cpu_load = lmm_constraint_get_usage(surf_host->cpu_->getConstraint()) / surf_host->cpu_->getPstateSpeedCurrent();
/** Divide by the number of cores here **/
- cpu_load /= surf_host->cpu_->getCoreCount();
+ cpu_load /= surf_host->cpu_->coreCount();
if (cpu_load > 1) // A machine with a load > 1 consumes as much as a fully loaded machine, not more
cpu_load = 1;
if (cpu_load > 0) { /* Something is going on, the machine is not idle */
double min_power = range.min;
double max_power = range.max;
- double power_slope = max_power - min_power;
- current_power = min_power + cpu_load * power_slope;
+
+ /**
+ * The min_power states how much we consume when only one single
+ * core is working. This means that when cpu_load == 1/coreCount, then
+ * current_power == min_power.
+ *
+ * The maximum must be reached when all cores are working (but 1 core was
+ * already accounted for by min_power)
+ * i.e., we need min_power + (maxCpuLoad-1/coreCount)*power_slope == max_power
+ * (maxCpuLoad is by definition 1)
+ */
+ double power_slope;
- int coreCount = host->coresCount();
++ int coreCount = host->coreCount();
+ if (coreCount > 1)
+ power_slope = (max_power - min_power) / (1 - 1 / coreCount);
+ else
+ power_slope = 0; // Should be 0, since max_power == min_power (in this case)
+
+ current_power = min_power + (cpu_load - (1 / coreCount)) * power_slope;
}
else { /* Our machine is idle, take the dedicated value! */
current_power = range.idle;