Approx-RM: Reducing Energy on Heterogeneous Multicore Processors under Accuracy and Timing Constraints

• DOI: 10.1145/3605214
• Publication Date: 2023-06-22T20:10:46Z
• Authors: Muhammad Waqar Azhar; Madhavan Manivannan; Per Stenström
• Keywords: Multi-core processor; Ranging

Reducing energy consumption while providing performance and quality guarantees is crucial for computing systems ranging from battery-powered embedded to data centers. This article considers approximate iterative applications executing on heterogeneous multi-core platforms under user-specified targets. We note that allowing a slight yet bounded relaxation in solution can considerably reduce the required iteration count thereby save significant amounts of energy. To this end, proposes Approx-RM , resource management scheme reduces expenditure guaranteeing specified as well accuracy target. predicts number iterations meet relaxed target at runtime. The time saved generates execution-time slack, which allows allocate fewer resources platform terms DVFS, core type, meeting contributes with lightweight methods predicting needed uses aforementioned predictions just enough comply service constraints Our evaluation shows savings 31.6%, average, compared Race-to-idle when only by 1%. incurs timing overheads less than 0.1%.