SBIR Phase I: Advanced Scientific Parallel Debugger

SBIR 第一阶段：高级科学并行调试器

• 批准号: 9661722
• 负责人: Michael Frese
• 金额: $ 7.5万
• 依托单位: MICHAEL H. FRESE
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1997
• 资助国家: 美国
• 起止时间: 1997-02-01 至 1997-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9661722&HistoricalAwards=false
• 关键词: SBIR; Phase; Advanced; Scientific; Parallel

*** 9661722 Giancola This Small Business Innovation Research Phase I is directed at developers working with high performance parallel computers. The primary commercial application codes for these computers are thus tools for software development. To date, these tools have not been designed with an eye to the requirements of the high performance computing user, but rather have been ported from workstations, where their targets are much more limited in scale. The most important of these tools, and to date the most inadequate, are the symbolic debuggers. Experience in the supercomputer development environment on both CTSS and UNICOS has shown that the most essential feature for an effective debugger to have in addition to the standard ones of process control and variable evaluation is the ability to save the target's state for later restoration. Macro processing is another required capability; it allows the user to record complex sequences of debugging commands for easy reuse. Also essential is acceptance of an incomplete set of symbol table information to enable the maximum possible function on optimized code. This project will develop software technology for debuggers which it is believed no such debugger for high performance parallel computing environment exists. This technology will be marketed to one or more parallel computer hardware vendors for incorporation into proprietary debuggers for bundled and unbundled sale with systems and to retain the rights to market a workstation version. ***

*** 9661722 Giancola这个小企业创新研究阶段是针对使用高性能并行计算机的开发人员的。因此，这些计算机的主要商业应用程序代码是软件开发的工具。到目前为止，这些工具的设计还没有考虑到高性能计算用户的需求，而是从工作站移植过来的，它们的目标在规模上要有限得多。这些工具中最重要的，也是迄今为止最不完善的是符号调试器。在CTSS和UNICOS上的超级计算机开发环境中的经验表明，除了标准的过程控制和变量评估之外，一个有效的调试器最重要的特性是能够保存目标状态以供以后恢复。宏处理是另一个必需的功能；它允许用户记录复杂的调试命令序列，以便于重用。同样重要的是接受一组不完整的符号表信息，以便在优化的代码上实现最大可能的功能。该项目将开发调试器的软件技术，据信目前还没有针对高性能并行计算环境的调试器。该技术将被销售给一个或多个并行计算机硬件供应商，用于合并到与系统捆绑或非捆绑销售的专有调试器中，并保留销售工作站版本的权利。＊＊＊