A memory-optimal buffering protocol for preservation of synchronous semantics under preemptive scheduling

Abstract

Recently, we have proposed a set of buffering schemes to preserve the semantics of a synchronous program when the latter is implemented as a set of multiple tasks running under preemptive scheduling. These schemes, however, are not optimal in terms of memory (buffer usage). In this paper we propose a new protocol which generalizes the previous schemes. The new protocol is not only semantics-preserving but also memory-optimal in two senses: first, in terms of the number of buffers required to preserve semantics in the worst case (i.e., for the "worst" possible arrival/execution pattern of the tasks); second, in terms of the number of buffers required to preserve semantics for any arrival/execution pattern and at any time, assuming no knowledge of future arrivals.

Optimizing the Implementation of Communication in Synchronous Reactive Models

“…First, we provide a better bound for the case of priority-scheduled tasks activated with a minimum interarrival time (sporadic) or periodically, but with unknown phases. The bound shows how it is possible to improve with respect to [8] by leveraging the additional information about the minimum inter-arrival times and the task priorities, and applies to a more general (and practically useful) task model than [7]. Furthermore, our bound also applies to a more general application model than the bounds presented up to now, namely by allowing the case of multiple active instances of tasks at the same time, and communication links with arbitrary delays.…”

“…This bound has been defined in [12], where the protocol in [11] was extended to a multi-reader asynchronous protocol for single writer and multiple readers systems. The bound can be further improved in the case of single processor systems [8] if additional information is available. In particular, if all reader tasks have priority lower than the writer, then only N+1 buffers are needed, N used by the readers and the one with the most recent update by the writer.…”

“…In [8] a semantics-preserving implementation of the single-writer to multiple-reader case is proposed and the buffer bound is further extended for the case of communication links with a unit delay. The proposed protocol is called the Dynamic Buffering Protocol (DBP).…”

Synchronous Programming

“…This method has been developed in a number of recent works [33,34,35]. We only sketch the main ideas of the method here, and refer the reader to the above publications for details.…”

“…This protocol is used when the writer has higher priority than the reader (for simplicity we assume a single reader, but the protocol can be extended to any number of readers, with the addition of extra buffers, see [35]). In this protocol, the reader τ j maintains a double buffer B[0,1].…”