> On a second look, the frac UnaryOp UGen doesn't use sc_frac, but the
> nova-simd stuff now. Still it seems it suffers from the same rounding
> error.
hm, maybe the issue is more complex ... the unary op ugen behaves
differently from the SimpleNumber
i just saw, frac_a is implemented as:
LOOP(inNumSamples,
float xa = ZXP(a);
ZXP(out) = xa - std::floor(xa);
);
so it is not the same operation as sc_frac, which is implemented as:
inline float32 sc_frac(float32 x)
{
return x - sc_trunc(x);
}
... so for negative input values, the fractional part will be negative
in sclang, but positive in scsynth (without the nova-simd patch).
this is inconsistent and imo should be changed. since it changes the
behavior, i would like to hear, what other people think about this
issue, though ...
thnx, tim
--
tim@xxxxxxxxxx
http://tim.klingt.org
Our products just aren't engineered for security.
Brian Valentine, Microsoft's vice president of Windows development
Attachment:
signature.asc
Description: OpenPGP digital signature