A Study of Type Analysis for Speculative Method Inlining in a JIT Environment

Method inlining is one of the most important optimizations for JIT compilers in Java virtual machines. In order to increase the number of inlining opportunities, a type analysis can be used to identify monomorphic virtual calls. In a JIT environment, the compiler and type analysis must also handle dynamic class loading properly because class loading can invalidate previous analysis results and invalidate some speculative inlining decisions. To date, a very simple type analysis, class hierarchy analysis (CHA), has been used successfully in JIT compilers for speculative inlining with invalidation techniques as backup. This paper seeks to determine if more powerful dynamic type analyses could further improve inlining opportunities in a JIT compiler. To achieve this goal we developed a general dynamic type analysis framework which we have used for designing and implementing dynamic versions of several well-known static type analyses, including CHA, RTA, XTA and VTA. Surprisingly, the simple dynamic CHA is nearly as good as an ideal type analysis for inlining virtual method calls. There is little room for further improvement. On the other hand, only a reachabilitybased interprocedural type analysis (VTA) is able to capture the majority of monomorphic interface calls. We also found that memory overhead is the biggest issue with dynamic whole-program analyses. To address this problem, we outline the design of a demand-driven interprocedural type analysis for inlining hot interface calls.