Towards a Qualifiable OpenMP Framework for Embedded Systems

Abstract

OpenMP is a very convenient programming model for critical real-time parallel applications due to its powerful tasking model and its proven time predictability. However, current implementations are not suitable for critical environments based on the intensive use of dynamically allocated memory needed to efficiently manage the parallel execution. This jeopardizes the qualification processes needed to ensure that the integrated software stack is compliant with system requirements. This paper proposes a novel OpenMP framework that statically allocates the data structures needed to efficiently manage the parallel execution of OpenMP tasks. Our framework is composed of a compiler that captures the environment of the OpenMP tasks instantiated along the parallel execution and bounds the exposed parallelism, and a runtime implementing a lazy task creation policy that significantly reduces the runtime memory requirements, whilst exploiting parallelism efficiently. The evaluation shows that our tool achieves the same performance as current OpenMP implementations, while bounds and drastically reduces the dynamic memory requirements at run-time.

Vector Extensions in COTS Processors to Increase Guaranteed Performance in Real-Time Systems

“…There are mainly two approaches for its adoption in high-integrity domains: 1. reusing the OpenMP specification and implementation for mainstream systems, i.e., as it is currently defined in its standard. However, this fails to cover many requirements for high-integrity systems [53,36].…”

“…For use in high-integrity systems, runtimes need to be qualified according to the applicable functional safety standard and support documents. Recently, there have been some research proposals in this direction [53,36]. However, to our knowledge, the code of those solutions is neither publicly available nor qualified for use in the industry yet.…”

“…Parallel programming models, including OpenMP, exhibit fundamental differences with respect to sequential programming, mainly in the capability of concurrently sharing the same data across runtime entities. The necessary synchronization/communication primitives are typically designed for average performance and high throughput, at the expense of predictability and analysability [45,46,36]. This results in programs difficult to analyze with either static or dynamic approaches (M2).…”

Enhancing OpenMP Tasking Model: Performance and Portability

“…This work strengthened the possibility of using OpenMP untied tasks (i.e., once such task is suspended by the initial thread, it can be correctly resumed by any idle thread within the same OpenMP team) on safety-critical embedded systems. A. Munera et al [10] showed how statically generated TDGs can reduce the dynamic memory usage of OpenMP tasks, so that the tasking model can be used on embedded systems conveniently, where the amount of dynamic memory is often limited by safety constraints. Taskgraph makes OpenMP more suitable for safety-critical embedded systems by reinforcing their work:…”

“…-Techniques used in [10] rely on the static generation of the TDG, the information from which can be leveraged by the taskgraph clause to enhance the performance. Therefore, by including the new clause in their method, the resulting framework should deliver better performance than the current OpenMP tasking model, and also use less dynamic memory throughout the execution.…”