[Bf-committers] Re: Re: Re: [Bf-blender-cvs] CVS commit:
gsr.b3d at infernal-iceberg.com
Mon Oct 10 21:41:02 CEST 2005
arne.schmitz at mmweg.rwth-aachen.de (2005-10-10 at 2132.52 +0200):
> Am Montag, 10. Oktober 2005 21:14 schrieb GSR:
> > Should not the result already be between that, and if not, then
> > something is really wrong? In other words, not clamp but return error.
> No, after a dot product, the value will sometimes slightly fall out of the
> [-1, 1] range, due to rounding errors. For performance reasons, you wouldn't
> want to clamp the values ALL the time, but only e.g. if you compute the acos
> of a value. The acos is of course *only* defined over [-1, 1].
Which is what I said below... and then pointed the zero test is left
free to suffer of similar problems.
More information about the Bf-committers