Research Initiation: A Language for Multiprocessors: Design, Implementation, Optimization

这个项目的目的是检查设计和实现问题的语言打算编程多处理器系统。这个项目的重点是SR语言，它最初是为分布式系统编程而开发的。该研究将研究如何修改SR以支持共享内存多处理器架构中的编程。计划通过允许全局声明变量和操作或在资源规范中声明变量和操作来扩展SR以利用共享内存。前者将提供真正的全局共享变量和操作；后者将提供特定于资源实例的共享对象。支持多处理器编程的其他语言特性也将被研究。修改后的语言的实现将以当前的SR实现为基础；它将提供关于语言特性的重要反馈。实现工作将调查可能在多处理器环境中使用的几种优化；例如，代价昂贵的进程会合可能被代价较低的信号量或监视器操作所取代。程序间分析技术将用于识别可优化的代码。这项研究将确定哪种语言机制最适合支持多处理器编程，证明这种机制可以合理有效地实现，并为进一步研究并发编程提供一个可用的工具。

The objective of this project is to examine design and implementation issues in languages intended for programming multiprocessor systems. The focus of this project is the SR language, which was originally developed for programming distributed systems. The research will study how to modify SR to support programming in shared memory multiprocessor architectures. It is planned to extend SR to exploit the shared memory by allowing variables and operations to be declared globally or in the specification of a resource. The former would provide truly global shared variables and operations; the latter would provide resource instance-specific shared objects. Other language features to support multiprocessor programming will also be investigated. The implementation of the modified language will build on the current SR implementation; it will provide important feedback on the language features. The implementation work will investigate several optimizations that might be employed in the mutiprocessor environment; e.g., expensive process rendezvous might be replaced by less expensive semaphore or monitor operations. Interprocedural analysis techniques will be used to identify optimizable code. The research will identify what language mechanisms are most appropriate to support multiprocessor programming, demonstrate that such mechanisms can be implemented reasonably efficiently, and provide a usable tool for further investigations into concurrent programming.