lock breaking preemptible kernel

Felipe Contreras al593181 at mail.mty.itesm.mx
Wed Dec 5 18:23:01 PST 2001


On Wed, Dec 05, 2001 at 09:20:49PM +0100, Matthias Benkmann wrote:
> On 5 Dec 2001, at 10:22, Ian Molton wrote:
> 
> > 
> > Actually, thats just not true - compiling kernels on a loaded box is often
> > quite a bit quicker with the pre-empt patch.
> > 
> > what it does do (other than improve latency) is keep one task from hogging
> > the CPU if its stuck in the kernel for a long time (perhaps in a new /
> > badly written / inefficient driver).
> 
> You're actually telling me there are drivers that do busy waiting in the 
> Linux kernel?  Because that is the only situation where the latency patch 
> would help batch jobs, because it would cause CPU time to be used that is 
> otherwise just wasted. I had assumed that any kind of i/o would put the 
> process into the blocked state and give the CPU to another process till 
> i/o is finished. I'm sure this is the case for disk i/o. So if you don't 
> use any exotic hardware in background jobs and just play an mp3 file and 
> in parallel compile a kernel, I'm sure the kernel will compile slower and 
> the mp3 will play smoother if the preempt patch is used. And in fact from 
> what I gather this behaviour is what gets people so hyped about the 
> preemption patches. No skipping for the mp3 is more important than a few 
> seconds additional compile time. After all, if you actually wanted to get 
> work done, you wouldn't be listening to mp3s would you ;-)

Well, IMHO it's very hard to predict how a kernel will work on what
circumstances. There are people working for the kernel preemption and
some other to reduce the lockups. Remember that lockups can occur
anytime, it can happend in the middle of compilation, IIRC the virtual
terminal code does bad lockups and so if you change terminals in the
middle of a compilation you'll get a small time lost. If all the lockups
are optimized then the kernel will be more responsive, withouth the
preepmt patch, and I can't be sure about combinations of the patches or
kernels, it's complex.

What I know is that the 2.4.17-pre2 kernel rules, simply it's great, and
withouth any patch. I've not tryied the patches, but I will, once I have
run this one a while, so I can compare.

Felipe Contreras
-- 
Unsubscribe: send email to listar at linuxfromscratch.org
and put 'unsubscribe blfs-support' in the subject header of the message



More information about the blfs-support mailing list