dilmuff

For distributed scenarios where the architecture, language and library are unspecified and effectively random, it is not. Not so long ago x87 80-bit precision was common while non x87 systems typically used single and double precision. Finally, IEEE-754 does not specify transcendental functions, so once you use those your results vary all over the place as approximations vary. The fact that Go does not use the the CRT for FP will demonstrate this in the real world.

This is just aggressive loop unrolling. Duff's device is specifically the syntactic curiosity in C/C++ that allows implementing unrolls with a switch and inner while statement.