Trailing zero factors in rational functions

[ghalimi]

What is the rationale for including trailing zero factors in rational functions such as s2, which is used to implement erfinv? In this function, the last three factors are equal to zero. Is it to facilitate code reading (at the expense of a minuscule degradation in performance)? Is it because the code was generated from metadata and no routine was implemented to get rid of these trailing zero factors? Or is there another reason for this?

[kgryte]

The coefficients originate from a script. Those coefficients are subject to change. Modern JITs can leverage constant folding to avoid unnecessary computation.

[ghalimi]

That's what I was expecting. Thanks!