Frank van der Linden wrote:
> 
> The results are at ftp://ftp.netbsd.org/pub/NetBSD/misc/fvdl/bench/
> 
> The tests were done using hbench. See the README in that directory,
> and the hbench source code for more info. These are simply
> the 'summary' files.

Thanks for including this in your README file:

  The kernel configuration used was GENERIC, but with the DIAGNOSTIC
  option removed (it causes some slowdown and interference, and I
  plan to remove it from -release kernels on the i386 port anyway).
  Also I put sb0 at isa at irq7 instead of irq5, to match reality.
  No other changes were made to the config file. 

I've been using DIAGNOSTIC on i386/NetBSD since I don't know when.
The kernel comment "# cheap kernel consistency checks" must be
judging some level of cheapness other than size and CPU time B^)

After commenting this option out, my kernels are smaller and a
"make build" is noticeably faster!

I would never have considered turfing this option had you not
mentioned that it had an impact on measuring performance.

Thanks!

-- Ken