DocumentCode :
3383087
Title :
Improvement on branch scheduling for VLIW architecture
Author :
Bao, Lidan ; Wang, Hongmei ; Zhang, Tiejun ; Wang, Donghui ; Hou, Chaohuan
Author_Institution :
Digital Syst. Integration Lab., Inst. of Acoust., Beijing, China
fYear :
2011
fDate :
25-28 Oct. 2011
Firstpage :
723
Lastpage :
726
Abstract :
Branch scheduling is an effective instruction scheduling technique to minimize branch penalty which is especially exhausting for VLIW architectures. In order to find out the most proper instructions for branch delay slots, the proposed algorithm built a cost model to evaluate candidate instructions for optimal choice. And a flow-sensitive data analysis framework is used to improve the accuracy of the scheme. The experimental result shows that, the proposed branch scheduling scheme reduces the executed cycles by 15% in average.
Keywords :
parallel architectures; scheduling; VLIW architecture; branch delay slots; branch penalty; branch scheduling; cost model; flow-sensitive data analysis framework; instruction scheduling; Delay; VLIW architecture; branch delay slots; branch scheduling; dependence analysis;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
ASIC (ASICON), 2011 IEEE 9th International Conference on
Conference_Location :
Xiamen
ISSN :
2162-7541
Print_ISBN :
978-1-61284-192-2
Electronic_ISBN :
2162-7541
Type :
conf
DOI :
10.1109/ASICON.2011.6157307
Filename :
6157307
Link To Document :
بازگشت