Re: [ecasound] Improving low latency performance

New Message Reply Date view Thread view Subject view Author view Other groups

Subject: Re: [ecasound] Improving low latency performance
From: Dave Phillips (
Date: Mon Nov 20 2000 - 16:38:11 EET

Kai Vehmanen wrote:
> Ecasound is used as an example in this Dave Phillips' article at O'Reilly
> Network. Even without the ecasound references, it'd still be a great
> arcticle! ;) Seriously, it's a good introduction to low latency issues
> including patching kernels, use of realtime scheduling, disk tuning, etc.

Thank you, Kai. A fair amount of research went into the article, and I
did indeed test everything as reported. Took a bit of time, and I
certainly got handy compiling and installing multiple kernel versions...
> On the 4th page of the article, Dave has noted one important thing
> concerning ecasound:
> [snip]
> "Ecasound was the most severely affected, even with the real-time priority
> flag set, probably due to the nature of the test (CD-quality stereo
> recording). The program reported multiple under-runs through the ALSA PCM
> device (/dev/pcmC0D0), and the resulting WAV file was ruined by sixteen
> dropouts in a three-minute recording."

I want to emphasize that ecasound works perfectly for my project, as
long as the LL patch is applied and the disks are tuned. The disk-tuning
alone may yield acceptable latencies for some purposes.

I also want to thank you again for such an excellent application. I've
made some good contacts now at Linux Journal and O'Reilly Network, so
there will eventually be a profile of ecasound in print. It will
probably appear in LJ, I'm hoping to update and expand various chapters
from my book, running them in LJ while I write the 2nd edition.

> This is the very reason why I'm currently concentrating my efforts to
> the new double-buffering system. My goal is to achieve reliable
> streaming even with standard (no ll-patches) kernel. This means
> ecasound has to manage both scheduling and disk i/o latency peaks
> upto 300-400ms.

Yes, the standard kernel has some real problems with scheduling, all
addressed by the LL patches. Linus Torvalds needs to be made better
aware of the real achievement in those patches, but my guess is that his
audio awareness extends to his CD player (and maybe an MP3 player).
Otherwise it's an area where he is utterly ignorant and capable of
making extremely stupid decisions (to use one of his favorite
descriptors). But as long as Ingo and Andrew continue to supply working
LL patches, I'm still happy. And God bless Linus anyway for starting
this wonderful operating system. :)

Best regards,

== Dave Phillips

To unsubscribe send message 'unsubscribe' in the body of the
message to <>.

New Message Reply Date view Thread view Subject view Author view Other groups

This archive was generated by hypermail 2b28 : Mon Nov 20 2000 - 16:36:20 EET