[linux-audio-dev] Fw: Re: Low Latency patches and kernel 2.6.x

Florian Schmidt mista.tapas at gmx.net
Mon Dec 22 21:42:49 UTC 2003


Oops i messed up the quoting a little bit. Here's Morton's answer:

Begin forwarded message:

Date: Mon, 22 Dec 2003 11:58:54 -0800
From: Andrew Morton <akpm at osdl.org>
To: Florian Schmidt <mista.tapas at gmx.net>
Cc: linux-audio-dev at music.columbia.edu,
linux-audio-user at music.columbia.edu
Subject: Re: Low Latency patches and kernel 2.6.x


Florian Schmidt <mista.tapas at gmx.net> wrote:
>
> 
> Hi Andrew,
> 
> i'm a user of a 2.4.22 kernel patched with the preemption patches plus
> your Low Latency patches. It works very nicely for audio applications
> [like jackd, ardour, etc...].

Cool.

> Looking into the config menu of the 2.6.0 kernel i only find a
> "preemptible kernel" config option. Your low latency patches seem not
> to be included. Do you have plans of including your patches in the
> 2.6.x kernel? I heard rumors about merging the preemptible and your LL
> patches since they seem to go very nicely together.. Any truth to
> that?

The objective in 2.6 is that the preemptible kernel achieve similar
worst-case latencies to the low-latency-patched kernel.  So 2.6 should
meet
your requirements out of the box.


That being said, last time I instrumented the 2.6 kernel it was not
achieving the targets.  The specific failure was occurring when the
machine
had a very large number of inodes in cache and the VM system was
reclaiming
those inodes.

It is unlikely that you will strike this problem in real-world usage, so
2.6 should work fine for you.  As ever, testing results would be
appreciated.

(The inode reclaim problem is fairly complex, but I just happen to see a
patch from Dipankar Sarma in my inbox this morning which is designed to
fix
it up).

End forwarded message

-- 
music: http://www.soundclick.com/bands/9/florianschmidt.htm





More information about the Linux-audio-dev mailing list