Dean Rasheed <dean.a.rasheed@gmail.com> writes:
> I think the hash map lookups could be made pretty quick, if we're
> prepared to write a bit of dedicated code there rather than relying on
> the more general-purpose caching code.
It's still going to be very significantly slower than what I'm
suggesting --- I'm *already* assuming that step 4 is using a hash
or something else smarter than just traversing a list. My step 3
is just a bit-array index operation (well, 2 of 'em).
(I'm pretty dubious of unsubstantiated claims that you can build a hash
table that's significantly faster than our existing hash code, anyway.)
regards, tom lane