The Experimental Profile Switcher moved from Park Control to Process Lasso in v18.2.0.50 (previous Park Control forum [topic=37678]).
Having upgraded to the this new way (Process Lasso v18.2.3.42 & Park Control v5.5.4.4) I have a few questions as the experimental version seemed to work better for my requirements:
1) The Experimental Park Control version had a most useful option which is no longer present:
"Check load of individual core" which was perfect for boosting single threaded apps on high core processors.
Is there a technical reason why this option is no longer available & if there is none can it be restored please ?
2) New option "Smoothing windows (samples)"
How is this connected with the "Minimum time in power plan" please ? (I understand there likely is a connection but better understanding may aid usage).
Kind regards
Nudge
The per-core load option was particularly useful for single-threaded workloads, so I'd also like to know whether its removal was technical or simply a design change, and how the smoothing sample window interacts with the minimum power-plan time in practice.
We are considering restoring this. The Governor doesn't otherwise need to track per-core loads (only the GUI), so that was the reason the option wasn't carried over to Process Lasso. However, we can certainly add per-core load checks to the Governor and only use them when a dependent feature is turned on. I'll get back with you on this ;).
The smoothing window option averages the samples over the last X ticks of the Governor (default 1 second) and uses that average to decide when to act. This is needed if you don't want it to act during a 1-second load spike, for instance. It's similar to the minimum time in power plan option in that it prevents frequent switching, but different in that it is using the average to decide when to act in the first place.