Some Challenges of Specifying Concurrent Program Components

Abstract

The purpose of this paper is to address some of the challenges of formally specifying components of shared-memory concurrent programs. The focus is to provide an abstract specification of a component that is suitable for use both by clients of the component and as a starting point for refinement to an implementation of the component. We present some approaches to devising specifications, investigating different forms suitable for different contexts. We examine handling atomicity of access to data structures, blocking operations and progress properties, and transactional operations that may fail and need to be retried.

Deriving Laws for Developing Concurrent Programs in a Rely-Guarantee Style

“…a buffer is non-empty). Handling resources and termination of operations that may wait are further extensions we are actively pursuing; initial ideas for incorporating these may be found in [Hay18]. The approach in the current paper assumes a sequentially consistent memory model and hence additional "fencing" is needed for use on a multi-processor with a weak memory model where caches and write buffers only give a sequentially consistent view for a single processor.…”

“…Additional work is needed to include the appropriate fencing to restore the desired behaviour. The use of a concept of a resource allows one to link control/locking variables with the data they control/lock [Hay18], thus allowing the generation of appropriate fencing.…”