翻訳と辞書 |
Testing high-performance computing applications : ウィキペディア英語版 | Testing high-performance computing applications High performance computing applications run on massively parallel supercomputers consist of concurrent programs designed using multi-threaded, multi-process models. The applications may consist of various constructs (threads, local processes, distributed processes, etc.) with varying degree of parallelism. Although high performance concurrent programs use similar design patterns, models and principles as that of sequential programs, unlike sequential programs, they typically demonstrate non-deterministic behavior. As the number of interactions between the various parallel constructs increase, the probability of bugs exponentially increases. Race conditions, data races, deadlocks, missed signals and live lock are common error types. == Challenges == Parallel programs can be divided into two general categories: explicitly and implicitly parallel. Using parallel language constructs defined for process creation, communication and synchronization make an application explicitly parallel. Using a tool or parallelizing compiler to convert a serial program into a parallel one, makes it implicitly parallel. Both categories are equally bug-prone.
抄文引用元・出典: フリー百科事典『 ウィキペディア(Wikipedia)』 ■ウィキペディアで「Testing high-performance computing applications」の詳細全文を読む
スポンサード リンク
翻訳と辞書 : 翻訳のためのインターネットリソース |
Copyright(C) kotoba.ne.jp 1997-2016. All Rights Reserved.
|
|