How far from point zero can we go without problems ?

Started by Kadri, January 04, 2010, 12:27:45 AM

Previous topic - Next topic

Kadri

This not much of a practical problem for me. There are many ways to deal with this and other software have this problem too.

But just out of curiosity is there a way to know how far away from point zero we can go  until we see problem in TG2 calculation rounding ?

Say you have a default planet and then make a new one , nearly 10 times in distance (the diameter of the default planet) .
Can we go to the new one and make all adjustment without worry or will there any  problems ?
Or are there problems right on the first planet too ?

I searched the forum but could not find anything .

Just curious  :)

Kadri.

cyphyr

Your absolutely right, the further away from the zero point the greater the rounding calculations are. I think in your example (x10 planetary diameters) it will be way off. I've had rounding errors from imported .chan data coming in from orbital range, a couple of hundred clicks up. In that instance I was able to fix the issue by letting terragen re-interpret the keyframe data say every 25th frame (the .chan format is effectively one keyframe every frame), any keening was simply absorbed in the large scale. However trying the same method on say an off-world anim of flying down a canyon would run into multiple problems.
Good luck
Richard
www.richardfraservfx.com
https://www.facebook.com/RichardFraserVFX/
/|\

Ryzen 9 5950X OC@4Ghz, 64Gb (TG4 benchmark 4:13)


Njen

Systems based around 32bit will have problems accurately portraying large numbers on either side of the decimal point. Once the (much hoped for) 64bit version of TG2 is released, this will allow for far greater precision in the handling of large numbers, which will reduce any calculation problems far away from the origin.

Kadri

Quote from: njen on January 04, 2010, 11:28:36 AM
Systems based around 32bit will have problems accurately portraying large numbers on either side of the decimal point. Once the (much hoped for) 64bit version of TG2 is released, this will allow for far greater precision in the handling of large numbers, which will reduce any calculation problems far away from the origin.

I thought so too , Njen. But then Matt said that TG2 internally handles all calculations already  in 64 bit .
Thus with the coming 64 bit version there will be no difference.

Kadri.

Hetzen

I'd like to know if that's true. I seem to remember reading that render calculations are already in the 64bit floating point arena. I maybe wrong.


Njen