ref: 35cf18e9d76b54f5a9f310fb5d5db3fc07ea758f
parent: 886dc594f087284eefa4c6947da2a35eceb4cef9
author: Jean-Marc Valin <jmvalin@jmvalin.ca>
date: Tue Jan 5 21:02:16 EST 2016
Making stereo theta decision based on minimizing distortion No point in minimizing the rate too since it's almost constant.
--- a/celt/bands.c
+++ b/celt/bands.c
@@ -1336,7 +1336,7 @@
const celt_ener *bandE, int *pulses, int shortBlocks, int spread,
int dual_stereo, int intensity, int *tf_res, opus_int32 total_bits,
opus_int32 balance, ec_ctx *ec, int LM, int codedBands,
- opus_uint32 *seed, int arch)
+ opus_uint32 *seed, int complexity, int arch)
{int i;
opus_int32 remaining_bits;
@@ -1344,6 +1344,8 @@
celt_norm * OPUS_RESTRICT norm, * OPUS_RESTRICT norm2;
VARDECL(celt_norm, _norm);
VARDECL(celt_norm, _lowband_scratch);
+ VARDECL(celt_norm, X_save);
+ VARDECL(celt_norm, Y_save);
int resynth_alloc;
celt_norm *lowband_scratch;
int B;
@@ -1352,10 +1354,11 @@
int update_lowband = 1;
int C = Y_ != NULL ? 2 : 1;
int norm_offset;
+ int theta_rdo = encode && Y_!=NULL && !dual_stereo && complexity>=8;
#ifdef RESYNTH
int resynth = 1;
#else
- int resynth = !encode;
+ int resynth = !encode || theta_rdo;
#endif
struct band_ctx ctx;
SAVE_STACK;
@@ -1381,6 +1384,8 @@
lowband_scratch = _lowband_scratch;
else
lowband_scratch = X_+M*eBands[m->nbEBands-1];
+ ALLOC(X_save, resynth_alloc, celt_norm);
+ ALLOC(Y_save, resynth_alloc, celt_norm);
lowband_offset = 0;
ctx.bandE = bandE;
@@ -1392,6 +1397,7 @@
ctx.spread = spread;
ctx.arch = arch;
ctx.resynth = resynth;
+ ctx.theta_round = 0;
for (i=start;i<end;i++)
{opus_int32 tell;
@@ -1441,7 +1447,7 @@
Y = norm;
lowband_scratch = NULL;
}
- if (i==end-1)
+ if (last && !theta_rdo)
lowband_scratch = NULL;
/* Get a conservative estimate of the collapse_mask's for the bands we're
@@ -1489,10 +1495,51 @@
} else {if (Y!=NULL)
{- ctx.theta_round = 0;
- x_cm = quant_band_stereo(&ctx, X, Y, N, b, B,
- effective_lowband != -1 ? norm+effective_lowband : NULL, LM,
- last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, x_cm|y_cm);
+ if (theta_rdo)
+ {+ ec_ctx ec_save;
+ struct band_ctx ctx_save;
+ opus_val32 dist0, dist1;
+ unsigned cm;
+ /* Make a copy. */
+ cm = x_cm|y_cm;
+ ec_save = *ec;
+ ctx_save = ctx;
+ OPUS_COPY(X_save, X, N);
+ OPUS_COPY(Y_save, Y, N);
+ /* Encode and round down. */
+ ctx.theta_round = -1;
+ x_cm = quant_band_stereo(&ctx, X, Y, N, b, B,
+ effective_lowband != -1 ? norm+effective_lowband : NULL, LM,
+ last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, cm);
+ dist0 = celt_inner_prod(X_save, X, N, arch) + celt_inner_prod(Y_save, Y, N, arch);
+ /* Restore */
+ *ec = ec_save;
+ ctx = ctx_save;
+ OPUS_COPY(X, X_save, N);
+ OPUS_COPY(Y, Y_save, N);
+ /* Encode and round up. */
+ ctx.theta_round = 1;
+ x_cm = quant_band_stereo(&ctx, X, Y, N, b, B,
+ effective_lowband != -1 ? norm+effective_lowband : NULL, LM,
+ last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, cm);
+ dist1 = celt_inner_prod(X_save, X, N, arch) + celt_inner_prod(Y_save, Y, N, arch);
+ /* Restore */
+ *ec = ec_save;
+ ctx = ctx_save;
+ OPUS_COPY(X, X_save, N);
+ OPUS_COPY(Y, Y_save, N);
+ /* Encode with best choice. */
+ ctx.theta_round = dist0 >= dist1 ? -1 : 1;
+ x_cm = quant_band_stereo(&ctx, X, Y, N, b, B,
+ effective_lowband != -1 ? norm+effective_lowband : NULL, LM,
+ last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, cm);
+ } else {+ ctx.theta_round = 0;
+ x_cm = quant_band_stereo(&ctx, X, Y, N, b, B,
+ effective_lowband != -1 ? norm+effective_lowband : NULL, LM,
+ last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, x_cm|y_cm);
+ }
} else {x_cm = quant_band(&ctx, X, N, b, B,
effective_lowband != -1 ? norm+effective_lowband : NULL, LM,
--- a/celt/bands.h
+++ b/celt/bands.h
@@ -105,7 +105,7 @@
const celt_ener *bandE, int *pulses, int shortBlocks, int spread,
int dual_stereo, int intensity, int *tf_res, opus_int32 total_bits,
opus_int32 balance, ec_ctx *ec, int M, int codedBands, opus_uint32 *seed,
- int arch);
+ int complexity, int arch);
void anti_collapse(const CELTMode *m, celt_norm *X_,
unsigned char *collapse_masks, int LM, int C, int size, int start,
--- a/celt/celt_decoder.c
+++ b/celt/celt_decoder.c
@@ -1005,7 +1005,7 @@
quant_all_bands(0, mode, start, end, X, C==2 ? X+N : NULL, collapse_masks,
NULL, pulses, shortBlocks, spread_decision, dual_stereo, intensity, tf_res,
- len*(8<<BITRES)-anti_collapse_rsv, balance, dec, LM, codedBands, &st->rng, st->arch);
+ len*(8<<BITRES)-anti_collapse_rsv, balance, dec, LM, codedBands, &st->rng, 0, st->arch);
if (anti_collapse_rsv > 0)
{--- a/celt/celt_encoder.c
+++ b/celt/celt_encoder.c
@@ -2079,7 +2079,7 @@
quant_all_bands(1, mode, start, end, X, C==2 ? X+N : NULL, collapse_masks,
bandE, pulses, shortBlocks, st->spread_decision,
dual_stereo, st->intensity, tf_res, nbCompressedBytes*(8<<BITRES)-anti_collapse_rsv,
- balance, enc, LM, codedBands, &st->rng, st->arch);
+ balance, enc, LM, codedBands, &st->rng, st->complexity, st->arch);
if (anti_collapse_rsv > 0)
{