A power-aware branch predictor by accessing the BTB selectively

Cheol Hong Kim, Sung Woo Chung, Chu Shik Jhon

Research output: Contribution to journalArticlepeer-review

3 Citations (Scopus)

Abstract

Microarchitects should consider power consumption, together with accuracy, when designing a branch predictor, especially in embedded processors. This paper proposes a power-aware branch predictor, which is based on the gshare predictor, by accessing the BTB (Branch Target Buffer) selectively. To enable the selective access to the BTB, the PHT (Pattern History Table) in the proposed branch predictor is accessed one cycle earlier than the traditional PHT if the program is executed sequentially without branch instructions. As a side effect, two predictions from the PHT are obtained through one access to the PHT, resulting in more power savings. In the proposed branch predictor, if the previous instruction was not a branch and the prediction from the PHT is untaken, the BTB is not accessed to reduce power consumption. If the previous instruction was a branch, the BTB is always accessed, regardless of the prediction from the PHT, to prevent the additional delay/accuracy decrease. The proposed branch predictor reduces the power consumption with little hardware overhead, not incurring additional delay and never harming prediction accuracy. The simulation results show that the proposed branch predictor reduces the power consumption by 29-47%.

Original languageEnglish
Pages (from-to)607-614
Number of pages8
JournalJournal of Computer Science and Technology
Volume20
Issue number5
DOIs
Publication statusPublished - 2005 Sep
Externally publishedYes

Keywords

  • BTB
  • Branch predictor
  • Embedded processor
  • Low power design
  • PHT

ASJC Scopus subject areas

  • Software
  • Theoretical Computer Science
  • Hardware and Architecture
  • Computer Science Applications
  • Computational Theory and Mathematics

Fingerprint Dive into the research topics of 'A power-aware branch predictor by accessing the BTB selectively'. Together they form a unique fingerprint.

Cite this