Output details
11 - Computer Science and Informatics
University of Edinburgh
Exploring and Predicting the Effects of Microarchitectural Parameters and Compiler Optimizations on Performance and Energy
<08> Originality: This presents a novel approach to designing embedded processors by considering compiler optimisations early on in the design space and by developing a machine-learning model that speeds up this process.
Significance: The design methodology of new embedded microprocessors does not take into account software optimisations leading to suboptimal designs. Designing new processors requires large numbers of simulations. Adding software optimisations on top of this would typically make this process even slower. The approach developed in this work alleviates this problem, resulting in more efficient microprocessors.
Rigour: Published in a world-class journal for embedded systems (ACM TECS).