GemStone: Hardware-Validated CPU Performance and Energy Modelling

Gemstone Power Modelling Tools for gem5

GemStone is a suite of five software tools that, together, identify and evaluate the sources of error in gem5 models against a reference hardware (HW) platform by utilising statistical and machine learning approaches. Independently, the tools serve many purposes:

  1. Automating the running of experiments and collection of Performance Monitoring Counters (and temperature and power on supported platforms) on Arm-based devices (ARMv7 and ARMv8);
  2. Automating the running of gem5 simulations and aggregating the results;
  3. Applying Powmon power models to the hardware results;
  4. Applying Powmon power models to gem5 simulation data;
  5. Combining the HW and gem5 model characterisations and applying statistical and machine learning techniques to identify and evaluate sources of error in the gem5 models;
  6. Conducting power analysis to both the gem5 data and HW data and evaluating the performance, power and energy scaling across various DVFS level and different HMP core types.

An interactive results visualiser dynamically creates detailed graphs and tables of the results when the results files are drag-and-dropped into the webpage – click on the links under ‘View Existing Results’ to see examples.

GemStone Tool Suite

GemStone Tool Suite

More details of our methodology can be found in the following publications:

M J. Walker, S. Diestelhorst, S. Bischoff, G. V. Merrett and B. Al-Hashimi Hardware-Validated CPU Performance and Energy Modelling in IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), April 2018.

For more information, and to download the tool, visit

Related Projects

Related Publications


Related News

Related Theses