Linus Torvalds [off-list ref] writes:
In other words, the 256-entry fan-out was meant to avoid the first eight
levels of binary lookup. But the thing is, we should be able to generally
do a *lot* better than any binary lookup by just doing a LINEAR SEARCH,
which is also more cache-friendly, and prefetches much better when it's
not in the cache.
...
- use newton-raphson to iterate closer. It should be a much faster way to
find the rough area for the entry we're searching for than binary
search. Two or three iterations should get us there, easily.
This is another egg of columbus moment that makes me feel it was
worth to have been in git community ;-).
I like it.