Performance Counter-Based Power and Temperature Prediction


Performance Counter-Based Power and Temperature Prediction

This project builds on performance counter-based power and temperature prediction work that is being performed in Cornell. The work consists of a validated the power prediction model for an AMD four-core Phenom and an Intel Core 2 Quad Q6600. The temperature model is being actively developed.

Proposed Work

The work for project has two directions

  1. Use of the continuous optimization system developed in Siena to see if we can leverage the information gathered from the performance counters
  2. The addition of similar performance counters to an ARM simulator to see if the approach works equally well for embedded systems

Participants
  • Sally McKee (Chalmers)
  • Sandro Bartolini (Siena)
  • Timothy Jones (Edinburgh)