Triple Floating Point Precision

Is it possible to create a triple floating point transform system? And if you can how hard would it be to do.

Generally speaking, “double precision” is as good as it gets because that’s what the programming languages at your disposal support natively. You could create your own type, but it would take some work and using it wouldn’t be as fast as native types. Also, you can’t redefine Unity’s built-in data structures, and that means object transforms, among other things.

“Double precision” actually means the binary representation of a number is twice as long (double numbers have 64 bits, floats have 32), effectively improving precision by multiple orders of magnitude, and not just by an order of two. You don’t really need more precision than that, unless you’re doing some extremely sophisticated calculations. You can also use decimal type if you need to perform accurate calculations in base 10 specifically (accounting with real money involved is the most common example).

The best way to keep numerical errors at bay is to be smart about how you perform your calculations. Managing numerical stability is a vast topic, though.