Subject: Re: Preparing callout(9) for a HZ-less kernel
To: None <email@example.com>
From: Joerg Sonnenberger <firstname.lastname@example.org>
Date: 10/18/2007 23:19:41
On Thu, Oct 18, 2007 at 01:54:33PM -0700, Jason Thorpe wrote:
> Are you planning on using this along with something like Soft Timers?
You mean: Soft Timers: Efficient Microsecond Software Timer Support for Network
I'm not sure yet. I have an idea of ensuring a minimum time when
reprogramming the timer to ensure that a certain rate threshold.
A check for pending interrupt timers is relatively cheap and the idea of
soft timers is to do that in a number of places. That is useful if you
have a high number of timers, which we don't have yet. It is also a
question of whether we want to separate the two classes into different
trees. But I think it is more important and helpful to have per-cpu
timers and to split the network timers apart. When that is done we can
evaluate the performance impact and look for optimising it.