<link rel="stylesheet" href="styles.cf390c6bfbe7555f.css"> Skip to main content

Low-power branch prediction

dc.citation.epage217en_US
dc.citation.spage211en_US
dc.contributor.authorHu, YCen_US
dc.contributor.authorChiao, WHen_US
dc.contributor.authorShann, JJJen_US
dc.contributor.authorChung, CPen_US
dc.contributor.authorChen, WFen_US
dc.contributor.department資訊工程學系zh_TW
dc.contributor.departmentDepartment of Computer Scienceen_US
dc.date.accessioned2014-12-08T15:25:28Z
dc.date.available2014-12-08T15:25:28Z
dc.date.issued2005en_US
dc.description.abstractLow-power design has gained much attention recently, especially for computing on battery-powered equipments. Reducing BTB (branch target buffer) accesses is an effective way to reduce processor power consumption, since BTB consumes a significant portion of power in a processor. In this paper, we propose two approaches to reduce BTB accesses. The first approach expects the distance of every two dynamic branch instructions to be a constant n, where n can be statically profiled, and forces BTB to repose for n instructions after a BTB hit. The second approach dynamically predicts the address of the next branch instruction, and accesses BTB only on the predicted address. Multimedia/DSP benchmarks are used in our evaluation. Experimental results show that these methods can potentially reduce 22-033% of all BTB accesses.en_US
dc.identifier.isbn1-932415-54-8en_US
dc.identifier.journalCDES '05: Proceedings of the 2005 International Conference on Computer Designen_US
dc.identifier.urihttps://ir.lib.nycu.edu.tw/handle/11536/17871
dc.identifier.wosnumberWOS:000236728000031
dc.language.isoen_USen_US
dc.subjectlow-poweren_US
dc.subjectbranch predictionen_US
dc.subjectbranch target bufferen_US
dc.titleLow-power branch predictionen_US
dc.typeProceedings Paperen_US

Files

License bundle

Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed to upon submission
Description: