Hello Tanu,
thanks for reviewing!
> > > + case PA_SAMPLE_S16NE:
> > > + {
> > > + int16_t *d = (int16_t *) dst, *s = (int16_t *) src;
> > > +
> > > + for (i = n>> 2; i> 0; i--) {
> > > + d[0] += (s[0] + s[1])/2;
> > > + d[1] += (s[2] + s[3])/2;
> > > + d[2] += (s[4] + s[5])/2;
> > > + d[3] += (s[6] + s[7])/2;
> >
> > You probably meant '=', not '+='?
> >
> > Also, s[0] + s[1] can overflow, so the inputs should be divided
> > individually before summing.
>
> Or the inputs could be cast to uint32_t, which I guess would be better
> than dividing multiple times.
it should be '=' obviously -- will fix and repost shortly
I think (s[0] + s[1])/2 is correct; at least as long as sizeof(int) >
sizeof(short); short gets promoted to int -- see 'integer promotion in
C99'; also my compiler thinks above is correct :)
dividing before summation looses precision and should be avoided (the
generic matrix remapping codes does multiplication and truncation,
probably for performance reasons; there should be an intermediate (larger)
datatype and rounding if we cared about precision)
regards, p.
--
Peter Meerwald
+43-664-2444418 (mobile)
