You may be aware of this folkloric shared divisor division trick. This trick generalizes to any Abelian group and is useful in any case where finding the inverse of a group element is significantly more expensive than the group operation. The reasoning is identical to the multiplication/division case, just generalized.

\[ab^{-1} = ad(bd)^{-1}\] \[cd^{-1} = cb(bd)^{-1}\]

We can transform the computation from a form requiring two multiplications and two inversions, to a form requiring five multiplications and one inversion.

We can similarly generalize the trick for optimizing three or more such products:

\[ab^{-1} = adf(bdf)^{-1}\] \[cd^{-1} = cbf(bdf)^{-1}\] \[ef^{-1} = ebd(bdf)^{-1}\]

This trades three multiplications and three inversions for eleven multiplications and one inversion. As we increase the number of products to compute, we can further optimize by sharing partial products. For example with four products:

\[ab^{-1} = ad(fh)(bdfh)^{-1}\] \[cd^{-1} = cb(fh)(bdfh)^{-1}\] \[ef^{-1} = e(bd)h(bdfh)^{-1}\] \[gh^{-1} = g(bd)f(bdfh)^{-1}\]

We can compute this with 15 multiplications and a single inversion, instead of the naive route with four multiplications and four inversions.