There is a Real-time Linux kernel, but way (years, I think) behind the HEAD.
Used in various industrial systems, maybe in weapon systems.
The article author seems to have a weak understanding of real-time systems in general, especially around hard vs soft.
Audio is clearly not a particularly difficult thing, IF you can buffer, which basically all do to some extent.
If you don't (i.e. <5 samples) buffer, not only is the deadline critical. No timing variance allowed.
I use STM32F405/F446 for audio processing in modular synthesizer products.
To get there, variance of 3-4 microseconds on the 20.8microsecond target is audible, and had to be clever to solve that.
The software doesn't do buffering (well, technically 1 sample).
https://github.com/awesomeaudioapparatus/demiurge/
The software/ is a bit gnarly to navigate, since several hardwares are to be supported (just gave up on CH32V307 Risc-V mcu), but the base audio processing is in components/demiurge
@lupyuen Interesting. In something like audio, you need sub 1ms timing. I don't think there's any Linux that will do this. I actually wonder if RISCOS would be good for real-time systems.
My big dream is to have some kind of RP2040/Pi combo, so one can have fast and proper real-time where necessary.