Fixes some overflows in the fixed-point pitch code
The overflows reported were:
- Syy in find_best_pitch() in pitch.c
- xy, xx and yy in remove_doubling() in pitch.c
The fixes are:
- Adaptive scaling in pitch_downsample() which also improves quality
- Adding a missing downshift in find_best_pitch()
- More conservative yshift when calling find_best_pitch()
diff --git a/celt/pitch.c b/celt/pitch.c
index b407902..bea0c3d 100644
--- a/celt/pitch.c
+++ b/celt/pitch.c
@@ -66,7 +66,7 @@
best_pitch[0] = 0;
best_pitch[1] = 1;
for (j=0;j<len;j++)
- Syy = MAC16_16(Syy, y[j],y[j]);
+ Syy = ADD32(Syy, SHR32(MULT16_16(y[j],y[j]), yshift));
for (i=0;i<max_pitch;i++)
{
if (xcorr[i]>0)
@@ -104,14 +104,23 @@
opus_val32 ac[5];
opus_val16 tmp=Q15ONE;
opus_val16 lpc[4], mem[4]={0,0,0,0};
+#ifdef FIXED_POINT
+ int shift;
+ opus_val32 maxabs = celt_maxabs32(x[0], len);
+ if (C==2)
+ maxabs = MAX32(maxabs, celt_maxabs32(x[1], len));
+ shift = IMAX(0,celt_ilog2(maxabs)-11);
+ if (C==2)
+ shift++;
+#endif
for (i=1;i<len>>1;i++)
- x_lp[i] = SHR32(HALF32(HALF32(x[0][(2*i-1)]+x[0][(2*i+1)])+x[0][2*i]), SIG_SHIFT+3);
- x_lp[0] = SHR32(HALF32(HALF32(x[0][1])+x[0][0]), SIG_SHIFT+3);
+ x_lp[i] = SHR32(HALF32(HALF32(x[0][(2*i-1)]+x[0][(2*i+1)])+x[0][2*i]), shift);
+ x_lp[0] = SHR32(HALF32(HALF32(x[0][1])+x[0][0]), shift);
if (C==2)
{
for (i=1;i<len>>1;i++)
- x_lp[i] += SHR32(HALF32(HALF32(x[1][(2*i-1)]+x[1][(2*i+1)])+x[1][2*i]), SIG_SHIFT+3);
- x_lp[0] += SHR32(HALF32(HALF32(x[1][1])+x[1][0]), SIG_SHIFT+3);
+ x_lp[i] += SHR32(HALF32(HALF32(x[1][(2*i-1)]+x[1][(2*i+1)])+x[1][2*i]), shift);
+ x_lp[0] += SHR32(HALF32(HALF32(x[1][1])+x[1][0]), shift);
}
_celt_autocorr(x_lp, ac, NULL, 0,
@@ -231,7 +240,7 @@
}
find_best_pitch(xcorr, y, len>>1, max_pitch>>1, best_pitch
#ifdef FIXED_POINT
- , shift, maxcorr
+ , shift+1, maxcorr
#endif
);