DocumentCode :
131301
Title :
Back-annotation of gate-level power properties into system level descriptions
Author :
Ghohroud, Najmeh Farajipour ; Navabi, Zainalabedin
Author_Institution :
Electr. & Comput. Eng, Univ. of Tehran, Tehran, Iran
fYear :
2014
fDate :
22-25 June 2014
Firstpage :
237
Lastpage :
240
Abstract :
It is important to accurately estimate power consumption in early stages of the design in order to avoid costly redesign. System level design consists of defining transaction level (TL) communications between various components that can be either processors or cores described at system level using high level languages such as SystemC. An important component of a design at the system level is a processor, the accurate power estimation of which is necessary. In this paper we propose a new power estimation method with high accuracy and relatively fast simulation for general purpose processors. The average accuracy achieved is about 90% while the performance is better than the state-based estimation methods. Our proposed method is a hybrid method which uses the concepts of instruction-based, state-based and activity-based power estimation.
Keywords :
integrated circuit design; microprocessor chips; power consumption; SystemC; TL communications; activity-based power estimation; back-annotation; gate-level power properties; general purpose processors; high level languages; instruction-based power estimation; power consumption; power estimation method; state-based estimation methods; state-based power estimation; system level descriptions; transaction level communications; Accuracy; Estimation error; Logic gates; Power demand; Program processors; Switches; General Purpose Processors; Keyboards: Measurement; Macro Modeling; Power Estimation; System Level Design; Transaction level Communication;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
New Circuits and Systems Conference (NEWCAS), 2014 IEEE 12th International
Conference_Location :
Trois-Rivieres, QC
Type :
conf
DOI :
10.1109/NEWCAS.2014.6934027
Filename :
6934027
Link To Document :
بازگشت