2005-01-11 01:49:58

by walt

[permalink] [raw]
Subject: Preemptible Big Kernel Lock?

I've been running the P-BKL all day and I've been very
happy with it so far.

I've found the machine to be very responsive and smooth
during compilation of large programs, playback of audio/
video, all the usual stuff I do on my desktop machine.

The problem is: the same machine was also great two
days ago, before the preemptible BKL -- the best it's
been for years. I can't tell the difference, and I
had no complaints even before this update.

Two questions:

Would I expect to see a difference on a uni-processor
machine? (That's all I have.)

What kind of testing could I do to demonstrate the
difference?


2005-01-11 07:47:01

by Arjan van de Ven

[permalink] [raw]
Subject: Re: Preemptible Big Kernel Lock?

On Mon, 2005-01-10 at 17:48 -0800, walt wrote:

> Would I expect to see a difference on a uni-processor
> machine? (That's all I have.)
>

humans in general don't notice things < 1 milisecond. The preemptable
BKL and other latency fixes are (most) there to reduce the existing
typical latency that is < 1 msec to something even far lower. (and the
maximum which is a bit above 1 msec to below it.. but that you hardly
ever trigger unless you try).
So I doubt that you as human will perceive this.
Professional audio applications might notice though..,. ;)