Subject: Re: [linux-audio-dev] Hmmm, interesting statement on latency
From: David Olofson (david_AT_gardena.net)
Date: Thu Jul 26 2001 - 02:16:53 EEST
On Tuesday 24 July 2001 00:31, Benno Senoner wrote:
> On Monday 23 July 2001 19:13, Richard C. Burnett wrote:
> > I found this article on www.prorec.com on Windows 2000, thought
> > you all would find it interesting:
> >
> >
> > "in the average PC Windows 2000 is able to meet worst case high
> > priority task latencies close to 1 ms, while Windows 9X
> > does not even meet 100 ms. Worst-case high priority task latency
> > is the single most important measure of a systems ability to
> > perform low latency real-time processing of massive data
> > amounts."
>
> I want to see this in action first before believing to "press
> releases". I'll believe it when someone will produce
> "latencytest-like" diagrams on a heavily stressed box.
^^^^^^^^^^^^^^^^
There's the catch. Some major driver cleaning up *might* work, but
I'm having some trouble trusting a non-RTOS with closed source
drivers...
> Of course to achieve this kind of latencies you do need to write
> windows kernel modules with all the related disadvantages ..
(That's like using RTLinux for audio - only on Win2k you can use the
existing audio drivers, and on RTL you get a worst case latency of
some 3-15 µs on any decent machine. Yes, microseconds. *heh*)
> See gigasampler: it's a really nice sampler and squeezes out the
> most of win95/98 but I seem to recall that they are having big
> troubles to port it to win2k/xp because it is programmed too close
> to the hardware.
I'm sure they're having *lots* of fun... ;-> But that's life if you
program for a living.
> > That seems like a HELL of an improvement! I like the 'in the
> > average' part though. What latency are we targeting for Linux??
>
> RELIABLE 2-3msec (under Benno's ultra-heavy typical load) is
> possible on todays boxes.
...and as a result, I've dropped RTLinux for audio. I haven't seen a
valid technical reason to pick it up again so far. (Might be because
I haven't started using this stuff for serious work, though...! ;-)
Seriously, I've *never* seen a single drop-out while testing a
properly working lowlatency system, so I'm not expecting any problems
of that kind.
//David
.- M A I A -------------------------------------------------.
| Multimedia Application Integration Architecture |
| A Free/Open Source Plugin API for Professional Multimedia |
`----------------------> http://www.linuxaudiodev.com/maia -'
.- David Olofson -------------------------------------------.
| Audio Hacker - Open Source Advocate - Singer - Songwriter |
`--------------------------------------> david_AT_linuxdj.com -'
This archive was generated by hypermail 2b28 : Thu Jul 26 2001 - 03:23:24 EEST