Optimizations in distributed run-time compilation

摘要: Distributed run-time code generation is a useful technique that can easily be implemented using the known as “compositional semantics.” In this dissertation, we describe new system, Jumbo, which adds features to Java. It allows generators from disparate sources combined dynamically. Jumbo creation of highly adaptable distributed in binary form. In order handle increasing complexity software, programmers have long been creating reusable function libraries. These libraries form without revealing implementation secrets. While allow sharing or sale code, granularity and inflexibility pre-compiled functions force either overly general prohibitively specific. Run-time systems present light-weight alternative calls, where fragments not necessarily whole functions, manipulated, higher-quality specialized code. However, currently existing RTCG operate at source level take non-compositional approach, thus are amenable distribution. Jumbo implements novel compilation, several run time efficient By compositional semantics, some previously unknown functionality. For instance, database systems, client create query interact directly with server's internal calls neither nor needing disclosed. Render farms let clients specify textures merges local, disclosing texture details. Anonymity, copyrights trade secrets retained sacrificing efficiency. also for non-distributed advantage staged compilation efficiency, it additionally used manner similar macros, providing way codify common programming practices. We consider possibilities optimizations, rewriting This general-purpose optimizer exploits nature aggressively optimize individual fragments.