Not just more natural, it also lends itself to better accuracy and/or efficiency in some cases too.
As a quick example, if you're writing a numerical approximation or even just a LUT for sin(x), you can get away with only approximating/storing values for a reduced range by using some bit twiddling. The top bit becomes sign of output if you XOR it out first, the next bit lets you take advantage of sin(pi/2-x) = sin(x+pi/2) to half the input range again, eek out symmetry for the next bit and you can nearly approximate sin(x) as a linear function for some use cases.
As a quick example, if you're writing a numerical approximation or even just a LUT for sin(x), you can get away with only approximating/storing values for a reduced range by using some bit twiddling. The top bit becomes sign of output if you XOR it out first, the next bit lets you take advantage of sin(pi/2-x) = sin(x+pi/2) to half the input range again, eek out symmetry for the next bit and you can nearly approximate sin(x) as a linear function for some use cases.