Compiler optimizations play a crucial role in enhancing software performance by improving execution speed and reducing resource consumption. However, these optimizations can also introduce variability in execution times, a significant concern for real-time systems where predictability is paramount. This paper investigates the relationship between GCC compiler optimizations and the execution time variability. By compiling a set of benchmark programs under different optimization levels (O0, O1, O2, O3), we analyze the impact on execution time variability using the WCET/BCET ratio, and the dispersion of the execution times around the WCET.

In mixed-criticality systems, classical models assume an estimated execution time budget for tasks in each possible criticality, and when a job presents a budget overrun, they tend to drop the lower criticality tasks to preserve the schedulability of the higher criticality tasks. However, few studies focus on how the different execution time budgets are calculated, and it has been emphasized that tasks of lower criticality must provide a minimum service for the system to function properly. In this paper, we present a greedy heuristic for calculating the execution time budget to be allocated to real-time tasks according to their criticality and the shape of their execution time distribution. This budget is calculated with the aim of minimizing potential budget overruns. We first introduce a new statistical dispersion parameter called the maximum coefficient of variation (VWCET). This parameter describes the tendency of execution times towards the worst-case execution time and can be adjusted according to the criticality of the task. We show through experiments that the proposed heuristic reduces the probability of exceeding the time budget allocated to the task, and improves the performance of low-criticality tasks.

Recently, deep neural networks (DNNs) have been utilized in real-time systems such as autonomous vehicles, where meeting temporal constraints is essential. However, executing such systems on CPU-GPU architectures can make scheduling analysis challenging due to the added delays caused by computing and memory contention. In addition, classic task models are not directly able to model accurately such systems. In this article, we propose a new task model called DNN Task Model (DTM). This model considers both DNN properties and GPU architecture at the same time. It allows us to distinguish between CPU and GPU tasks, provides information about the DNN application and give more accurate execution time analysis through consideration of data quality. We compute DTM from a source CUDA file and a set of real-time specifications of the system. The proposed model is extensible enough to be adopted to various DNN type applications allowing designer to compare candidate software and GPU architectures. Furthermore, we propose a graph optimization inspired by Tensor-RT.

In this paper, we prove that a mean system utilization smaller than one is a necessary condition for the feasibility of real-time systems. Such systems are defined asstable. Stable systems have two distinct states: a transient state, followed by a steady-state where the same distribution of response times is repeated infinitely for each task. We prove that the Liu and Layland theorem holds for stable probabilistic real-time systems with implicit deadlines, we provide an analytical approximation of response times for each of those two states and a bound of the instant when a real-time system becomes steady.

The real-time systems community is facing the lack of benchmarks adapted to measurement-based worst-case execution time (WCET) estimators. We provide in this paper first steps towards such benchmarks by proposing them for single core microcontrollers, while we leave as future work the migration to multicore microcontrollers. The considered benchmarks are the programs of an open source drone autopilot. We conclude the paper by underlining the main difficulties of such migration.

In this paper we present a probabilistic response time analysis for mixed criticality real-time systems running on a single processor according to a fixed priority pre-emptive scheduling policy. The analysis extends the existing state of the art probabilistic analysis to the case of mixed criticalities, taking into account both the level of assurance at which each task needs to be certified, as well as the possible criticalities at which the system may execute. The proposed analysis is formally presented as well as explained with the aid of an illustrative example.

The paper addresses the real-time fixed-priority scheduling problem for battery-powered embedded systems whose energy storage unit is replenished by an environmental energy source. In this context, a task may meet its deadline only if its cost of energy can be satisfied early enough. Hence, a scheduling policy for such a system should account for properties of the source of energy, capacity of the energy storage unit and tasks cost of energy. Classical fixed-priority schedulers are no more suitable for this model. Based on these motivations, we propose PFPASAPan optimal scheduling algorithm that handles both energy and timing constraints. Furthermore, we state the worst case scenario for non concrete task sets a non concrete task set is a set of real-time tasks whose offsets are known only at run-time scheduled with this algorithm and build a necessary and sufficient feasibility condition for non concrete task sets. Moreover, a minimal bound of the storage unit capacity that keeps a task set schedulable with PFPASAPis also proposed. Finally, we validate the proposed theory with large scale simulations and compare our algorithm with other existing ones.

In this paper, we propose feasibility and schedulability tests for a real-time scheduling problem under energy constraints. We first introduce the problem and show how to model it using timed automata. We then propose a feasibility test based on CTL model checking and schedulability tests for EDF and Preemptive Fixed Priority algorithms (PFP). Our approach also permits to generate a feasible schedule if one exists or otherwise to find how to correct battery characteristics to make the problem feasible. It is finally possible to generate schedules that optimize some criteria, such as the number of context switches between the battery recharging and discharging modes, the minimal and the maximal energy levels reached during the execution, or the number of preemptions. The approach is illustrated by some experiments using the model checking tool UPAAL [1].

Energy harvesting is the process of generating electrical energy from environmental sources such as solar panels. In recent years, this term has been frequently applied in the context of small autonomous devices such as wireless sensor nodes. The classical scheduling theory is insufficient for this kind of systems and new scheduling problems arise in this context. Until now, the research on this area focused in trying to improve the efficiency of existing algorithms. Our approach is to complete these efforts by a feasibility theory allowing us to understand why classical optimal algorithms are not efficient anymore with energy constraints.In this paper, we try to establish a schedulability test for a fixed priority real-time scheduling problem with energy constraints. We first introduce the problem and describe the model. Then, to illustrate the difficulty of the problem, we focus on a preemptive fixed priority scheduling policy where all the executions are postponed as long as possible. This policy lets the harvester the maximal amount of time to refill the battery. We call this policy PFPALAPfor As Late As Possible. We try to define sufficient and/or necessary schedulability conditions and discuss its potential optimality under some additional assumptions. Then, through simple counter examples, we show that intuitive assumptions are wrong for this scheduling problem, making it very interesting to study.

We propose a subclass of timed game automata(TGA), called Task TGA, representing networks of communicating tasks where the system can choose when to start the task and the environment can choose the duration of the task. We search to solve finite-horizon reachability games on Task TGA by building strategies in the form of Simple Temporal Networks with Uncertainty (STNU). Such strategies have the advantage of being very succinct due to the partial order reduction of independent tasks.We show that the existence of such strategies is an NP-complete problem. A practical consequence of this result is a fully forward algorithm for building STNU strategies.Potential applications of this work are planning and scheduling under temporal uncertainty.