Determining the potential targets of virtual method invocations is essential for inter-procedural optimizations of objectoriented programs. It is generally hard to determine such targets accurately. The problem is especially difficult for dynamic languages such as Java, because additional targets of virtual calls may appear at runtime. Current mechanisms that enable inter-procedural optimizations for dynamic languages, repeatedly validate the optimizations at runtime. This paper addresses this predicament by proposing a novel technique for conservative devirtualization analysis, which applies to a significant number of virtual calls in Java programs. Unlike previous work, our technique requires neither whole program analysis nor runtime information, and incurs no runtime overhead. Our solution is very efficient to compute and is based on a newly introduced, seemingly unrelated security feature of Java file archives. On average, our analysis "seals" (safely devirtualizes) about 39% of the virtual calls (to non-final methods) that appear in SPECjvm98 programs, and about 29% of the calls invoked while executing these programs. In the runtime library rt.jar, about 10% of the packages contain a significant percentage (20-60%) of sealed calls, with a total average of about 8.5%. Most of these calls are also shown to be monomorphic, a fact which can be safely exploited by aggressive inter-procedural optimizations such as direct inlining. These results indicate that our technique has a strong potential for enhancing the analysis and optimization of Java programs.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.