Subject: Re: [linux-audio-dev] [ANN] A novel approach to real-time free software
From: David Olofson (david_AT_gardena.net)
Date: Mon Jun 10 2002 - 03:32:48 EEST
On Saturday 08 June 2002 09.28, Vincent Touquet wrote:
> On Sat, Jun 08, 2002 at 01:11:10AM +0200, David Olofson wrote:
> (cut)
>
> >>From the application POV, both approaches give the same result:
> >> Two
> >
> >"worlds" with different schedulers and services - and
> > unfortunately, different drivers. That is, if you want real time
> > I/O, you still have to port Linux drivers to RTAI/ADEOS, just
> > like you have to do with the current RTAI, or RTL.
>
> True, but at least it will be real realtime,
> something that cannot be achieved using any
> lowlatency patch.
Not quite true. The definition "hard real time" is not to be
considered absolute, as long as you're dealing with the real world -
and there are no implicit figures either.
Hard Real Time just means "never miss a deadline", and that's all
there is to it.
The reliability for any worst case latency requirement is a property
of the hardware and the OS, and as long as it meets the requirements
for your application, you're fine.
For professional audio, a digital latency of 2-3 ms seems to be
sufficient, and as to reliability; as long as it's more reliable than
a properly maintained analogue multitrack, we're fine.
> Its great that we will be able to run
> Linux and RTAI side to side :)
Uhm, what difference would that make to applications? Currently,
Linux runs under RTAI instead, but the net result is the same: You
get Linux and you get RTAI, and you can have applications running
under both "at the same time".
> But, you're right, the road is long.
> At least there is a road ;)
Sure, but there has been one for several years. I started playing
with RTLinux (which has the same issues as RTAI or RTAI/ADEOS) before
there was a lowlatency patch. I'm not going down that route again,
unless I *really* have to. (*Maybe* some mmap() based hybrid, if it
works well, but no porting ALSA to RTAI...)
I think future Linux kernels (perhaps with a "real time" config
option) will be quite capable of guaranteeing <.5 ms latencies with
sufficiently high reliability for any applications where you would
dare using magnetic tape or non-mirrored hard drives.
And besides, not even RTAI or RTLinux can guarantee any minimum
latency on any hardware. There are single board computers with
chipsets that cause latencies in the ms range regardless of OS. In
some cases, it's almost as hard to track down as latency sources in a
Linux/lowlatency kernel - mostly because you can't directly see
what's causing it.
If you wan't *real* guarantees and µs range latencies, build your own
hardware, using only chips you know inside out, and don't use an
operating system at all. (Considering that any audio ADC or DAC needs
a few samples for proper filtering, there are *very* few real
applications for that kind of real time systems in audio.)
//David
.- M A I A -------------------------------------------------.
| Multimedia Application Integration Architecture |
| A Free/Open Source Plugin API for Professional Multimedia |
`---------------------------> http://www.linuxdj.com/maia/ -'
.- David Olofson -------------------------------------------.
| Audio Hacker - Open Source Advocate - Singer - Songwriter |
`-------------------------------------> http://olofson.net -'
This archive was generated by hypermail 2b28 : Tue Jun 11 2002 - 04:03:02 EEST