Thermal aware method improves computing speed without losing accuracy
AMAX: Accuracy-Aware Multi-Loop Asymmetric Perforation for Heterogeneous Architectures
Hardware Architecture
Summary
Many computers with different types of processor cores can slow down when they get too hot because usual ways to cool down reduce performance. The authors show a method that skips some repeated calculations in certain parts of a program more smartly, so the computer runs faster but still gives good answers. They check different parts of a program to see how much skipping can be done without making mistakes and change this while the program is running. Their tests show this method gives better accuracy compared to skipping the same amount everywhere.
What this means in practice
- •For system architects: Design thermal-aware multi-core processors that use adaptive loop skipping to improve speed under heat limits without large accuracy loss.
- •For embedded systems developers: Implement dynamic loop perforation techniques in resource-constrained devices to balance performance and accuracy under thermal constraints.
Authors
Amna Hassan, Paul Danielse, Anuj Pathania, Andy Pimentel, Semeen Rehman
Abstract
Thermal constraints in heterogeneous multi-core systems can significantly degrade performance, as conventional techniques such as dynamic voltage and frequency scaling (DVFS) and task migration introduce overhead under high utilization. Approximate computing, particularly loop perforation, can mitigate this overhead by reducing computation at the cost of accuracy loss. However, existing approaches typically apply a uniform perforation rate across all loops, ignoring loop-specific sensitivity. We propose AMAX, an accuracy-aware multi-loop asymmetric perforation framework that reduces the combinational design space through loop-wise profiling and multi-objective configuration selection, and dynamically adapts the resulting configurations at runtime according to performance requirements. Experiments on PARSEC benchmarks show that {\em AMAX} achieves up to 0.85 improvement in normalized accuracy over symmetric perforation for the same performance targets.