/* (C) 2007-2008 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.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include "mathops.h"
#include "cwrs.h"
#include "vq.h"
#include "arch.h"
#include "os_support.h"
/** Takes the pitch vector and the decoded residual vector, computes the gain
that will give ||p+g*y||=1 and mixes the residual with the pitch. */
static void mix_pitch_and_residual(int * restrict iy, celt_norm_t * restrict X, int N, int K, const celt_norm_t * restrict P)
{
int i;
celt_word32_t Ryp, Ryy, Rpp;
celt_word32_t g;
VARDECL(celt_norm_t, y);
#ifdef FIXED_POINT
int yshift;
#endif
SAVE_STACK;
#ifdef FIXED_POINT
yshift = 13-celt_ilog2(K);
#endif
ALLOC(y, N, celt_norm_t);
/*for (i=0;i0)
signx[j]=1;
else
signx[j]=-1;
iy[j] = 0;
y[j] = 0;
sum = MAC16_16(sum, P[j],P[j]);
} while (++j 0)
{
int pulsesAtOnce=1;
int best_id;
celt_word16_t magnitude;
#ifdef FIXED_POINT
int rshift;
#endif
/* Decide on how many pulses to find at once */
pulsesAtOnce = (pulsesLeft*N_1)>>9; /* pulsesLeft/N */
if (pulsesAtOnce<1)
pulsesAtOnce = 1;
#ifdef FIXED_POINT
rshift = yshift+1+celt_ilog2(K-pulsesLeft+pulsesAtOnce);
#endif
magnitude = SHL16(pulsesAtOnce, yshift);
best_id = 0;
/* The squared magnitude term gets added anyway, so we might as well
add it outside the loop */
yy = ADD32(yy, MULT16_16(magnitude,magnitude));
/* Choose between fast and accurate strategy depending on where we are in the search */
if (pulsesLeft>1)
{
/* This should ensure that anything we can process will have a better score */
celt_word32_t best_num = -VERY_LARGE16;
celt_word16_t best_den = 0;
j=0;
do {
celt_word16_t Rxy, Ryy;
/* Select sign based on X[j] alone */
s = signx[j]*magnitude;
/* Temporary sums of the new pulse(s) */
Rxy = EXTRACT16(SHR32(xy + MULT16_16(s,X[j]),rshift));
/* We're multiplying y[j] by two so we don't have to do it here */
Ryy = EXTRACT16(SHR32(yy + MULT16_16(s,y[j]),rshift));
/* Approximate score: we maximise Rxy/sqrt(Ryy) (we're guaranteed that
Rxy is positive because the sign is pre-computed) */
Rxy = MULT16_16_Q15(Rxy,Rxy);
/* The idea is to check for num/den >= best_num/best_den, but that way
we can do it without any division */
/* OPT: Make sure to use conditional moves here */
if (MULT16_16(best_den, Rxy) > MULT16_16(Ryy, best_num))
{
best_den = Ryy;
best_num = Rxy;
best_id = j;
}
} while (++j MULT16_16(Ryy, best_num))
{
best_den = Ryy;
best_num = num;
best_id = j;
}
} while (++j= (Nmax>>1))
{
for (i=0;i