Loop bounds computation for multilevel tiling

M. Jiménez, J. Llabería, Agustín Fernández
{"title":"Loop bounds computation for multilevel tiling","authors":"M. Jiménez, J. Llabería, Agustín Fernández","doi":"10.1109/EMPDP.1998.647232","DOIUrl":null,"url":null,"abstract":"The paper focuses on the complexity of computing exact loop bounds in multilevel tiling. Conventional tiling techniques implement tiling using first strip mining and afterwards loop interchange. Multilevel tiling has typically been implemented applying tiling level by level. We present a new way to implement multilevel tiling that consists in applying first strip mining at each dimension in all levels and, afterwards, performing once a loop interchange transformation of the strip mined loops. Our algorithm computes exact loop bounds, that is, loops in the generated code never execute empty iterations. We evaluate the complexity of this algorithm and show that it depends doubly exponentially only an the number of loops (space dimensions) in the original loop nest rather than in the number of loops in the final tiled code. The algorithm is based on the Fourier-Motzkin elimination, that generates redundant bounds. We also present an implementation of the technique that reduces the number of redundant bounds in the tiled code.","PeriodicalId":265140,"journal":{"name":"Proceedings of the Sixth Euromicro Workshop on Parallel and Distributed Processing - PDP '98 -","volume":"23 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1998-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"7","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the Sixth Euromicro Workshop on Parallel and Distributed Processing - PDP '98 -","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/EMPDP.1998.647232","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 7

Abstract

The paper focuses on the complexity of computing exact loop bounds in multilevel tiling. Conventional tiling techniques implement tiling using first strip mining and afterwards loop interchange. Multilevel tiling has typically been implemented applying tiling level by level. We present a new way to implement multilevel tiling that consists in applying first strip mining at each dimension in all levels and, afterwards, performing once a loop interchange transformation of the strip mined loops. Our algorithm computes exact loop bounds, that is, loops in the generated code never execute empty iterations. We evaluate the complexity of this algorithm and show that it depends doubly exponentially only an the number of loops (space dimensions) in the original loop nest rather than in the number of loops in the final tiled code. The algorithm is based on the Fourier-Motzkin elimination, that generates redundant bounds. We also present an implementation of the technique that reduces the number of redundant bounds in the tiled code.
多层平铺的循环边界计算
本文主要研究多层平铺中精确循环边界计算的复杂性。传统的平铺技术首先采用条带开采,然后进行环路交换。多层平铺通常是通过逐层应用平铺来实现的。我们提出了一种实现多层平铺的新方法,该方法包括首先在所有级别的每个维度上应用条带挖掘,然后对条带挖掘的环路执行一次循环交换变换。我们的算法计算精确的循环边界,也就是说,生成代码中的循环永远不会执行空迭代。我们评估了该算法的复杂性,并表明它只与原始循环巢中的循环数量(空间维度)呈双指数关系,而与最终平铺代码中的循环数量无关。该算法是基于傅立叶-莫兹金消去,产生冗余界。我们还提供了该技术的实现,减少了平铺代码中冗余边界的数量。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信
小红书