Output-Pattern Directed Decomposition for Low Power Design

碩士 === 國立清華大學 === 資訊工程學系 === 91 === In this thesis, we present an output-pattern directed circuit decomposition to reduce power consumption. We observed that in some circuits, highly active output-pattern falls into only a few patterns.Based on this observation, we propose a decomposition...

Full description

Bibliographic Details
Main Authors: Chi-Wei Hu, 胡琦偉
Other Authors: TingTing Hwang
Format: Others
Language:en_US
Published: 2003
Online Access:http://ndltd.ncl.edu.tw/handle/61134944501846040388
id ndltd-TW-091NTHU0392038
record_format oai_dc
spelling ndltd-TW-091NTHU03920382016-06-22T04:26:24Z http://ndltd.ncl.edu.tw/handle/61134944501846040388 Output-Pattern Directed Decomposition for Low Power Design 輸出導向電路分割之低功率設計 Chi-Wei Hu 胡琦偉 碩士 國立清華大學 資訊工程學系 91 In this thesis, we present an output-pattern directed circuit decomposition to reduce power consumption. We observed that in some circuits, highly active output-pattern falls into only a few patterns.Based on this observation, we propose a decomposition architecture and an algorithm to synthesize the decomposed logic. In the decomposed architecture, one small part is used to compute a few highly occurred output-patterns, and the other large part is used for all the other infrequently occurred output-patterns. Consequently, most of time, only the small part of the circuit is active so as to reduce power consumption. In addition, we propose to use an OBDD-based heuristic algorithm to compute the output-pattern frequency in a very short CPU time. Several MCNC benchmark circuits were tested to show the performance. The results shows that the proposed method can achieve 64.9% reduction in power with little area increase as compared to circuits without decomposition. TingTing Hwang 黃婷婷 2003 學位論文 ; thesis 35 en_US
collection NDLTD
language en_US
format Others
sources NDLTD
description 碩士 === 國立清華大學 === 資訊工程學系 === 91 === In this thesis, we present an output-pattern directed circuit decomposition to reduce power consumption. We observed that in some circuits, highly active output-pattern falls into only a few patterns.Based on this observation, we propose a decomposition architecture and an algorithm to synthesize the decomposed logic. In the decomposed architecture, one small part is used to compute a few highly occurred output-patterns, and the other large part is used for all the other infrequently occurred output-patterns. Consequently, most of time, only the small part of the circuit is active so as to reduce power consumption. In addition, we propose to use an OBDD-based heuristic algorithm to compute the output-pattern frequency in a very short CPU time. Several MCNC benchmark circuits were tested to show the performance. The results shows that the proposed method can achieve 64.9% reduction in power with little area increase as compared to circuits without decomposition.
author2 TingTing Hwang
author_facet TingTing Hwang
Chi-Wei Hu
胡琦偉
author Chi-Wei Hu
胡琦偉
spellingShingle Chi-Wei Hu
胡琦偉
Output-Pattern Directed Decomposition for Low Power Design
author_sort Chi-Wei Hu
title Output-Pattern Directed Decomposition for Low Power Design
title_short Output-Pattern Directed Decomposition for Low Power Design
title_full Output-Pattern Directed Decomposition for Low Power Design
title_fullStr Output-Pattern Directed Decomposition for Low Power Design
title_full_unstemmed Output-Pattern Directed Decomposition for Low Power Design
title_sort output-pattern directed decomposition for low power design
publishDate 2003
url http://ndltd.ncl.edu.tw/handle/61134944501846040388
work_keys_str_mv AT chiweihu outputpatterndirecteddecompositionforlowpowerdesign
AT húqíwěi outputpatterndirecteddecompositionforlowpowerdesign
AT chiweihu shūchūdǎoxiàngdiànlùfēngēzhīdīgōnglǜshèjì
AT húqíwěi shūchūdǎoxiàngdiànlùfēngēzhīdīgōnglǜshèjì
_version_ 1718319344228761600