Outsourced by viewtopic.php?p=121705#p121705:
According to this thread (*1), many of the computer hardware / (programming) techniques used back then no longer exist today.
So it makes no sense to use these optimizations anymore.
Can we get together the current state of what is needed to make a good LAW (Audio RT OS), collect and collate evidence?
Who is familiar with this?
Ideally, this should also be updated in https://wiki.linuxaudio.org/wiki/system_configuration. or generally on linuxaudio and linuxmusicians.
What do you think about this?
How - if necessary - do we update our knowledge?
Single fragments of @x42 (*1):
set permissions to the audio group to /dev/hpet e /dev/rtc;
Which apps use those? It should not be needed anymore, nor should any app use those for audio timing on a modern system (kernel newer than 2.6.39).
I wrote most of the content on this page in 2006/7 and sadly most of it is out of date and has since been frankensteined by various wiki editors. Take it with a huge grain of salt.
[...] Installing jackd asks to enable reatime-permissions. I don’t use jack, I only install the package to conveniently setup rt-permissons and groups. After that setting up rtirq-init (and the threadirqs kernel boot option) is sufficient. – The only manual change I made is to re-apply rtirq settings after suspend/resume cycle (similar to https://github.com/jhernberg/udev-rtirq 2).
That’s all.On this i7, I don’t even need to change the CPU governor. CPU frequency change are orders or magnitude faster than audio process cycles.