Refine
Document Type
Is part of the Bibliography
- no (6)
Institute
Begutachtungsstatus
- peer-reviewed (2)
Forschungsbericht 2013
(2014)
The Logical Execution Time (LET) has recently been integrated in multi-core automotive systems to ensure timing and dataflow determinism. Although buffering mechanisms are introduced to incorporate LET semantics, they do not guarantee that tasks are executed within their LET frames. In fact, LET and buffering semantics are violated if scheduling is not designed to execute all tasks within their LET frames and in a specific order. In this paper, we describe a scheduling synthesis technique for Fixed-Priority Scheduling (FPS) to achieve resource-efficient execution of LET systems. The proposed approach considers LET semantics, scheduling overheads, and delays caused by operating system operations and provides the possibility to optimize the schedule with respect to aspects like scheduling overheads. Our performance and feasibility evaluation shows that the proposed algorithm provides results in a reasonable amount of time for models of complex industrial applications. Thus, the integration of the proposed algorithm into an automated process is of high benefit to accelerate the development of vehicle applications.
We present in this paper a new lock-based resource sharing protocol PWLP (Preemptable Waiting Locking Protocol) for embedded multi-core processors. It is based on the busy-wait model and works with non-preemptive critical sections while task may be preempted by tasks with a higher priority when waiting for resources. Our protocol can be applied in partitioned as well as global scheduling scenarios, in which task-fix, job-fix or dynamically assigned priorities may be used. Furthermore, the PWLP permits nested requests to shared resources. Finally, we present a case study based on event-based simulations in which the FMLP (Flexible Multiprocessor Locking Protocol) and the proposed PWLP are compared.
With multi-core controllers entering the area of automotive control ECUs, strategies for parallelizing the control- algorithms come into focus. This paper deals with a special part of automotive powertrain software, called state transitions. Since dependencies between runnables executed there are weak, the transitions provide a good basis for parallelization. We present a strategy of how to distribute efficiently the execution of runnables to different cores while taking care of inner and outer dependencies. The strategy is accompanied by two case studies demonstrating the performance of the concept. The first one is carried out to find the most efficient strategies of parallelize state transitions based on randomly generated, simulated state transitions. In the second one, the developed partitioning strategies are applied to a real software project for an automotive powertrain system.
The amount of safety-critical embedded systems in automotive development is heavily growing. Ensuring their reliability not only increases the complexity of functions but also requires determinism at design and execution time, which is considerably challenging to fulfill and verify for multi-core processors. The Logical Execution Time (LET) is recently recognized in automotive industry as an approach for ensuring deterministic functional behavior. However, to decrease the manual design effort and time for deploying such complex systems to multi-core platforms and for ensuring their strict timing and safety requirements, automatic solutions are needed. This work presents a solution for allocating tasks to multi-core processors and generating a time-triggered schedule for embedded systems considering safety, timing, and LET semantics. The approach we propose solves both challenges by defining them as a Constraint Satisfaction Problem (CSP). To examine our CSP formulation, we use MiniZinc, which is a solver-independent constraint modeling language that can employ a variety of solvers. In a case study, we explore optimizations of an industrial system that are enabled by scheduling and task allocation design decisions. Further, the performance of the proposed solutions is evaluated based on large set of synthetically generated system models.