Then why didn't you follow those exactly?

The XInt library was designed before I knew about them. I adapted it to them later.

There are only a few notable places where XInt differs from them:

• It does not provide separate mathematical primitive functions for long long, as suggested in n1744. This is partly because long long is not yet in the C++ standard, but mostly because I don't see a need for it. The conversion constructors are efficient enough for smaller values that there wouldn't be any noticeable benefit from it.
• It does not, at present, implement a multiplication scheme with better performance than the "basecase" mentioned in section 3.4 of n1692, primarily because I haven't had the time to research all of them yet. I suspect that most people using the library will usually be using numbers below the threshold where the more exotic algorithms are faster, but I plan to add at least one of them in the future, maybe several.