Newton-Raphson algorithm

[Earl Killian]

I am looking for references to Newton-Raphson algorithms for divide
and square root.  I know that it is difficult (but possible, as TI/Sun
proved) to get a correctly rounded result when you round at each
intermediate step, but I've heard that if you have a double-width
datapath available (e.g. on a machine with a fused multiply/add
instruction like the RS6000), that it is fairly easy.  If you can
provide any references for the latter approach, I would appreciate it.