Correctness guarantees of erfinv

[Damien Desfontaines]

Hi folks,

Thanks a lot for your work on mpmath!

I had a question about the erfinv function: is it covered by the correctness guarantees that mpmath provides for "typical" transcendental functions? Meaning, if I pass an exact input to this function with mpmath's default parameters, and then convert the result to a 64-bit float, is this going to be rounded to the best possible value?

Thanks a lot in advance! =)

Damien