On Thu, 25 Jun 1998 23:24:08 +0200, Erik Corry <erik@arbat.com> said:
> In article <199806251132.MAA00848@dax.dcs.ed.ac.uk> you wrote:
>> Hi,
>> Ingo has had a swap prediction algorithm for a while, and we talked
>> at Usenix about a number of other things we can do to tune vm
>> performance automatically. 2.3 ought to be a great deal better.
>> madvise() may still have merit,
> Definitely does. I have an app (at work) with a very large
> tiled image file. The access pattern is something that the
> OS is never going to guess, but with madvise on Solaris it
> really flies. Performance without madvise would be (is!)
> disastrous. There are simply cases where the app knows
> better.
Ingo's stuff, if applied that way, will learn the access pattern after
the first pass over the data set. Should still be pretty good!
What sort of madvise() patterns are you supplying to see the best
performance benefits, btw?
--Stephen
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu