|
||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||
ABSTRACT
Debugging parallel programs containing thousands of processes and threads across hundreds or thousands of nodes by individually looking at each process and thread is possible. Possible, however, does not mean practical. The only workable approach is to organize processes and threads into groups and then debug the program by using these groups. This paper discusses the ways in which the TotalView Technologies Debugger (TVD) automatically organizes a program's processes and threads, the way developers can manually organize them when additional grouping capabilities are needed, and the implications of executing processes, threads, and groups asynchronously under debugger control. REFERENCES
Note: OCR errors may be found in this Reference List extracted from the full text article. ACM has opted to expose the complete List rather than only correct and linked references.
INDEX TERMS
Primary Classification:
|
||||||||||||||||||||||||||||