I am interested in making a space game in which the universe moves around the ship and i would use double for objects matrix/pos . (i know double is not big enough but still way better than just floats).
Can you tell please if there is a way to make the engine use by default doubles for it's pos/rotation matrixes for objects ? I know that at rendering phase it will switch to floats anyway due to hardware constrains.

VecD // Vector 3D (double precision)
MatrixD : MatrixD3 // Matrix 4x3 (orientation + scale + position, double precision)

Keep in mind that AFAIK, the physics engine(PhysX) does not support double precision. You would need to work out some kind of translation system to fool the physics into behaving, or use a custom implementation of physics. Maybe you could force Bullet or some other physics engine that supports double precision.

I suppose this would be a good read.
http://www.udellgames.com/posts/size-mat...ision-too/

Thank you for your answers.
Esenthel, can you tell please if it would be possible to use bullet and double floats in engine in order to have a space bigger than 100km ? ( float limitation) ?

Hello,

To use Bullet you can look into "headers.h"

To use double in Bullet you'd need to look into the Bullet sources.

Thank you for the fast reply. I think for now i will start with normal floats and down on the road i will switch to doubles. Is good to know that the engine will allow me.
There are very few engines that can use double precision. Maybe if is not too much work for you, you can make it an advertised feature (even if not out of the box).