A question about the PREEMPT_RT patch set

A question about the PREEMPT_RT patch set

[余华兵(研发部 福州)]

Hello, everyone.
I have a question. If I apply the PREEMPT_RT patch set to the Linux kernel, is the real-time supported by the kernel hard real-time or soft real-time?

Re: A question about the PREEMPT_RT patch set

[Ralf Mardorf]

On Tue, 9 Nov 2021 02:27:37 +0000:
>I have a question. If I apply the PREEMPT_RT patch set to the Linux
>kernel, is the real-time supported by the kernel hard real-time or
>soft real-time?

Hi,

when I programmed my C64 in Assembly I got no audible MIDI jitter,
using software written by other, the same was true for my Atari ST, on
my Linux PCs there very often is audible MIDI jitter and the amount of
jitter varies depending on the used MIDI interface. USB MIDI devices
are less accurate than PCI/PCIe devices.

What I try to point out is, that a lot of layers are involved, when
using a PC and apart from those layers some things are even not
real-time processed at all.

If you are using the PC to process audio signals (instead of MIDI)
there at least will be a minimal amount of a delay (latency) already
produced by the audio device's hardware.

In short, I don't know what it is called done by a rt patched Linux
kernel, but the result, including the communication with interfaces,
when using a PC, in my opinion doesn't fit the definition of "hard
real-time. OTOH some users have got success to use patched kernels for
CNC, but I don't know what patch is used.

To my understanding "hard real-time" is a squishy term when using a
multi-tasking, multi-user PC. Such machines are probably more suitable
as remote control for independent hard-realtime machines equipped with
microcontrollers.

Regards,
Ralf