Publikation

Zurück zur Übersicht

JP2: Call-site aware calling context profiling for the Java Virtual Machine

Publikationsart Peer-reviewed
Publikationsform Originalbeitrag (peer-reviewed)
Publikationsdatum 2014
Autor/in Sarimbekov Aibek, Sewe Andreas, Binder Walter, Moret Philippe, Mezini Mira,
Projekt FAN: Foundations of dynamic program ANalysis
Alle Daten anzeigen

Originalbeitrag (peer-reviewed)

Zeitschrift Science of Computer Programming
Volume (Issue) 79
Seite(n) 146 - 157
Titel der Proceedings Science of Computer Programming
DOI 10.1016/j.scico.2011.11.003

Abstract

Calling context profiling fulfills programmers’ information needs to obtain a complete picture of a program’s inter- and intra-procedural control flow, which are important for workload characterization, debugging, profiling, program comprehension, and reverse engineering. Many existing calling context profilers for Java, however, resort to sampling or other incomplete instrumentation techniques; thus, they collect incomplete profiles only. In this article we present JP2, a new calling context profiler for the Java Virtual Machine, which collects profiles that are not only complete but also call-site aware; that is, JP2 is able to distinguish between multiple call sites within a single method. JP2 supports selective profiling of the dynamic extent of chosen methods and supports profiling of native method invocations. Moreover, produced profiles contain execution statistics at the level of individual basic blocks of code, thereby preserving the intra-procedural control flow of the profiled applications. We rigorously evaluate the overhead incurred by JP2. This overhead is acceptable in tasks such as workload characterization. JP2 is freely available under an Open Source license.
-