Redundant computation and its application in software performance analysis.

摘要

A significant research has been done in the area of software performance analysis. In this research, we have concentrated on the dynamic oriented performance analysis of the procedural sequential programs. Redundant computation is time-related performance deficiency for which the actual execution of statement(s) during program execution does not influence the program output. In this dissertation, we investigate the automated detection of redundant computation in programs and the application of redundant computation in software performance analysis.; The objectives of this research are to develop a new algorithm that efficiently detects redundant computation in any procedural program, develop a redundant computation analysis tool (RCAT) that helps programmer to detect performance deficiencies related to redundant computation in C programs, and conduct an experimental study. The goal of the experimental study was to validate the redundant computation approach in detecting performance deficiencies for “semi-commercial” programs and compare this approach with the existing execution profiling methods.; The experimental study showed that the redundant computation is a commonly occurring phenomenon in the software, and in many cases, it is a source of performance deficiency. In addition, significant improvements in the programs' performance can be achieved because of redundant computation elimination or reduction.