[FEATURE REQUEST] Using double-precision floating-point numbers in some parts of the BEPU engine

[ahmetenesay]

I appreciate the BEPU project; I'm incredibly pleased that a low-level physics engine has been written in pure C#! Honestly, I don't know why I really want what I mentioned in the title, but having FP64 precision in some parts of the engine sounds appealing. While GPUs for home users are terrible at handling double-precision dotted numbers, CPUs are quite comfortable and powerful enough for FP64 (usually half as good as, or sometimes equal to, FP32). What do you think?

[RossNordby]

No plans for full f64, some plans for position level f64 because it is indeed relatively cheap. I deliberately left the pose structure open to it. The only tricky part is realistically the broad phase. Inflating the entire tree to f64 would be a bit ouchy, but required for truly f64-scale worlds. A hybrid midpoint would be to estimate maximum error and expand bounds by f32, but that only goes so far. Might prototype some things at some point.

It is on the roadmap, though.