[LAU] kernel rt-patches future

Joakim Hernberg jhernberg at alchemy.lu
Wed Oct 22 08:32:42 UTC 2014

On Tue, 21 Oct 2014 14:05:23 -0700 (PDT)
Len Ovens <len at ovenwerks.net> wrote:

> In my opinion, Much of the use for RT kernels in audio is to try and
> fix problems with system HW tuning. In my trials, I have found that
> using only a "lowlatency" kernel, I can get clean latency as low as
> my card is able to be set anyway (32 samples). The only xruns I get
> are from applications that do not shut down correctly or take to long
> starting up (that is they enable ports in jack before they are
> stable).

IME/O, the one thing the -rt kernel brings to audio processing on
linux is reliable kernel scheduling. Provided that all the hardware
plays ball (which most does), you can expect a scheduling delay of
under 100usecs with a -rt patched kernel.

With vanilla kernels (even low latency ones), you can expect to see the
occasional latency peak into the msec range, each such peak will mean
an xrun if the kernel scheduling and audio processing latencies exceed
the deadline as given by samplerate and buffer sizes.  They might be
just occasional and not a problem, but they will be there.

> But yes, the standard RT keneral development is seeing the money dry
> up.

There never was any money to dry up.  Most of the development have been
carried by https://www.linutronix.de, Red Hat and private individuals.



More information about the Linux-audio-user mailing list