On Mon, Mar 25, 2002 at 11:02:31AM -0500, Ben Collins wrote: > On Mon, Mar 25, 2002 at 03:51:24PM +0100, Wichert Akkerman wrote: > > Funny, Jason Gunthorpe said he made the exact same benchmark a while > > ago and claimed it did not improve things at all. Looks like we do > > have to look into the hashing. > > I wonder if Colin took into account that the data was cached on the > second run? An average of 3-5 runs for each would give a better result.
I did run the simple test a few times, and there didn't appear to be a significant difference. findpackage() is called quite a lot even in just a single --status query, but admittedly I haven't tried to profile full dpkg runs. Making the hash larger than the number of items it's going to have to contain does seem like an "obviously correct" optimization though. -- Colin Watson [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
