961 lines
26 KiB
C
961 lines
26 KiB
C
/* Copyright (c) 2007-2008 CSIRO
|
|
Copyright (c) 2007-2009 Xiph.Org Foundation
|
|
Copyright (c) 2008-2009 Gregory Maxwell
|
|
Written by Jean-Marc Valin and Gregory Maxwell */
|
|
/*
|
|
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 <math.h>
|
|
#include "bands.h"
|
|
#include "modes.h"
|
|
#include "vq.h"
|
|
#include "cwrs.h"
|
|
#include "stack_alloc.h"
|
|
#include "os_support.h"
|
|
#include "mathops.h"
|
|
#include "rate.h"
|
|
|
|
|
|
#ifdef FIXED_POINT
|
|
/* Compute the amplitude (sqrt energy) in each of the bands */
|
|
void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bank, int _C, int M)
|
|
{
|
|
int i, c, N;
|
|
const celt_int16 *eBands = m->eBands;
|
|
const int C = CHANNELS(_C);
|
|
N = M*m->eBands[m->nbEBands+1];
|
|
for (c=0;c<C;c++)
|
|
{
|
|
for (i=0;i<m->nbEBands;i++)
|
|
{
|
|
int j;
|
|
celt_word32 maxval=0;
|
|
celt_word32 sum = 0;
|
|
|
|
j=M*eBands[i]; do {
|
|
maxval = MAX32(maxval, X[j+c*N]);
|
|
maxval = MAX32(maxval, -X[j+c*N]);
|
|
} while (++j<M*eBands[i+1]);
|
|
|
|
if (maxval > 0)
|
|
{
|
|
int shift = celt_ilog2(maxval)-10;
|
|
j=M*eBands[i]; do {
|
|
sum = MAC16_16(sum, EXTRACT16(VSHR32(X[j+c*N],shift)),
|
|
EXTRACT16(VSHR32(X[j+c*N],shift)));
|
|
} while (++j<M*eBands[i+1]);
|
|
/* We're adding one here to make damn sure we never end up with a pitch vector that's
|
|
larger than unity norm */
|
|
bank[i+c*m->nbEBands] = EPSILON+VSHR32(EXTEND32(celt_sqrt(sum)),-shift);
|
|
} else {
|
|
bank[i+c*m->nbEBands] = EPSILON;
|
|
}
|
|
/*printf ("%f ", bank[i+c*m->nbEBands]);*/
|
|
}
|
|
}
|
|
/*printf ("\n");*/
|
|
}
|
|
|
|
/* Normalise each band such that the energy is one. */
|
|
void normalise_bands(const CELTMode *m, const celt_sig * restrict freq, celt_norm * restrict X, const celt_ener *bank, int _C, int M)
|
|
{
|
|
int i, c, N;
|
|
const celt_int16 *eBands = m->eBands;
|
|
const int C = CHANNELS(_C);
|
|
N = M*m->eBands[m->nbEBands+1];
|
|
for (c=0;c<C;c++)
|
|
{
|
|
i=0; do {
|
|
celt_word16 g;
|
|
int j,shift;
|
|
celt_word16 E;
|
|
shift = celt_zlog2(bank[i+c*m->nbEBands])-13;
|
|
E = VSHR32(bank[i+c*m->nbEBands], shift);
|
|
g = EXTRACT16(celt_rcp(SHL32(E,3)));
|
|
j=M*eBands[i]; do {
|
|
X[j+c*N] = MULT16_16_Q15(VSHR32(freq[j+c*N],shift-1),g);
|
|
} while (++j<M*eBands[i+1]);
|
|
} while (++i<m->nbEBands);
|
|
}
|
|
}
|
|
|
|
#else /* FIXED_POINT */
|
|
/* Compute the amplitude (sqrt energy) in each of the bands */
|
|
void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bank, int _C, int M)
|
|
{
|
|
int i, c, N;
|
|
const celt_int16 *eBands = m->eBands;
|
|
const int C = CHANNELS(_C);
|
|
N = M*m->eBands[m->nbEBands+1];
|
|
for (c=0;c<C;c++)
|
|
{
|
|
for (i=0;i<m->nbEBands;i++)
|
|
{
|
|
int j;
|
|
celt_word32 sum = 1e-10;
|
|
for (j=M*eBands[i];j<M*eBands[i+1];j++)
|
|
sum += X[j+c*N]*X[j+c*N];
|
|
bank[i+c*m->nbEBands] = sqrt(sum);
|
|
/*printf ("%f ", bank[i+c*m->nbEBands]);*/
|
|
}
|
|
}
|
|
/*printf ("\n");*/
|
|
}
|
|
|
|
/* Normalise each band such that the energy is one. */
|
|
void normalise_bands(const CELTMode *m, const celt_sig * restrict freq, celt_norm * restrict X, const celt_ener *bank, int _C, int M)
|
|
{
|
|
int i, c, N;
|
|
const celt_int16 *eBands = m->eBands;
|
|
const int C = CHANNELS(_C);
|
|
N = M*m->eBands[m->nbEBands+1];
|
|
for (c=0;c<C;c++)
|
|
{
|
|
for (i=0;i<m->nbEBands;i++)
|
|
{
|
|
int j;
|
|
celt_word16 g = 1.f/(1e-10f+bank[i+c*m->nbEBands]);
|
|
for (j=M*eBands[i];j<M*eBands[i+1];j++)
|
|
X[j+c*N] = freq[j+c*N]*g;
|
|
}
|
|
}
|
|
}
|
|
|
|
#endif /* FIXED_POINT */
|
|
|
|
void renormalise_bands(const CELTMode *m, celt_norm * restrict X, int _C, int M)
|
|
{
|
|
int i, c;
|
|
const celt_int16 *eBands = m->eBands;
|
|
const int C = CHANNELS(_C);
|
|
for (c=0;c<C;c++)
|
|
{
|
|
i=0; do {
|
|
renormalise_vector(X+M*eBands[i]+c*M*eBands[m->nbEBands+1], Q15ONE, M*eBands[i+1]-M*eBands[i], 1);
|
|
} while (++i<m->nbEBands);
|
|
}
|
|
}
|
|
|
|
/* De-normalise the energy to produce the synthesis from the unit-energy bands */
|
|
void denormalise_bands(const CELTMode *m, const celt_norm * restrict X, celt_sig * restrict freq, const celt_ener *bank, int _C, int M)
|
|
{
|
|
int i, c, N;
|
|
const celt_int16 *eBands = m->eBands;
|
|
const int C = CHANNELS(_C);
|
|
N = M*m->eBands[m->nbEBands+1];
|
|
if (C>2)
|
|
celt_fatal("denormalise_bands() not implemented for >2 channels");
|
|
for (c=0;c<C;c++)
|
|
{
|
|
celt_sig * restrict f;
|
|
const celt_norm * restrict x;
|
|
f = freq+c*N;
|
|
x = X+c*N;
|
|
for (i=0;i<m->nbEBands;i++)
|
|
{
|
|
int j, end;
|
|
celt_word32 g = SHR32(bank[i+c*m->nbEBands],1);
|
|
j=M*eBands[i];
|
|
end = M*eBands[i+1];
|
|
do {
|
|
*f++ = SHL32(MULT16_32_Q15(*x, g),2);
|
|
x++;
|
|
} while (++j<end);
|
|
}
|
|
for (i=M*eBands[m->nbEBands];i<M*eBands[m->nbEBands+1];i++)
|
|
*f++ = 0;
|
|
}
|
|
}
|
|
|
|
int compute_pitch_gain(const CELTMode *m, const celt_sig *X, const celt_sig *P, int norm_rate, int *gain_id, int _C, celt_word16 *gain_prod, int M)
|
|
{
|
|
int j, c;
|
|
celt_word16 g;
|
|
celt_word16 delta;
|
|
const int C = CHANNELS(_C);
|
|
celt_word32 Sxy=0, Sxx=0, Syy=0;
|
|
int len = M*m->pitchEnd;
|
|
int N = M*m->eBands[m->nbEBands+1];
|
|
#ifdef FIXED_POINT
|
|
int shift = 0;
|
|
celt_word32 maxabs=0;
|
|
|
|
for (c=0;c<C;c++)
|
|
{
|
|
for (j=0;j<len;j++)
|
|
{
|
|
maxabs = MAX32(maxabs, ABS32(X[j+c*N]));
|
|
maxabs = MAX32(maxabs, ABS32(P[j+c*N]));
|
|
}
|
|
}
|
|
shift = celt_ilog2(maxabs)-12;
|
|
if (shift<0)
|
|
shift = 0;
|
|
#endif
|
|
delta = PDIV32_16(Q15ONE, len);
|
|
for (c=0;c<C;c++)
|
|
{
|
|
celt_word16 gg = Q15ONE;
|
|
for (j=0;j<len;j++)
|
|
{
|
|
celt_word16 Xj, Pj;
|
|
Xj = EXTRACT16(SHR32(X[j+c*N], shift));
|
|
Pj = MULT16_16_P15(gg,EXTRACT16(SHR32(P[j+c*N], shift)));
|
|
Sxy = MAC16_16(Sxy, Xj, Pj);
|
|
Sxx = MAC16_16(Sxx, Pj, Pj);
|
|
Syy = MAC16_16(Syy, Xj, Xj);
|
|
gg = SUB16(gg, delta);
|
|
}
|
|
}
|
|
#ifdef FIXED_POINT
|
|
{
|
|
celt_word32 num, den;
|
|
celt_word16 fact;
|
|
fact = MULT16_16(QCONST16(.04f, 14), norm_rate);
|
|
if (fact < QCONST16(1.f, 14))
|
|
fact = QCONST16(1.f, 14);
|
|
num = Sxy;
|
|
den = EPSILON+Sxx+MULT16_32_Q15(QCONST16(.03f,15),Syy);
|
|
shift = celt_zlog2(Sxy)-16;
|
|
if (shift < 0)
|
|
shift = 0;
|
|
if (Sxy < MULT16_32_Q15(fact, MULT16_16(celt_sqrt(EPSILON+Sxx),celt_sqrt(EPSILON+Syy))))
|
|
g = 0;
|
|
else
|
|
g = DIV32(SHL32(SHR32(num,shift),14),ADD32(EPSILON,SHR32(den,shift)));
|
|
|
|
/* This MUST round down so that we don't over-estimate the gain */
|
|
*gain_id = EXTRACT16(SHR32(MULT16_16(20,(g-QCONST16(.5f,14))),14));
|
|
}
|
|
#else
|
|
{
|
|
float fact = .04f*norm_rate;
|
|
if (fact < 1)
|
|
fact = 1;
|
|
g = Sxy/(.1f+Sxx+.03f*Syy);
|
|
if (Sxy < .5f*fact*celt_sqrt(1+Sxx*Syy))
|
|
g = 0;
|
|
/* This MUST round down so that we don't over-estimate the gain */
|
|
*gain_id = floor(20*(g-.5f));
|
|
}
|
|
#endif
|
|
/* This prevents the pitch gain from being above 1.0 for too long by bounding the
|
|
maximum error amplification factor to 2.0 */
|
|
g = ADD16(QCONST16(.5f,14), MULT16_16_16(QCONST16(.05f,14),*gain_id));
|
|
*gain_prod = MAX16(QCONST32(1.f, 13), MULT16_16_Q14(*gain_prod,g));
|
|
if (*gain_prod>QCONST32(2.f, 13))
|
|
{
|
|
*gain_id=9;
|
|
*gain_prod = QCONST32(2.f, 13);
|
|
}
|
|
|
|
if (*gain_id < 0)
|
|
{
|
|
*gain_id = 0;
|
|
return 0;
|
|
} else {
|
|
if (*gain_id > 15)
|
|
*gain_id = 15;
|
|
return 1;
|
|
}
|
|
}
|
|
|
|
void apply_pitch(const CELTMode *m, celt_sig *X, const celt_sig *P, int gain_id, int pred, int _C, int M)
|
|
{
|
|
int j, c, N;
|
|
celt_word16 gain;
|
|
celt_word16 delta;
|
|
const int C = CHANNELS(_C);
|
|
int len = M*m->pitchEnd;
|
|
|
|
N = M*m->eBands[m->nbEBands+1];
|
|
gain = ADD16(QCONST16(.5f,14), MULT16_16_16(QCONST16(.05f,14),gain_id));
|
|
delta = PDIV32_16(gain, len);
|
|
if (pred)
|
|
gain = -gain;
|
|
else
|
|
delta = -delta;
|
|
for (c=0;c<C;c++)
|
|
{
|
|
celt_word16 gg = gain;
|
|
for (j=0;j<len;j++)
|
|
{
|
|
X[j+c*N] += SHL32(MULT16_32_Q15(gg,P[j+c*N]),1);
|
|
gg = ADD16(gg, delta);
|
|
}
|
|
}
|
|
}
|
|
|
|
static void stereo_band_mix(const CELTMode *m, celt_norm *X, celt_norm *Y, const celt_ener *bank, int stereo_mode, int bandID, int dir, int N)
|
|
{
|
|
int i = bandID;
|
|
int j;
|
|
celt_word16 a1, a2;
|
|
if (stereo_mode==0)
|
|
{
|
|
/* Do mid-side when not doing intensity stereo */
|
|
a1 = QCONST16(.70711f,14);
|
|
a2 = dir*QCONST16(.70711f,14);
|
|
} else {
|
|
celt_word16 left, right;
|
|
celt_word16 norm;
|
|
#ifdef FIXED_POINT
|
|
int shift = celt_zlog2(MAX32(bank[i], bank[i+m->nbEBands]))-13;
|
|
#endif
|
|
left = VSHR32(bank[i],shift);
|
|
right = VSHR32(bank[i+m->nbEBands],shift);
|
|
norm = EPSILON + celt_sqrt(EPSILON+MULT16_16(left,left)+MULT16_16(right,right));
|
|
a1 = DIV32_16(SHL32(EXTEND32(left),14),norm);
|
|
a2 = dir*DIV32_16(SHL32(EXTEND32(right),14),norm);
|
|
}
|
|
for (j=0;j<N;j++)
|
|
{
|
|
celt_norm r, l;
|
|
l = X[j];
|
|
r = Y[j];
|
|
X[j] = MULT16_16_Q14(a1,l) + MULT16_16_Q14(a2,r);
|
|
Y[j] = MULT16_16_Q14(a1,r) - MULT16_16_Q14(a2,l);
|
|
}
|
|
}
|
|
|
|
|
|
int folding_decision(const CELTMode *m, celt_norm *X, celt_word16 *average, int *last_decision, int _C, int M)
|
|
{
|
|
int i, c, N0;
|
|
int NR=0;
|
|
celt_word32 ratio = EPSILON;
|
|
const int C = CHANNELS(_C);
|
|
const celt_int16 * restrict eBands = m->eBands;
|
|
|
|
N0 = M*m->eBands[m->nbEBands+1];
|
|
|
|
for (c=0;c<C;c++)
|
|
{
|
|
for (i=0;i<m->nbEBands;i++)
|
|
{
|
|
int j, N;
|
|
int max_i=0;
|
|
celt_word16 max_val=EPSILON;
|
|
celt_word32 floor_ener=EPSILON;
|
|
celt_norm * restrict x = X+M*eBands[i]+c*N0;
|
|
N = M*eBands[i+1]-M*eBands[i];
|
|
for (j=0;j<N;j++)
|
|
{
|
|
if (ABS16(x[j])>max_val)
|
|
{
|
|
max_val = ABS16(x[j]);
|
|
max_i = j;
|
|
}
|
|
}
|
|
#if 0
|
|
for (j=0;j<N;j++)
|
|
{
|
|
if (abs(j-max_i)>2)
|
|
floor_ener += x[j]*x[j];
|
|
}
|
|
#else
|
|
floor_ener = QCONST32(1.,28)-MULT16_16(max_val,max_val);
|
|
if (max_i < N-1)
|
|
floor_ener -= MULT16_16(x[(max_i+1)], x[(max_i+1)]);
|
|
if (max_i < N-2)
|
|
floor_ener -= MULT16_16(x[(max_i+2)], x[(max_i+2)]);
|
|
if (max_i > 0)
|
|
floor_ener -= MULT16_16(x[(max_i-1)], x[(max_i-1)]);
|
|
if (max_i > 1)
|
|
floor_ener -= MULT16_16(x[(max_i-2)], x[(max_i-2)]);
|
|
floor_ener = MAX32(floor_ener, EPSILON);
|
|
#endif
|
|
if (N>7)
|
|
{
|
|
celt_word16 r;
|
|
celt_word16 den = celt_sqrt(floor_ener);
|
|
den = MAX32(QCONST16(.02f, 15), den);
|
|
r = DIV32_16(SHL32(EXTEND32(max_val),8),den);
|
|
ratio = ADD32(ratio, EXTEND32(r));
|
|
NR++;
|
|
}
|
|
}
|
|
}
|
|
if (NR>0)
|
|
ratio = DIV32_16(ratio, NR);
|
|
ratio = ADD32(HALF32(ratio), HALF32(*average));
|
|
if (!*last_decision)
|
|
{
|
|
*last_decision = (ratio < QCONST16(1.8f,8));
|
|
} else {
|
|
*last_decision = (ratio < QCONST16(3.f,8));
|
|
}
|
|
*average = EXTRACT16(ratio);
|
|
return *last_decision;
|
|
}
|
|
|
|
static void interleave_vector(celt_norm *X, int N0, int stride)
|
|
{
|
|
int i,j;
|
|
VARDECL(celt_norm, tmp);
|
|
int N;
|
|
SAVE_STACK;
|
|
N = N0*stride;
|
|
ALLOC(tmp, N, celt_norm);
|
|
for (i=0;i<stride;i++)
|
|
for (j=0;j<N0;j++)
|
|
tmp[j*stride+i] = X[i*N0+j];
|
|
for (j=0;j<N;j++)
|
|
X[j] = tmp[j];
|
|
RESTORE_STACK;
|
|
}
|
|
|
|
static void deinterleave_vector(celt_norm *X, int N0, int stride)
|
|
{
|
|
int i,j;
|
|
VARDECL(celt_norm, tmp);
|
|
int N;
|
|
SAVE_STACK;
|
|
N = N0*stride;
|
|
ALLOC(tmp, N, celt_norm);
|
|
for (i=0;i<stride;i++)
|
|
for (j=0;j<N0;j++)
|
|
tmp[i*N0+j] = X[j*stride+i];
|
|
for (j=0;j<N;j++)
|
|
X[j] = tmp[j];
|
|
RESTORE_STACK;
|
|
}
|
|
|
|
static void haar1(celt_norm *X, int N0, int stride)
|
|
{
|
|
int i, j;
|
|
N0 >>= 1;
|
|
for (i=0;i<stride;i++)
|
|
for (j=0;j<N0;j++)
|
|
{
|
|
celt_norm tmp = X[stride*2*j+i];
|
|
X[stride*2*j+i] = MULT16_16_Q15(QCONST16(.7070678f,15), X[stride*2*j+i] + X[stride*(2*j+1)+i]);
|
|
X[stride*(2*j+1)+i] = MULT16_16_Q15(QCONST16(.7070678f,15), tmp - X[stride*(2*j+1)+i]);
|
|
}
|
|
}
|
|
|
|
/* This function is responsible for encoding and decoding a band for both
|
|
the mono and stereo case. Even in the mono case, it can split the band
|
|
in two and transmit the energy difference with the two half-bands. It
|
|
can be called recursively so bands can end up being split in 8 parts. */
|
|
static void quant_band(int encode, const CELTMode *m, int i, celt_norm *X, celt_norm *Y,
|
|
int N, int b, int spread, int tf_change, celt_norm *lowband, int resynth, void *ec,
|
|
celt_int32 *remaining_bits, int LM, celt_norm *lowband_out, const celt_ener *bandE, int level)
|
|
{
|
|
int q;
|
|
int curr_bits;
|
|
int stereo, split;
|
|
int imid=0, iside=0;
|
|
int N0=N;
|
|
int N_B=N;
|
|
int N_B0;
|
|
int spread0=spread;
|
|
int time_divide=0;
|
|
int recombine=0;
|
|
|
|
if (spread)
|
|
N_B /= spread;
|
|
N_B0 = N_B;
|
|
|
|
split = stereo = Y != NULL;
|
|
|
|
/* Special case for one sample */
|
|
if (N==1)
|
|
{
|
|
int c;
|
|
celt_norm *x = X;
|
|
for (c=0;c<1+stereo;c++)
|
|
{
|
|
int sign=0;
|
|
if (b>=1<<BITRES && *remaining_bits>=1<<BITRES)
|
|
{
|
|
if (encode)
|
|
{
|
|
sign = x[0]<0;
|
|
ec_enc_bits((ec_enc*)ec, sign, 1);
|
|
} else {
|
|
sign = ec_dec_bits((ec_dec*)ec, 1);
|
|
}
|
|
*remaining_bits -= 1<<BITRES;
|
|
b-=1<<BITRES;
|
|
}
|
|
if (resynth)
|
|
x[0] = sign ? -NORM_SCALING : NORM_SCALING;
|
|
x = Y;
|
|
}
|
|
if (lowband_out)
|
|
lowband_out[0] = X[0];
|
|
return;
|
|
}
|
|
|
|
/* Band recombining to increase frequency resolution */
|
|
if (!stereo && spread > 1 && level == 0 && tf_change>0)
|
|
{
|
|
while (spread>1 && tf_change>0)
|
|
{
|
|
spread>>=1;
|
|
N_B<<=1;
|
|
if (encode)
|
|
haar1(X, N_B, spread);
|
|
if (lowband)
|
|
haar1(lowband, N_B, spread);
|
|
recombine++;
|
|
tf_change--;
|
|
}
|
|
spread0=spread;
|
|
N_B0 = N_B;
|
|
}
|
|
|
|
/* Increasing the time resolution */
|
|
if (!stereo && level==0)
|
|
{
|
|
while ((N_B&1) == 0 && tf_change<0 && spread <= (1<<LM))
|
|
{
|
|
if (encode)
|
|
haar1(X, N_B, spread);
|
|
if (lowband)
|
|
haar1(lowband, N_B, spread);
|
|
spread <<= 1;
|
|
N_B >>= 1;
|
|
time_divide++;
|
|
tf_change++;
|
|
}
|
|
spread0 = spread;
|
|
N_B0 = N_B;
|
|
}
|
|
|
|
/* Reorganize the samples in time order instead of frequency order */
|
|
if (!stereo && spread0>1 && level==0)
|
|
{
|
|
if (encode)
|
|
deinterleave_vector(X, N_B, spread0);
|
|
if (lowband)
|
|
deinterleave_vector(lowband, N_B, spread0);
|
|
}
|
|
|
|
/* If we need more than 32 bits, try splitting the band in two. */
|
|
if (!stereo && LM != -1 && b > 32<<BITRES && N>2)
|
|
{
|
|
if (LM>0 || (N&1)==0)
|
|
{
|
|
N >>= 1;
|
|
Y = X+N;
|
|
split = 1;
|
|
LM -= 1;
|
|
spread = (spread+1)>>1;
|
|
}
|
|
}
|
|
|
|
if (split)
|
|
{
|
|
int qb;
|
|
int itheta=0;
|
|
int mbits, sbits, delta;
|
|
int qalloc;
|
|
celt_word16 mid, side;
|
|
int offset, N2;
|
|
offset = m->logN[i]+(LM<<BITRES)-QTHETA_OFFSET;
|
|
|
|
/* Decide on the resolution to give to the split parameter theta */
|
|
N2 = 2*N-1;
|
|
if (stereo && N>2)
|
|
N2--;
|
|
qb = (b+N2*offset)/(N2<<BITRES);
|
|
if (qb > (b>>(BITRES+1))-1)
|
|
qb = (b>>(BITRES+1))-1;
|
|
|
|
if (qb<0)
|
|
qb = 0;
|
|
if (qb>14)
|
|
qb = 14;
|
|
|
|
qalloc = 0;
|
|
if (qb!=0)
|
|
{
|
|
int shift;
|
|
shift = 14-qb;
|
|
|
|
if (encode)
|
|
{
|
|
if (stereo)
|
|
stereo_band_mix(m, X, Y, bandE, qb==0, i, 1, N);
|
|
|
|
mid = renormalise_vector(X, Q15ONE, N, 1);
|
|
side = renormalise_vector(Y, Q15ONE, N, 1);
|
|
|
|
/* theta is the atan() of the ration between the (normalized)
|
|
side and mid. With just that parameter, we can re-scale both
|
|
mid and side because we know that 1) they have unit norm and
|
|
2) they are orthogonal. */
|
|
#ifdef FIXED_POINT
|
|
/* 0.63662 = 2/pi */
|
|
itheta = MULT16_16_Q15(QCONST16(0.63662f,15),celt_atan2p(side, mid));
|
|
#else
|
|
itheta = floor(.5f+16384*0.63662f*atan2(side,mid));
|
|
#endif
|
|
|
|
itheta = (itheta+(1<<shift>>1))>>shift;
|
|
}
|
|
|
|
/* Entropy coding of the angle. We use a uniform pdf for the
|
|
first stereo split but a triangular one for the rest. */
|
|
if (stereo || qb>9 || spread>1)
|
|
{
|
|
if (encode)
|
|
ec_enc_uint((ec_enc*)ec, itheta, (1<<qb)+1);
|
|
else
|
|
itheta = ec_dec_uint((ec_dec*)ec, (1<<qb)+1);
|
|
qalloc = log2_frac((1<<qb)+1,BITRES);
|
|
} else {
|
|
int fs=1, ft;
|
|
ft = ((1<<qb>>1)+1)*((1<<qb>>1)+1);
|
|
if (encode)
|
|
{
|
|
int j;
|
|
int fl=0;
|
|
j=0;
|
|
while(1)
|
|
{
|
|
if (j==itheta)
|
|
break;
|
|
fl+=fs;
|
|
if (j<(1<<qb>>1))
|
|
fs++;
|
|
else
|
|
fs--;
|
|
j++;
|
|
}
|
|
ec_encode((ec_enc*)ec, fl, fl+fs, ft);
|
|
} else {
|
|
int fl=0;
|
|
int j, fm;
|
|
fm = ec_decode((ec_dec*)ec, ft);
|
|
j=0;
|
|
while (1)
|
|
{
|
|
if (fm < fl+fs)
|
|
break;
|
|
fl+=fs;
|
|
if (j<(1<<qb>>1))
|
|
fs++;
|
|
else
|
|
fs--;
|
|
j++;
|
|
}
|
|
itheta = j;
|
|
ec_dec_update((ec_dec*)ec, fl, fl+fs, ft);
|
|
}
|
|
qalloc = log2_frac(ft,BITRES) - log2_frac(fs,BITRES) + 1;
|
|
}
|
|
itheta <<= shift;
|
|
}
|
|
|
|
if (itheta == 0)
|
|
{
|
|
imid = 32767;
|
|
iside = 0;
|
|
delta = -10000;
|
|
} else if (itheta == 16384)
|
|
{
|
|
imid = 0;
|
|
iside = 32767;
|
|
delta = 10000;
|
|
} else {
|
|
imid = bitexact_cos(itheta);
|
|
iside = bitexact_cos(16384-itheta);
|
|
/* This is the mid vs side allocation that minimizes squared error
|
|
in that band. */
|
|
delta = (N-1)*(log2_frac(iside,BITRES+2)-log2_frac(imid,BITRES+2))>>2;
|
|
}
|
|
|
|
/* This is a special case for N=2 that only works for stereo and takes
|
|
advantage of the fact that mid and side are orthogonal to encode
|
|
the side with just one bit. */
|
|
if (N==2 && stereo)
|
|
{
|
|
int c, c2;
|
|
int sign=1;
|
|
celt_norm v[2], w[2];
|
|
celt_norm *x2, *y2;
|
|
mbits = b-qalloc;
|
|
sbits = 0;
|
|
if (itheta != 0 && itheta != 16384)
|
|
sbits = 1<<BITRES;
|
|
mbits -= sbits;
|
|
c = itheta > 8192 ? 1 : 0;
|
|
*remaining_bits -= qalloc+sbits;
|
|
|
|
x2 = X;
|
|
y2 = Y;
|
|
if (encode)
|
|
{
|
|
c2 = 1-c;
|
|
|
|
if (c==0)
|
|
{
|
|
v[0] = x2[0];
|
|
v[1] = x2[1];
|
|
w[0] = y2[0];
|
|
w[1] = y2[1];
|
|
} else {
|
|
v[0] = y2[0];
|
|
v[1] = y2[1];
|
|
w[0] = x2[0];
|
|
w[1] = x2[1];
|
|
}
|
|
/* Here we only need to encode a sign for the side */
|
|
if (v[0]*w[1] - v[1]*w[0] > 0)
|
|
sign = 1;
|
|
else
|
|
sign = -1;
|
|
}
|
|
quant_band(encode, m, i, v, NULL, N, mbits, spread, tf_change, lowband, resynth, ec, remaining_bits, LM, lowband_out, NULL, level+1);
|
|
if (sbits)
|
|
{
|
|
if (encode)
|
|
{
|
|
ec_enc_bits((ec_enc*)ec, sign==1, 1);
|
|
} else {
|
|
sign = 2*ec_dec_bits((ec_dec*)ec, 1)-1;
|
|
}
|
|
} else {
|
|
sign = 1;
|
|
}
|
|
w[0] = -sign*v[1];
|
|
w[1] = sign*v[0];
|
|
if (c==0)
|
|
{
|
|
x2[0] = v[0];
|
|
x2[1] = v[1];
|
|
y2[0] = w[0];
|
|
y2[1] = w[1];
|
|
} else {
|
|
x2[0] = w[0];
|
|
x2[1] = w[1];
|
|
y2[0] = v[0];
|
|
y2[1] = v[1];
|
|
}
|
|
} else
|
|
{
|
|
/* "Normal" split code */
|
|
celt_norm *next_lowband2=NULL;
|
|
celt_norm *next_lowband_out1=NULL;
|
|
int next_level=0;
|
|
|
|
/* Give more bits to low-energy MDCTs than they would otherwise deserve */
|
|
if (spread>1 && !stereo)
|
|
delta >>= 1;
|
|
|
|
mbits = (b-qalloc/2-delta)/2;
|
|
if (mbits > b-qalloc)
|
|
mbits = b-qalloc;
|
|
if (mbits<0)
|
|
mbits=0;
|
|
sbits = b-qalloc-mbits;
|
|
*remaining_bits -= qalloc;
|
|
|
|
if (lowband && !stereo)
|
|
next_lowband2 = lowband+N;
|
|
if (stereo)
|
|
next_lowband_out1 = lowband_out;
|
|
else
|
|
next_level = level+1;
|
|
|
|
quant_band(encode, m, i, X, NULL, N, mbits, spread, tf_change, lowband, resynth, ec, remaining_bits, LM, next_lowband_out1, NULL, next_level);
|
|
quant_band(encode, m, i, Y, NULL, N, sbits, spread, tf_change, next_lowband2, resynth, ec, remaining_bits, LM, NULL, NULL, level);
|
|
}
|
|
|
|
} else {
|
|
/* This is the basic no-split case */
|
|
q = bits2pulses(m, m->bits[LM][i], N, b);
|
|
curr_bits = pulses2bits(m->bits[LM][i], N, q);
|
|
*remaining_bits -= curr_bits;
|
|
|
|
/* Ensures we can never bust the budget */
|
|
while (*remaining_bits < 0 && q > 0)
|
|
{
|
|
*remaining_bits += curr_bits;
|
|
q--;
|
|
curr_bits = pulses2bits(m->bits[LM][i], N, q);
|
|
*remaining_bits -= curr_bits;
|
|
}
|
|
|
|
if (encode)
|
|
alg_quant(X, N, q, spread, lowband, resynth, (ec_enc*)ec);
|
|
else
|
|
alg_unquant(X, N, q, spread, lowband, (ec_dec*)ec);
|
|
}
|
|
|
|
/* This code is used by the decoder and by the resynthesis-enabled encoder */
|
|
if (resynth)
|
|
{
|
|
int k;
|
|
|
|
if (split)
|
|
{
|
|
int j;
|
|
celt_word16 mid, side;
|
|
#ifdef FIXED_POINT
|
|
mid = imid;
|
|
side = iside;
|
|
#else
|
|
mid = (1.f/32768)*imid;
|
|
side = (1.f/32768)*iside;
|
|
#endif
|
|
for (j=0;j<N;j++)
|
|
X[j] = MULT16_16_Q15(X[j], mid);
|
|
for (j=0;j<N;j++)
|
|
Y[j] = MULT16_16_Q15(Y[j], side);
|
|
}
|
|
|
|
if (!stereo && spread0>1 && level==0)
|
|
{
|
|
interleave_vector(X, N_B, spread0);
|
|
if (lowband)
|
|
interleave_vector(lowband, N_B, spread0);
|
|
}
|
|
|
|
/* Undo time-freq changes that we did earlier */
|
|
N_B = N_B0;
|
|
spread = spread0;
|
|
for (k=0;k<time_divide;k++)
|
|
{
|
|
spread >>= 1;
|
|
N_B <<= 1;
|
|
haar1(X, N_B, spread);
|
|
if (lowband)
|
|
haar1(lowband, N_B, spread);
|
|
}
|
|
|
|
for (k=0;k<recombine;k++)
|
|
{
|
|
haar1(X, N_B, spread);
|
|
if (lowband)
|
|
haar1(lowband, N_B, spread);
|
|
N_B>>=1;
|
|
spread <<= 1;
|
|
}
|
|
|
|
if (lowband_out && !stereo)
|
|
{
|
|
int j;
|
|
celt_word16 n;
|
|
n = celt_sqrt(SHL32(EXTEND32(N0),22));
|
|
for (j=0;j<N0;j++)
|
|
lowband_out[j] = MULT16_16_Q15(n,X[j]);
|
|
}
|
|
|
|
if (stereo)
|
|
{
|
|
stereo_band_mix(m, X, Y, bandE, 0, i, -1, N);
|
|
renormalise_vector(X, Q15ONE, N, 1);
|
|
renormalise_vector(Y, Q15ONE, N, 1);
|
|
}
|
|
}
|
|
}
|
|
|
|
void quant_all_bands(int encode, const CELTMode *m, int start, celt_norm *_X, celt_norm *_Y, const celt_ener *bandE, int *pulses, int shortBlocks, int fold, int *tf_res, int resynth, int total_bits, void *ec, int LM)
|
|
{
|
|
int i, balance;
|
|
celt_int32 remaining_bits;
|
|
const celt_int16 * restrict eBands = m->eBands;
|
|
celt_norm * restrict norm;
|
|
VARDECL(celt_norm, _norm);
|
|
int B;
|
|
int M;
|
|
int spread;
|
|
celt_norm *lowband;
|
|
int update_lowband = 1;
|
|
int C = _Y != NULL ? 2 : 1;
|
|
SAVE_STACK;
|
|
|
|
M = 1<<LM;
|
|
B = shortBlocks ? M : 1;
|
|
spread = fold ? B : 0;
|
|
ALLOC(_norm, M*eBands[m->nbEBands+1], celt_norm);
|
|
norm = _norm;
|
|
|
|
balance = 0;
|
|
lowband = NULL;
|
|
for (i=start;i<m->nbEBands;i++)
|
|
{
|
|
int tell;
|
|
int b;
|
|
int N;
|
|
int curr_balance;
|
|
celt_norm * restrict X, * restrict Y;
|
|
int tf_change=0;
|
|
|
|
X = _X+M*eBands[i];
|
|
if (_Y!=NULL)
|
|
Y = _Y+M*eBands[i];
|
|
else
|
|
Y = NULL;
|
|
N = M*eBands[i+1]-M*eBands[i];
|
|
if (encode)
|
|
tell = ec_enc_tell((ec_enc*)ec, BITRES);
|
|
else
|
|
tell = ec_dec_tell((ec_dec*)ec, BITRES);
|
|
|
|
if (i != start)
|
|
balance -= tell;
|
|
remaining_bits = (total_bits<<BITRES)-tell-1;
|
|
curr_balance = (m->nbEBands-i);
|
|
if (curr_balance > 3)
|
|
curr_balance = 3;
|
|
curr_balance = balance / curr_balance;
|
|
b = IMIN(remaining_bits+1,pulses[i]+curr_balance);
|
|
if (b<0)
|
|
b = 0;
|
|
/* Prevents ridiculous bit depths */
|
|
if (b > C*16*N<<BITRES)
|
|
b = C*16*N<<BITRES;
|
|
|
|
if (M*eBands[i]-N >= M*eBands[start])
|
|
{
|
|
if (update_lowband)
|
|
lowband = norm+M*eBands[i]-N;
|
|
} else
|
|
lowband = NULL;
|
|
|
|
tf_change = tf_res[i];
|
|
quant_band(encode, m, i, X, Y, N, b, spread, tf_change, lowband, resynth, ec, &remaining_bits, LM, norm+M*eBands[i], bandE, 0);
|
|
|
|
balance += pulses[i] + tell;
|
|
|
|
/* Update the folding position only as long as we have 2 bit/sample depth */
|
|
update_lowband = (b>>BITRES)>2*N;
|
|
}
|
|
RESTORE_STACK;
|
|
}
|
|
|