Dynamic Voltage Scaling (DVS)
Power consumption has become a critical design parameter as more transistors are integrated in a single chip
thanks to the rapid advance of process technology.

DVS aims at reducing energy consumption by scheduling voltage / frequency pairs for a task (or tasks) running
on a computing unit to be completed in and as close as to a given deadline.



One of critical points in DVS is the prediction accuracy of the task completion time being executed.
- The prediction accuracy becomes more critical when the workload shows a large variation and/or non-stationary property.

A run-time workload estimation technique for DVS with adaptive filters
- Adaptive filters can continuously reduce estimation errors using feedback error correction.

Adaptive Filters for run-time estimation
- Previous adaptive filters : Moving Average (MA), Weighted Mean (WM), PID controller
- Hidden Markov Model (HMM)
- The Kalman Filter (KF)
- The Particle Filter (PF)

The goal is to obtain High-speed FPD power estimation/optimization simulator.




Interpolation or/and extrapolation between panels of different size.