MasterDuke17

Yeah, I didn't imagine it was a five-min fix. If we do end up using DwarFS in the near future we'd very likely do something pretty similar to what you...

See #328

Now that I’m thinking about it, a lot of functions have the name of the algorithm in them. Maybe make it internal and rename to s_mp_todecimal_barrett? Sent from my iPhone...

> These could either branch to a fast variant if available or branch to the generic toradix. What do you mean, "if available"? Should I change this PR to also...

@minad > I would prefer if we use a radix size function and don't grow the buffer. I can switch to assuming the passed in buffer is the correct size,...

A style/correctness question. Should I move the 3 arrays of `mp_int`s that are passed around outside the functions so they don't have to be passed in every call? So ```...

Timings with the new version that expects to be passed a pre-allocated buffer (i.e., similar API to `mp_toradix`). It got a bit faster, now the crossover point with `mp_toradix` is...

@minad #262 hasn't been merged. Are you saying wait until it has? Or rebase this branch onto `feature-detection2` and use its `MP_HAS`?

> > > Snprintf should not be used. > > > > > > Instead you should convert the number to a string "by hand". It does not make sense...

> > Snprintf should not be used. > > You can use `mp_to_radix` which also allows for bigger chunks at the tree-leaves to speed it up. At the point in...