Algebraic program semantics for supercomputing | |
Chen, Yifeng | |
2013 | |
英文摘要 | The competition for higher performance/price ratio is pushing processor chip design into the manycore age. Existing multicore technologies such as caching no longer scale up to dozens of processor cores. Even moderate level of performance optimisation requires direct handling of data locality and distribution. Such architectural complexity inevitably introduces challenges to programming. A better approach for abstraction than completely relying on compiler optimization is to expose some performance-critical features of the system and expect the programmer to handle them explicitly. This paper studies an algebra-semantic programming theory for a performance-transparent level of parallel programming of array-based data layout, distribution, transfer and their affinity with threads. The programming model contributes to the simplification of system-level complexities and the answering of crucial engineering questions through rigorous reasoning. ? 2013 Springer-Verlag.; EI; 0 |
语种 | 英语 |
DOI标识 | 10.1007/978-3-642-39698-4_8 |
内容类型 | 其他 |
源URL | [http://ir.pku.edu.cn/handle/20.500.11897/411735] |
专题 | 信息科学技术学院 |
推荐引用方式 GB/T 7714 | Chen, Yifeng. Algebraic program semantics for supercomputing. 2013-01-01. |
个性服务 |
查看访问统计 |
相关权益政策 |
暂无数据 |
收藏/分享 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论