search for: ovflw

Hi Jean-Marc, The original celt_fir() is a little bit messy. It has 2 branches chosen by #ifdef SMALL_FOOTPRINT. For floating-point, the 2 branches are identical (except the operation sequence of accumulating x[i] to sum, which is not a big deal). For fixed-point, the 2 branches are different. I separate them into 2 functions: the new celt_fir(), and celt_fir_permit_overflow() which is the

...erence lies with the signed integer overflow suppression. silk_LPC_analysis_filter() relies on the knowledge that signed integer overflows can occur during the accumulation, but they are guaranteed to cancel each other (i.e. equal wrap-arounds in each direction). For that reason, we use silk_SMLABB_ovflw() which casts to unsigned to avoid the undefined behaviour and thus the ubsan warnings. In celt_fir() and the pitch correlation code it uses, we know there should not be signed overflows, so we would like to detect any problem when using ubsan. I believe the solution would be to always have either...

> > I believe the solution would be to always have either: > 1) USE_CELT_FIR=1 and use ovflw() macros in the xcorr code; or > 2) USE_CELT_FIR=0 and no ovflw() in the xcorr code > I prefer to create a function named silk_fir() with optimization to do the calculation when USE_CELT_FIR=0. xcorr_kernel() itself is great and provides many gains. The only issue is that calling it in a fo...