Near-Critical Path Analysis of Program Activity Graphs

Program activity graphs can be constructed from timestamped traces of appropriate execution events. Information about the activities on the k longest execution paths is useful in the analysis of parallel program performance. In this paper, four algorithms for finding the near–critical paths of program activity graphs are presented and compared, including an efficient new algorithm that utilizes slack values calculated by the critical path method to perform a best–first search in linear space. The worst–case time and memory requirements of the new algorithm are in O(ke) and O(k+e), where e is the number of edges in the graph. Results confirming the efficiency of the algorithm are presented for five application programs. A framework for utilizing the near–critical path information is also described. The framework includes both statistical summaries and visualization capabilities.