P
Paul M. Carpenter
Researcher at Barcelona Supercomputing Center
Publications - 64
Citations - 801
Paul M. Carpenter is an academic researcher from Barcelona Supercomputing Center. The author has contributed to research in topics: Efficient energy use & Energy consumption. The author has an hindex of 13, co-authored 60 publications receiving 673 citations. Previous affiliations of Paul M. Carpenter include Birmingham–Southern College & Polytechnic University of Catalonia.
Papers
More filters
Proceedings ArticleDOI
Supercomputing with commodity CPUs: are mobile SoCs ready for HPC?
TL;DR: The trend in mobile SoC performance is analyzed, comparing it with the similar trend in the 1990s, and a first answer as to whether mobile SoCs are ready for HPC is given.
Proceedings ArticleDOI
Hipster: Hybrid Task Manager for Latency-Critical Cloud Workloads
TL;DR: Hipster is introduced, a technique that combines heuristics and reinforcement learning to manage latency-critical workloads and achieves its goal by exploring heterogeneous multi-cores and dynamic voltage and frequency scaling (DVFS).
Proceedings ArticleDOI
EUROSERVER: Energy Efficient Node for European Micro-Servers
Yves Durand,Paul M. Carpenter,Stefano Adami,Angelos Bilas,Denis Dutoit,Alexis Farcy,Georgi Gaydadjiev,John Goodacre,Manolis Katevenis,Manolis Marazakis,Emil Matus,Iakovos Mavroidis,John Thomson +12 more
TL;DR: The EUROSERVER device will embed multiple silicon "chiplets" on an active silicon interposer, which is pioneering a system architecture approach that allows specialized silicon devices to be built even for low-volume markets where NRE costs are currently prohibitive.
Proceedings ArticleDOI
Twig: Multi-Agent Task Management for Colocated Latency-Critical Cloud Services
TL;DR: The results show that Twig outperforms prior works in reducing energy usage by up to 38% while achieving up to 99% QoS guarantee for latency-critical services.
Proceedings ArticleDOI
Mapping stream programs onto heterogeneous multiprocessor systems
TL;DR: A novel definition of connectedness is introduced that enables the algorithm to model the capabilities of the compiler, and the algorithm uses convexity and connectedness constraints to produce partitions that are easier to compile and require short pipelines.