I noticed a couple of minor issues in mpn_divexact_1 (dive1.c) that are
not in my own implementation, so I thought I'd pass them on.
If the quotient will be one word shorter than the dividend then set
the top word to zero:
if ((divisor 1) == 0)
{
if ((dst != src) (src[size - 1]
Mark Sofroniou ma...@wolfram.com writes:
If the quotient will be one word shorter than the dividend then set
the top word to zero:
if ((divisor 1) == 0)
{
if ((dst != src) (src[size - 1] divisor)) dst[size - 1] = 0;
...
I don't understand. Doesn't the current loop always
I agree with Niels (don't understand + not appropriate for
low-level...).
We should replace mpn_divexact_1 with code that:
(1) Uses Jebelean's trick with a Euclidean division working left-to-
right and a simultaneous Hensel division working right-to-left.
This is faster in the
Torbjörn reminded me of the work on more clever div_qr_1 we discussed a
few years ago, which seemed promising but never got completed.
At the time, I think one reason it stalled was the somewhat unwieldy
divrem_1 primitive. The plan (see https://gmplib.org/devel/) is to
replace divrem_1 by a
