algorithm.txt

measurement cycles:
- STATISTICS_INTERVAL (ms)
- MEASUREMENT_INTERVAL (ms)
	
STATISTICS_INTERVAL
	- Summary: gather (method) activity
	- for all running threads, from top / newest of stacktrace
		- first element rated as computation time of 1 STATISTICS_INTERVAL / MEASUREMENT_INTERVAL
		- for prefixed report:
			- first element which matches at least one prefix rated as computation time of 1 STATISTICS_INTERVAL / MEASUREMENT_INTERVAL
	- increment METHOD_ACTIVITY accordingly
			
MEASUREMENT_INTERVAL
	- summary
		- gather power consumption
		- aggregate method activity and power consumption to energy consumption per method
	- POWER_TOTAL (W) - get current CPU power per MEASUREMENT_INTERVAL
		- call e.g. Libre Hardware Monitor for current CPU power usage
		- assume power was the same for the entirety of MEASUREMENT_INTERVAL
	- get CPU times
		- THREAD_TIME (ns) - total CPU time of each running threads -> should time since last measurement?
		- APPLICATION_TIME (ns) - total CPU time of all running threads -> should be sum of should time since last measurement?
	- computation -> as separate task?
		- allocate power usage to threads per MEASUREMENT_INTERVAL
			- POWER_THREAD (W) = POWER_TOTAL * (THREAD_TIME / APPLICATION_TIME)
		- determine power usage of each method per MEASUREMENT_INTERVAL
			- POWER_METHOD (W) = POWER_THREAD * (METHOD_ACTIVITY * STATISTICS_INTERVAL / MEASUREMENT_INTERVAL) -> currently fix ration of 1 / 100
		- determine energy usage of each method in MEASUREMENT_INTERVAL (current assumption: MEASUREMENT_INTERVAL always 1000ms, so W = J)
			- ENERGY_METHOD (J) = POWER_METHOD * MEASUREMENT_INTERVAL / 1000