| 摘要: |
| 讨论如何完全地调试不确定性MPI／PVM并行程序．在循环调试过程中，不确定性导致前次遇到的错误在以后的执行中很可能无法再现．基于MPI／PVM的FIFO通信模型，给出一种记录－重放技术的实现．通过可控制的重放，用户可以覆盖所有可能的程序执行路径，从而达到完全调试的目的．和其它方法相比，所提供的方法所需时空开销要小得多．此技术已在两种消息传递体系结构上得到实现：一种是曙光－2000超级服务器（由国家智能计算机研究中心开发），它由单处理器（PowerPC）结点经MESH网互联而成；另一种是国家高性能计算中心（ |
| 关键词: 并行调试器 记录－重放 消息传递 MPI PVM |
| DOI： |
| 分类号: |
| 基金项目:This project is supported by the NationalHigh Technology Development Program of China under Grant No．863－306－ZD01－02－3（国家863高科技发展计划基金）and the Youth Science Foundation of the University of Science and Technology of China un－der Grant No．98－1101（中国科学技术大学青年科 |
|
| Completely Debugging Indeterminate MPI／PVM Programs |
|
WANG Feng,AN Hong,CHEN Zhi－hui,CHEN Guo－liang
|
| Abstract: |
| This paper discusses how to completely debug indeterminate MPI／PVM parallelprograms．Due tothe indeterminacy，the previous bugs may be non－repeatable in successive executions during a cyclic debuggingsession．Based on the FIFO communication model of MPI／PVM，an implementation of record and replay tech－nique is presented．Moreover，users are provided with an easy way to completely debug their programs by cover－ing all possible execution paths through controllable replay．Comparied with other solutions，the proposedmethod produces much less temporaland spatialoverhead．The implementation has been completed on two kindsof message passing architectures：one is Dawning－2000 super server（that was developed by the National Re－search Center for Intelligent Computing Systems ofChina）with single－processor（PowerPC）nodes which are in－terconnected by a custom－built wormhole mesh network；the other is a cluster ofworkstations（PowerPC／AIX）which has been built in NationalHigh Performance Computing Center at Hefei． |
| Key words: paralleldebugger record and replay message passing MPI PVM |
