Power Profiling: From Measurements to Simulation Models
Energy efficiency is currently one of the biggest challenges in
modern computer design. High power density limits further performance
growth, and energy efficiency affects both the power bill for supercomputers
and battery lifetime for embedded devices. A better understanding of energy
efficiency during the design stage eases development of better architectures.
In this thesis, we investigate energy consumption and architectural
properties of an ARM Cortex-A9 processor. Further, this information is used
to create a tool for estimating its power consumption through simulation.
Instruction level energy consumption is determined through measurements and
experiments on real hardware, which are further mapped to certain
architectural events found in the gem5 simulator. The tool utilizes these
events together with a simulator trace log and outputs a representation of
energy consumption over time.
This method can be applied during the development process at the simulator
level, while traditional methods typically involves hardware synthesis. The
results show that this tool can estimate energy consumption with margin of
error of 5 % on general workloads, and is able to identify power
consumption trends throughout a program.
Place, publisher, year, edition, pages
Institutt for datateknikk og informasjonsvitenskap , 2014. , 85 p.
IdentifiersURN: urn:nbn:no:ntnu:diva-26806Local ID: ntnudaim:11481OAI: oai:DiVA.org:ntnu-26806DiVA: diva2:751714
Tufte, Gunnar, FørsteamanuensisDjupdal, Asbjørn