It calculates it by doing... trigonometric operations! The remainder is implementation specific, depending on the language (or software), etc. It might use polynomials, delegating all floating point calculations to the FPU.

I guess the calc use some formula (same cos a = a/h)

And i about this:

1. cos(x)=1-x^2/2!+x^4/4!-x^6/6!+..+(-1)^n x^(2n)/(2n)!+..