Re: [xwax-devel] kernel for Audio 4 DJ?

[Mark H. <ma...@po...>]

On Sat, 19 Dec 2009, Martin Horn wrote:

> Hi Mark,
> thanks for your answer, it took me a while to reply...
> 
> On Fri, Dec 18, 2009 at 12:04 AM, Mark Hills <ma...@po...> wrote:
> 
> > On Thu, 17 Dec 2009, Martin Horn wrote:
> >
> > > Hi Mark, hi All,
> > >
> > > finally I decided to buy the Audio 4 DJ and now that I got it I am to
> > > recognize that the 2.6.32 kernel which is higly recommended by Mark has
> > > no rt-patch yet.
> > >
> > > What kernels are you using and what are the concrete drawbacks of 2.6.31
> > > compared to .32? I read the git commits but do not understand much of
> > > that.
> >
> > The concrete drawbacks of v2.6.31 are described in the commit messages
> > that apply to both Audio 8 DJ and Audio 4 DJ:
> >
> >  http://www.pogo.org.uk/~mark/linuxdj/#audio8dj<http://www.pogo.org.uk/%7Emark/linuxdj/#audio8dj>
> >
> > Hopefully you can understand "Fix a bug which can result in white noise
> > from the driver after stream start or unpause." :-)
> >
> 
> Shure, but I was not shure what "stream start or unpause" means, I guess the
> raw audio stream is meant, right?! Is this error only to occur when starting
> xwax or everytime a track is loaded/ playback is started?

Yes, it's the raw output audio stream. So when you start xwax, or if you 
get an xrun. You wouldn't want a single xrun to result in white noise.

We could be splitting hairs and wasting time here. I wouldn't want to use 
A4DJ with v2.6.31 for live performance, but if you're just messing with 
xwax then you may just go with that version for now.
 
> > I'm not especially familiar with the RT patches, but perhaps you should
> > look at whether you actually need them. I find that the 2ms buffer size
> > which I can use on a regular kernel with the A4DJ is ample for vinyl
> > control.
> >
> 
> What distro and hardware are you using for this and can you tell me wether
> the below mentioned CONFIG_PREEMPT_VOLUNTARY option applies for you?

I have an Arch Linux laptop with the stock kernel which works well:

$ zcat /proc/config.gz | grep PREEMPT
# CONFIG_TREE_PREEMPT_RCU is not set
CONFIG_PREEMPT_NOTIFIERS=y
# CONFIG_PREEMPT_NONE is not set
# CONFIG_PREEMPT_VOLUNTARY is not set
CONFIG_PREEMPT=y
# CONFIG_DEBUG_PREEMPT is not set

CONFIG_HZ may also be another relevant option, I am not sure though.

> > I did some reading and found on the RT wiki:
> >
> >
> > http://rt.wiki.kernel.org/index.php/Frequently_Asked_Questions#What_are_real-time_capabilities_of_the_stock_2.6_linux_kernel.3F
> >
> >  "The 2.6 Linux kernel has an additional configuration option,
> >  CONFIG_PREEMPT, [...] With this option, worst case latency drops to
> >  (around) single digit milliseconds, [...] If a real-time Linux
> >  application requires latencies smaller than single-digit milliseconds,
> >  use of the CONFIG_PREEMPT_RT patch is highly recommended."
> >
> 
> It is said before your quote that the generic kernel can produce some
> hundreds of ms to seconds latency, but a  CONFIG_PREEMPT_VOLUNTARY option
> can be used which decreases these to single-digit ms. But as far as I
> understand CONFIG_PREEMPT_VOLUNTARY is not included in the generic kernel,
> right?

I think you misread, thi sis actually the other way around. CONFIG_PREEMPT 
goes further than CONFIG_PREEMPT_VOLUNTARY in reducing latency.
 
> Whatever I am downloading the latest ubuntu alpha which has 2.6.32 to test
> it, any SUGGESTIONS for other distros who are already on kernel 2.6.32 are
> welcome!
> 
> Martin
[...]

-- 
Mark