🚀 go-pugleaf

In article <nevin-0709950221230001@mac-nevin.cs.arizona.edu>,
Nevin ":-]" Liber <nevin@CS.Arizona.EDU> wrote:
>I have a little secret for you:  the only difference between CMT and PMT
>is how tasks get swapped; ie, the whether scheduling is under application
>control or OS control.  Neither solution is perfect.  CMT apps have to
>guess about what the rest of the system is doing; PMT OSes have to guess
>about what all the apps are doing.  Some PMT schedulers, such as the one
>in Copland, will be better at doing this, since the apps will cooperate by
>telling the OS exactly what kind of task it is (UI, CPU-bound, etc.);
>schedulers rarely get this right using heuristics alone (such as the Unix
>scheduler).

However, PMT at least has a *chance* of getting it right adaptively,
because the OS can track the activities of the processes.  CMT has no
chance at this.  CMT programs must *literally* guess blindly about
what to do because the OS provides them no information about what the
other processes are doing.  Therefore, PMT will be better than CMT in
all but the most pathological of cases.

The question isn't whether PMT is always better than CMT.  The
question is how often PMT is better than CMT versus how often CMT is
better than PMT.  I argue that PMT is better than CMT *far* more than
the reverse.

>> You may have them but they are not being run Preemtively, there is a BIG
>> difference.
>
>Here we go, another assertion without even a shred of evidence to support it.
>
>Right now on my Mac, I am doing two concurrent file transfers via
>Anarchie, printing a document, letting Eudora fetch my email, playing an
>audio CD, surfing the Web with Netscape, and writing this message waiting
>for some of the other stuff to finish (I need the files to do some work). 
>Oh, and did I mention that I have a clock in the menubar merrily going
>along, and it gets resynced via NTP every hour, just like those big PMT
>Unix boxes?  

In short, you're doing *nothing* but I/O-intensive tasks.  Well, in
that case, CMT is going to work reasonably well because the calls to
do I/O typically cause a task switch anyway.

Where PMT works well is when you have a mix of I/O-intensive and
CPU-intensive tasks.  In that environment, neither type of task will
cause significant delays for the other tasks.

>I suppose I could go into my office and do this on a PMT
>system with 3x as much horsepower as my Mac at about 5x the initial cost
>(not counting the maintenance / lab staff), but the UI response is
>pathetic compared to my Mac (and I'm the only one on the damn
>machine),

...because of the design of the UI, and possibly because the scheduler
is a fair scheduler and designed for multiuser use.

>and the tools take much more work on my part to do the same job.

...which has absolutely nothing at all to do with PMT vs CMT.

>So, why don't you tell us *how* PMT is going to make our lives oh so much
>better, instead of emptily asserting it without proof or justification
>that it will.

PMT will make it possible for multiple people to use your machine at
the same time without having a significant impact on your own work.
You may not consider this to be particularly important, but I believe
it will become more important as the net becomes more popular and as
more client machines become server machines as well.

-- 

				Kevin Brown
                      SGI Customer Support Engineering