Code transformations for energy efficiency; a decoupled access-execute approach
Speaker: Konstantinos Koukos, Postdoc, KTH
The end of Dennard scaling is threatening the effectiveness of DVFS in future architectures. This necessitates new approaches to exploit the non-linear relationship between performance and energy efficiency of memory-bound application-regions. For that, we propose the decoupled access-execute (DAE) paradigm. DAE transforms regions of interest in two coarse-grain phases: the access-phase and the execute-phase, which we can independently DVFS. The access-phase is intended to prefetch the data in the cache, and is therefore expected to be predominantly memory-bound, while the execute-phase runs immediately after the access-phase (that has warmed-up the cache) and is therefore expected to be compute-bound. Our approach, achieves good energy savings (on average 25% lower EDP) without performance degradation, as opposed to contemporary DVFS techniques that trade-off performance for energy-savings. Furthermore, DAE increases the memory level parallelism (MLP) of memory-bound regions, which results in performance improvements of memory-bound applications.
Konstantinos is currently a PostDoc researcher at KTH, working on the Model-based Computing Systems (MCS) group. He has a PhD from Uppsala University on "efficient execution paradigms for heterogeneous architectures", and a master specialization on parallel programming and code optimization for heterogeneous systems. His research interests focus on code optimizations for energy efficiency, and optimizations to better exploit the memory hierarchy of heterogeneous systems.
Slides: K.Koukos (pdf 5,1 MB)