Output details
11 - Computer Science and Informatics
University of Edinburgh
A Predictive Model for Dynamic Microarchitectural Adaptivity Control
<08> Originality: This paper is the first to show how the internal microarchitecture can be adapted on the fly to meet any unseen application's requirement. It uses any offline learnt predictor; is remarkably accurate and delivers significant savings over non-adaptive schemes.
Significance: With the rise of dark silicon, the ability to adapt hardware energy and performance is critical. Micro is the premier conference in computer microarchitecture. The acceptance rate is 17%: this is one of very few papers from the UK. HiPEAC Paper Award recipient.
Rigour: This paper relied on a massive number of simulations. Cross-validation provided statistically significant results.