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

Fixes fixed-point overflow in stereo_merge()

Browse source
This commit is contained in:
Jean-Marc Valin 2010-09-14 16:02:47 -04:00
parent d60e0db0e2
commit 3c15659c39
1 changed files with 13 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

17
libcelt/bands.c
View file

@ -264,8 +264,10 @@ static void stereo_merge(celt_norm *X, celt_norm *Y, celt_word16 mid, celt_word1
for (j=0;j<N;j++)
xp = MAC16_16(xp, X[j], Y[j]);
/* mid and side are in Q15, not Q14 like X and Y */
El = MULT16_16(mid, mid) + MULT16_16(side, side) - 2*SHL32(xp,2);
Er = MULT16_16(mid, mid) + MULT16_16(side, side) + 2*SHL32(xp,2);
mid = SHR32(mid, 1);
side = SHR32(side, 1);
El = MULT16_16(mid, mid) + MULT16_16(side, side) - 2*xp;
Er = MULT16_16(mid, mid) + MULT16_16(side, side) + 2*xp;
if (Er < EPSILON)
Er = EPSILON;
if (El < EPSILON)
@ -280,13 +282,20 @@ static void stereo_merge(celt_norm *X, celt_norm *Y, celt_word16 mid, celt_word1
t = VSHR32(Er, (kr-7)<<1);
rgain = celt_rsqrt_norm(t);
#ifdef FIXED_POINT
if (kl < 7)
kl = 7;
if (kr < 7)
kr = 7;
#endif
for (j=0;j<N;j++)
{
celt_norm r, l;
l = X[j];
r = Y[j];
X[j] = EXTRACT16(PSHR32(MULT16_16(lgain, l-r), kl));
Y[j] = EXTRACT16(PSHR32(MULT16_16(rgain, l+r), kr));
X[j] = EXTRACT16(PSHR32(MULT16_16(lgain, SUB16(l,r)), kl+1));
Y[j] = EXTRACT16(PSHR32(MULT16_16(rgain, ADD16(l,r)), kr+1));
}
}