Article: Increasing the D Compiler Speed by Over 75%

bearophile bearophileHUGS at lycos.com
Thu Jul 25 13:00:21 PDT 2013


Walter Bright:

> It's not the hashing that's slow. It's the lookup that is.

By "different hashing scheme" I meant different strategies in 
resolving hash collisions, likes double hashing, internal 
hashing, cuckoo hashing, and so on and on. Maybe one of such 
alternative strategies is more fit for the needs of dmd 
compilation. (I think that currently the Python dicts are using a 
hashing strategy different from the built-in dictionaries of D. 
The Python style of hashing was implemented in D some months ago, 
but I don't remember what happened to that project later).

Bye,
bearophile


More information about the Digitalmars-d-announce mailing list