SaaR: a strategy for energy efficiency in DVFS computing platforms beyond the scaling axiomatic approach

Abstract Dynamic Voltage and Frequency Scaling (DVFS) computing platforms are highly effective in reducing energy consumption by dynamically adjusting the operating frequency and voltage of processing units within predefined operating pairs. By selectively scaling down the execution frequency of application tasks, significant energy savings can be achieved while preserving timing constraints. For applications composed of dependent tasks, energy-aware frequency scaling is predominantly addressed through the Scaling Axiomatic Approach ( SAA ), which exploits task slack to enable safe frequency reduction but incurs a high computational cost due to repeated global timing recalculations. To mitigate this limitation, the GinGa approach was proposed to reduce the computational complexity, albeit with a degradation in energy optimization effectiveness. This paper introduces the Scaling Axiomatic Approach Replacement ( SaaR ), a low-complexity, compile-time mechanism designed as a principled replacement for SAA . While preserving the axiomatic foundation of slack-based frequency scaling, SaaR restructures the computation through bounded and localized timing-update mechanisms and a dedicated time-updating criterion, thereby eliminating repeated global recomputation. As a result, SaaR achieves energy savings comparable to those of SAA while significantly reducing the computational complexity. Experimental results confirm that SaaR outperforms GinGa and provides an effective balance between energy optimization and execution efficiency on DVFS-enabled computing platforms.