[jefftk]

> Float32 should only be used when Float64 computation is too expensive

Or when you're bottlenecked on memory and want to store each number in four bytes instead of eight.