Re: Switch statement code generation

Ray" <bear@sonic.net> wrote in message news:09-11-031...
> nathan.mccauley@gmail.com wrote:
>
> > On Nov 5, 12:12 pm, Walter Banks <wal...@bytecraft.com> wrote:
> >> Joshua Cranmer wrote:
> >> > I'm trying to put together a list of various methods to do code
> >> > generate for switch statements. ...
> >
> > Has anyone seen tries used for the string-based switch statements?
>
> I think the winner for string switches is a string representation with
> a precomputed hash code, used in combination with a hashtable of
> jump addresses.


I'm surprised nobody mentioned perfect hashing. This is ideal
when the set of values is known in advance. No rehashing.


(I'll put this in my ToTry list for our parlanse compiler :)


-- IDB
[I gather that finding a perfect hash function that runs quickly is
not always easy, and that a slightly imperfect hash, e.g., into a
hash table with no collisions but a few empty slots can often be
a lot faster. -John]

Return to the comp.compilers page.
Search the comp.compilers archives again.