onFri, 27 Mar 1998 10:54:27 -0800 Jason Thorpe writes:

>On Fri, 27 Mar 98 10:50:03 -0800 
> "Chris G. Demetriou" <cgd@pa.dec.com> wrote:
>
> > Showing some numbers for FreeBSD and saying that he's never gotten
> > close to them with any NetBSD system isn't particularly great
> > argument, but neither is showing no numbers whatsoever and asserting
> > that NetBSD's performance must be as good.
>
>My intent is not to assert that NetBSD's performance in this case "must
>be as good".  My intent is to get Jonathan to back up his claim with
>real numbers derived from an identical measurement on identical hardware.

Jason, that is an unfair misrepresntation of what i actually said.
What I said was:

>I did say that FreeBSD pays more attention to low-level performance
>than NetBSD does.  I happen to think that's true. It shows up in all
>sorts of ways, like looking at the assembler output by gcc and
>tweaking their source code to match; paying more attention to tuning
>of critical code fragments like kernel bcopy() and copyin(), interrupt
>dispatch code, and similar hackery.  NetBSD doesn't do that.

That is _not_ as simple as the single head-to-head number for
interrupt latency that you insist on.

As I said to you on the phone earlier this week, my lab book documents
that I can get substantial tcp speedups on a pentium simply by
dropping in FreeBSD's kernel bcopy.

I've given you the other numbers I have. THey substantiate the claim I
made.  That case is entirely consistent with somewhat dated, but
published work, by Baker and Lai, Larry McVoy, and Aaron Brown and
Margo Seltzer.  You can compare Aaron's numbers for lmbench to
the published  lmbench numbers. 

Those numbers and my own reruns of hbench on an identical Endevour
system back up my claim. If you dont like that, tough.  It's up to
*you* to prove prove the situation has changed since then.  
Maybe it has, but for these specific tests, my numbers say it hasn't.

For you to say I'm being unprofesional is garbabge.