Subject: Benchmarks (was: Please Revert newlock2)
To: None <tech-kern@NetBSD.org>
From: Bernd Ernesti <firstname.lastname@example.org>
Date: 02/21/2007 08:34:06
On Tue, Feb 20, 2007 at 11:02:40PM -0800, Bucky Katz wrote:
> We have two choices: 1) we can make all of our stuff work in -current
> and then find out if 1:1 kills performance or not, followed by which
> we get to come along for the ride while you make the scheduler and
> interrupt handler change or 2) we can do what dozens of people here
> and in current have been telling us to do: use 4.0.
My personal way would be to try to get a benchmark of -current after
it compiles and works and see how good or badly it is now.
And if it does not compile or work I would send a mail either where
the compile fails or a small test case which makes it possible to
reproduce a threading problem.
Which doesn't mean I wouldn't send a pr about a kernel crash or core