Installing a Liquorix Kernel

[lilith]

No single xrun in almost 4 hours now! This would be too good to be true! Never would have believed that a different kernel would help. But let's wait and hope...

[khz]

The instructions on their HP are correct. Thanks for the test and the feedback!
(So I made a mistake back then.)

[studio32]

It also installs 'irqbalance'. Never came across that package before afaik.

[studio32]

something

[khz]

It also installs 'irqbalance'. Never came across that package before afaik.

Hmm - irqbalance 2019

[lilith]

I also read that it's not necessary to set the CPU governor to performance and that it even becomes worse.

https://techpatterns.com/forums/about2786.html

@studio32 Was it better with the Liquorix kernel or did you get the xruns with it?

[studio32]

something

[lilith]

I also read that it's not necessary to set the CPU governor to performance and that it even becomes worse.

https://techpatterns.com/forums/about2786.html

@studio32 Was it better with the Liquorix kernel or did you get the xruns with it?

The problem is probably the LV2 plugin of Zynaddsubfx.
I didn't find significant differences between the kernels. The rt debian kernel has little left to wish for I think.

I also read that it's not necessary to set the CPU governor to performance and that it even becomes worse.

Are you sure you got that right? I read something else. Liquorix uses performance by default, so don't put your governer on something else. That's how I read it.

@khz After your post, I disabled irqbalance.

You are right, I thought he had it send to performance.
edit: Also got a xrun again. Sometimes something has happening in the background causing it. Maybe when Debian is looking for updates, cause it happens only sporadically.

That's the xrun reported by Jack in verbose mode

Code: Select all

Sat Jun 13 19:16:39 2020: Jack: **** alsa_pcm: xrun of at least -362921.600 msecs
Sat Jun 13 19:16:39 2020: Jack: ALSA XRun wait_status = 0

362000 msec makes no sense to me. I guess there's something completely wrong. (?)

The number is quite large.

[milo]

I got it working on Linux Mint, through the damentz ppa (deb http://ppa.launchpad.net/damentz/liquorix/ubuntu bionic main). I had to disable secure boot in the BIOS, though, because the system complained that it was an unsigned kernel. Should I be worried about this?

So far the kernel seems to be running well, but I haven't put it under stress. I'll let you know if it starts spitting out xruns when I record layered vocals. That's usually when it happens.

[lilith]

I got it working on Linux Mint, through the damentz ppa (deb http://ppa.launchpad.net/damentz/liquorix/ubuntu bionic main). I had to disable secure boot in the BIOS, though, because the system complained that it was an unsigned kernel. Should I be worried about this?

So far the kernel seems to be running well, but I haven't put it under stress. I'll let you know if it starts spitting out xruns when I record layered vocals. That's usually when it happens.

Hi Milo,

the strange thing in Bitwig is that it does not depend on the DSP load and I even get occasional xruns when the transport is stopped. Maybe a plugin? I have to do more test to be sure.

[lilith]

I deinstalled it again as it was using a lot more of CPU compared to the Debian stock kernel. 10-15% more when running Bitwig.

[milo]

I've been recording vocals in Ardour this afternoon with Liquorix enabled, 26 layers of voices, with active reverb and compressors on various tracks, and have had no xruns.

But I notice a similar thing with DSP and CPU load. Compared to the lowlatency kernel, Liquorix uses about 1-2% more DSP load. According to top the CPU load drops from 48-56% with Liquorix to 42-48% with lowlatency.

As an aside, I also noticed that the linux-lowlatency dummy package that depends on the latest version of the lowlatency kernel won't trigger the install of anything higher than the 4.15-lowlatency kernel (at least in Mint 19.3). I manually marked the 5.4-lowlatency kernel for installation, and I think it does a bit better then the 4.15. CPU usage ranges maybe 2-3% lower with the 5.4-lowlatency kernel compared to 4.15. All of this was playing the same song in Ardour. Not a scientific benchmark study, though.

I think I will also drop Liquorix for now, but may return in the future.

[sysrqer]

I believe the CPU scheduler it uses is more hungry but is meant to be more responsive overall...or something like that.

[GMaq]

Hi,

Liquorix is a very tuned kernel but it doesn't really seem to translate into a terrific kernel for Audio work in many cases, Years ago in AV Linux I tried making Liquorix the default because it solved a lot of problems with installing 3rd party Video Drivers (Liquorix and SGFXI Video driver scripts work closely together). But I just couldn't get consistency in Audio performance so had to go back to rolling our own Kernels.

[studio32]

something