On Mon, 2009-01-12 at 10:25 +1100, Paul Wankadia wrote: > On Mon, Jan 12, 2009 at 3:42 AM, Ian Kent <ra...@themaw.net> wrote: > > > It's just that the additive hash function and the tiny, > non-prime hash > > table size offended my delicate sensibilities. ;) > > Oh .. right, looks like v4 uses a prime to me but I did mean > to use a > prime in v5, oops! > > $ factor 27 77 > 27: 3 3 3 > 77: 7 11 > $
Yeah, yeah, I realized how stupid I was just after I made the post, double oops! > > > > Actually, we discovered that reloading a large file map took > well over > > ten seconds. Profiling revealed that the daemon spent most > of its time > > calling strcmp(3). > > > > Oh, and I should note that the problem was observed in > autofs v4. I'll > > be looking more closely at autofs v5 in the near future. > > It's much the same, the hash size is not much bigger and it > can well do > with a healthy increase. I don't know what sort of > distribution the > simple additive hash gives and if we are going to change it > then we > should do some checking with representative strings to ensure > that what > we use is in fact better. > > How could it be worse? :P > > I'll check our large file map just to satisfy my curiosity. > _______________________________________________ autofs mailing list autofs@linux.kernel.org http://linux.kernel.org/mailman/listinfo/autofs
