Float or Fix

DWM: Interesting, but important, detail: In general, for world modelling, do you use floating point or fixed point numbers (for position, velocity, acceleration, force, etc.)? Traditional wisdom has it that floating point is plagued by slowness and complex accuracy considerations. The first objection may no longer be true (and perhaps everyone but old geezers like me know it already), or it may depend on the nature of your dominant calculations. (I have a hard time believe that floating point addition & subtraction approaches that of integers even today.) The second may be manageable?

I have not found any concrete guidance on this on the net. A lot of the examples I'm seeing on the net seem to use floating point.

Much later... Yes, it's true. Almost everyone uses floating point nowadays for simulations. I can only assume that modern FPUs are fast enough to make this attractive. In any case, I'm hedging my bets by making sure that the base scalar data type used as a building for representing positional and kinetic quantities is easily changed in the source code. (Hmm, I've been using templates for this, but it occurs to me that I should probably be using a typedef instead. The small added complexity of templates isn't needed in application-specific code.) With careful coding, it should be possible to experiment with single vs. double fp, and maybe even try plugging in a multi-word or infinite-precision integral type. (There are some C++ libraries around that define such, or else it's pretty easy to roll your own.)

Much later yet... Yes, it's very true. All the graphics hardware works with floats. The era of integer arithmetic for 3D graphics is long past, and the model software has followed suit.