Re: KAIO vps, and bufwaits

Juan R. Guzman wrote: > > Any special reason your suggest prime? Just to best guarantee that you will not hit another nodal value in the interaction function between the number of LRUS and the LRU rehash algorithm. I figure a prime should be safe even if the problem is that some bozo (I will not insult a fine clown by capitalizing bozo) in R&D used a non-prime for the rehash constant. Not that I think the bug is cause by anything as simple and obvious as a non-prime hash constant (I've never actually had reason to think of anyone in Informix R&D as a bozo) but I like to eliminate the simple and obvious first. The technique has saved me many times from banging my head against the wall. > Art S. Kagel wrote in message <6bq77p$k3r$1@news.xmission.com>... > > > >On Tue, 10 Feb 1998 davin_harvey@dsti.co.uk wrote: > > > >> Art > >> > >> Just a small question. Is there any reason for keeping the number > of > >> LRU queues to a power of two? Everybody seems to use 8,16,32, etc. > >> Just habit, or what? > > > >Naw, just dumb programmer habits. I even suggested in the original > >posting that if someone wanted to try numbers other than 32 and 128 to > >avoid the bug I mentioned that a prime number would probably be safe. > > > >Art S. Kagel, kagel@bloomberg.com > >