Re: Suggestion For mpz_sizeinbase() Function

[Kevin Ryde]

"David T. Ashley" <address@hidden> writes:
> It depends on the assumptions about the platforms.  Can they all divide
> integers?  And the least capable platform--what sizes of integers can it
> divide?

Alas division as such is usually slow, and sometimes doesn't exist at
all natively (alpha, itanium).

A multiply-by-inverse is often best if an inverse can be
pre-calculated (as in this case), or will be used a few times.  It
ends up roughly equivalent to 2 muls, and finds good use in
mpn_divrem_1 in particular.  See "Division by Invariant Integers"
paper in the references in the manual, and udiv_qrnnd_preinv in
gmp-impl.h.