lowlatency kernel status?

[David Henningsson]

On 2011-01-27 17:13, Scott Lavender wrote:
>
>
> On Wed, Jan 26, 2011 at 9:25 AM, David Henningsson
> <david.henningsson at canonical.com
> <mailto:david.henningsson at canonical.com>> wrote:
>
>     What is really the current status of the lowlatency kernel? Are we
>     planning to have it in universe for Natty? If so, is any work going
>     on in that area?
>
>
> Alessio answered most of these questions, however I would like to add a
> few more comments.
>
> Alessio, would you please upload your -lowlatency kernel to the
> repository since there will be a delay in the documentation?  I believe
> it is paramount to have the -lowlatency kernel in place for user
> experience and functionality.
>
> I also would like to note that a few items about Alessio's kernel that
> have been explained to me:
>   * it is based on the -generic Ubuntu kernel
>   * it does NOT have an invasive patch as others have declared or believed
>   * changes in build time flags ONLY are required to build Alessio's kernel
>
> I apologize if I am being pedantic but sometimes I feel like this is an
> continually uphill battle, but it is one that I am committed.
>
>
>     I talked to Andy Whitcroft yesterday about providing "derivative
>     flavour" documentation, but he is currently busy with other tasks so
>     this documentation will have to wait (no prognosis given). So I
>     assume that waiting on that documentation is not an option.
>
>
> Again, I want to reiterate my desire to get the -lowlatency kernel into
> the repositories for a Natty release.  Given that it is based on the
> -generic Ubuntu kernel and ONLY requires compile flags I would like to
> believe that this is possible.
>
>     Also, do we have any signs of this kernel performing better than the
>     generic one? (As for bug 690010, that's fixed upstream now, and will
>     be fixed in the generic kernel on the next Natty kernel upload.)
>
>
> My experience is that the -lowlatency performs in a far superior manner
> than the -generic one for audio work.
>
> Using my Dell P4, 2.8ghz machine with 3 gigs memory and an MAudio Delta
> 44 card I test the -generic kernel using JACK and Ardour and made test
> recordings of guitar.  I found that the -generic kernel provided stable
> performance (i.e. no xruns) at slightly over 22msecs.
>
> Using the same machine I installed Alessio's -lowlatency kernel and
> testing it in the same manner.  I found that the -lowlatency kernel
> provided stable performance (i.e. no xruns) at just under 3msecs.
>
> I hope all find this information useful.

Thanks, it's the first time I've actually seen some numbers, that's 
great and could point as confirmation that there is actually a need for 
this kernel.

Would you (and others) mind redoing the test with the 2.6.38-1-generic 
kernel? It is not yet installed by default, but you can test it with

sudo apt-get install linux-image-2.6.38-1-generic

(then reboot and choose the right kernel in grub.)

And to be fair, I guess it should be tested against a 2.6.38 version of 
the lowlatency kernel as well.

I have confirmed that bug #690010 is fixed in this kernel.

-- 
David Henningsson, Canonical Ltd.
http://launchpad.net/~diwic