Efficiently identifying object production sites

Abstract

Most programming environments are shipped with accurate memory profilers. Although efficient in their analyses, memory profilers traditionally output textual listing reports, thus reducing the memory profile exploration as a set of textual pattern-matching operations. Memory blueprint visually reports the memory consumption of a program execution. A number of simple visual cues are provided to identify direct and indirect object production sites, key ingredients to efficiently address memory issues. Scalability is addressed by restricting the scope of interest both in the call graph and the considered classes. Memory blueprint has been implemented in the Pharo programming language, and is available under the MIT license.