Low Latency Patch

Robert Dinse (nanook@eskimo.com)
Thu, 29 Jun 2000 18:59:27 -0700 (PDT)


I've been watching the debate over low latency issues for audio and video
applications where jitter is important and I wanted to make a comment with
respect to the patch.

I played with it a bit on various machines just to out of curiousity,
wanted to see how it impacted normal applications.

What I found was that with the patches applied the interactive response on
a loaded machine was HUGELY better. I have servers doing IRC, shell, mail,
news, web, everything I tried it on responded better when it was heavily
loaded.

The only downside I ran into was that it made the spin_lock deadlock
problems on Sparc SMP worse. But in general terms, when things got loaded,
it made the systems a lot more usable. None of the 2 seconds until you see
a keystroke echo stuff.

I realize this wasn't the intent of the patch, that the intent was to
guarantee low jitter to allow things like audio processing to work reliably,
but that's sure not an unpleasant side effect.

I haven't kept them in, despite the improvements in overall
responsiveness, because I didn't want to keep applying them for each kernel
release, and that's the problem with not including capabilities, like the Solar
Design security enhancements, or the low latency patch, in the main kernel
tree. It's too much of a headache to keep patching for each version.

Why can't this be made into a configuration option?

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/