Re: [BENCHMARK] 2.5.73-bk3, -bk4, -mjb1 regression test results

[Mark Peloquin <peloquin@austin.ibm.com>]

Dave Hansen wrote:

>On Fri, 2003-06-27 at 11:34, Mark Peloquin wrote:
>  
>
>>Nightly Regression Summary
>>for
>>    
>>
>...
>
>Have you ever had these regression tests run against the same kernel?  
>say, 2.5.73 vs 2.5.73
>
>How often do you see improvements or regressions reported then?
>  
>

Its benchmark specific. Currently dbench, specjbb, and lmbench have a 
few problematic data points that aren't terribly repeatable. So on the 
same kernel, one can see improvements and regressions. That said, the 
majority of data points are very repeatable and give good indications of 
real change. We will continue to tweak things here and there to reduce 
the number of false positives.

Mark