Date: Tue, 8 Jun 2004 23:08:43 -0400 From: Ken Smith <kensmith@cse.Buffalo.EDU> To: John Baldwin <jhb@freebsd.org> Cc: freebsd-doc@freebsd.org Subject: Re: Fwd: Re: RFC: additions to the Glossary Message-ID: <20040609030843.GC4439@electra.cse.Buffalo.EDU> In-Reply-To: <200406081659.06684.jhb@FreeBSD.org> References: <p0620010abceb7a5479be@[129.85.219.160]> <200406081659.06684.jhb@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Jun 08, 2004 at 04:59:06PM -0400, John Baldwin wrote:
> I'm not sure the use of 'Symmetric' vs. 'Asymmetric' is correct, either, but
> I'd need to recheck my Schimmel SMP book. Also, Giant is not a hold over
> from previous BSD releases. FreeBSD 3.x and 4.x used a single mutual
> exclusion lock (not the same as the Giant mutex in 5.x) around most of the
> kernel (a few places like fast interrupt handlers have their own spin locks).
> FreeBSD 5 has a sleep mutex called 'Giant' that protects most of the kernel.
> FreeBSD <= 3.x and BSD in general do not have mutexes at all but use spl +
> lockmgr locks.
You're right - I probably shouldn't have tried to define SMP versus
Asymmetric SMP. The definitions seem to depend on who you ask (even
if you just ask people you otherwise think are credible). The one I
gave is the one I've seen the most but of the three books I have here
in the office at the moment there are three different definitions for
what constitutes Asymmetric Multiprocessing:
- The one I gave - only one processor allowed into the kernel
at a time but different processors can be running different
threads of control in user-land at the same time.
- Divide the system up into subsystems (network, disk I/O,
etc) and dediate one CPU per subsystem. This seems like
the most arcane definition...
- Only one *specific* processor (e.g. CPU-0) is allowed into
the kernel, and it is the only one that fields interrupts.
This is a bit more restrictive case than the first one...
Given those odds it seems likely there are even more definitions...
And sorry if it had seemed like I was attributing Giant to BSD, that
wasn't intended. AFAIK the CSRG folks never considered MP systems at
all. The first work on SMP I was aware of actually came from companies
like Encore doing DARPA grant work though I could be mistaken. We had
one of their computers for a while (I named it sybil after the
schizophrenic lady with several personalities... :-).
--
Ken Smith
- From there to here, from here to | kensmith@cse.buffalo.edu
there, funny things are everywhere. |
- Theodore Geisel |
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040609030843.GC4439>