Subject: Re: [linux-audio-dev] an open letter to Linus re: low latency
From: Jun Sun (jsun_AT_mvista.com)
Date: Mon Jun 26 2000 - 20:24:53 EEST
Benno Senoner wrote:
>> some of the strategies are found here:
> http://www.mvista.com/realtime/rtsched_wp.html
>
> Jun, perhaps there are some other experts at your firm that could tell
> us about alternative solutions ? ( of course only those which Linus would like
> :-) )
>
> cheers,
> Benno.
The idea is pretty simple : the SMP kernel is preemptable and
reentrant. We can, at least in theory, transform the SMP version
preemptivity into a general UP kernel preemptivity.
This implies the worst-case preemption latency would be equal to the
longest spin lock holding time in the current SMP kernel, which
certainly is shorter than a nearly complete syscall in the standard
linux. (BTW, does anybody know what this number is in 2.2.x kernel? I
heard SGI has a tool to measure this.)
Due to the resource constraints we have not complete this work yet.
However, I did finish measuring the preemption latency in the standard
kernel and the kernel with Ingo's patch - using a more conservative
method (which I believe more accurate). The report should be out soon.
Here are a couple numbers on a Pentinum 266 machine : standard kernel
has 30-50 ms "typical" worst-case latency. Ingo's patch reduces it to
around 1ms.
Jun
This archive was generated by hypermail 2b28 : Mon Jun 26 2000 - 20:52:30 EEST