Towards an OpenMP Specification for Critical Real-Time Systems

Abstract

OpenMP is increasingly being considered as a convenient parallel programming model to cope with the performance requirements of critical real-time systems. Recent works demonstrate that OpenMP enables to derive guarantees on the functional and timing behavior of the system, a fundamental requirement of such systems. These works, however, focus only on the exploitation of fine grain parallelism and do not take into account the peculiarities of critical real-time systems, commonly composed of a set of concurrent functionalities. OpenMP allows exploiting the parallelism exposed within real-time tasks and among them. This paper analyzes the challenges of combining the concurrency model of real-time tasks with the parallel model of OpenMP. We demonstrate that OpenMP is suitable to develop advanced critical real-time systems by virtue of few changes on the specification, which allow the scheduling behavior desired (regarding execution priorities, preemption, migration and allocation strategies) in such systems.

Multi-core Devices for Safety-critical Systems

“…In this context, OpenMP has been regarded as a potentially appropriate solution, enabling functional verification, and being already supported by Keystone II and MPPA multi-core devices [189]. Regarding timing verification for real-time systems, two different OpenMP models have been considered so far in the literature, namely fork-join [190] and tasking [191] models, being the latter the one gaining more popularity due to its ability to materialize parallelism even at fine granularity, and not being restricted to only structured parallelism. Irregular task graphs can also be parallelized with the tasking model, and research initiatives aim at leveraging support for heterogeneous and asynchronous parallelism to deliver time predictability [192].…”

The Cooperative Parallel: A Discussion About Run-Time Schedulers for Nested Parallelism

“…On the other hand, timing analysis for work non-conserving schedulers (i.e., there may be idle threads while there is still work to be done) have been proven to be very complex, and hence lead to unacceptable pessimistic results [35]. As a result, timing analysis techniques impose the real-time system to use a single team of OpenMP threads to execute all real-time tasks [36]. The reason lies in the black-box nature of concurrent parallel regions: the execution of each parallel region is governed by the team associated to that region, and each team has access only to the tasks associated to that team.…”

“…OpenMP, widely used in the High Performance Computing (HPC) domain, is increasingly gaining attention in others domains [23,22,15,36] due to its efficient parallel execution model in shared memory systems, and also its support for heterogeneous computing. This is the case of critical real-time embedded systems, in which new computational intensive functionalities are being developed (e.g., autonomous driving).…”

“…Here, OpenMP allows to efficiently exploit the performance capabilities of the newest highly parallel and heterogeneous embedded architectures, while benefiting from its programmability and portability capabilities. Moreover, OpenMP has been proven to be time predictable [38,35,36] OpenMP implements a fork-join model in which the parallel execution is initiated when a parallel construct is encountered. Then, a new team of threads (and implicit tasks) is created, associated to the corresponding parallel region.…”

“…As a result, a team can have idle OpenMP threads waiting in a barrier, and occupying computing resources, while there is another team with pending work. The blackbox problem in critical real-time systems was already identified [36], so the use of a unique team of threads was proposed to parallelize such systems.…”