Marco Platzner

Operating systems for reconfigurable embedded platforms: online scheduling of real-time tasks

Todays reconfigurable hardware devices have huge densities and are partially reconfigurable, allowing for the configuration and execution of hardware tasks in a true multitasking manner. This makes reconfigurable platforms an ideal target for many modern embedded systems that combine high computation demands with dynamic task sets. A rather new line of research is engaged in the construction of operating systems for reconfigurable embedded platforms. Such an operating system provides a minimal programming model and a runtime system. The runtime system performs online task and resource management. In this paper, we first discuss design issues for reconfigurable hardware operating systems. Then, we focus on a runtime system for guarantee-based scheduling of hard real-time tasks. We formulate the scheduling problem for the 1D and 2D resource models and present two heuristics, the horizon and the stuffing technique, to tackle it. Simulation experiments conducted with synthetic workloads evaluate the performance and the runtime efficiency of the proposed schedulers. The scheduling performance for the 1D resource model is strongly dependent on the aspect ratios of the tasks. Compared to the 1D model, the 2D resource model is clearly superior. Finally, the runtime overhead of the scheduling algorithms is shown to be acceptably low.

Fast online task placement on FPGAs: free space partitioning and 2D-hashing

Partial reconfiguration allows for mapping and executing several tasks on an FPGA during runtime. Multitasking on FPGAs raises a number of questions on the management of the reconfigurable resource which leads to the concept of a reconfigurable operating system. A major aspect of such an operating system is task placement. Online placement methods are required that achieve a high placement quality and lead to efficient implementations. This paper presents placement methods that rely on efficient algorithms for the partitioning of the reconfigurable resource and a hash matrix data structure to maintain the free space. Given n as the number of currently placed tasks, previously known placers find a feasible location in O(n) time. Our approach is able to find a feasible location in constant time. Additionally, simulations show that our methods improve the placement quality by up to 70%.

Virtualizing Hardware with Multi-context Reconfigurable Arrays

In contrast to processors, current reconfigurable devices totally lack programming models that would allow for device independent compilation and forward compatibility. The key to overcome this limitations is hardware virtualization. In this paper, we resort to a macro-pipelined execution model to achieve hardware virtualization for data streaming applications. As a hardware implementation we present a hybrid multi-context architecture that attaches a coarse-grained reconfigurable array to a host CPU. A co-simulation framework enables cycle-accurate simulation of the complete architecture. As a case study we map an FIR filter to our virtualized hardware model and evaluate different designs. We discuss the impact of the number of contexts and the feature of context state on the speedup and the CPU load.

Online Scheduling for Block-Partitioned Reconfigurable Devices

This paper presents our work toward an operating system that manages the resources of a reconfigurable device in a multitasking manner. We propose an online scheduling system that allocates tasks to a block-partitioned reconfigurable device. The blocks are statically-fixed but can have different widths, which allows the matching of the computational resources with the task requirements. We implement several non-preemptive and preemptive schedulers as well as different placement strategies. Finally, we present a simulation environment that allows the experimental investigation of the effects of specific partitioning, placement, and scheduling methods.

Online scheduling and placement of real-time tasks to partially reconfigurable devices

This paper deals with online scheduling of tasks to partially reconfigurable devices. Such devices are able to execute several tasks in parallel. All tasks share the reconfigurable surface as a single resource which leads to highly dynamic allocation situations. To manage such devices at runtime, we propose a reconfigurable operating system that splits into three main modules: scheduler, placer, and loader. The main characteristic of the resulting online scheduling problem is the strong nexus between scheduling and placement. We discuss a fast online placement technique and then focus on scheduling real-time tasks. We devise guarantee-based schedulers for two scenarios, namely tasks with arbitrary and synchronous arrival times. The schedulers exploit the knowledge about task properties to improve the systems performance. The experiments show that the developed schedulers lead to substantial performance gains at an acceptable runtime overhead.

A Runtime Environment for Reconfigurable Hardware Operating Systems

We present a runtime environment that partially reconfigures and executes hardware tasks on Xilinx Virtex. To that end, the FPGA’s reconfigurable surface is split into a varying number of variable-sized vertical task slots that can accommodate the hardware tasks. A bus-based communication infrastructure allows for task communication and I/O. We discuss the design of the runtime system and its prototype implementation on an reconfigurable board architecture that was specifically tailored to reconfigurable hardware operating system research.

ReconOS: An Operating System Approach for Reconfigurable Computing

The ReconOS operating system for reconfigurable computing offers a unified multithreaded programming model and OS services for threads executing in software and threads mapped to reconfigurable hardware. The OS interface lets hardware threads interact with software threads using well-known mechanisms such as semaphores, mutexes, condition variables, and message queues. By semantically integrating hardware accelerators into a standard OS environment, ReconOS allows for rapid design-space exploration, supports a structured application development process, and improves the portability of applications between different reconfigurable computing systems.

Fluctuating emg signals: Investigating long-term effects of pattern matching algorithms

In this paper, we investigate the behavior of state-of-the-art pattern matching algorithms when applied to electromyographic data recorded during 21 days. To this end, we compare the five classification techniques k-nearest-neighbor, linear discriminant analysis, decision trees, artificial neural networks and support vector machines. We provide all classifiers with features extracted from electromyographic signals taken from forearm muscle contractions, and try to recognize ten different hand movements. The major result of our investigation is that the classification accuracy of initially trained pattern matching algorithms might degrade on subsequent data indicating variations in the electromyographic signals over time.

Partially Reconfigurable Cores for Xilinx Virtex

Recent generations of high-density and high-speed FPGAs provide a sufficient capacity for implementing complete configurable systems on a chip (CSoCs). Hybrid CPUs that combine standard CPU cores with reconfigurable coprocessors are an important subclass of CSoCs. With partially reconfigurable FPGAs, coprocessors can be loaded on demand while the CPU remains running. However, the lack of high-level design tools for partial reconfiguration makes practical implementations a challenging task.In this paper, we introduce a design flow to implement hybrid processors on Xilinx Virtex. The design flow is based on two techniques, virtual sockets and feed-through components, and can efficiently generate partial configurations from industry-quality cores. We discuss the design flow and present a fully operational audio streaming prototype to demonstrate its feasibility.

Heuristics for Online Scheduling Real-Time Tasks to Partially Reconfigurable Devices

Partially reconfigurable devices allow to configure and execute tasks in a true multitasking manner. The main characteristics of mapping tasks to such devices is the strong nexus between scheduling and placement. In this paper, we formulate a new online real-time scheduling problem and present two heuristics, the horizon and the stuffing technique, to tackle it. Simulation experiments evaluate the performance and the runtime efficiency of the schedulers. Finally, we discuss our prototyping work toward an integration of scheduling and placement into an operating system for reconfigurable devices.