Please help me bust some myths about really large worlds

[dumbo2007]

So what are my current options for very large worlds for a spaceflight sim. After going through many threads here this is the conclusion I came to:

1. I cannot use multiple bullet simulations in a single program if I am updating them in parallel, to say simulate 2 different bases on say earth and mars. This is because the Bullet memory allocator is not thread safe.

2. Bullet can support huge worlds, but since floating point precision drops below 0.01 beyond say about 5 km, I effectively can use a dynamics world of size only +5000 by -5000.
(Ok maybe its not 5km but 10 km, but a single base in my space sim is 3 km, so since I have to keep them apart from each other I can support at most 3 bases, which is too less !!)

3. The MultiSAP broadphase is not ready and I cannot use it yet to distribute objects which are in 2 different bases(so very far from each other) into 2 different broadphases.

4. I cannot use multithreading with MultiSAP.

Please help me bust these myths

[Ebon]

I'm just brainstorming here, but would it be possible simulating both bases on top of each other, using masks to separate the collisions?

[CookieMonster]

Are you defining btScalar as a 64 bit float?

[dumbo2007]

Ebon,

Yes thats the solution I am leaning towards right now. Have you tried the MultiSAP yet by the way ? Hadnt thought of the mask though, thats a good idea


CookieMonster,
You mean the double type ? No I am using single precision for now. I will probably shift to double even though it may slow things down. Something tells me the speed hit wont make a
difference on current multi-core CPUs


Whats the current opinion on multiple sims updating parallely , by the way, in case anyone tried it ?