On Thu, 23 Jul 2009, Mel Gorman wrote:
> oprofile_start.sh has the option of scaling the sampling of PMU events
> by some factor. The reasoning is that oprofile introduces high overhead
> when sampling at it's maximum rate - somewhere between 8% and 14%
> depending on the workload. However, the mi
oprofile_start.sh has the option of scaling the sampling of PMU events
by some factor. The reasoning is that oprofile introduces high overhead
when sampling at it's maximum rate - somewhere between 8% and 14%
depending on the workload. However, the minimum sampling rate for CPU
cycles and events su
