On Friday 20 May 2005 22:17, Mike Melanson wrote: > Could we mitigate this problem by "warming up" the cache by running the > function 1000 times and then profiling run #1001? Perhaps profile a few runs and take the fastest time?