bigfloat

bigfloat copied to clipboard

Panic in Pow for negative base with integer exponent

1

Hi, Thanks for this great little package! The `bigfloat.Pow()` function seems a little too restrictive relative to the `math.Pow()` standard library equivalent for negative bases with integer exponents (which are...

vsivsi

Pow(Inf, -1) == 0 not Inf. True for all negative exponents

2

Hi again, this one is more of an edge case... But for consistency with `math.Pow()` Also, for `-Inf`: `Pow(-Inf, -1) == -0`, but this also doesn't work because of #35...

vsivsi

ALTree

ALTree

e.g. in `sqrtInverse`: introduce temp and test perf.

ALTree

If the requested prec is much smaller than the current size of the pi cache, it could be cheaper to just compute pi from scratch.

ALTree

Sometimes we set `prec := z.Prec()` and then use `prec + 64`, in other places we set `prec := z.Prec() + 64`. Uniform.

ALTree

i.e. remove the last step if `prec` is much bigger then `dPrec` and just do a final one with `prec + guard` precision.

ALTree

Allow only the last bit to be off(?)

ALTree

I.E. what happens if in `Exp(z)`, `z.Float64()` returns `+Inf`?

ALTree