Paper Info
Title
Complexity Adaptive Branch Predictor for Thermal-Aware 3D Multi-core Processors.
Abstract
High-performance processors usually perform speculative executions for improving the instruction-level parallelism. In the speculative executions, the most important factor determining the performance is the accuracy of the branch predictor. As this accuracy becomes higher, thereby enhancing the instruction-level parallelism, the hardware complexity of the branch prediction logic increases. In the conventional 2D multi-core processor, the branch predictor is not classified as the hot unit. However, in the 3D multi-core processor where the power density is substantially high, the temperature of the complicated branch predictor also becomes high, which causes a negative impact on the processor reliability. To reduce the temperature of the branch predictor in the 3D multi-core processor, we propose the thermal management technique which determines the hardware complexity of the branch predictor in each layer by considering the temperature. According to our experimental results, the proposed complexity adaptive branch predictor technique shows a superior thermal efficiency than the conventional dynamic thermal management technique.
Year
Venue
Keywords
2011
Communications in Computer and Information Science
3D Multi-core Processor,Branch Predictor,Hotspot,Dynamic Thermal Management
DocType
Volume
ISSN
Conference
256
1865-0929
Citations 
PageRank 
References 
0
0.34
0
Authors
5
Name
Order
Citations
PageRank
Jin Woo Ahn100.68
Dong Oh Son2214.19
Hyung Gyu Jeon300.34
Jong-Myon Kim49125.99
Cheol Hong Kim57324.39