#New WB Multithread?
9 messages in this thread
Read the message again. I don't know if WB is multithreaded. — Art
Art,
From what I've been told WB 2.0 is _sort of_ multithreaded. More I
cannot say, since this stuff is still under lock & key (so to speak).
-Dave
David:
What does multithreaded and _sort of_ multithreaded mean?
Bill Perez
It is for luanching programs, tried that out today on our demo unit.
Ben
Leigh's Computers
Starting a program is multithreaded, but not reading a directory. At
least not as of beta 3. Of course, that doesn't mean that it won't happen,
just that as of beta 3 it wasn't in there yet.
— Ethan
> Starting a program is multithreaded, but not reading a directory.
You WANT to processes doing a directory search at the same time? Sounds
like thrash-time to me…
OK, that's not so bad. After all that would require additional technology
that would have slowed the release of the 3000, which Commodore could NOT
have afforded at this point. You have to have more intelligent arbitration
of Hard disk requests than the OS now supplies.
What would be neat is something like what Novell does with its Network OS,
where all requests are queued in what they call an "elevator seeking"
algorithm. They are positioned by the controller in order of position on
the disk relative to where the heads are currently positioned. If the
request is towards the center and the armature is going outward, it is
placed on a queue that is to be satisfied when the arm starts back inward;
If the arm is already on the sweep inward, it gets queued for the next
sweep.
–Gabe
If Novell is indeed using 'elevator seeking' then they are at least 6 years
behind the technical papers being published on reducing disk access times.
The algorithm which seems to be winning (by only a small margin,
admittedly) all the 'benchmarks' (yeah, I know….lies, damn lies,
statistics, and benchmarks) is a simple 'sweep' algorithm.
That HAD been my impression. My apologies if I am wrong. Can anybody
substantiate this one way or another?
–Gabe