Reducing pipeline energy demands with local DVS and dynamic retiming

Authors:
Seokwoo Lee;Shidhartha Das;Toan Pham;Todd Austin;David Blaauw;Trevor Mudge
Affiliations:
The University of Michigan, Ann Arbor, MI;The University of Michigan, Ann Arbor, MI;The University of Michigan, Ann Arbor, MI;The University of Michigan, Ann Arbor, MI;The University of Michigan, Ann Arbor, MI;The University of Michigan, Ann Arbor, MI
Venue:
Proceedings of the 2004 international symposium on Low power electronics and design
Year:
2004

Citing 12
Cited 9

The simulation and evaluation of dynamic voltage scaling algorithms

ISLPED '98 Proceedings of the 1998 international symposium on Low power electronics and design
High speed CMOS design styles

High speed CMOS design styles
Clock rate versus IPC: the end of the road for conventional microarchitectures

Proceedings of the 27th annual international symposium on Computer architecture
High-efficiency multiple-output DC-DC conversion for low-voltage systems

IEEE Transactions on Very Large Scale Integration (VLSI) Systems - Special section on low-power electronics and design
Skew-tolerant circuit design

Skew-tolerant circuit design
Automatically characterizing large scale program behavior

Proceedings of the 10th international conference on Architectural support for programming languages and operating systems
Power: A First-Class Architectural Design Constraint

Computer
SimpleScalar: An Infrastructure for Computer System Modeling

Computer
Dynamic frequency and voltage control for a multiple clock domain microarchitecture

Proceedings of the 35th annual ACM/IEEE international symposium on Microarchitecture
Razor: A Low-Power Pipeline Based on Circuit-Level Timing Speculation

Proceedings of the 36th annual IEEE/ACM International Symposium on Microarchitecture
Circuit-aware architectural simulation

Proceedings of the 41st annual Design Automation Conference
Statistical Timing Analysis for Intra-Die Process Variations with Spatial Correlations

Proceedings of the 2003 IEEE/ACM international conference on Computer-aided design

Razor: Circuit-Level Correction of Timing Errors for Low-Power Operation

IEEE Micro
Low-power design methodology for module-wise dynamic voltage and frequency scaling with dynamic de-skewing systems

ASP-DAC '06 Proceedings of the 2006 Asia and South Pacific Design Automation Conference
Energy-efficient embedded software implementation on multiprocessor system-on-chip with multiple voltages

ACM Transactions on Embedded Computing Systems (TECS)
ReCycle:: pipeline adaptation to tolerate process variation

Proceedings of the 34th annual international symposium on Computer architecture
A self-adjusting clock tree architecture to cope with temperature variations

Proceedings of the 2007 IEEE/ACM international conference on Computer-aided design
Process variation tolerant pipeline design through a placement-aware multiple voltage island design style

Proceedings of the conference on Design, automation and test in Europe
A new compensation mechanism for environmental parameter fluctuations in CMOS digital ICs

Microelectronics Journal
Overscaling-friendly timing speculation architectures

Proceedings of the 20th symposium on Great lakes symposium on VLSI
SACTA: a self-adjusting clock tree architecture for adapting to thermal-induced delay variation

IEEE Transactions on Very Large Scale Integration (VLSI) Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

The quadratic relationship between voltage and energy has madedynamic voltage scaling (DVS) one of the most powerful techniquesto reduce system power demands. Recently, techniques such as RazorDVS, voltage overscaling, and Intelligent Energy Management haveemerged as approaches to further reduce voltage by eliminatingcostly voltage margins inserted into traditional designs to ensurealways-correct operation. The degree to which a global voltagecontroller can shave voltage margins is limited by imbalances inpipeline stage latency. Since all pipeline stages share the samevoltage, the stage exercising the longest critical path will definethe overall voltage of the system, even if other stages couldpotentially run at lower voltages. In this paper, we evaluate twolocal tuning mechanisms in the context of Razor DVS, a localvoltage controller scheme that allows each pipeline stages it's ownvoltage level, and a lower cost dynamic retiming scheme thatincorporates per-stage clock delay elements to allow longer-latencypipeline 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 DVSenergy savings. A Razor pipeline design with dynamic retiming findsan additional 12% energy savings over global voltage control(resulting in an overall energy savings of more than 28% comparedto fully-margined DVS). Second, we demonstrate that imbalancesarise not only from design factors, but also from run-timecharacteristics. As the program (or program phase) changes, we seedifferent logic paths in multiple stages exercised frequently,necessitating a dynamic fine-tuning of local control. This resultsuggests that even well-balanced pipelines could benefit fromdynamic retiming.