Sam Emrick wrote: > > FWIW - I looked at this very briefly and concluded that most of the time > was actually spent in constructing the hash key, rather than in actual > lookup. The problem seemed to be frequent use of very long strings used for > hash key and terrible slow ns char string parsing routines. Yeah but what this optimization is reducing is a cid->factory lookup. Still the nsCStringKey creation exists. So either my timing was off or the cid hash table is really working poorly. The nsIDKey seems to be using the top 4 bytes bytes of the CID as the hash code. Do you have results of your analysis still. dp
