Synchronous Dataflow Graphs Research Articles

Efficient Pipelining of Synchronous Dataflow Graphs Via Graph Conversion

Efficient Pipelining of Synchronous Dataflow Graphs Via Graph Conversion

Efficient Computation of the Max-Plus Semantics of Synchronous Dataflow Graphs

Streaming systems are naturally modeled with synchronous dataflow graphs (SDFGs). The max-plus semantics of an SDFG is a compact matrix representation of its timing behavior. The max-plus semantics enables us to analyze and control timing properties of the systems, such as the obtainable minimum guaranteed throughput and maximum latency. Deriving the max-plus semantics, and consequently, performance analysis may be computationally expensive since the state-of-the-art method simulates one iteration of an SDFG. This holds, in particular, for systems whose components operate at different levels of granularity, as this results in many executions of some components in one iteration. This paper aims at efficiently calculating the max-plus semantics of SDFGs. The paper proposes an optimization framework exploring decompositions of a given SDFG and finding a composition sequence whose computational effort for compositionally obtaining the max-plus semantics is minimal. Not only does our proposed technique accelerate the performance analysis of multiscale streaming systems, but it also allows us to compute the max-plus semantics of some systems for which the state-of-the-art method does not succeed because of memory limitations.

IoT based dynamic Bayesian prediction of crop evapotranspiration in soilless cultivations

As water resources become scarce around the globe, a sustainable irrigation management is essential for increasing the crop use efficiency. Crop EvapoTranspiration (ET) is considered as one of the fundamental variables for correct irrigation management. Current ET prediction models adjust their system parameters: i) during time consuming training cycles; ii) by (destructive) leaf-area measurements; iii) only once during training for a particular meteorological scenario. These parameters are static. They cannot learn over time. On top of that, these models fail when some of the input variables are missing. To overcome these shortcomings, we follow an Internet of Things (IoT) approach to crop water use modeling and prediction for soilless cultivations. First, a stochastic crop coefficient confined to global radiation, Kr, is defined. Then, the evolution of Kr is modeled as state-space system with the daily temperature as input data, and global radiation and crop weight as measurement data. The latter data may also be missing at random. We derive a predictor for the posteriori probability of Kr and show that predicting crop ET via Kr it is immune to propagation of uncertainty over time due to erroneous weather forecasting. Our predictor is implemented in the cloud server of the IoT network. The sensor nodes, connected to a lightweight IoT client, supply our model with information in (soft) real-time. A web interface allows for visualization of the time series in real-time. To discuss Quality of Service (QoS) issues, we model the IoT network as synchronous data flow graph revealing its workflow and timings of the individual nodes.Sweet basil (Ocimum basilicum L.) is used to test the performance in four climatic environments and two different sampling methods: i) sample daily from the same environment in parallel; ii) sample daily from one environment and move on to the next in round robin fashion, emulating reuse of the same set of measurement equipment at a quarter of the costs. It turns out that in both cases, our Kr-based crop ET predictor outperforms the competing Baille technique. The performance of our IoT network is assessed by means of the QoS metrics response time, jitter, and availability.

Iterative Probabilistic Performance Prediction for Multiple IoT Applications in Contention

Internet of Things (IoT) has become omnipresent in many applications, such as healthcare, vehicles, and precision farming. They sense data from dozens of sensors scheduled periodically in a synchronous fashion on mobile CPUs that are forwarded to the cloud or other IoT devices via an essentially stochastic wireless channel. Hence, the task response time becomes stochastic, preventing optimization at compile time. On the other hand, knowing response time at compile time along with jitter, availability, and scalability is crucial to ensure a certain level of Quality of Service. This contribution presents a stochastic framework for performance analyses of multiapplications on a possible multiprocessor platform. When annotated with (stochastic) execution time, a traditional synchronous dataflow (SDF) graph can be transformed into a directed acyclic workflow graph, revealing the timing of individual actors. A generalized version of the rejection sampling Monte Carlo algorithm explores the properties of the workflow graph, to determine the distribution of the response time in a single application as well as a multiapplication multiple access scenario. Mean and jitter are the moments of the distribution. An IoT toy example with a number of distributed smart sensors was deployed in real environments to assess the performance of the proposed framework. Our analysis framework works at compile time of the code, scales with the number of things, and has low computational complexity.

DFSynthesizer: Dataflow-based Synthesis of Spiking Neural Networks to Neuromorphic Hardware

Spiking Neural Networks (SNNs) are an emerging computation model that uses event-driven activation and bio-inspired learning algorithms. SNN-based machine learning programs are typically executed on tile-based neuromorphic hardware platforms, where each tile consists of a computation unit called a crossbar, which maps neurons and synapses of the program. However, synthesizing such programs on an off-the-shelf neuromorphic hardware is challenging. This is because of the inherent resource and latency limitations of the hardware, which impact both model performance, e.g., accuracy, and hardware performance, e.g., throughput. We propose DFSynthesizer, an end-to-end framework for synthesizing SNN-based machine learning programs to neuromorphic hardware. The proposed framework works in four steps. First, it analyzes a machine learning program and generates SNN workload using representative data. Second, it partitions the SNN workload and generates clusters that fit on crossbars of the target neuromorphic hardware. Third, it exploits the rich semantics of the Synchronous Dataflow Graph (SDFG) to represent a clustered SNN program, allowing for performance analysis in terms of key hardware constraints such as number of crossbars, dimension of each crossbar, buffer space on tiles, and tile communication bandwidth. Finally, it uses a novel scheduling algorithm to execute clusters on crossbars of the hardware, guaranteeing hardware performance. We evaluate DFSynthesizer with 10 commonly used machine learning programs. Our results demonstrate that DFSynthesizer provides a much tighter performance guarantee compared to current mapping approaches.

Improved Scheduling Algorithm for Synchronous Data Flow Graphs on a Homogeneous Multi-Core Systems

In order to accelerate the execution of streaming applications on multi-core systems, this article studies the scheduling problem of synchronous data flow graphs (SDFG) on homogeneous multi-core systems. To describe the data flow computation process, we propose the SDAG (Super DAG) computation model based on the DAG model combined with data-driven thoughts. Further, we analyze the current common SDFG scheduling algorithms and propose an improved SDFG scheduling algorithm, LSEFT (level-shortest-first-earliest-finish time). The LSEFT algorithm uses an inverse traversal algorithm to calculate the priority of tasks in the task-selection phase; the shortest-job-priority earliest-finish-time policy is used in the processor selection phase to replace the original long job priority policy. In the experimental part, we designed an SDFG random generator and generated 958 SDFGs with the help of the random generator as test cases to verify the scheduling algorithm. The experimental results show that our improved algorithm performs well for different numbers of processor cores, especially for 8 cores, where the speedup of our improved algorithm improves by 10.17% on average.

Optimal Offloading for Streaming Applications in Mobile Edge Computing

With the rapid development of smart mobile devices, mobile applications are becoming more and more popular. Since mobile devices usually have constrained computing capacity, computation offloading to mobile edge computing (MEC) to achieve a lower latency is a promising paradigm. In this paper, we focus on the optimal offloading problem for streaming applications in MEC. We present solutions to find offloading policies of streaming applications to achieve an optimal latency. Streaming applications are modeled with synchronous data flow graphs. Two architecture assumptions are considered — with sufficient processors on both the local device and the MEC server, and with a limited number of processors on both sides. The problem is generally NP-complete. We present an exact algorithm and a heuristic algorithm for the former architecture assumption and a heuristic method for the latter. We carry out our experiments on a practical application and thousands of synthetic graphs to comprehensively evaluate our methods. The experimental results show that our methods are effective and computationally efficient.

Evaluation of the Exact Throughput of a Synchronous DataFlow Graph

Synchronous DataFlow Graph (SDFG in short) is a formalism frequently considered in electronic design and software compilers to model communications between components with different rates. The development of efficient algorithms to evaluate the maximum throughput of SDFGs is a challenging question. This paper presents a mathematical framework to perform schedulability analysis and to compute the maximum throughput of SDFGs. This work focuses on strictly K-Periodic schedules for which a fixed set of execution times coupled with a period are associated with each task and define a schedule of every task executions. This class of schedules can always reach maximal throughput; we present an algorithm that computes the exact maximum throughput by iteratively generating K-periodic schedules until we reach optimality. The complexity of this iterative algorithm is studied by using the well-established benchmarking suite SDF3, and compared against the most common throughput analysis techniques. We show several orders of magnitude improvement over state-of-the-art both in terms of computation time, and size of the final schedules.

Code-size-aware Scheduling of Synchronous Dataflow Graphs on Multicore Systems

Synchronous dataflow graphs are widely used to model digital signal processing and multimedia applications. Self-timed execution is an efficient methodology for the analysis and scheduling of synchronous dataflow graphs. In this article, we propose a communication-aware self-timed execution approach to solve the problem of scheduling synchronous dataflow graphs on multicore systems with communication delays. Based on this communication-aware self-timed execution approach, four communication-aware scheduling algorithms are proposed using different allocation rules. Furthermore, a code-size-aware mapping heuristic is proposed and jointly used with a proposed scheduling algorithm to reduce the code size of SDFGs on multicore systems. The proposed scheduling algorithms are experimentally evaluated and found to perform better than existing algorithms in terms of throughput and runtime for several applications. The experiments also show that the proposed code-size-aware mapping approach can achieve significant code size reduction with limited throughput degradation in most cases.

Cooperative Coevolution-based Design Space Exploration for Multi-mode Dataflow Mapping

Some signal processing and multimedia applications can be specified by synchronous dataflow (SDF) models. The problem of SDF mapping to a given set of heterogeneous processors has been known to be NP-hard and widely studied in the design automation field. However, modern embedded applications are becoming increasingly complex with dynamic behaviors changes over time. As a significant extension to the SDF, the multi-mode dataflow (MMDF) model has been proposed to specify such an application with a finite number of behaviors (or modes) and each behavior (mode) is represented by an SDF graph. The multiprocessor mapping of an MMDF is far more challenging as the design space increases with the number of modes. Instead of using traditional genetic algorithm (GA)-based design space exploration (DSE) method that encodes the design space as a whole, this article proposes a novel cooperative co-evolutionary genetic algorithm (CCGA)-based framework to efficiently explore the design space by a new problem-specific decomposition strategy in which the solutions of node mapping for each individual mode are assigned to an individual population. Besides, a problem-specific local search operator is introduced as a supplement to the global search of CCGA for further improving the search efficiency of the whole framework. Furthermore, a fitness approximation method and a hybrid fitness evaluation strategy are applied for reducing the time consumption of fitness evaluation significantly. The experimental studies demonstrate the advantage of the proposed DSE method over the previous GA-based method. The proposed method can obtain an optimization result with 2×−3× better quality using less (1/2−1/3) optimization time.

Formal verification of timed synchronous dataflow graphs using Lustre

The timed synchronous dataflow graph model is a graphical model of computation that allows concurrency between processes. This model is widely used due to its expressive power, semantic simplicity and predefined properties. However, the lack of support for formal verification makes it difficult to manually verify user-defined properties, even for small-sized graphs. This paper presents a formal verification solution of timed synchronous dataflow graphs based on the Lustre language and model checking. The solution consists in an automatic translation of synchronous dataflow graphs into Lustre code and in coding the user-defined properties as Lustre expressions. Use cases are presented showing the applicability of the method.

Parallel Scheduling of Multiple SDF Graphs Onto Heterogeneous Processors

Parallel scheduling of multiple real-time applications onto heterogeneous processors is needed in the emerging embedded systems such as self-driving cars, smart cameras, and smartphones. Assuming that an embedded application is specified as a synchronous dataflow (SDF) graph or its extension, we propose a novel parallel scheduling methodology based on an evolutionary algorithm where the mapping of tasks onto processors is evolved to optimize a given objective function in an iterative fashion. In each iteration, we use an existing worst-case response time (WCRT) analysis tool to check if all applications satisfy their real-time requirements by translating each SDF graph into a directed acyclic graph (DAG) that is assumed in the WCRT analysis tool. Since the WCRT analysis must be performed in each iteration of evolution, we propose a clustering technique to reduce drastically the analysis time that depends on the number of nodes and their dependency. We formally prove that the proposed clustering technique does not change the estimated WCRT of each application. The effectiveness of the proposed scheduling methodology with the clustering technique is verified with extensive experiments using real-life benchmarks, randomly generated graphs, and the comparison with the existing technique.

Monitoring Practitioner's Skills in Pure-Tone Audiometry

So far, there exists no standard, to evaluate a practitioner's skills in pure-tone audiometry. To narrow the gap, this article presents an artificial patient (AP) emulating various types of hearing impairment. In contrast to other solutions, the AP autonomously listens to real pure-tones in soft real-time, while taking into account elements from psycho-acoustics. The emulated patient profiles are reproducible. New profiles can be easily added. The AP is able to recover from error. In this contribution, the authors develop software requirements specifications and derive a modular system architecture. To analyze the performance, the article proposes a stochastic extension to existing synchronous data flow graphs, taking into account the unbounded nature of the tasks' worst case response time. Maximization and summation over the graph reveals the joint distribution of the response time with first and second central moments corresponding to, respectively, the expected response time and the jitter of the task. The theoretical results have finally been validated by measurements on the target.

Latency Evaluation of SDFGs on Heterogeneous Processors Using Timed Automata

Synchronous Data Flow (SDF) is a graphical computation model used for analyzing digital signal processing and real time multimedia applications. In general, these applications have two primary performance metrics - throughput and latency. Latency is important in multimedia processing applications such as video-conferencing, Internet telephony and games since latency surpassing a specific limit results in poor quality of service (QoS). Past work had focused on computing the latency of SDF graphs on homogeneous multiprocessor platforms. In this paper, we present an approach to compute the latency of a static schedule for a given unfolding factor with an optimal throughput for an SDF graph on a heterogeneous multiprocessor platform using timed automata. We use timed automata as a semantic model to represent the system model, which includes a synchronous data flow graph and an execution platform. We use the UPPAAL model-checker to specify the resulting network of timed automata and compute the latency.

Resource Optimization for Real-Time Streaming Applications Using Task Replication

In this paper, we study the problem of exploiting parallelism in a hard real-time streaming application modeled as an acyclic synchronous data flow (SDF) graph and scheduled on a heterogeneous multiprocessor system-on-chip platform to alleviate the capacity fragmentation due to partitioned scheduling algorithms and reduce the number of required processors when a throughput requirement is satisfied. As the main contribution in this paper, we propose a method to determine a replication factor for each task in an acyclic SDF graph such that by distributing the workloads among more parallel tasks with lower utilization in the obtained transformed graph, the left capacity on the processors can be efficiently exploited, hence reducing the number of required processors. The experimental results, on a set of real-life streaming applications, demonstrate that our approach can reduce the minimum number of processors required to schedule an application and considerably improve the memory requirements and application latency compared to related approaches while meeting the same throughput constraint.

Compositionality in scenario-aware dataflow: a rendezvous perspective

Finite-state machine-based scenario-aware dataflow (FSM-SADF) is a dynamic dataflow model of computation that combines streaming data and finite-state control. For the most part, it preserves the determinism of its underlying synchronous dataflow (SDF) concurrency model and only when necessary introduces the non-deterministic variation in terms of scenarios that are represented by SDF graphs. This puts FSM-SADF in a sweet spot in the trade-off space between expressiveness and analyzability. However, FSM-SADF supports no notion of compositionality, which hampers its usability in modeling and consequent analysis of large systems. In this work we propose a compositional semantics for FSM-SADF that overcomes this problem. We base the semantics of the composition on standard composition of processes with rendezvous communication in the style of CCS or CSP at the control level and the parallel, serial and feedback composition of SDF graphs at the dataflow level. We evaluate the approach on a case study from the multimedia domain.

Uniprocessor scheduling of real-time synchronous dataflow tasks

The synchronous dataflow graph (SDFG) model is widely used today for modeling real-time applications in safety-critical application domains. Schedulability analysis techniques that are well understood within the real-time scheduling community are applied to the analysis of recurrent real-time workloads that are represented using this model. An enhancement to the standard SDFG model is proposed, which supports the specification of a real-time latency constraint between a specified input and a specified output of an SDFG. A polynomial-time algorithm is derived for representing the computational requirement of each such enhanced SDFG task in terms of the notion of the demand bound function (dbf), which is widely used in real-time scheduling theory for characterizing computational requirements of recurrent processes represented by, e.g., the sporadic task model. By so doing, the extensive dbf-centered machinery that has been developed in real-time scheduling theory for the hard-real-time schedulability analysis of systems of recurrent tasks may be applied to the analysis of systems represented using the SDFG model as well. The applicability of this approach is illustrated by applying prior results from real-time scheduling theory to construct an exact preemptive uniprocessor schedulability test for collections of independent recurrent processes that are each represented using the enhanced SDFG model.

A Novel Methodology on Optimizing the Performance of Multi-core Processor Using FPGA

With rapid progress in microelectronic technology and higher performance requirements, multi-core processors (MP) have become widely used in various applications. Among them, multiprocessor System-on-Chip (MPSoC) received focus because of its high integration level and low power solution. This work proposed more efficient way to implement MPSoC with higher performance. Synchronous data flow (SDF) is used to model high-speed MPSoC systems, and to simulate the system level functional algorithms. After building a suitable FPGA platform architecture, we prototype these SDF graphs onto FPGA based MPSoC platform. Simulation results show that the MPSoC platform has higher performance than a single processor platform, with novel architecture and scheduling strategy used. Furthermore, different scheduling strategies, including static order and run-time order, are compared on MPSoC. The results have shown that the optimized static order strategy can achieve higher performance than run-time order strategy, with less buffer cost.

Worst-Case Response Time Analysis of a Synchronous Dataflow Graph in a Multiprocessor System with Real-Time Tasks

In this article, we propose a novel technique that estimates a tight upper bound of the worst-case response time (WCRT) of a synchronous dataflow (SDF) graph when the SDF graph shares processors with other real-time tasks. When an SDF graph is executed at runtime under a self-timed or static assignment scheduling policy on a multi-processor system, static scheduling of the SDF graph does not guarantee the satisfaction of latency constraints since changes to the schedule may result in timing anomalies. To estimate the WCRT of an SDF graph with a given mapping and scheduling result, we first construct a task instance dependency graph that depicts the dependency between node executions in a static schedule. The proposed technique combines two techniques in a novel way: schedule time bound analysis and response time analysis. The former is used to consider the interference between task instances in the same SDF graph, and the latter is used to consider the interference from other real-time tasks. Through extensive experiments with synthetic examples and benchmarks, we verify the superior performance of the proposed technique compared to other existent techniques.

Multiprocessor Scheduling of a Multi-Mode Dataflow Graph Considering Mode Transition Delay

The Synchronous Data Flow (SDF) model is widely used for specifying signal processing or streaming applications. Since modern embedded applications become more complex with dynamic behavior changes at runtime, several extensions of the SDF model have been proposed to specify the dynamic behavior changes while preserving static analyzability of the SDF model. They assume that an application has a finite number of behaviors (or modes), and each behavior (mode) is represented by an SDF graph. They are classified as multi-mode dataflow models in this article. While there exist several scheduling techniques for multi-mode dataflow models, no one allows task migration between modes. By observing that the resource requirement can be additionally reduced if task migration is allowed, we propose a multiprocessor scheduling technique of a multi-mode dataflow graph considering task migration between modes. Based on a genetic algorithm, the proposed technique schedules all SDF graphs in all modes simultaneously to minimize the resource requirement. To satisfy the throughput constraint, the proposed technique calculates the actual throughput requirement of each mode and the output buffer size for tolerating throughput jitter. We compare the proposed technique with a method that analyzes SDF graphs in each execution mode separately, a method that does not allow task migration, and a method that does not allow mode-overlapped schedule for synthetic examples and five real applications: H.264 decoder, lane detection, vocoder, MP3 decoder, and printer pipeline.