More de-interleaving: denormalised MDCT no longer stored with interleaved

channels. Fixed two stereo bugs in the process: one in the handling of
mdct_weight_shift and one in the PLC.
This commit is contained in:
Jean-Marc Valin 2009-06-14 14:05:19 -04:00
parent 66fb246219
commit 08a82ffb45
2 changed files with 34 additions and 27 deletions

View file

@ -49,9 +49,10 @@ const celt_word16_t sqrtC_1[2] = {QCONST16(1.f, 14), QCONST16(1.414214f, 14)};
/* Compute the amplitude (sqrt energy) in each of the bands */
void compute_band_energies(const CELTMode *m, const celt_sig_t *X, celt_ener_t *bank)
{
int i, c;
int i, c, N;
const celt_int16_t *eBands = m->eBands;
const int C = CHANNELS(m);
N = FRAMESIZE(m);
for (c=0;c<C;c++)
{
for (i=0;i<m->nbEBands;i++)
@ -61,16 +62,16 @@ void compute_band_energies(const CELTMode *m, const celt_sig_t *X, celt_ener_t *
celt_word32_t sum = 0;
j=eBands[i]; do {
maxval = MAX32(maxval, X[j*C+c]);
maxval = MAX32(maxval, -X[j*C+c]);
maxval = MAX32(maxval, X[j+c*N]);
maxval = MAX32(maxval, -X[j+c*N]);
} while (++j<eBands[i+1]);
if (maxval > 0)
{
int shift = celt_ilog2(maxval)-10;
j=eBands[i]; do {
sum = MAC16_16(sum, EXTRACT16(VSHR32(X[j*C+c],shift)),
EXTRACT16(VSHR32(X[j*C+c],shift)));
sum = MAC16_16(sum, EXTRACT16(VSHR32(X[j+c*N],shift)),
EXTRACT16(VSHR32(X[j+c*N],shift)));
} while (++j<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 */
@ -87,9 +88,10 @@ void compute_band_energies(const CELTMode *m, const celt_sig_t *X, celt_ener_t *
/* Normalise each band such that the energy is one. */
void normalise_bands(const CELTMode *m, const celt_sig_t * restrict freq, celt_norm_t * restrict X, const celt_ener_t *bank)
{
int i, c;
int i, c, N;
const celt_int16_t *eBands = m->eBands;
const int C = CHANNELS(m);
N = FRAMESIZE(m);
for (c=0;c<C;c++)
{
i=0; do {
@ -100,7 +102,7 @@ void normalise_bands(const CELTMode *m, const celt_sig_t * restrict freq, celt_n
E = VSHR32(bank[i+c*m->nbEBands], shift);
g = EXTRACT16(celt_rcp(SHL32(E,3)));
j=eBands[i]; do {
X[j*C+c] = MULT16_16_Q15(VSHR32(freq[j*C+c],shift-1),g);
X[j*C+c] = MULT16_16_Q15(VSHR32(freq[j+c*N],shift-1),g);
} while (++j<eBands[i+1]);
} while (++i<m->nbEBands);
}
@ -110,9 +112,10 @@ void normalise_bands(const CELTMode *m, const celt_sig_t * restrict freq, celt_n
/* Compute the amplitude (sqrt energy) in each of the bands */
void compute_band_energies(const CELTMode *m, const celt_sig_t *X, celt_ener_t *bank)
{
int i, c;
int i, c, N;
const celt_int16_t *eBands = m->eBands;
const int C = CHANNELS(m);
N = FRAMESIZE(m);
for (c=0;c<C;c++)
{
for (i=0;i<m->nbEBands;i++)
@ -120,7 +123,7 @@ void compute_band_energies(const CELTMode *m, const celt_sig_t *X, celt_ener_t *
int j;
celt_word32_t sum = 1e-10;
for (j=eBands[i];j<eBands[i+1];j++)
sum += X[j*C+c]*X[j*C+c];
sum += X[j+c*N]*X[j+c*N];
bank[i+c*m->nbEBands] = sqrt(sum);
/*printf ("%f ", bank[i+c*m->nbEBands]);*/
}
@ -131,9 +134,10 @@ void compute_band_energies(const CELTMode *m, const celt_sig_t *X, celt_ener_t *
#ifdef EXP_PSY
void compute_noise_energies(const CELTMode *m, const celt_sig_t *X, const celt_word16_t *tonality, celt_ener_t *bank)
{
int i, c;
int i, c, N;
const celt_int16_t *eBands = m->eBands;
const int C = CHANNELS(m);
N = FRAMESIZE(m);
for (c=0;c<C;c++)
{
for (i=0;i<m->nbEBands;i++)
@ -141,7 +145,7 @@ void compute_noise_energies(const CELTMode *m, const celt_sig_t *X, const celt_w
int j;
celt_word32_t sum = 1e-10;
for (j=eBands[i];j<eBands[i+1];j++)
sum += X[j*C+c]*X[j*C+c]*tonality[j];
sum += X[j*C+c]*X[j+c*N]*tonality[j];
bank[i+c*m->nbEBands] = sqrt(sum);
/*printf ("%f ", bank[i+c*m->nbEBands]);*/
}
@ -153,9 +157,10 @@ void compute_noise_energies(const CELTMode *m, const celt_sig_t *X, const celt_w
/* Normalise each band such that the energy is one. */
void normalise_bands(const CELTMode *m, const celt_sig_t * restrict freq, celt_norm_t * restrict X, const celt_ener_t *bank)
{
int i, c;
int i, c, N;
const celt_int16_t *eBands = m->eBands;
const int C = CHANNELS(m);
N = FRAMESIZE(m);
for (c=0;c<C;c++)
{
for (i=0;i<m->nbEBands;i++)
@ -163,7 +168,7 @@ void normalise_bands(const CELTMode *m, const celt_sig_t * restrict freq, celt_n
int j;
celt_word16_t g = 1.f/(1e-10+bank[i+c*m->nbEBands]);
for (j=eBands[i];j<eBands[i+1];j++)
X[j*C+c] = freq[j*C+c]*g;
X[j*C+c] = freq[j+c*N]*g;
}
}
}
@ -188,9 +193,10 @@ void renormalise_bands(const CELTMode *m, celt_norm_t * restrict X)
/* De-normalise the energy to produce the synthesis from the unit-energy bands */
void denormalise_bands(const CELTMode *m, const celt_norm_t * restrict X, celt_sig_t * restrict freq, const celt_ener_t *bank)
{
int i, c;
int i, c, N;
const celt_int16_t *eBands = m->eBands;
const int C = CHANNELS(m);
N = FRAMESIZE(m);
if (C>2)
celt_fatal("denormalise_bands() not implemented for >2 channels");
for (c=0;c<C;c++)
@ -200,12 +206,12 @@ void denormalise_bands(const CELTMode *m, const celt_norm_t * restrict X, celt_s
int j;
celt_word32_t g = SHR32(bank[i+c*m->nbEBands],1);
j=eBands[i]; do {
freq[j*C+c] = SHL32(MULT16_32_Q15(X[j*C+c], g),2);
freq[j+c*N] = SHL32(MULT16_32_Q15(X[j*C+c], g),2);
} while (++j<eBands[i+1]);
}
for (i=eBands[m->nbEBands];i<eBands[m->nbEBands+1];i++)
freq[i+c*N] = 0;
}
for (i=C*eBands[m->nbEBands];i<C*eBands[m->nbEBands+1];i++)
freq[i] = 0;
}

View file

@ -312,7 +312,7 @@ static void compute_mdcts(const CELTMode *mode, int shortBlocks, celt_sig_t * re
mdct_forward(lookup, x, tmp, mode->window, overlap);
/* Interleaving the sub-frames */
for (j=0;j<N;j++)
out[C*j+c] = tmp[j];
out[j+c*N] = tmp[j];
}
RESTORE_STACK;
} else {
@ -336,7 +336,7 @@ static void compute_mdcts(const CELTMode *mode, int shortBlocks, celt_sig_t * re
mdct_forward(lookup, x, tmp, mode->window, overlap);
/* Interleaving the sub-frames */
for (j=0;j<N;j++)
out[C*(j*B+b)+c] = tmp[j];
out[(j*B+b)+c*N*B] = tmp[j];
}
}
RESTORE_STACK;
@ -367,7 +367,7 @@ static void compute_inv_mdcts(const CELTMode *mode, int shortBlocks, celt_sig_t
ALLOC(tmp, N, celt_word32_t);
/* De-interleaving the sub-frames */
for (j=0;j<N;j++)
tmp[j] = X[C*j+c];
tmp[j] = X[j+c*N];
/* Prevents problems from the imdct doing the overlap-add */
CELT_MEMSET(x+N4, 0, N);
mdct_backward(lookup, tmp, x, mode->window, overlap);
@ -397,7 +397,7 @@ static void compute_inv_mdcts(const CELTMode *mode, int shortBlocks, celt_sig_t
{
/* De-interleaving the sub-frames */
for (j=0;j<N2;j++)
tmp[j] = X[C*(j*B+b)+c];
tmp[j] = X[(j*B+b)+c*N2*B];
mdct_backward(lookup, tmp, x+N4+N2*b, mode->window, overlap);
}
if (transient_shift > 0)
@ -613,6 +613,7 @@ int celt_encode_float(CELTEncoder * restrict st, const celt_sig_t * pcm, celt_si
ALLOC(bandLogE,st->mode->nbEBands*C, celt_word16_t);
/* Compute MDCTs */
compute_mdcts(st->mode, shortBlocks, in, freq);
if (shortBlocks && !transient_shift)
{
celt_word32_t sum[4]={1,1,1,1};
@ -622,7 +623,7 @@ int celt_encode_float(CELTEncoder * restrict st, const celt_sig_t * pcm, celt_si
m=0;
do {
celt_word32_t tmp=0;
for (i=m*C+c;i<N;i+=C*st->mode->nbShortMdcts)
for (i=m+c*N;i<(c+1)*N;i+=st->mode->nbShortMdcts)
tmp += ABS32(freq[i]);
sum[m++] += tmp;
} while (m<st->mode->nbShortMdcts);
@ -645,7 +646,7 @@ int celt_encode_float(CELTEncoder * restrict st, const celt_sig_t * pcm, celt_si
{
for (c=0;c<C;c++)
for (m=mdct_weight_pos+1;m<st->mode->nbShortMdcts;m++)
for (i=m*C+c;i<N;i+=C*st->mode->nbShortMdcts)
for (i=m+c*N;i<(c+1)*N;i+=st->mode->nbShortMdcts)
freq[i] = SHR32(freq[i],mdct_weight_shift);
}
#else
@ -665,7 +666,7 @@ int celt_encode_float(CELTEncoder * restrict st, const celt_sig_t * pcm, celt_si
{
for (c=0;c<C;c++)
for (m=mdct_weight_pos+1;m<st->mode->nbShortMdcts;m++)
for (i=m*C+c;i<N;i+=C*st->mode->nbShortMdcts)
for (i=m+c*N;i<(c+1)*N;i+=st->mode->nbShortMdcts)
freq[i] = (1./(1<<mdct_weight_shift))*freq[i];
}
#endif
@ -888,7 +889,7 @@ int celt_encode_float(CELTEncoder * restrict st, const celt_sig_t * pcm, celt_si
int m;
for (c=0;c<C;c++)
for (m=mdct_weight_pos+1;m<st->mode->nbShortMdcts;m++)
for (i=m*C+c;i<N;i+=C*st->mode->nbShortMdcts)
for (i=m+c*N;i<(c+1)*N;i+=st->mode->nbShortMdcts)
#ifdef FIXED_POINT
freq[i] = SHL32(freq[i], mdct_weight_shift);
#else
@ -1242,7 +1243,7 @@ static void celt_decode_lost(CELTDecoder * restrict st, celt_word16_t * restrict
while (offset+len >= MAX_PERIOD)
offset -= pitch_index;
compute_mdcts(st->mode, 0, st->out_mem+offset*C, freq);
for (i=0;i<N;i++)
for (i=0;i<C*N;i++)
freq[i] = ADD32(EPSILON, MULT16_32_Q15(QCONST16(.9f,15),freq[i]));
#endif
@ -1412,7 +1413,7 @@ int celt_decode_float(CELTDecoder * restrict st, const unsigned char *data, int
int m;
for (c=0;c<C;c++)
for (m=mdct_weight_pos+1;m<st->mode->nbShortMdcts;m++)
for (i=m*C+c;i<N;i+=C*st->mode->nbShortMdcts)
for (i=m+c*N;i<(c+1)*N;i+=st->mode->nbShortMdcts)
#ifdef FIXED_POINT
freq[i] = SHL32(freq[i], mdct_weight_shift);
#else