Double-Precision Rendering of Large Objects in T3D

[JackStone]

As I mentioned in my previous post, I am attempting to render planet-sized objects in T3D. I was trying to implement a solutions based on scaling the planet, and dividing it into chunks, each of which was less than 10,000 units, but this hasnt worked.

I am now considering changing T3D's position and scaling variables for my custom "spherical terrain" object type to use F64's instead of F32's. I found this thread on the old forum:

https://www.garagegames.com/community/forums/viewthread/82413

Which seems to indicate that this involves interacting closely with the rendering layer.

My question is, how complex would this be to do? I am trying to simply render an object the size of a planet, I dont want to change the entire engine to 64-bit. This is the effect when I try to do this with the current engine:

http://phoenixgamedevelopment.com/blog/wp-content/uploads/2016/06/PhoenixGameDevelopment-12_06_2016-08_10_21-PM.jpg

http://phoenixgamedevelopment.com/blog/wp-content/uploads/2016/06/PhoenixGameDevelopment-12_06_2016-08_08_36-PM.jpg

Clearly these are 32-bit precision issues. Can anyone shed any light on the right approach here? Is it just a matter of going through the rendering system and changing F32/Point3F's to F64's/Point3D's?

[LukasPJ]

Looks like z-fighting to me, but I've only looked at it in my phone

[buckmaster]

In case you haven't seen it already, I highly recommend this series of articles on Gamasutra to anyone trying to to large-scale rendering. I don't think you've having exactly the issue I thought of first (the author uses double-precision for storing object locations, but still renders using floats), but they definitely solved a lot of issues.

[JackStone]

Sorry, I only just noticed this. I have looked at those tutorials in detail, thanks Daniel, they are hugely useful.