Free CPU modelling tool captures mobile device power data
Multi-core systems, with several processor cores embedded on a single silicon die, are common in a wide range of markets, such as smart phones and other mobile devices. It is anticipated that future ‘many-core’ systems will use hundreds or even thousands of embedded processing cores.
Accurately estimating central processing unit (CPU) power consumption is a key requirement for controlling the CPUs – for implementing energy saving techniques for example – and also for exploring the CPU design space and architecture.
Researchers from Electronics and Computer Science at Southampton, working on the UK’s EPSRC-funded PRiME research programme, have developed a tool that uses models built and validated from real, measured data from an actual device which means that the tool accuracy is known and the power figures can be trusted.
“Obtaining accurate data from mobile devices can be challenging and more time-consuming than using a simulator or desktop or server devices. For this reason, we are making available our experimental platform software tools which allow workloads to be automatically run on a mobile device. Metrics such as Performance Monitoring Counters (PMCs), temperature, CPU utilisation, CPU power and CPU voltage are collected and analyzed to produce accurate power models,” explained Gerry Scott, PRiME Programme Impact and Collaboration Manager.
The PRiME power modelling software allows users to automate the model building methodology for their hardware of choice, producing accurate and stable models. The software also provides power models for specific CPUs, for example quad core Cortex-A7 and quad core Cortex-A15, which can be used as reference models for this hardware.