Strange Performance Tests

[Lars Andersen]

On Wed, 2002-07-10 at 20:59, Justin Chapweske wrote:
> A 10ms benchmark??  This is too small of a time interval to be able to 
> make any meaningful performance observations.  A number of things can 
> happen in the kernel in that amount of time that can affect performance.

As i wrote in the original post this is just an extract. The test has
been running for many hours. Every measure point last approx 10ms, thats
true, but where talking thousands and thousands of measurements. For
obvious reasons I had to limit the sample size to fit in a post ;)

And its not a single run either. The results are always the same.
Kernel events have nothing to do with this.

/La