Reducing pipeline energy demands with local DVS and dynamic retiming

ABSTRACT

The quadratic relationship between voltage and energy has made dynamic voltage scaling (DVS) one of the most powerful techniques to reduce system power demands. Recently, techniques such as Razor DVS, voltage overscaling, and Intelligent Energy Management have emerged as approaches to further reduce voltage by eliminating costly voltage margins inserted into traditional designs to ensure always-correct operation. The degree to which a global voltage controller can shave voltage margins is limited by imbalances in pipeline stage latency. Since all pipeline stages share the same voltage, the stage exercising the longest critical path will define the overall voltage of the system, even if other stages could potentially run at lower voltages. In this paper, we evaluate two local tuning mechanisms in the context of Razor DVS, a local voltage controller scheme that allows each pipeline stages it's own voltage level, and a lower cost dynamic retiming scheme that incorporates per-stage clock delay elements to allow longer-latency pipeline stages to "borrow" time from shorter-latency stages.

Using simulation, we draw two key insights from our study. First, mitigating pipeline stage imbalances renders additional DVS energy savings. A Razor pipeline design with dynamic retiming finds an additional 12% energy savings over global voltage control (resulting in an overall energy savings of more than 28% compared to fully-margined DVS). Second, we demonstrate that imbalances arise not only from design factors, but also from run-time characteristics. As the program (or program phase) changes, we see different logic paths in multiple stages exercised frequently, necessitating a dynamic fine-tuning of local control. This result suggests that even well-balanced pipelines could benefit from dynamic retiming.

REFERENCES

• 1: Dan Ernst , Nam Sung Kim , Shidhartha Das , Sanjay Pant , Rajeev Rao , Toan Pham , Conrad Ziesler , David Blaauw , Todd Austin , Krisztian Flautner , Trevor Mudge, Razor: A Low-Power Pipeline Based on Circuit-Level Timing Speculation, Proceedings of the 36th annual IEEE/ACM International Symposium on Microarchitecture, p.7, December 03-05, 2003
• 2: Todd Austin , Eric Larson , Dan Ernst, SimpleScalar: An Infrastructure for Computer System Modeling, Computer, v.35 n.2, p.59-67, February 2002 [doi>10.1109/2.982917]
• 3: Seokwoo Lee , Shidhartha Das , Valeria Bertacco , Todd Austin , David Blaauw , Trevor Mudge, Circuit-aware architectural simulation, Proceedings of the 41st annual conference on Design automation, June 07-11, 2004, San Diego, CA, USA [doi>10.1145/996566.996656]
• 4: David Harris, Skew-tolerant circuit design, Morgan Kaufmann Publishers Inc., San Francisco, CA, 2000
• 5: Kerry Bernstein , Keith M. Carrig , Christopher M. Durham , Patrick R. Hansen , David Hogenmiller , Edward J. Nowak , Norman J. Rohrer, High speed CMOS design styles, Kluwer Academic Publishers, Norwell, MA, 1998
• 6: Synopsis Corporation, "PrimeTime", http://www.synopsys.com/products/analysis/analysis.html.
• 7: Trevor Pering , Tom Burd , Robert Brodersen, The simulation and evaluation of dynamic voltage scaling algorithms, Proceedings of the 1998 international symposium on Low power electronics and design, p.76-81, August 10-12, 1998, Monterey, California, United States [doi>10.1145/280756.280790]
• 8: Trevor Mudge, Power: A First-Class Architectural Design Constraint, Computer, v.34 n.4, p.52-58, April 2001 [doi>10.1109/2.917539]
• 9: T. Burd. et, al "A Dynamic Voltage Scaled Microprocessor System", IEEE Journal of Solid-State Circuits, Vol 35, No. 11, November 2000.
• 10: Abram P. Dancy , Rajeevan Amirtharajah , Anantha P. Chandrakasan, High-efficiency multiple-output DC-DC conversion for low-voltage systems, IEEE Transactions on Very Large Scale Integration (VLSI) Systems, v.8 n.3, p.252-263, June 2000 [doi>10.1109/92.845892]
• 11: K. Sakallah, T. Mudge, and O. Olukotun. "checkTc and MinTc: Timing Veri-fication and Optimal Clocking of Synchronous Digital Circuits", 1990 IEEE.
• 12: T. Njølstad. et.al "A Socket Interface For GALS Using Locally Dynamic Voltage Scaling For Rate-Adaptive Energy Saving", IEEE 2001.
• 13: Greg Semeraro , David H. Albonesi , Steven G. Dropsho , Grigorios Magklis , Sandhya Dwarkadas , Michael L. Scott, Dynamic frequency and voltage control for a multiple clock domain microarchitecture, Proceedings of the 35th annual ACM/IEEE international symposium on Microarchitecture, November 18-22, 2002, Istanbul, Turkey
• 14: Aseem Agarwal , David Blaauw , Vladimir Zolotov, Statistical Timing Analysis for Intra-Die Process Variations with Spatial Correlations, Proceedings of the 2003 IEEE/ACM international conference on Computer-aided design, p.900, November 09-13, 2003 [doi>10.1109/ICCAD.2003.130]
• 15: Vikas Agarwal , M. S. Hrishikesh , Stephen W. Keckler , Doug Burger, Clock rate versus IPC: the end of the road for conventional microarchitectures, Proceedings of the 27th annual international symposium on Computer architecture, p.248-259, June 2000, Vancouver, British Columbia, Canada
• 16: Timothy Sherwood , Erez Perelman , Greg Hamerly , Brad Calder, Automatically characterizing large scale program behavior, Proceedings of the 10th international conference on Architectural support for programming languages and operating systems, October 05-09, 2002, San Jose, California
• 17: R. Gonzalez, B. Gordon, and M. Horowitz, "Supply and Threshold Voltage Scaling for Low Power CMOS", IEEE JSSC, 32 (8), August 1997.

INDEX TERMS

Primary Classification:
  C. Computer Systems Organization
  C.0 GENERAL
      Subjects: System architectures

General Terms:
Design

Keywords:
dynamic retiming with global DVS, local DVS, razor