Re: Threads (was Re: MSI interrupts)

In article <fm6sujprvvtaq33pnliab0jl6bofbjvhm9@4ax.com>,
George Neuner  <gneuner2@comcast.net> wrote:

On Tue, 1 Apr 2025 16:44:11 +0200, Terje Mathisen
<terje.mathisen@tmsw.no> wrote:
>
>

In my programmer world view, a pure user software thread isn't that
interesting, it typically allows the programmer to control (more or
less) when it can be interrupted.

>
Done properly, user space threads can be pre-emptively scheduled and
otherwise interrupted.

This is true, but the details on contemporary machines and
operating systems are super intricate.  Goroutines can be
preemptively scheduled, though, as an example.

Also, consider "scheduler activations", wherein the kernel provides
only virtual cores, and all threading is done in user space.
>
https://homes.cs.washington.edu/~tom/pubs/sched_act.pdf
https://www.cs.ucr.edu/~heng/teaching/cs202-sp18/lec7.pdf
>
There is an implementation of this idea on NetBSD
http://web.mit.edu/nathanw/www/usenix/freenix-sa/freenix-sa.html
>
>
It's not a terribly popular idea, but IMO it is an interesting one.
YMMV.

Scheduler activations was an interesting idea, but came with
significant drawbacks.  Lack of support for deferring activation
events during (user) critical sections and lack of a virtual
core abstraction to complement its virtual multiprocessor were
probably the biggest.

Psyche provided a virtual core abstraction, but not a VMP.

Akaros combined the two in a novel way that gave an M:C model,
but never got past the research stage.

- Dan C.