Ces travaux s'inscrivent dans le cadre du projet AROS (Automotive Robust Operating Services). Il a pour objectif de proposer un outil de prototypage rapide d'applications dynamiques distribuées temps-réel, principalement dans le domaine de l'automobile et de la robotique. Les applications distribuées temps-réel sont traditionnellement développées selon deux approches. La première, l'ordonnancement temporel, est basée sur l'analyse du pire temps d'exécution (worst execution time). Un partage du temps entre les différentes tâches de l'application est établi de façon statique. Cette technique offre une grande sureté de fonctionnement au prix d'une analyse temporelle parfois difficile à mener. La seconde, l'ordonnancement par priorité, est basée sur l'attribution à chaque tâche d'un niveau de priorité qui permet d'établir l'ordre d'exécution en fonction des évènements reçus par le système. Cette seconde technique, plus souple à mettre en œuvre, offre moins de garanties et conduit à un comportement non déterministe de l'application. La structure des applications AROS étant dynamique, l'approche temporelle est exclue car elle demande une analyse statique qu'il est impossible de produire. L'approche basée sur les priorités d'exécution est également exclue à cause de son non déterminisme comportemental. Nous proposons une approche basée sur un ordonnancement évènementiel causal inspirée des techniques d'ordonnancement des simulateurs évènementiels distribués. Tout en étant relativement simple à utiliser pour le concepteur d'application, cette technique produit des applications dont le comportement est parfaitement déterministe. Deux principales difficultés sont à surmonter : la synchronisation en temps-réel du moteur d'exécution et le respect des contraintes temps-réel.

Nowadays, there is a high demand for computer vision applications in numerous application areas, such as autonomous driving or unmanned aerial vehicles. However, the application areas and scenarios are becoming increasingly complex, and their data requirements are growing. To meet these requirements, it needs increasingly powerful computing systems. FPGA-based heterogeneous systems offer an excellent solution in terms of energy efficiency, flexibility, and performance, especially in the field of computer vision. Due to complex applications and the use of FPGAs in combination with other architectures, efficient programming is becoming increasingly difficult. Thus, developers need a comprehensive framework with efficient automation, good usability, reasonable abstraction, and seamless integration of tools. It should provide an easy entry point, and reduce the effort to learn new concepts, programming languages and tools. Additionally, it needs optimized libraries for the user to focus on developing applications without getting involved with the underlying details. These should be well integrated, easy to use, and cover a wide range of possible use cases. The framework needs efficient algorithms to execute applications on heterogeneous architectures with maximum performance. These algorithms should distribute applications across various nodes with low fragmentation and communication overhead and find a near-optimal solution in a reasonable amount of time. This thesis addresses the research problem of an efficient implementation of object detection applications, their distribution across FPGA-based heterogeneous systems, and methods for automation and integration using toolchains. Within this, the three contributions are the HiFlipVX object detection library, the DECISION framework, and the APARMAP application distribution algorithm. HiFlipVX is an open-source HLS-based FPGA library optimized for performance and resource efficiency. It contains 66 highly parameterizable computer vision functions including neural networks, ideally for design space exploration. It extends the OpenVX standard for feature extraction, which is challenging due to unknown element size at design time. All functions are streaming capable to achieve maximum performance by increasing parallelism and reducing off-chip memory access. It does not require external or vendor libraries, which eases project integration, device coverage, and vendor portability, as shown for Intel. The library consumed on average 0.39% FFs and 0.32% LUTs for a set of image processing functions compared to a vendor library. A HiFlipVX implementation of the AKAZE feature detector computes between 3.56 and 4.13 times more pixels per second than the related work, while its resource consumption is comparable to optimized VHDL designs. Its neural network extension achieved a speedup of 3.23 for an AlexNet layer in comparison to a related work, while consuming 73% less on-chip memory. Furthermore, this thesis proposes an improved feature extraction implementation that achieves a repeatability of 72.57% when weighting complex cases, while the next best algorithm only achieves 62.99 %. DECISION is a framework consisting of two toolchains for the efficient programming of FPGA-based heterogeneous systems. Both integrate HiFlipVX and use a joint OpenVXbased frontend to implement computer vision applications. It abstracts the underlying hardware and algorithm details while covering a wide range of architectures and applications. The first toolchain targets x86-based systems consisting of CPUs, GPUs, and FPGAs using OpenCL (Open Computing Language). To create a heterogeneous schedule, it considers device profiles, kernel profiles and estimates, and FPGA dataflow characteristics. It manages synchronization, memory transfers and data coherence at design time. It creates a runtime optimized program which excels by its high parallelism and a low overhead. Additionally, this thesis looks at the integration of OpenCL-based libraries, automatic OpenCL kernel generation, and OpenCL kernel optimization and comparison for different architectures. The second toolchain creates an application specific and adaptive NoC-based architecture. The streaming-optimized architecture enables the reusability of vision functions by multiple applications to improve the resource efficiency while maintaining high performance. For a set of example applications, the resource consumption was more than halved, while its overhead was only 0.015% in terms of performance. APARMAP is an application distribution algorithm for partition-based and mesh-like FPGA topologies. It uses a NoC (Network-on-Chip) as communication infrastructure to connect reconfigurable regions and generate an application-specific hardware architecture. The algorithm uses load balancing techniques to find reasonable solutions within a predictable and scalable amount of time. It optimizes solutions using various heuristics, such as Simulated Annealing and Tabu Search. It uses a multithreaded grid-based approach to prevent threads from calculating the same solution and getting stuck in local minimums. Its constraints and objectives are the FPGA resource utilization, NoC bandwidth consumption, NoC hop count, and execution time of the proposed algorithm. The evaluation showed that the algorithm can deal with heterogeneous and irregular host graph topologies. The algorithm showed a good scalability in terms of computation time for an increasing number of nodes and partitions. It was able to achieve an optimal placement for a set of example graphs up to a size of 196 nodes on host graphs of up to 49 partitions. For a real application with 271 nodes and 441 edges, it was able to achieve a distribution with low resource fragmentation in an average time of 149 ms.

In a large class of distributed embedded systems, most of the code generation models in use today target at object-oriented applications. Distributed computing using procedural language applications is challenging because there are no compatible code generators to test the partitioned programs mapped on to the multi-processor system. In this thesis, we design a code generator to produce procedural language code for a distributed embedded system. Unpartitioned programs along with the partition primitives are converted into independently executable concrete implementations. The process consists of two steps, first translating the primitives of the unpartitioned program into equivalent code clusters, and then scheduling the implementations of these code clusters according to the data dependency inherent in the unpartitioned program. Communication and scheduling of the partitioned programs require the original source code to be reverse engineered. A reverse engineering tool is used for the creation of a metadata table describing the program elements and dependency trees. This data gathered, is used by Parallel Virtual Machine message passing system for communication between the partitioned programs in the distributed environment. The proposed Code Generation Model has been implemented using C and the experimental results are presented for various test cases. Further, metrics are developed for testing the quality of the results taking into consideration the correctness and the execution of time.

The diploma thesis answers set questions that relate to two different distribution environments, the two options for developers to directly offer their mobile applications to the end users. Distribution is being conducted in the App Store and Google Play application stores, where only such applications that oblige basic requirements can be released and provided to customers. The requirements, official sources of information and support services of the two providers are analyzed in the theoretical part of the thesis. Processes revolving around the application distribution itself are described in the following practical part. Distribution consists of processes from the developer account registration to submitting the application for a review or a check. Sufficient knowledge base could be gathered not only due to a proper exploration of the issue, but also due to an actual application release accomplished by the author. A cross-platform mobile application to be released has been developed solely for the purposes of this diploma thesis. The application has been built in two platform specific versions. Whilst publishing it, an emphasis was laid on keeping the conditions as similar as possible to achieve viable case study results. The application release represents the case that produced all data that enabled a direct comparison and overall evaluation of the different developer settings at the end of the thesis. Findings and conclusions of this thesis can be used by developers and other interested parties that are looking for an integrated source of information that regards releasing a cross-platform mobile application on iOS and Android platform.