- 【Updated on May 12, 2025】 Integration of CiNii Dissertations and CiNii Books into CiNii Research
- Trial version of CiNii Research Knowledge Graph Search feature is available on CiNii Labs
- Suspension and deletion of data provided by Nikkei BP
- Regarding the recording of “Research Data” and “Evidence Data”
A loop restructuring technique to optimize memory access locality
-
- TSUDA TAKESHI
- Division of Communications and Computer Engineering, Graduate School of Infomatics, Kyoto University
-
- YAMAMOTO TAKANOBU
- Division of Communications and Computer Engineering, Graduate School of Infomatics, Kyoto University
-
- TANAKA TOSHIHIKO
- Division of Communications and Computer Engineering, Graduate School of Infomatics, Kyoto University
-
- GOSHIMA MASAHIRO
- Division of Communications and Computer Engineering, Graduate School of Infomatics, Kyoto University
-
- MORI SHIN-ICHIRO
- Division of Communications and Computer Engineering, Graduate School of Infomatics, Kyoto University
-
- TOMITA SHINJI
- Division of Communications and Computer Engineering, Graduate School of Infomatics, Kyoto University
Bibliographic Information
- Other Title
-
- メモリ・アクセスの局所性を最適化するループ再構成法
Search this article
Description
It is known that a loop optimize technique, Tiling can be applied to each class of memory hierarchy, In this paper, we speak about the technique to apply tiling for the size of each class of memory hierarchy and decide the size of the tile and the order of loop so that hit ratio of each memory hierarchy is improved, The best size of a certain level of memory hierarchy is contray to that of former level. In this technique, we know that tile size can be decided by only slow memory hierarchy. We decide the order of loop to make use of the size of memory hierarchy. When we apply this technique to matrix product and lu decomposition, the performance is not decreased when the size is larger. In lu decomposition, we get 26% better result than the former technique.
Journal
-
- IPSJ SIG Notes
-
IPSJ SIG Notes 99 (21), 133-138, 1999-03-04
Information Processing Society of Japan (IPSJ)
- Tweet
Details 詳細情報について
-
- CRID
- 1573105976798792960
-
- NII Article ID
- 110004752365
-
- NII Book ID
- AN10463942
-
- Text Lang
- ja
-
- Data Source
-
- CiNii Articles