The Community for Technology Leaders
Green Image
Issue No. 02 - July-Dec. (2012 vol. 11)
ISSN: 1556-6056
pp: 37-40
Hu Chen , National University of Defence Technology, PR.China.
Xiaowen Chen , National University of Defence Technology, PR.China.
Jianghua Wan , National University of Defence Technology, PR.China.
Kai Zhang , National University of Defence Technology, PR.China.
Shuming Chen , National University of Defence Technology, PR.China.
Yaohua Wang , National University of Defence Technology, PR.China.
Haibo Wang , National University of Defence Technology, PR.China.
ABSTRACT
SIMD architectures are less efficient for applications with the diverse control-flow behavior, which can be mainly attributed to the requirement of the identical control-flow. In this paper, we propose a novel instruction shuffle scheme that features an efficient control-flow handling mechanism. The cornerstones are composed of a shuffle source instruction buffer array and an instruction shuffle unit. The shuffle unit can concurrently deliver instructions of multiple distinct control-flows from the instruction buffer array to eligible SIMD lanes. Our instruction shuffle scheme combines the best attributes of both the SIMD and MIMD execution paradigms. Experimental results show that, an average performance improvement of 86% can be achieved, at a cost of only 5.8% area overhead.
INDEX TERMS
Instruction sets, Process control, Resource management, Vectors, Scalability, SIMD, Kernel, Process control, Resource management, Vectors, Scalability, Arrays, data dependent control-flow, instruction shuffle, instruction buffer array
CITATION
Hu Chen, Xiaowen Chen, Jianghua Wan, Kai Zhang, Shuming Chen, Yaohua Wang, Haibo Wang, "Instruction Shuffle: Achieving MIMD-like Performance on SIMD Architectures", IEEE Computer Architecture Letters, vol. 11, no. , pp. 37-40, July-Dec. 2012, doi:10.1109/L-CA.2011.34
165 ms
(Ver )