It’s curious to see the rdiv instruction specified with only a vague idea about it’s specification. May be it is just the public information that is unclear about the specifics, but from Ivan’s comments it looks like this really is uncharted territory for the Mill team.
On the wiki, at LarryP’s page on division I posted a small comment about an alternative algorithm, Goldschmidt division. It could very well be that when that is used as a software div implementation, the instruction providing the initial guess should be another function or could be omitted altogether.
Anyway, it’s probably time for a separate topic on integer division.