Macro tree transformations of linear size increase achieve cost-optimal parallelism

Akimasa Morihata

研究成果: Conference contribution

2 被引用数 (Scopus)

抄録

This paper studies parallel evaluation of tree transformations, in particular accumulative ones. Accumulation is a ubiquitous programming pattern. However, since accumulation usually imposes restrictions on evaluation orders, accumulative tree transformations appear to be unsuitable for parallel evaluation. We propose a parallel evaluation method for a large class of tree-to-tree recursive functions, which may contain accumulations, higher-order terms, and function compositions. Our parallel evaluation method achieves optimal parallel speedup if the transformation is of linear size increase, namely, the size of each output is linearly bounded by the size of the corresponding input. Our result is based on the theory of macro tree transducers and that of parallel tree contractions. The main contribution is to reveal a good collaboration between them.

本文言語English
ホスト出版物のタイトルProgramming Languages and Systems - 9th Asian Symposium, APLAS 2011, Proceedings
ページ204-219
ページ数16
DOI
出版ステータスPublished - 2011 12月 26
イベント9th Asian Symposium on Programming Languages and Systems, APLAS 2011 - Kenting, Taiwan, Province of China
継続期間: 2011 12月 52011 12月 7

出版物シリーズ

名前Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
7078 LNCS
ISSN(印刷版)0302-9743
ISSN(電子版)1611-3349

Other

Other9th Asian Symposium on Programming Languages and Systems, APLAS 2011
国/地域Taiwan, Province of China
CityKenting
Period11/12/511/12/7

ASJC Scopus subject areas

  • 理論的コンピュータサイエンス
  • コンピュータ サイエンス(全般)

フィンガープリント

「Macro tree transformations of linear size increase achieve cost-optimal parallelism」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル