Unified architecture level energy-efficiency metric
Victor V. Zyuban
Abstract
Victor V. Zyuban
Abstract
The development of power-efficient microprocessors presents the need to consider power consumption at early stages of design, particularly at the ISA and microarchitecture definition stages, where the potential for power savings is more significant than at lower-level stages, and the opportunity for making power-performance tradeoffs is the largest. Design modifications to the ISA and microarchitecture, however, affect most (if not all) parameters of the design, including architectural speed, code density, clocking rate and power. A reliable metric is required to make knowledgeable power-performance tradeoffs in this multi-dimensional space. This paper derives a unified energy-efficiency metric for evaluating ISA and microarchitecture features, which subsumes other commonly used power-performance metrics as special cases of a more general equation. This new metric is derived based on an analysis of a multi-dimensional power optimization problem, and the resulting formula involves only relative changes in the characteristics of a processor, enabling its application at the early stages of the design.
OpenAlex reports 18 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
The development of power-efficient microprocessors presents the need to consider power consumption at early stages of design, particularly at the ISA and microarchitecture definition stages, where the potential for power savings is more significant than at lower-level stages, and the opportunity for making power-performance tradeoffs is the largest. Design modifications to the ISA and microarchitecture, however, affect most (if not all) parameters of the design, including architectural speed, code density, clocking rate and power. A reliable metric is required to make knowledgeable power-performance tradeoffs in this multi-dimensional space. This paper derives a unified energy-efficiency metric for evaluating ISA and microarchitecture features, which subsumes other commonly used power-performance metrics as special cases of a more general equation. This new metric is derived based on an analysis of a multi-dimensional power optimization problem, and the resulting formula involves only relative changes in the characteristics of a processor, enabling its application at the early stages of the design.
Key concepts: Microarchitecture, Metric (unit), Computer science, Performance metric, Power (physics), Design space exploration, Efficient energy use, Energy consumption