Visualizing and Quantifying Memory-Access Locality of Linux Processes

Over time, computers have become more and more performant through improvements to their hardware and software. The software can contribute significantly to an improvement in performance if it cooperates efficiently with the computer's memory hierarchy. The locality of reference is useful in this context to make statements about the efficiency of the program. It describes whether computer programs access the same memory location more often within a specific time. This behavior usually favors faster access times, due to the use of caches. In this thesis a concept for a tool to analyze memory accesses of a Linux process is designed. It should offer the ability to help developers improve the performance of their software or to help lecturers teach the topic of memory accesses. The memory accesses of processes are visualized, and metrics for determining the locality of these are introduced. These metrics are also provided on the basis of the visualization. The practicability of this design is proven by a prototypical implementation. Part of this is also the evaluation of the implementation by using it with example programs. Based on this, it is shown how the analysis of programs looks like with the implemented solution and how scalable it is, respectively where the limits of it are. At the end of this thesis, possible extensions and improvements for future work are discussed.

(Bachelor Defense)