On Mon, 4 Jun 2001, David Brownlee wrote:

> 	(Picking back..)
> 
> 	I'm not aware of any system that will give "a factor of N
> 	speed up, for each Nth processor". I do not believe it
> 	possible for inherent reasons of memory and i/o contention
> 	alone.

Probably correct. But you want to keep the line as close to a linear curve
as possible.

> 	So... if percentage speed up is not a valid metric for SMP,
> 	how ahout "re-entrant on all processors with fine-grain
> 	locking".  How re-entrant and fine-grain do you need to be
> 	to qualify.  NetBSD is trivially re-entrant for one syscall
> 	- getpid(), but should it be all syscalls? - all i/o?
> 	locking on every structure access?

For me, SMP definitely means that all processors should be able to do I/O,
within the limitations of what processors are physically connected to
different devices. *Very* little should be bound to a specific
processor. The only thing I can think of straight away is that only one
processor should update the system time.

I cannot think of any one system call that should only be executed by one
specific processor, well perhaps shutdown could be restricted to the
master CPU...

As for the granularity of locking, that don't have anything to do with SMP
or not, only performance.

Only my view on symmetric in relation to MP.
Symmetric is a pretty simple word. Unless there is a *very* good reason
for it, all processors should be able to do anything.

	Johnny

Johnny Billquist                  || "I'm on a bus
                                  ||  on a psychedelic trip
email: bqt@update.uu.se           ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol