Compiler Optimization to Reduce Cache Power with Victim Cache

碩士 === 國立中正大學 === 資訊工程所 === 96 === This thesis proposes an approach to improve memory hierarchy by reducing miss rate and miss penalty for saving power. a victim cache is added for this purpose. The power consumption is reduced by decreasing the access times to level-2 cache caused by the victim cac...

Full description

Bibliographic Details
Main Authors: Cheng-yu Lee, 李承諭
Other Authors: Rong-Guey Chang
Format: Others
Language:en_US
Published: 2008
Online Access:http://ndltd.ncl.edu.tw/handle/42885585653096631889
id ndltd-TW-096CCU05392131
record_format oai_dc
spelling ndltd-TW-096CCU053921312016-05-04T04:25:45Z http://ndltd.ncl.edu.tw/handle/42885585653096631889 Compiler Optimization to Reduce Cache Power with Victim Cache 以編譯器最佳化犧牲者快取之低必v快取技術 Cheng-yu Lee 李承諭 碩士 國立中正大學 資訊工程所 96 This thesis proposes an approach to improve memory hierarchy by reducing miss rate and miss penalty for saving power. a victim cache is added for this purpose. The power consumption is reduced by decreasing the access times to level-2 cache caused by the victim cache. There are two proposed approaches to improve programs: First, cache block controlling can reduce the unnecessary access. Second, victim cache controlling can make a trade off between L2 cache and victim cache on power. The power and performance is improved with a compiler-control and the miss rate is also reduced. The result that s small and simple cache performs better is shown in the experiment. It can improve about 6.5% of execution time and reduce 32.66% of cache miss and decrease 4.19% of power consumption with 2k direct-mapped cache. The approach is implemented on SUIF and Machine SUIF compiler for the analyzing and inserting the controlling instructions and running on the Wattch simulator. Rong-Guey Chang 張榮貴 2008 學位論文 ; thesis 40 en_US
collection NDLTD
language en_US
format Others
sources NDLTD
description 碩士 === 國立中正大學 === 資訊工程所 === 96 === This thesis proposes an approach to improve memory hierarchy by reducing miss rate and miss penalty for saving power. a victim cache is added for this purpose. The power consumption is reduced by decreasing the access times to level-2 cache caused by the victim cache. There are two proposed approaches to improve programs: First, cache block controlling can reduce the unnecessary access. Second, victim cache controlling can make a trade off between L2 cache and victim cache on power. The power and performance is improved with a compiler-control and the miss rate is also reduced. The result that s small and simple cache performs better is shown in the experiment. It can improve about 6.5% of execution time and reduce 32.66% of cache miss and decrease 4.19% of power consumption with 2k direct-mapped cache. The approach is implemented on SUIF and Machine SUIF compiler for the analyzing and inserting the controlling instructions and running on the Wattch simulator.
author2 Rong-Guey Chang
author_facet Rong-Guey Chang
Cheng-yu Lee
李承諭
author Cheng-yu Lee
李承諭
spellingShingle Cheng-yu Lee
李承諭
Compiler Optimization to Reduce Cache Power with Victim Cache
author_sort Cheng-yu Lee
title Compiler Optimization to Reduce Cache Power with Victim Cache
title_short Compiler Optimization to Reduce Cache Power with Victim Cache
title_full Compiler Optimization to Reduce Cache Power with Victim Cache
title_fullStr Compiler Optimization to Reduce Cache Power with Victim Cache
title_full_unstemmed Compiler Optimization to Reduce Cache Power with Victim Cache
title_sort compiler optimization to reduce cache power with victim cache
publishDate 2008
url http://ndltd.ncl.edu.tw/handle/42885585653096631889
work_keys_str_mv AT chengyulee compileroptimizationtoreducecachepowerwithvictimcache
AT lǐchéngyù compileroptimizationtoreducecachepowerwithvictimcache
AT chengyulee yǐbiānyìqìzuìjiāhuàxīshēngzhěkuàiqǔzhīdībìvkuàiqǔjìshù
AT lǐchéngyù yǐbiānyìqìzuìjiāhuàxīshēngzhěkuàiqǔzhīdībìvkuàiqǔjìshù
_version_ 1718256923763015680