You can test it against the results of BigFloats:
d(f, x, y) = f(x, y) - f(big(x), big(y))
I got deviations in the order of 10^-17 for both methods and Float64.
x*1/sqrt(y) is by the way identical to x/sqrt(y).
You can test it against the results of BigFloats:
d(f, x, y) = f(x, y) - f(big(x), big(y))
I got deviations in the order of 10^-17 for both methods and Float64.
x*1/sqrt(y) is by the way identical to x/sqrt(y).