| Presentation | 2021-12-17 An LSTM-based prefetcher exploiting delta correlation Hiroki Taniai, Tomoki Nakamura, Toru Koizumi, Yuya Degawa, Hidetsugu Irie, Shuichi Sakai, Ryota Shioya, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Prefetching is one of the major hardware techniques to improve the execution performance of programs in modern processors (CPUs). Prefetching predicts future memory access patterns and looks ahead to them in order to reduce cache misses. For prefetching, various rule-based algorithms have been proposed in the past to correctly predict more complex memory access patterns. In recent years, researchers have begun to explore algorithms based on machine learning to further improve performance. However, many machine learning-based algorithms proposed so far do not work well and often do not achieve high performance compared to simple rule-based prefetchers. In this paper, we analyze memory access patterns that rule-based prefetchers can handle well, and propose a new LSTM model and a pre-training method for it by utilizing the domain knowledge obtained from the analysis. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Time series forecasting / Prefetch |
| Paper # | PRMU2021-53 |
| Date of Issue | 2021-12-09 (PRMU) |
| Conference Information | |
| Committee | PRMU |
|---|---|
| Conference Date | 2021/12/16(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Seiichi Uchida(Kyushu Univ.) |
| Vice Chair | Masakazu Iwamura(Osaka Pref. Univ.) / Mitsuru Anpai(Denso IT Lab.) |
| Secretary | Masakazu Iwamura(NTT) / Mitsuru Anpai(Tottori Univ.) |
| Assistant | Kouta Yamaguchi(CyberAgent) / Yusuke Matsui(Univ. of Tokyo) |
| Paper Information | |
| Registration To | Technical Committee on Pattern Recognition and Media Understanding |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | An LSTM-based prefetcher exploiting delta correlation |
| Sub Title (in English) | |
| Keyword(1) | Time series forecasting |
| Keyword(2) | Prefetch |
| 1st Author's Name | Hiroki Taniai |
| 1st Author's Affiliation | The University of Tokyo(Tokyo Univ.) |
| 2nd Author's Name | Tomoki Nakamura |
| 2nd Author's Affiliation | The University of Tokyo(Tokyo Univ.) |
| 3rd Author's Name | Toru Koizumi |
| 3rd Author's Affiliation | The University of Tokyo(Tokyo Univ.) |
| 4th Author's Name | Yuya Degawa |
| 4th Author's Affiliation | The University of Tokyo(Tokyo Univ.) |
| 5th Author's Name | Hidetsugu Irie |
| 5th Author's Affiliation | The University of Tokyo(Tokyo Univ.) |
| 6th Author's Name | Shuichi Sakai |
| 6th Author's Affiliation | The University of Tokyo(Tokyo Univ.) |
| 7th Author's Name | Ryota Shioya |
| 7th Author's Affiliation | The University of Tokyo(Tokyo Univ.) |
| Date | 2021-12-17 |
| Paper # | PRMU2021-53 |
| Volume (vol) | vol.121 |
| Number (no) | PRMU-304 |
| Page | pp.pp.160-164(PRMU), |
| #Pages | 5 |
| Date of Issue | 2021-12-09 (PRMU) |