Subject: Re: [linux-audio-dev] request to use latencytest (OSDL)
From: Roger Larsson (roger.larsson_AT_norran.net)
Date: Tue Jul 09 2002 - 11:29:09 EEST
On Tuesday 09 July 2002 04.10, cliffw_AT_easystreet.com wrote:
> I've been a grateful user of latencytest for audio setup.
> I've just started working for the Open Source Development Lab
> (www.osdl.org) which is a non-profit supporting kernel development
> by providing hardware and a Scalable Test Platform (www.osdl.org/stp)
>
> I'm working on adding tests to the STP, and latencytest looks like
> it would be very useful, since we don't have many other scheduler focused
> tests (yet).
You might make it even more useful by using it twice...
One time with RT prio - testing latency for RT processes
And once without - testing latency for ordinary processes
Make sure to minimize the idle burning of CPU in both cases - it is not
interesting and make the concurrently run testcase look bad...
>
> I'd like to include it in the platform. At a minimum this would mean putting
a
> copy
> in our CVS ( cvs.stp.sourceforge.net:/cvsroot/stp )
> and tweaking up a set of automated scripts for the test engine.
>
> I'm not sure of all the etiquette when borrowing/seizing GPL code
> for another GPL project, so i would appreciate help/advice.
Asking is nice but not strictly necessary...
>
> I've written Takashi and benno, but no response yet. I would
> have to receive the blessing and permission of the authors before doing
> anything with the code - are there any other authors, or concerned citizens?
>
My guess is that they will be happy to see it in more regular use...
An alternative would be Andrew Mortons amlat - it is simpler, possibly better
in an automated environment.
> Also, more general:
> Is this code really useful from the kernel perspective?
> In other words, is there a reason _not to use this test?
Latency monitoring is good to do during all the other test cases...
since a change that gets you better throughput might destroy interactive
behavior...
But those tests are probably better done by adding some code in kernel.
(The simplest way that I have used is to start a timer when the need_reshed
flag is set, and stop it when schedule is called...
/RogerL
-- Roger Larsson Skellefteċ Sweden
This archive was generated by hypermail 2b28 : Tue Jul 09 2002 - 11:38:15 EEST