Making a dynamically-linked root

[Robert Watson]

On Tue, 3 Jun 2003, Alexey Dokuchaev wrote:

> > I don't want to sound harsh, and I do appreciate your work.  However,
> > I think the last thing FreeBSD needs now is to get slower.  We're
> > already far slower than that other free OS.  Shouldn't we consider
> 
> Can you show any evidence of how slow is RELENG_5 (and _4) compared to
> those "other free OS"?  Some folks make such statements occasionally,
> but I haven't heard of any decent benchmarks from them.  That would be
> interesting to know though.  Thank you. 

One of the reasons we don't yet have a RELENG_5 is the performance issue: 
you'll notice we're cutting the 5.x releases off of the HEAD while
stability and maturity are still in the works.  If you take a look at the
5.2 TODO list, you'll see a number of interesting performance-related
activities that are intended to restore the higher interrupt latency to
lower levels, lower context switch costs, improve parallelism, etc.  In
the long term, we're going for both higher levels of parallelism and a
more explicit synchronization model in the kernel (which will have its own
benefits architecturally).  Any help you or others can give us on the way
to those goals would be much appreciated :-).

Robert N M Watson             FreeBSD Core Team, TrustedBSD Projects
robert at fledge.watson.org      Network Associates Laboratories