Presentation 2001/4/6
Multi-Path Execution on SMT for Next Trace Prediction
Masamichi TAKAGI, Kei HIRAKI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Branch misprediction penalty of a processor grows as the pipeline deepens. We employ multi-path execution to reduce the penalty. To reduce the cost of speculative fork or join, we use Simultaneous Multithreading (SMT). To boost the effective fetch rate which decreases on multi-path execution, we use trace as an unit of multi-path execution. Because the processor can follow only limited number of paths simultaneously, we need to choose paths which are most likely to be on the correct path. For this purpose, we introduce a fork policy which predicts the execution probability of a path candidate dynamically. We evalutate the proposed architecture under the proposed policy and others with a simulator. In the best case, the architecture yields 16.0% speedup over the case without multi-path execution. The proposed policy produces more stable speedup than other simple policies on the programs with different characteristics, and the policy produces almost the same speedup as the ideal policies.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) multiple path execution / simultaneous multithreading / SMT / trace cache / trace / confidence
Paper # CPSY2001-10,FTS2001-10
Date of Issue

Conference Information
Committee CPSY
Conference Date 2001/4/6(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Computer Systems (CPSY)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Multi-Path Execution on SMT for Next Trace Prediction
Sub Title (in English)
Keyword(1) multiple path execution
Keyword(2) simultaneous multithreading
Keyword(3) SMT
Keyword(4) trace cache
Keyword(5) trace
Keyword(6) confidence
1st Author's Name Masamichi TAKAGI
1st Author's Affiliation Department of Information Science, Graduate School of Science, The University of Tokyo()
2nd Author's Name Kei HIRAKI
2nd Author's Affiliation Department of Information Science, Graduate School of Science, The University of Tokyo
Date 2001/4/6
Paper # CPSY2001-10,FTS2001-10
Volume (vol) vol.101
Number (no) 2
Page pp.pp.-
#Pages 8
Date of Issue