search for: f5adc48

...float sum = 0; > + float *yi = _y+i; > + for (j = 0; j < len; j++) > + sum += _x[i]*yi[i]; > + xcorr[i] = sum; > + } This loop can still be largely vectorized. Any reason not to do so? > +} > diff --git a/celt/arm/pitch_arm.h b/celt/arm/pitch_arm.h > index a07f8ac..f5adc48 100644 > --- a/celt/arm/pitch_arm.h > +++ b/celt/arm/pitch_arm.h > @@ -52,6 +52,19 @@ opus_val32 celt_pitch_xcorr_edsp(const opus_val16 *_x, const opus_val16 *_y, > ((void)(arch),PRESUME_NEON(celt_pitch_xcorr)(_x, _y, xcorr, len, max_pitch)) > # endif > > -# endif > +...

...xcorr[i] = sum; >> + } > > This loop can still be largely vectorized. Any reason not to do so? Should have put TBD in the comments..I will try to get this done in the re-spin. > >> +} >> diff --git a/celt/arm/pitch_arm.h b/celt/arm/pitch_arm.h >> index a07f8ac..f5adc48 100644 >> --- a/celt/arm/pitch_arm.h >> +++ b/celt/arm/pitch_arm.h >> @@ -52,6 +52,19 @@ opus_val32 celt_pitch_xcorr_edsp(const opus_val16 *_x, const opus_val16 *_y, >> ((void)(arch),PRESUME_NEON(celt_pitch_xcorr)(_x, _y, xcorr, len, max_pitch)) >> # endif >&...

Hello, I received feedback from engineers working on NE10 [1] that it would be better to use NE10 [1] for FFT optimizations for opus use cases. However, these FFT patches are currently in review and haven't been integrated into NE10 yet. While the FFT functions in NE10 are getting baked, I wanted to optimize the celt_pitch_xcorr (floating point only) and use it to introduce ARM NEON

...+ + /* In case max_pitch isn't multiple of 4, do unrolled version */ + for (; i < max_pitch; i++) { + float sum = 0; + float *yi = _y+i; + for (j = 0; j < len; j++) + sum += _x[i]*yi[i]; + xcorr[i] = sum; + } +} diff --git a/celt/arm/pitch_arm.h b/celt/arm/pitch_arm.h index a07f8ac..f5adc48 100644 --- a/celt/arm/pitch_arm.h +++ b/celt/arm/pitch_arm.h @@ -52,6 +52,19 @@ opus_val32 celt_pitch_xcorr_edsp(const opus_val16 *_x, const opus_val16 *_y, ((void)(arch),PRESUME_NEON(celt_pitch_xcorr)(_x, _y, xcorr, len, max_pitch)) # endif -# endif +#else /* Start !FIXED_POINT */ +/* Float...