
void mdct_forward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar * restrict out, const celt_word16_t *window, int overlap)
{
int i;
int N, N2, N4;
VARDECL(kiss_fft_scalar, f);
SAVE_STACK;
N = l->n;
N2 = N>>1;
N4 = N>>2;
ALLOC(f, N2, kiss_fft_scalar);
/* Consider the input to be compused of four blocks: [a, b, c, d] */
/* Window, shuffle, fold */
{
/* Temp pointers to make it really clear to the compiler what we're doing */
const kiss_fft_scalar * restrict xp1 = in+(overlap>>1);
const kiss_fft_scalar * restrict xp2 = in+N2-1+(overlap>>1);
kiss_fft_scalar * restrict yp = out;
const celt_word16_t * restrict wp1 = window+(overlap>>1);
const celt_word16_t * restrict wp2 = window+(overlap>>1)-1;
for(i=0;i<(overlap>>2);i++)
{
/* Real part arranged as -d-cR, Imag part arranged as -b+aR*/
*yp++ = MULT16_32_Q15(*wp2, xp1[N2]) + MULT16_32_Q15(*wp1,*xp2);
*yp++ = MULT16_32_Q15(*wp1, *xp1) - MULT16_32_Q15(*wp2, xp2[-N2]);
xp1+=2;
xp2-=2;
wp1+=2;
wp2-=2;
}
wp1 = window;
wp2 = window+overlap-1;
for(;i>2);i++)
{
/* Real part arranged as a-bR, Imag part arranged as -c-dR */
*yp++ = *xp2;
*yp++ = *xp1;
xp1+=2;
xp2-=2;
}
for(;itrig[0];
for(i=0;ikfft, out, f, N4);
/* Post-rotate */
{
/* Temp pointers to make it really clear to the compiler what we're doing */
const kiss_fft_scalar * restrict fp = f;
kiss_fft_scalar * restrict yp1 = out;
kiss_fft_scalar * restrict yp2 = out+N2-1;
kiss_fft_scalar *t = &l->trig[0];
/* Temp pointers to make it really clear to the compiler what we're doing */
for(i=0;in;
N2 = N>>1;
N4 = N>>2;
ALLOC(f, N2, kiss_fft_scalar);
ALLOC(f2, N2, kiss_fft_scalar);
/* Pre-rotate */
{
/* Temp pointers to make it really clear to the compiler what we're doing */
const kiss_fft_scalar * restrict xp1 = in;
const kiss_fft_scalar * restrict xp2 = in+N2-1;
kiss_fft_scalar * restrict yp = f2;
kiss_fft_scalar *t = &l->trig[0];
for(i=0;ikfft, f2, f, N4);
/* Post-rotate */
{
kiss_fft_scalar * restrict fp = f;
kiss_fft_scalar *t = &l->trig[0];
for(i=0;i