Re: interesting test of hash-table vs a-list

From: Harvey J. Stein <hjstein_at_bfr.co.il>
Date: 03 May 1999 15:37:37 +0300

Lars Thomas Hansen <lth_at_ccs.neu.edu> writes:

> The best compromise between performance and the distribution given by
> hashing on the characters of the print name is probably obtained by
> computing the hash value when the symbol is created and storing it in
> the symbol for future use. Both MacScheme and Larceny use this trick.
> The code for symbol-hash is then simple: a type test and a lookup.
>
> In Larceny, the break-even point of hash tables vs. alists is at about
> 25 elements, according to Will Clinger.

What about scrambling the bits of the address, or using the address
modulo hash table size with the hash table size being a prime number?

-- 
Harvey J. Stein
BFM Financial Research
hjstein_at_bfr.co.il
Received on Mon May 03 1999 - 15:40:29 CEST

This archive was generated by hypermail 2.3.0 : Mon Jul 21 2014 - 19:38:59 CEST