The BEEBS Benchmark Suite
Introduction to the BEEBS suite
In the course of our energy efficiency research at the University of Bristol, in partnership with Embecosm, we discovered that no suitable benchmark suite exists for exploring the relative energy consumptions of embedded systems. These systems exhibit a wide variety of ISAs, programming paradigms and selections of single- or super-scalar execution, inclusion or lack of floating-point support, and simple or complex memory heirarchies.
In common, however, is a relatively limited memory store and the lack of a guarantee of operating system support.
Therefore, in 2012 as part of a project aimed at identifying energy efficient compiler options, we developed BEEBS, the Bristol/Embecosm Embedded Energy Benchmark Suite. BEEBS is explicity designed to support as wide a range of embedded systems as possible, whilst allowing the efficient exploration of their energy efficiency. Along with BEEBS, we have also developed energy montoring hardware and software infrastructures, that can be replicated by users of BEEBS.
BEEBS was designed primarily to be an open source suite. All of the benchmarks from which it is derived are either free and open or have been re-licensed for BEEBS in an open and free manner. BEEBS is licensed under the GPLv3 license to encourage all users to contribute any additions to BEEBS to the wider community.
The BEEBS suite is stable and is currently (2013-14) being maintained and updated as part of the MAGEEC research project, a collaboration between the University of Bristol and Embecosm Ltd.
Academic staff (University of Bristol)
Industrial staff (Embecosm)
- "BEEBS: Open Benchmarks for Energy Measurements on Embedded Platforms", James Pallister, Simon Hollis, Jeremy Bennett, 23rd August 2013, arXiv:1308.5174 [cs.PF]
- "Identifying Compiler Options to Minimise Energy Consumption for Embedded Platforms", James Pallister, Simon Hollis, Jeremy Bennett, 26th March 2013, arXiv:1303.6485 [cs.PF]
Downloads of BEEBS code, designs and benchmark results
- The BEEBS benchmark sources are available via GitHub repository.
- Measurement Framework
- Designs for the associated energy measurement hardware (GitHub)
- Raw result data files from running BEEBS on the following platforms: ARM Cortex-M0, ARM Cortex-M3, ARM Cortex A8, Adapteva Epiphany (warning: large files)