eden-emu/opus
19
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Re-organize spreading/folding constants.

Browse source 
These were stored internally in one order and in the bitstream in a
 different order.
Both used bare constants, making it unclear what either actually
 meant.
This changes them to use the same order, gives them named constants,
 and renames all the "fold" decision stuff to "spread" instead,
 since that is what it is really controlling.
This commit is contained in:
Timothy B. Terriberry 2010-12-17 05:52:06 -08:00 • committed by Jean-Marc Valin
parent cd84e3d0f4
commit 320cf2e2cd
4 changed files with 37 additions and 41 deletions
Download patch file Download diff file Expand all files Collapse all files

29
libcelt/bands.c
View file

@ -282,7 +282,7 @@ static void stereo_merge(celt_norm *X, celt_norm *Y, celt_word16 mid, celt_word1
} }
/* Decide whether we should spread the pulses in the current frame */ /* Decide whether we should spread the pulses in the current frame */
int folding_decision(const CELTMode *m, celt_norm *X, int *average, int *last_decision, int end, int _C, int M) int spreading_decision(const CELTMode *m, celt_norm *X, int *average, int last_decision, int end, int _C, int M)
{ {
int i, c, N0; int i, c, N0;
int sum = 0, nbBands=0; int sum = 0, nbBands=0;
@ -293,7 +293,7 @@ int folding_decision(const CELTMode *m, celt_norm *X, int *average, int *last_de
N0 = M*m->shortMdctSize; N0 = M*m->shortMdctSize;
if (M*(eBands[end]-eBands[end-1]) <= 8) if (M*(eBands[end]-eBands[end-1]) <= 8)
return 0; return SPREAD_NONE;
c=0; do { c=0; do {
for (i=0;i<end;i++) for (i=0;i<end;i++)
{ {
@ -327,23 +327,18 @@ int folding_decision(const CELTMode *m, celt_norm *X, int *average, int *last_de
sum = (sum+*average)>>1; sum = (sum+*average)>>1;
*average = sum; *average = sum;
/* Hysteresis */ /* Hysteresis */
sum = (3*sum + ((*last_decision<<7) + 64) + 2)>>2; sum = (3*sum + (((3-last_decision)<<7) + 64) + 2)>>2;
/* decision and last_decision do not use the same encoding */
if (sum < 80) if (sum < 80)
{ {
decision = 2; decision = SPREAD_AGGRESSIVE;
*last_decision = 0;
} else if (sum < 256) } else if (sum < 256)
{ {
decision = 1; decision = SPREAD_NORMAL;
*last_decision = 1;
} else if (sum < 384) } else if (sum < 384)
{ {
decision = 3; decision = SPREAD_LIGHT;
*last_decision = 2;
} else { } else {
decision = 0; decision = SPREAD_NONE;
*last_decision = 3;
} }
return decision; return decision;
} }
@ -828,7 +823,7 @@ static void quant_band(int encode, const CELTMode *m, int i, celt_norm *X, celt_
int j; int j;
if (lowband != NULL && resynth) if (lowband != NULL && resynth)
{ {
if (spread==2 && B<=1) if (spread==SPREAD_AGGRESSIVE && B<=1)
{ {
/* Noise */ /* Noise */
for (j=0;j<N;j++) for (j=0;j<N;j++)
@ -903,7 +898,7 @@ static void quant_band(int encode, const CELTMode *m, int i, celt_norm *X, celt_
void quant_all_bands(int encode, const CELTMode *m, int start, int end, void quant_all_bands(int encode, const CELTMode *m, int start, int end,
celt_norm *_X, celt_norm *_Y, const celt_ener *bandE, int *pulses, celt_norm *_X, celt_norm *_Y, const celt_ener *bandE, int *pulses,
int shortBlocks, int fold, int dual_stereo, int intensity, int *tf_res, int resynth, int shortBlocks, int spread, int dual_stereo, int intensity, int *tf_res, int resynth,
int total_bits, void *ec, int LM, int codedBands) int total_bits, void *ec, int LM, int codedBands)
{ {
int i, balance; int i, balance;
@ -1028,14 +1023,14 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
} }
if (dual_stereo) if (dual_stereo)
{ {
quant_band(encode, m, i, X, NULL, N, b/2, fold, B, intensity, tf_change, quant_band(encode, m, i, X, NULL, N, b/2, spread, B, intensity, tf_change,
effective_lowband != -1 ? norm+effective_lowband : NULL, resynth, ec, &remaining_bits, LM, effective_lowband != -1 ? norm+effective_lowband : NULL, resynth, ec, &remaining_bits, LM,
norm+M*eBands[i], bandE, 0, &seed, Q15ONE, lowband_scratch); norm+M*eBands[i], bandE, 0, &seed, Q15ONE, lowband_scratch);
quant_band(encode, m, i, Y, NULL, N, b/2, fold, B, intensity, tf_change, quant_band(encode, m, i, Y, NULL, N, b/2, spread, B, intensity, tf_change,
effective_lowband != -1 ? norm2+effective_lowband : NULL, resynth, ec, &remaining_bits, LM, effective_lowband != -1 ? norm2+effective_lowband : NULL, resynth, ec, &remaining_bits, LM,
norm2+M*eBands[i], bandE, 0, &seed, Q15ONE, lowband_scratch); norm2+M*eBands[i], bandE, 0, &seed, Q15ONE, lowband_scratch);
} else { } else {
quant_band(encode, m, i, X, Y, N, b, fold, B, intensity, tf_change, quant_band(encode, m, i, X, Y, N, b, spread, B, intensity, tf_change,
effective_lowband != -1 ? norm+effective_lowband : NULL, resynth, ec, &remaining_bits, LM, effective_lowband != -1 ? norm+effective_lowband : NULL, resynth, ec, &remaining_bits, LM,
norm+M*eBands[i], bandE, 0, &seed, Q15ONE, lowband_scratch); norm+M*eBands[i], bandE, 0, &seed, Q15ONE, lowband_scratch);
} }

7
libcelt/bands.h
View file

@ -64,7 +64,12 @@ void normalise_bands(const CELTMode *m, const celt_sig * restrict freq, celt_nor
*/ */
void denormalise_bands(const CELTMode *m, const celt_norm * restrict X, celt_sig * restrict freq, const celt_ener *bands, int end, int _C, int M); void denormalise_bands(const CELTMode *m, const celt_norm * restrict X, celt_sig * restrict freq, const celt_ener *bands, int end, int _C, int M);
int folding_decision(const CELTMode *m, celt_norm *X, int *average, int *last_decision, int end, int _C, int M); #define SPREAD_NONE (0)
#define SPREAD_LIGHT (1)
#define SPREAD_NORMAL (2)
#define SPREAD_AGGRESSIVE (3)
int spreading_decision(const CELTMode *m, celt_norm *X, int *average, int last_decision, int end, int _C, int M);
#ifdef MEASURE_NORM_MSE #ifdef MEASURE_NORM_MSE
void measure_norm_mse(const CELTMode *m, float *X, float *X0, float *bandE, float *bandE0, int M, int N, int C); void measure_norm_mse(const CELTMode *m, float *X, float *X0, float *bandE, float *bandE0, int M, int N, int C);

31
libcelt/celt.c
View file

@ -77,7 +77,7 @@ struct CELTEncoder {
#define ENCODER_RESET_START frame_max #define ENCODER_RESET_START frame_max
celt_word32 frame_max; celt_word32 frame_max;
int fold_decision; int spread_decision;
int delayedIntra; int delayedIntra;
int tonal_average; int tonal_average;
int lastCodedBands; int lastCodedBands;
@ -155,7 +155,7 @@ CELTEncoder *celt_encoder_init(CELTEncoder *st, const CELTMode *mode, int channe
st->force_intra = 0; st->force_intra = 0;
st->delayedIntra = 1; st->delayedIntra = 1;
st->tonal_average = 256; st->tonal_average = 256;
st->fold_decision = 1; st->spread_decision = SPREAD_NORMAL;
st->complexity = 5; st->complexity = 5;
if (error) if (error)
@ -719,7 +719,6 @@ int celt_encode_with_ec_float(CELTEncoder * restrict st, const celt_sig * pcm, i
#endif #endif
int i, c, N; int i, c, N;
int bits; int bits;
int has_fold=1;
ec_byte_buffer buf; ec_byte_buffer buf;
ec_enc _enc; ec_enc _enc;
VARDECL(celt_sig, in); VARDECL(celt_sig, in);
@ -950,17 +949,17 @@ int celt_encode_with_ec_float(CELTEncoder * restrict st, const celt_sig * pcm, i
{ {
if (st->complexity == 0) if (st->complexity == 0)
{ {
has_fold = 0; st->spread_decision = SPREAD_NONE;
st->fold_decision = 3;
} else { } else {
has_fold = 1; st->spread_decision = SPREAD_NORMAL;
st->fold_decision = 1;
} }
} else { } else {
has_fold = folding_decision(st->mode, X, &st->tonal_average, &st->fold_decision, effEnd, C, M); st->spread_decision = spreading_decision(st->mode, X, &st->tonal_average, st->spread_decision, effEnd, C, M);
} }
ec_enc_bit_prob(enc, has_fold>>1, 8192); /* Probs: NONE: 21.875%, LIGHT: 6.25%, NORMAL: 65.625%, AGGRESSIVE: 6.25% */
ec_enc_bit_prob(enc, has_fold&1, (has_fold>>1) ? 32768 : 49152); ec_enc_bit_prob(enc, st->spread_decision>>1, 18432);
ec_enc_bit_prob(enc, st->spread_decision&1,
(st->spread_decision>>1) ? 5699 : 14564);
ALLOC(offsets, st->mode->nbEBands, int); ALLOC(offsets, st->mode->nbEBands, int);
@ -1144,7 +1143,7 @@ int celt_encode_with_ec_float(CELTEncoder * restrict st, const celt_sig * pcm, i
/* Residual quantisation */ /* Residual quantisation */
quant_all_bands(1, st->mode, st->start, st->end, X, C==2 ? X+N : NULL, quant_all_bands(1, st->mode, st->start, st->end, X, C==2 ? X+N : NULL,
bandE, pulses, shortBlocks, has_fold, dual_stereo, intensity, tf_res, resynth, bandE, pulses, shortBlocks, st->spread_decision, dual_stereo, intensity, tf_res, resynth,
nbCompressedBytes*8, enc, LM, codedBands); nbCompressedBytes*8, enc, LM, codedBands);
quant_energy_finalise(st->mode, st->start, st->end, bandE, oldBandE, error, fine_quant, fine_priority, nbCompressedBytes*8-ec_enc_tell(enc, 0), enc, C); quant_energy_finalise(st->mode, st->start, st->end, bandE, oldBandE, error, fine_quant, fine_priority, nbCompressedBytes*8-ec_enc_tell(enc, 0), enc, C);
@ -1388,7 +1387,7 @@ int celt_encoder_ctl(CELTEncoder * restrict st, int request, ...)
((char*)&st->ENCODER_RESET_START - (char*)st)); ((char*)&st->ENCODER_RESET_START - (char*)st));
st->vbr_offset = 0; st->vbr_offset = 0;
st->delayedIntra = 1; st->delayedIntra = 1;
st->fold_decision = 1; st->spread_decision = SPREAD_NORMAL;
st->tonal_average = QCONST16(1.f,8); st->tonal_average = QCONST16(1.f,8);
} }
break; break;
@ -1695,7 +1694,7 @@ int celt_decode_with_ec_float(CELTDecoder * restrict st, const unsigned char *da
{ {
#endif #endif
int c, i, N; int c, i, N;
int has_fold; int spread_decision;
int bits; int bits;
ec_dec _dec; ec_dec _dec;
ec_byte_buffer buf; ec_byte_buffer buf;
@ -1824,8 +1823,8 @@ int celt_decode_with_ec_float(CELTDecoder * restrict st, const unsigned char *da
ALLOC(tf_res, st->mode->nbEBands, int); ALLOC(tf_res, st->mode->nbEBands, int);
tf_decode(st->start, st->end, C, isTransient, tf_res, LM, dec); tf_decode(st->start, st->end, C, isTransient, tf_res, LM, dec);
has_fold = ec_dec_bit_prob(dec, 8192)<<1; spread_decision = ec_dec_bit_prob(dec, 18432)<<1;
has_fold |= ec_dec_bit_prob(dec, (has_fold>>1) ? 32768 : 49152); spread_decision |= ec_dec_bit_prob(dec, (spread_decision>>1) ? 5699 : 14564);
ALLOC(pulses, st->mode->nbEBands, int); ALLOC(pulses, st->mode->nbEBands, int);
ALLOC(offsets, st->mode->nbEBands, int); ALLOC(offsets, st->mode->nbEBands, int);
@ -1868,7 +1867,7 @@ int celt_decode_with_ec_float(CELTDecoder * restrict st, const unsigned char *da
/* Decode fixed codebook */ /* Decode fixed codebook */
quant_all_bands(0, st->mode, st->start, st->end, X, C==2 ? X+N : NULL, quant_all_bands(0, st->mode, st->start, st->end, X, C==2 ? X+N : NULL,
NULL, pulses, shortBlocks, has_fold, dual_stereo, intensity, tf_res, 1, NULL, pulses, shortBlocks, spread_decision, dual_stereo, intensity, tf_res, 1,
len*8, dec, LM, codedBands); len*8, dec, LM, codedBands);
unquant_energy_finalise(st->mode, st->start, st->end, bandE, oldBandE, unquant_energy_finalise(st->mode, st->start, st->end, bandE, oldBandE,

11
libcelt/vq.c
View file

@ -39,6 +39,7 @@
#include "vq.h" #include "vq.h"
#include "arch.h" #include "arch.h"
#include "os_support.h" #include "os_support.h"
#include "bands.h"
#include "rate.h" #include "rate.h"
#ifndef M_PI #ifndef M_PI
@ -71,6 +72,7 @@ static void exp_rotation1(celt_norm *X, int len, int stride, celt_word16 c, celt
static void exp_rotation(celt_norm *X, int len, int dir, int stride, int K, int spread) static void exp_rotation(celt_norm *X, int len, int dir, int stride, int K, int spread)
{ {
static const int SPREAD_FACTOR[3]={5,10,15};
int i; int i;
celt_word16 c, s; celt_word16 c, s;
celt_word16 gain, theta; celt_word16 gain, theta;
@ -84,14 +86,9 @@ static void exp_rotation(celt_norm *X, int len, int dir, int stride, int K, int
X[14] = 1; X[14] = 1;
K=5; K=5;
}*/ }*/
if (2*K>=len || spread==0) if (2*K>=len || spread==SPREAD_NONE)
return; return;
if (spread==1) factor = SPREAD_FACTOR[spread-1];
factor=10;
else if (spread==2)
factor=5;
else
factor=15;
gain = celt_div((celt_word32)MULT16_16(Q15_ONE,len),(celt_word32)(len+factor*K)); gain = celt_div((celt_word32)MULT16_16(Q15_ONE,len),(celt_word32)(len+factor*K));
/* FIXME: Make that HALF16 instead of HALF32 */ /* FIXME: Make that HALF16 instead of HALF32 */