DocumentCode
453634
Title
Codes reallocation and prediction for power efficiency in I-cache memory
Author
Xiaoping, Zhu ; Tiow, Tay Teng
Author_Institution
Dept. of Electr. & Comput. Eng., Singapore Nat. Univ.
Volume
1
fYear
2005
fDate
24-0 Oct. 2005
Firstpage
102
Lastpage
105
Abstract
To reduce energy consumption of cache memories is an important problem since this component spends a large portion of energy in a microprocessor. This paper proposes a compiler predicted strategy that dynamically turns off those unused cache lines to save power in architecture level. In our algorithm, the object codes of programs are reallocated in memory address map according to the I-cache structure so that the working sets are reduced in the cache memory. In addition, a few special cache-scaling instructions (CSIs) are added to the object codes to track the working set size. With the information from CSIs and the current system state, a hardware controller implements the decision of replacement algorithm and switching the power of each refill-line. Our experimental results indicate that the compiler-predicted algorithm can reduce 58.9% of energy in a 32K I-cache, while the average performance loss is only 2.7%
Keywords
cache storage; microprocessor chips; program compilers; storage allocation; I-cache memory; cache-scaling instructions; code prediction; code reallocation; compiler-predicted algorithm; hardware controller; memory address map; microprocessor circuit; power efficiency; power switching; replacement algorithm decision; Algorithms; CMOS technology; Cache memory; Energy consumption; Hardware; Microprocessors; Performance loss; Runtime; Switches; System performance;
fLanguage
English
Publisher
ieee
Conference_Titel
ASIC, 2005. ASICON 2005. 6th International Conference On
Conference_Location
Shanghai
Print_ISBN
0-7803-9210-8
Type
conf
DOI
10.1109/ICASIC.2005.1611276
Filename
1611276
Link To Document