Isolation for nested task parallelism

Abstract

Isolation-the property that a task can access shared data without interference from other tasks-is one of the most basic concerns in parallel programming. While there is a large body of past work on isolated task-parallelism, the integration of isolation, task-parallelism, and nesting of tasks has been a difficult and unresolved challenge. In this paper, we present a programming and execution model called Otello where isolation is extended to arbitrarily nested parallel tasks with irregular accesses to heap data. At the same time, no additional burden is imposed on the programmer, who only exposes parallelism by creating and synchronizing parallel tasks, leaving the job of ensuring isolation to the underlying compiler and runtime system.Otello extends our past work on Aida execution model and the delegated isolation mechanism [22] to the setting of nested parallelism. The basic runtime construct in Aida and Otello is an assembly: a task equipped with a region in the shared heap that it owns. When an assembly A conflicts with an assembly B, A transfers-or delegates-its code and owned region to a carefully selected assembly C in a way that will ensure isolation with B, leaving the responsibility of re-executing task A to C. The choice of C depends on the nesting relationship between A and B.We have implemented Otello on top of the Habanero Java (HJ) parallel programming language [8], and used this implementation to evaluate Otello on collections of nested taskparallel benchmarks and non-nested transactional benchmarks from past work. On the nested task-parallel bench-, and the relative overhead of Otello is lower than that of many published data-race detection algorithms that detect the isolation violations (but do not enforce isolation). For the transactional benchmarks, Otello incurs lower overhead than a state-of-the-art software transactional memory system (Deuce STM).

Efficient and reasonable object-oriented concurrency

“…Therefore the underlying mechanisms are fundamentally different, where Aida requires efficient heap ownership and conflict resolution via a parallel union-find algorithm, SCOOP/Qs requires efficient communication which is attained via novel and nested uses of specialized queue structures. Otello [28] extends the isolation found in Aida to include support for nested tasks. Another object-oriented approach which, like SCOOP, associates threads of execution with areas of the heap is JCoBox [23].…”

Friendly barriers

“…Habanero-Java is another such project, which is inspired from X10 and is actively being used [7,13,14,23,34,37]. To determine the performance of our system, we now do a speedup comparison with Fork-Join and Habanero-Java on all our benchmarks.…”

PaniniJ: adding the capsule programming abstraction to Java to provide linguistic support for modular reasoning in concurrent program design

“…This property simplifies reasoning about concurrent processes in a shared memory setting. However, reasoning about dynamically created, nested processes requires reasoning about nested regions, which is known to be a difficult problem [11,76].…”