Performance localisation

Abstract

Performance becomes an issue particularly when execution cost hinders the functionality of a program. Typically a profiler can be used to find program code execution which represents a large portion of the overall execution cost of a program. Pinpointing where a performance issue exists provides a starting point for tracing cause back through a program.While profiling shows where a performance issue manifests, we use mutation analysis to show where a performance improvement is likely to exist. We find that mutation analysis can indicate locations within a program which are highly impactful to the overall execution cost of a program yet are executed relatively infrequently. By better locating potential performance improvements in programs we hope to make performance improvement more amenable to automation.

A journey among Java neutral program variants

“…All variants should pass the test suite of the original program. Existing work in this domain rely on random program transformations to search for program variants: Schulte and colleagues [32] exploit mutational robustness to reduce energy consumption; Langdon et al [20] add, delete, replace lines in C, C++, CUDA program sources to improve performance; Cody-Kenny et al [10] add, delete, replace AST nodes, to profile program performance; López and colleagues [22] explore program mutations to optimize source code. Manotas and al [23] replace java collections to optimize energy consumption.…”

Inferring Performance Bug Patterns from Developer Commits

“…Despite the relatively large number of subjects in our data set, there are applications that require even larger numbers of defects to obtain statistically significant evaluation results. For these scenarios we envision the creation of performance mutants, along the lines proposed in recent work [10] and proposals [11], and provide insights to support the creation of performance mutants that resemble performance bugs fixed in real world projects.…”