Phil R. Karn (firstname.lastname@example.org)
Tue, 18 Aug 87 23:04:14 edt
I thought this probably wasn't original, but I'm glad to see somebody
else considers it worthwhile.
I too was concerned about the speed of the algorithm. I originally
implemented it in a totally brute-force fashion. A "for" loop searches
the hash table 32 times starting with the widest possible mask and
working down until it hits a match. I assumed I would have to place a
cache on top of this to get decent performance. To my delight, however,
I discovered that going all the way down to the default entry still took
only 6 milliseconds -- on a Taiwanese PC/XT turbo clone! Given the
channel speeds involved in the environment I wrote my software for, I
didn't consider further optimization to be a high priority task.
Sometimes brute force is good enough...
This archive was generated by hypermail 2.0b3 on Thu Mar 09 2000 - 14:39:14 GMT