
Build Power Profiling Tool for Modern CPUs
Author(s) -
Naif Aljabri and Osama Abulnaja Naif Aljabri and Osama Abulnaja
Publication year - 2019
Language(s) - English
Resource type - Journals
ISSN - 1658-6336
DOI - 10.4197/comp.8-1.2
Subject(s) - profiling (computer programming) , power consumption , computer science , code (set theory) , power demand , power (physics) , embedded system , real time computing , operating system , programming language , physics , set (abstract data type) , quantum mechanics
. Reduce the application power consumption is one of the main challenges for the HPCcommunity. Code power profilers are very important for researchers to identify the performancebottlenecks and power consumption for their code. Most of the modern CPUs are equipped with abuilt-in sensor to allow researchers and HPC engineers to estimate the power consumption of therunning applications. To estimate the power consumption for any piece of code running on CPU,you need to eliminate the confounding factors as possible and run the code many times until theaverage converge. The reason for that is the environment, which has the OS and other processesand services running at the same time with your code and may report incorrect power readings. Inthis paper, we build a power profiler tool, which saves the researcher time by running andprofiling different pieces of code with different types of workloads, and keeps running until theaverage converge. Furthermore, we identify and eliminate the environment confounding factorwhich saves the researcher time and gives a realistic result for power consumption experiments.