<greywolf@starwolf.com> (Greywolf) writes:

> I understand that threads/SMP will be going into release 2.0...or at
> least that is the rumour.

It's more rumour than anything else, and I think it's been said the
other way around: "When we get threads/SMP, we'll call that 2.0".

> I have a question:  Are we not going to at least run single-processor
> kernel threads for a while to make sure they work before attempting
> SMP?  or am I just being too cautious on this front?

As Hubert noted, we have "kernel threads" of a sort already: multiple
execution contexts within the kernel. They're not *preemptable*
threads, and linkage between these kernel threads and preemptable
kernel threads or process-level LWP stuff or any other
thread-kinda-thing is probably misplaced.

Also, we may well get SMP of the big-lock variety before we get
preemptable kernel threads (or we might not. Truth is stranger than
fiction). So there are at least three separate user-visible milestones
lumped into the phrase "threads/SMP".

I think the theory behind the 2.0 rumour was that converting the
kernel to use preemptable threads would be a sufficently large
re-write to warrant a major-version bump. IMO, version numbers are
"just" marketing fodder, and it's not clear to me that any of these
milestones are the "right" reasons for such a bump.

        - Nathan