[SunRescue] OT: dual PPro mb's...

[Joshua D. Boyd]

On Fri, 30 Mar 2001, Greg A. Woods wrote:

> Rumour has it (I'm not a Linux user) even the 2.4 kernel has far
> "grainier" data structure locking for SMP than would be desirable for
> really good CPU utilisation.  I'm not sure how many CPUs it's designed
> to handle well either.  IIRC Linus said something about the locking
> being a little more coarse than he'd like in his rushed announcement of
> the 2.4 kernel.

Well, I was under the impression that Linus' comments were to be
interpreted as meaning that it was fine for 2way SMP, but not as good as
possible on 4, 8, or 16 way SMP.  SGI has a team working on tuning linux
for 32 and more CPU machines BTW.  They are working ontop of origin 2000
machines last I heard.  Most of the changes they make should also benefit
Intel users.  

I can't really comment too much on network performance though since that
isn't my area of specialty.  Heck, I'm still running 10mbit everywhere in
my house rather than 100mbit.  I can tell you that dual processors get
twice the performance on kernel compiles.  I've also found that some
programs (like WindowMaker) can't be compiled using the -j2 flag.  Bummer.
I don't realy have anything else that would push my CPUs.  I haven't had
time to work on any threaded stuff since I got the machine running linux
(for the first few monthes of owning it, it ran windows to play DVDs).

--
Joshua Boyd


> 
> (I really don't know what the problem is -- the Bell Labs folks did a
> dual-CPU implementation of V7 way back in the early 80's and there have
> been lots of people doing it even better ever since and many of them
> have documented their methods and the various tricks and traps they've
> encountered along the way.  I.e. the knowledge is out there and it
> should be trivial enough to apply it, especially in Linux where the main
> target still seems to be the Intel processor family.  It should merely
> be a matter of engineering at this point.  The NetBSD folks have been
> doing a "clean" implementation and it'll probably pull up into the
> public source in the next few months, and it's supposed to have fully
> N-way CPU support and completely hardware independent.  I'll bet it gets
> fine-grained locking far more quickly once it's public than any of the
> other systems have apparently been managing.  That said though I've
> heard very promising things from the FreeBSD folks, especially ever
> since they joined forces with BSDi.)
> 
> -- 
> 							Greg A. Woods
> 
> +1 416 218-0098      VE3TCP      <gwoods at acm.org>      <robohack!woods>
> Planix, Inc. <woods at planix.com>; Secrets of the Weird <woods at weird.com>
> _______________________________________________
> rescue maillist  -  rescue at sunhelp.org
> http://www.sunhelp.org/mailman/listinfo/rescue
>