To write efficient df1b2variable code for moderate to large size
matrices M with
y=solve(M,x)
one needs to write the df1b2variable equivalent of adjoint code for the
LU decomposiiton. This is not too difficult but not the most
interesting exercise in the world.