Dynamic thermal management via architectural adaptation

ABSTRACT

Exponentially rising cooling/packaging costs due to high power density call for architectural and software-level thermal management. Dynamic thermal management (DTM) techniques continuously monitor the on-chip processor temperature. Appropriate mechanisms (e.g., dynamic voltage or frequency scaling (DVFS), clock gating, fetch gating, etc.) are engaged to lower the temperature if it exceeds a threshold. However, all these mechanisms incur significant performance penalty. We argue that runtime adaptation of micro-architectural parameters, such as instruction window size and issue width, is a more effective mechanism for DTM. If the architectural parameters can be tailored to track the available instruction-level parallelism of the program, the temperature is reduced with minimal performance degradation. Moreover, synergistically combining architectural adaptation with DVFS and fetch gating can achieve the best performance under thermal constraints. The key difficulty in using multiple mechanisms is to select the optimal configuration at runtime for time varying workloads. We present a novel software-level thermal management framework that searches through the configuration space at regular intervals to find the best performing design point that is thermally safe. The central components of our framework are (1) a neural-network based classifier that filters the thermally unsafe configurations, (2) a fast performance prediction model for any configuration, and (3) an efficient configuration space search algorithm. Experimental results indicate that our adaptive scheme achieves 59% reduction in performance overhead compared to DVFS and 39% reduction in overhead compared to DVFS combined with fetch gating.

REFERENCES

• 1: Matlab Neural Network Toolbox. www.mathworks.com/access/helpdesk/help/pdf_doc/nnet/nnet.pdf.
• 2: T. Austin, E. Larson, and D. Ernst. SimpleScalar: An Infrastructure for Computer System Modeling. IEEE Computer, 35(2), 2002.
• 3: D. Brooks, V. Tiwari, and M. Martonosi. Wattch: A Framework for Architectural-level Power Analysis and Optimizations. In ISCA 2000.
• 4: A. Buyuktosunoglu et al. A Circuit Level Implementation of an Adaptive Issue Queue for Power-Aware Microprocessors. In GLSVLSI, 2001.
• 5: J. Choi et al. Thermal-aware task scheduling at the system software level. In ISLPED 2007.
• 6: A. Cohen et al. On Estimating Optimal Performance of CPU Dynamic Thermal Management. IEEE Computer Architecture Letters, 2, 2003.
• 7: H. Hanson et al. Thermal response to DVFS: Analysis with an Intel Pentium M. In ISLPED 2007.
• 8: H. Jung, P. Rong, and M. Pedram. Stochastic modeling of a thermally-managed multi-core system. In DAC 2008.
• 9: T. S. Karkhanis and J. E. Smith. A First-Order Superscalar Processor Model. In ISCA 2004.
• 10: A. Kumar et al. HybDTM: A Coordinated Hardware-Software Approach for Dynamic Thermal Management. In DAC 2006.
• 11: E. Perelman, G. Hamerly, and B. Calder. Picking Statistically Valid and Early Simulation Points. In PACT 2003.
• 12: D. Ponomarev, G. Kucuk, and K. Ghose. Reducing Power Requirements of Instruction Scheduling through Dynamic Allocation of Multiple Datapath Resources. In MICRO 2001.
• 13: R. Rao, S. Vrudhula, and K. Berezowski. Analytical results for design space exploration of multi-core processors employing thread migration. In ISLPED 2008.
• 14: N. N. Schraudolph. A Fast, Compact Approximation of the Exponential Function. In Technical Report INDISA-07-98.
• 15: K. Skadron. Hybrid Architectural Dynamic Thermal Management. In DATE 2004.
• 16: K. Skadron et al. Temperature-aware Microarchitecture: Modeling and Implementation. ACM TACO, 1(1), 2004.
• 17: J. Srinivasan and S. V. Adve. Predictive Dynamic Thermal Management for Multimedia Applications. In ICS 2003.
• 18: X. Zhou, C. Yu, and P. Petrov. Compiler-driven register re-assignment for register file power-density and temperature reduction. In DAC 2008.

INDEX TERMS

Primary Classification:
  C. Computer Systems Organization
  C.1 PROCESSOR ARCHITECTURES
      C.1.0 General

General Terms:
Design, Performance, Reliability

Keywords:
architecture adaptation, dynamic thermal management