Energy characterization based on clustering

Huzefa Mehta, Robert Michael Owens, Mary Jane Irwin

Research output: Contribution to journalArticle

60 Citations (Scopus)

Abstract

We illustrate a new method to characterize the energy dissipation of circuits by collapsing closely related input transition vectors and energy patterns into capacitive coefficients. Energy characterization needs to be done only once for each module (ALU, multiplier etc.,) in order to build a library of these capacitive coefficients. A direct high-level energy simulator or profiler can then use the library of pre-characterized modules and a sequence of input vectors to compute the total energy dissipation. A heuristic algorithm which performs energy clustering under objective constraints has been devised. The worst case running time of this algorithm is O(m3n), where m is the number of simulation points and n is the number of inputs of the circuit. The designer can experiment with the criterion function by setting the appropriate relative error norms to control the `goodness' of the clustering algorithm and the sampling error and confidence level to maintain the sufficiency of representation of each cluster. Experiments on circuits show a significant reduction of the energy table size under a specified criterion function, cluster sampling error and confidence levels.

Original languageEnglish (US)
Pages (from-to)702-707
Number of pages6
JournalProceedings - Design Automation Conference
StatePublished - 1996

Fingerprint

Networks (circuits)
Energy dissipation
Sampling
Heuristic algorithms
Clustering algorithms
Electron energy levels
Simulators
Experiments

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture
  • Control and Systems Engineering

Cite this

Mehta, H., Owens, R. M., & Irwin, M. J. (1996). Energy characterization based on clustering. Proceedings - Design Automation Conference, 702-707.
Mehta, Huzefa ; Owens, Robert Michael ; Irwin, Mary Jane. / Energy characterization based on clustering. In: Proceedings - Design Automation Conference. 1996 ; pp. 702-707.
@article{7b15052750f746c08d8788d418e24da5,
title = "Energy characterization based on clustering",
abstract = "We illustrate a new method to characterize the energy dissipation of circuits by collapsing closely related input transition vectors and energy patterns into capacitive coefficients. Energy characterization needs to be done only once for each module (ALU, multiplier etc.,) in order to build a library of these capacitive coefficients. A direct high-level energy simulator or profiler can then use the library of pre-characterized modules and a sequence of input vectors to compute the total energy dissipation. A heuristic algorithm which performs energy clustering under objective constraints has been devised. The worst case running time of this algorithm is O(m3n), where m is the number of simulation points and n is the number of inputs of the circuit. The designer can experiment with the criterion function by setting the appropriate relative error norms to control the `goodness' of the clustering algorithm and the sampling error and confidence level to maintain the sufficiency of representation of each cluster. Experiments on circuits show a significant reduction of the energy table size under a specified criterion function, cluster sampling error and confidence levels.",
author = "Huzefa Mehta and Owens, {Robert Michael} and Irwin, {Mary Jane}",
year = "1996",
language = "English (US)",
pages = "702--707",
journal = "Proceedings - Design Automation Conference",
issn = "0738-100X",

}

Energy characterization based on clustering. / Mehta, Huzefa; Owens, Robert Michael; Irwin, Mary Jane.

In: Proceedings - Design Automation Conference, 1996, p. 702-707.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Energy characterization based on clustering

AU - Mehta, Huzefa

AU - Owens, Robert Michael

AU - Irwin, Mary Jane

PY - 1996

Y1 - 1996

N2 - We illustrate a new method to characterize the energy dissipation of circuits by collapsing closely related input transition vectors and energy patterns into capacitive coefficients. Energy characterization needs to be done only once for each module (ALU, multiplier etc.,) in order to build a library of these capacitive coefficients. A direct high-level energy simulator or profiler can then use the library of pre-characterized modules and a sequence of input vectors to compute the total energy dissipation. A heuristic algorithm which performs energy clustering under objective constraints has been devised. The worst case running time of this algorithm is O(m3n), where m is the number of simulation points and n is the number of inputs of the circuit. The designer can experiment with the criterion function by setting the appropriate relative error norms to control the `goodness' of the clustering algorithm and the sampling error and confidence level to maintain the sufficiency of representation of each cluster. Experiments on circuits show a significant reduction of the energy table size under a specified criterion function, cluster sampling error and confidence levels.

AB - We illustrate a new method to characterize the energy dissipation of circuits by collapsing closely related input transition vectors and energy patterns into capacitive coefficients. Energy characterization needs to be done only once for each module (ALU, multiplier etc.,) in order to build a library of these capacitive coefficients. A direct high-level energy simulator or profiler can then use the library of pre-characterized modules and a sequence of input vectors to compute the total energy dissipation. A heuristic algorithm which performs energy clustering under objective constraints has been devised. The worst case running time of this algorithm is O(m3n), where m is the number of simulation points and n is the number of inputs of the circuit. The designer can experiment with the criterion function by setting the appropriate relative error norms to control the `goodness' of the clustering algorithm and the sampling error and confidence level to maintain the sufficiency of representation of each cluster. Experiments on circuits show a significant reduction of the energy table size under a specified criterion function, cluster sampling error and confidence levels.

UR - http://www.scopus.com/inward/record.url?scp=0029695157&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0029695157&partnerID=8YFLogxK

M3 - Article

SP - 702

EP - 707

JO - Proceedings - Design Automation Conference

JF - Proceedings - Design Automation Conference

SN - 0738-100X

ER -