This chapter describes the knowledge-based automatic performance analysis of parallel programs with Knowledge based Automatic Parallel Program Analyzer for Performance Improvement (KAPPA-PI). The actual behavior of parallel programs is of capital importance for the development of an application. The KAPPA-PI is a performance automatic analyzer. This automatic tool analyzes the inactivity intervals found in the trace files. The KAPPA-PI tool uses a monitoring approach to analyze the performance of a program. It analyses a trace file obtained directly from the execution of the application. This means that the analysis is carried out on the low level events of the trace file. The tool, consulting the actual source files of the application, builds a series of explanations of the causes of the main problems found. The tool bases the feedback to the user in the names and structures of the application; that is to say, the programmer's view of the application. © 1998 Elsevier B.V.
|Title of host publication||Advances in Parallel Computing|
|Number of pages||3|
|Publication status||Published - 1 Jan 1998|