Re: Darwin & SMT?
Re: Darwin & SMT?
- Subject: Re: Darwin & SMT?
- From: Bernie Zenis <email@hidden>
- Date: Tue, 10 Sep 2002 14:05:17 -0400
On Tuesday, September 10, 2002, at 01:41 PM, Jim Magee wrote:
On Tuesday, September 10, 2002, at 01:16 PM, Bernie Zenis wrote:
Is anyone here familiar with Simultaneous Multithreading
(http://www.cs.washington.edu/research/smt/)? Does anyone know if
Darwin would lend itself well to SMT? For example, if a SMT PowerPC
chip came out, would it be easy to port Darwin to it? I realize that
might depend on what kind of SMT architecture was implemented.
I have always felt that Darwin works best on an SMP machine. There are
enough system threads, and multi-threaded applications to want two CPUs
in many "peak" CPU utilization situations (my Mac OS X system typically
has ~200 threads). I think the real world anecdotal evidence bears
this out as well (most people with dual-ies are quite happy with the
responsiveness of their machines, even when compared to single
processor machines of quite higher clock rates).
All of this is one of the reasons Apple ships all their PowerMacs as
dual processor machines right now.
The CPI numbers under most of those situations indicate that an SMT
(others call it Hyper-threading) processor might do reasonably well
compared to an SMP machine. But as you say, it all depends on the
particular SMT design and other architectural factors on any given
processor.
I've always know that modern computers are doing many things at once;
but, I didn't really realize (because I lived in Mac OS 8/9 land :-) how
many things that was until I got X installed and looked at outputs of
programs like top. Obviously, most of those threads are not demanding a
lot of the machine during much of the time (otherwise my CPU Monitor
load would be much higher). I had thought many years ago that computers
would benefit from a processor design like SMT, I just didn't know that
it had a name and that people we're working on it until a few years ago.
(You would think that a college professor teaching an MP class would
know.)
I see that real SMT processors are finally being built which makes me
happy. I just didn't know if Darwin had its foot in any of it. I'm
thinking that for an SMT processor to become a real possibility for
replacing (S)MP designs that the "thread" in SMT would have to refer to
a full fledged process (eg, own address space, etc.). Would it be easier
to port Darwin to a SMT processor in which its "threads" were light
weight or full fledged? Or both equally well?
Thanks,
Bernie
PS I'm really happy Apple decided to finally follow the dual processor
route. (It is about time! :-)
_______________________________________________
darwin-kernel mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/darwin-kernel
Do not post admin requests to the list. They will be ignored.