Stream*: Fast, Flexible Data-Parallel I/O

PARCO(1995)

引用 25|浏览10
暂无评分
摘要
Data-parallel programming languages, including C* and High-Performance Fortran do not have constructs for parallel I/O. Parallel I/O, if done at all, is performed via nonstandard vendor-specific facilities. To alleviate this problem, we propose Stream*, based on the notion of a data-parallel file, in which each virtual processor accesses its own stream. The programmer''s view of a virtual processor''s I/O operations is then identical to the more familiar, unambiguous operations found in a sequential program. Further, we show how Stream* can be implemented using three different modes. The Stream* mode most commonly used by Grand Challenge and scientific problems provides the programmer with I/O performance equal to what vendor-specific systems provide.
更多
查看译文
关键词
nonstandard vendor-specific facility,o performance,flexible data-parallel,o operation,own stream,data-parallel file,high-performance fortran,virtual processor,data-parallel programming language,grand challenge,different mode
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要