Energy Consumption of double vs. single precision with Matlab Mex-files
Abstract
The context is sustainbable software development. We try to find coding alternatives that are economical with regard to energy (electricity) consumption. We want to measure the difference in energy consumption between double and single numeric precision. For that we “outsorce” the main computation kernel from Matlab to C-code (mex-functions). Mex functions are C-compiled functions, readable from within Matlab code.
Supervision
- Jan-Andrea, Bard
Motivation
As a basis serves a numerical computation algorithm for pension funds liabilities. The student will then translate the core computation kernel to C-code and create mex-functions, either using the C-coder toolbox from Matlab or by compiling mex-files directly from C-code. Finally, we measure the difference in energy consumption between a double precision algorithm and its single precision counterpart. The technology to do the energy measurements on the CPU is the RAPL interface.
Goal
- Writing Matlab/C-code programm fragments.
- Executing experiments on a dedicated server from the university of Bern.
- Analyse results and data.
- Write a report, presenting data, analysis and insights.
- Read articles about related work.
Requirements
- (Basic) knowledge in Matlab and C
- Basics in Linux