Consider extracting the first limb into a separate field

[konsumlamm]

Currently, BigInt is defined as follows:

type
  BigInt* = object
    limbs: seq[uint32]
    isNegative: bool

It has the invariant that limbs.len >= 1.

It might be worth it changing that to

type
  BigInt* = object
    limbs: seq[uint32]
    first: uint32
    isNegative: bool

This has several advantages:

• The invariant is now always satisfied (the first limb always exists).
• Small BigInts require no allocation & pointer indirection.
• It would avoid special casing 0 (see https://github.com/nim-lang/bigints/issues/26).

Disadvantages:

• A bit more special handling is needed.
• The first limb not being next to the other limbs in memory may make a difference (but I doubt it would be noticable).

[pietroppeter]

Another option would be to consider Zero the one with empty limbs (which is what https://github.com/SciNim/megalo does)

[narimiran]

I did try some "small BigInt optimization", but when I benchmarked it, I haven't noticed any measurable improvements. But there's a possibility that my tests weren't great and/or my implementation wasn't good enough.