What Hodgman said. On G70 & G60 hardware (and some Intel chips) it will faster (up to 2x faster).
But beware that if you're using new hardware, you may miss a few bugs. I've had a case where a half was overflowing it's 16-bit precision, but on new HW it worked correctly because it was just being calculated in 32-bit precision.
When we tried it on old HW (an nForce 6150 & an Intel Mobile 4 Series to be exact), all hell break loose; and had to change a few variables from half to float (and add a clamp statement for a few halfs).
Okay, thanks. I'll take thorough thought about it. Another question is do you guys know where the shadowmap function (the functor is pointing at) is? I guess it is in 3ds max, but can't find it anywhere
Edited by lucky6969b, 11 November 2012 - 04:50 AM.