The global quiescence of a distributed computation (or distributed termination detection) is an important problem. Some concurrent programming languages and systems provide global quiescence detection as a built-in feature so that programmers do not need to write special synchronization code to detect quiescence. This paper introduces partial quiescence (PQ), which generalizes quiescence detection to a specified part of a distributed computation. Partial quiescence is useful, for example, when two independent concurrent computations that both rely on global quiescence need to be combined into a single program. The paper describes how we have designed and implemented a PQ mechanism within an experimental version of the JR concurrent programming language. Our early results are promising qualitatively and quantitatively.

No abstract

The JR concurrent programming language extends Java with additional concurrency mechanisms, which are built upon JR's operations and capabilities. JR operations generalize methods in how they can be invoked and serviced. JR capabilities act as reference to operations. Recent changes to the Java language and implementation, especially generics, necessitated corresponding changes to the JR language and implementation. This paper describes the new JR language features (known as JR2) of generic operations and generic capabilities. These new features posed some interesting implementation challenges. The paper describes our initial implementation (JR21) of generic operations and capabilities, which works in many, but not all, cases. It then describes the approach our improved implementation (JR24) uses to fully implement generic operations and capabilities. The paper also describes the benchmarks used to assess the compilation and execution time performances of JR21 and JR24. The JR24 implementation reduces compilation times, mainly due to reducing the number of files generated during JR program translation, without noticeably impacting execution times.

The global quiescence (GQ) of a distributed computation (or distributed termination detection) is an important problem. Some concurrent programming languages and systems provide GQ detection as a built-in feature so that programmers do not need to write special synchronization code to detect quiescence. This paper introduces partial quiescence (PQ), which generalizes quiescence detection to a specified part of a distributed computation. PQ is useful, for example, when two independent concurrent computations that both rely on GQ need to be combined into a single program. The paper describes how we have designed and implemented a PQ mechanism within an experimental version of the JR concurrent programming language, and have gained experience with several representative applications. Our early results are promising qualitatively and quantitatively.deadlocked. This quiescence problem is challenging because each process has only local information, but to solve the problem requires information about all processes (i.e. global state information). More formally, global quiescence (GQ) is defined as the state in which each process has terminated or deadlocked and there are no messages in the communication channels [2]. Quiescence detection, then, is the mechanism used to detect such a state in a distributed system. Many solutions to this problem have been proposed (e.g. [3][4][5][6][7][8]).Some programming languages and systems provide GQ detection as a built-in feature. That is, programmers do not need to write special synchronization code to detect quiescence. Instead, they can focus on writing application code. When quiescence is reached, the program can perform various actions such as simply terminating the program, outputting final results, gathering statistics from the overall computation, or initiating a new phase of the program, which might involve a new, corresponding phase of quiescence detection.Although useful, GQ is limited to dealing with the state of all processes in a program. A more general, but more difficult to detect, property would define when a specified part of the program has become quiescent. For example, suppose we have two programs that use GQ and we want to combine them into a single program in which we want to perform different actions when each part of it becomes quiescent. This motivation led us to explore partial quiescence (PQ).This paper proposes possible ways of defining PQ. It then discusses the particular definition selected and implemented in an experimental version of the JR concurrent programming language.(JR extends Java [9] with a richer concurrency model [10][11][12].) The definition of PQ and its implementation in JR apply to truly distributed programs. The paper also shows how PQ leads to a different programming style for some problems. We compare the performance of using PQ detection and GQ detection. PQ might be a useful feature for other languages and libraries that define process or thread groups, as many do, and especially useful for those languages and libraries that alre...