udiv_qr_3by2 vs divappr

[Torbjörn Granlund]

paul zimmermann <Paul.Zimmermann at inria.fr> writes:

  this looks quite interesting, I'm looking forward for cycle numbers.
  (I guess you target the case where the divisor is not invariant, otherwise
  you might do some preprocessing to speed up the partial quotient
  approximation.)

The divisor is invariant in some sense when the quotient is more than
one limb, right?

Are you thinking of Svoboda division here?  That's another possibility
for speeding up schoolbook division.

-- 
Torbjörn
Please encrypt, key id 0xC8601622