The rise of Java, C#, and other explicitly parallel languages has increased the importance of compiling for different software memory models. This paper describes co-operating escape, thread structure, and delay set analyses that enable high performance for sequentially consistent programs.We compare the performance of a set of Java programs compiled for sequential consistency (SC) with the performance of the same programs compiled for weak consistency. For SC, we observe a slowdown of 10% on average for an architecture based on the Intel Xeon processor, and 26% on average for an architecture based on the IBM Power3.

No abstract

No abstract

Abstract. The widespread popularity of languages allowing explicitly parallel, multi-threaded programming, e.g. Java and C#, have focused attention on the issue of memory model design. The Pensieve Project is building a compiler that will enable both language designers to prototype different memory models, and optimizing compilers to adapt to different memory models. Among the key analyses required to implement this system are thread escape analysis, i.e. detecting when a referenced object is accessible by more than one thread, delay set analysis, and synchronization analysis. In this paper, we evaluate the impact of different escape analysis algorithms on the effectiveness of the Pensieve system when both delay set analysis and synchronization analysis are used. Since both analyses make use of results of escape analyses, their precison and cost is dependent on the precision of the escape analysis algorithm. It is the goal of this paper to provide a quantitative evalution of this impact.