Matthias Scheler wrote:

>In article <44875CE2.8090801@netbsd.org>,
>	Frank Kardel <kardel@NetBSD.org> writes:
>  
>
>>Timecounters will be supported on a per port basis. With
>>the commit following ports make use of timecounters:
>>    i386
>>    amd64
>>    sparc
>>    sparc64
>>    
>>
>
>That will fix the time keeping problems on MP machine?
>  
>
Thats what I expect. My local tests on an AMD64 X2
where fine. Dave Tyson had success on a 3 CPU XEON
machine acting as a primary name server.

And the code is now in -current so we can smoke out
the last bugs and corner cases.
I have seen very solid performance on MP machines
with timecounters. Please test!

Some graphs on the time counter status web page
are taken from the AMD64 X2 machine. Be careful
when looking at the graphs - the scale of the y-axis
differs up to a factor of 21000. Basically fleas and
elephants are shown there at the same size - sort of :-)

Frank