Game Development Community

Doing the unthinkable...

by Jacob Dankovchik · in Torque Game Engine Advanced · 09/26/2005 (6:56 pm) · 27 replies

This time, I'm totally bringing Atlas to its knees. The Massive, a 32,768 square meter terrain has nothing on what im planning next. This of cource is going to be a very long scale work, several months if I'm lucky.

What I'm doing is I plan to make a terrain for my mission that has the same area as the planet Earth. Thats right, I hope to get my game to have as much area as the entire Earth. I'm not totally sure if Torque can handle all this at once, so right now I'm putting it to the test. I'm using 4096 heightmaps stretched to 40 meters per pixel for a 163,840 meter terrain. I'm simply using flat planes for the moment, just placeing them all out to see if it'll work right. So far I'm up to 1,310,720 meters, no problems. I need to reach about 40,000,000. If Torque can handle this much terrain all at once, then the battle is half won.

After I get all my terrains placed and see Torque can take it, I'll begin replacing the world, segment by segment, with the real data. This is the part that will take forever and a day. After this there will then be a few engine things that need adjustment such as the water physics bug and effects like sectional fog and rain. After that I need to pray for a dynamic shadowing across the terrain so my dynamic sunlight looks right. And I can only pray all the interiors work out right.

But if this works, it will be truley glorious. Never before has a game had an entire world all at once, the same size of the planet!! And to think, such a feat may be accomplished by a little indie game designer. :)

If anyone would like to offer support, I'll be more then happy to take it! Let me know and I'm sure I could use help to get this to form together easier.

Wish me luck! By Friday I should have the full area of the world placed out in flat planes. If Torque holds up that is...
Page«First 1 2 Next»
#21
10/20/2005 (7:33 pm)
It sounds like you've gone so far out that any and all floating-point percision has been lost. Maybe this would still be possible with 64 bit positioning?
#22
10/21/2005 (3:24 am)
Ya, but how would I go about doing that? :/ And would it still work right?
#23
10/21/2005 (6:02 pm)
Well, it looks like TSE already has a "high precision", "Point3D" class that uses 64 bit floats for it's coords. Now trying to integrate that into the current systems that all use Point3Fs is going to be a pain.

Edit: the class is located in engine/math/mPoint.h
#24
10/21/2005 (6:21 pm)
Yes, the Point3D class does exist, but FYI isn't used in world coordinates. I'm honestly not sure why...
#25
10/21/2005 (6:39 pm)
Because it would double the memory requirements for all positional data and slow down processing. Going to doubles also hides the problem - it doesn't solve it. You need to go to a frame of reference system or something similarly esoteric if you want to be numerically stable for very large areas.
#26
10/25/2005 (2:03 pm)
64-bit floating point coordinates should not be necessary for planet-sized play boxes. As a point of comparison, 32-bit integer coordinates (ECEF, for the map geeks out there) gives you 2cm resolution from the center of the Earth out to geosynchronous orbit.
#27
10/25/2005 (8:45 pm)
Well floating point numbers don't provide a consistent level of accuracy. But +/- 20 million (?) doesn't seem out of the question...

And the model exploding. The game transforms the matrix to the object's position, and then renders near the origin. I'm not sure if this would actually circumvent the problem or not, but all in all this seems wierd to be causing problems.

I'm curious what your coordinates were Jacob.
Page«First 1 2 Next»