data flow graphs are a popular program representation in machine learning, big data analytics, signal processing, and, increasingly, networking, where graph nodes correspond to processing primitives and graph edges de...
详细信息
ISBN:
(纸本)9781939133137
data flow graphs are a popular program representation in machine learning, big data analytics, signal processing, and, increasingly, networking, where graph nodes correspond to processing primitives and graph edges describe control flow. To improve CPU cache locality and exploit data-level parallelism, nodes usually process data in batches. Unfortunately, as batches are split across dozens or hundreds of parallel processing pathways through the graph they tend to fragment into many small chunks, leading to a loss of batch efficiency. We present Batchy, a scheduler for run-to-completion packet processing engines, which uses controlled queuing to efficiently reconstruct fragmented batches in accordance with strict service-level objectives (SLOs). Batchy comprises a runtime profiler to quantify batch-processing gain on different processing functions, an analytical model to fine-tune queue backlogs, a new queuing abstraction to realize this model in run-to-completion execution, and a one-step receding horizon controller that adjusts backlogs across the pipeline. We present extensive experiments on five networking use cases taken from an official 5G benchmark suite to show that Batchy provides 2-3x the performance of prior work while accurately satisfying delay SLOs.
A novel robust watermarking methodology for intellectual property core protection of control data flow graphs during behavioural synthesis based on user specified resource constraints and loop unrolling factor is prop...
详细信息
A novel robust watermarking methodology for intellectual property core protection of control data flow graphs during behavioural synthesis based on user specified resource constraints and loop unrolling factor is proposed. The proposed methodology incorporates robust multi-variable signature encoding scheme for securing against typical attacks such as false claim of ownership and tampering. Some of the features of the watermarking include greater robustness, lower embedding cost, lower storage overhead, fault tolerance and faster encoding/decoding. Experimental results suggest that the watermarked solution generated by the proposed approach produces lower embedding cost than previous approaches, as well as is more secured due to its multi-variable signature encoding rules.
data flow graphs are widely used for modeling and analysis of real-time streaming applications in which having a predictable and reliable implementation is an essential requirement. In this paper, we consider scheduli...
详细信息
ISBN:
(纸本)9783319390833;9783319390826
data flow graphs are widely used for modeling and analysis of real-time streaming applications in which having a predictable and reliable implementation is an essential requirement. In this paper, we consider scheduling a set of data flow graphs such that liveness and boundedness properties are guaranteed, which leads to a predictable and correct behavior of the application. A formal translation method is proposed to map a given set of data flow graphs to a set of graph-based real-time tasks. Additionally, sufficient conditions are derived under which the obtained task set provides a semantically correct implementation of the given data flow graphs. It is shown that the proposed approach provides a higher level of design flexibility compared to the existing methods which use a simpler, i.e. periodic, task model.
Simultaneous design space exploration (DSE) of k(c)-cycle transient fault-secured datapath and loop unrolling factor (UF) for control data flow graphs (CDFGs) during high-level synthesis is an unsolved problem in the ...
详细信息
Simultaneous design space exploration (DSE) of k(c)-cycle transient fault-secured datapath and loop unrolling factor (UF) for control data flow graphs (CDFGs) during high-level synthesis is an unsolved problem in the literature. The aforementioned problems are solved with the following specific contributions: (i) a novel DSE methodology that explores an optimal combination of transient fault-secured datapath configuration and loop UF using particle swarm optimisation for CDFG, and (ii) the novel k(c)-cycle fault secured algorithm. Finally, the proposed approach when tested on standard benchmarks yielded an optimal solution which minimises the fitness function and satisfies user constraints, as well as achieving significantly reduced cost of the final solution when compared with a recent fault-secured approach.
Synchronous data flow graphs (SDFGs) are widely used to model digital signal processing and streaming media applications. In this paper, we present exact methods for static optimal scheduling and mapping of SDFGs on a...
详细信息
ISBN:
(纸本)9783319192499;9783319192482
Synchronous data flow graphs (SDFGs) are widely used to model digital signal processing and streaming media applications. In this paper, we present exact methods for static optimal scheduling and mapping of SDFGs on a heterogenous multiprocessor platform. The optimization criteria we consider are throughput and energy consumption, taking into account the combination of various constraints such as auto-concurrency and buffer sizes. We present a concise and flexible (priced) timed automata semantics of system models, which include an SDFG and a multiprocessor platform, and formulate the optimization goals as temporal logic formulas. The optimization and scheduling problems are then transformed to model checking problems, which are solved by UPPAAL (CORA). Thanks to the exhaustive exploration nature of model checking and the facility of the tools, we obtain two pareto-optimal schedules, one with an optimal throughput and a best energy consumption and another with an optimal energy consumption and a best throughput. The approach is applied to two real applications with different parameters. The case studies show that our approach can deal with moderate models within reasonable execution time and reveal the impacts of different constraints on optimization goals.
data flow graphs can conveniently model embedded streaming applications (ESAs) that are typically implemented as networks of concurrent tasks having an iterative pipelined execution, where the activation of each task ...
详细信息
ISBN:
(纸本)9781479963072
data flow graphs can conveniently model embedded streaming applications (ESAs) that are typically implemented as networks of concurrent tasks having an iterative pipelined execution, where the activation of each task may be conditioned by intra-and inter-iteration data dependencies. We propose a novel analysis approach for preemptive Fixed Priority Scheduling (FPS) of multiple ESAs assuming a fixed mapping of tasks onto the processors of the underlying Heterogeneous Multi-Processor System-on-Chip (HMPSoC). The tasks of an ESA are event activated, have varying execution times, and participate in cyclic dependency chains such that they may not have an activation pattern that can be depicted using traditional periodic / sporadic event models. Instead we propose to characterize the data flow graphs of ESAs to upper bound the load they impose on a processor and use it to compute the worst-case response time of an actor executing on that processor at a lower priority. We show that ours is a generic approach for analyzing FPS of data flow graphs. We also propose a refinement of our technique for graphs with a dominant periodic source. We demonstrate our improvement over the state-of-the-art FPS analysis for dataflow in our experiments.
作者：
Chen, YiliWan, ZhengLi, YangyangHe, XiaoWei, XianHan, JunFujian Normal Univ
Coll Comp & Cyber Secur Fuzhou 350117 Peoples R China East China Normal Univ
Shanghai Engn Res Ctr Mol Therapeut & New Drug Dev Shanghai Frontiers Sci Ctr Mol Intelligent Synth Sch Chem & Mol Engn Shanghai 200062 Peoples R China Chinese Acad Sci
Acad Math & Syst Sci Beijing 100190 Peoples R China East China Normal Univ
Chongqing Key Lab Precis Opt Chongqing Inst Chongqing 401120 Peoples R China New York Univ Shanghai
New York Univ Ctr Computat Chem Sch Chem & Mol EngnEast China Normal Univ Shanghai 200062 Peoples R China East China Normal Univ
MoE Engn Res Ctr Hardware Software Codesign Techno Shanghai 200062 Peoples R China Chinese Acad Sci
Quanzhou Inst Equipment Mfg Fujian Inst Res Struct Matter Quanzhou 362216 Peoples R China
Graph neural networks (GNNs) have revolutionized drug discovery in chemistry and biology, enhancing efficiency and reducing resource demands. However, classical GNNs often struggle to capture long-range dependencies d...
详细信息
Graph neural networks (GNNs) have revolutionized drug discovery in chemistry and biology, enhancing efficiency and reducing resource demands. However, classical GNNs often struggle to capture long-range dependencies due to challenges like oversmoothing and oversquashing. Graph Transformers address these issues by employing global self-attention mechanisms that allow direct information exchange between any pair of nodes, enabling the modeling of long-range interactions. Despite this, Graph Transformers often face difficulties in capturing the nuanced structural information on graphs. To overcome these challenges, we introduce the Curvflow-Transformer, a novel graph Transformer model incorporating a curvature flow-based masked attention mechanism. By leveraging a topologically enhanced mask matrix, the attention layer can effectively detect subtle structural differences within graphs, balancing the focus between global mutual information and local structural details of molecules. The Curvflow-Transformer demonstrates superior performance on the MoleculeNet data set, surpassing several state-of-the-art models across various tasks. Moreover, the model provides unique insights into the relationship between molecular structure and chemical properties by analyzing the attention heat coefficients of individual atoms.
This paper proposes a mechanized formal semantics for dataflow circuits: rather than following a predetermined, static schedule, the execution of the circuit components is constrained solely by the availability of the...
详细信息
In this article, a model-driven structured approach is adopted to develop a decision support system for the capacitated vehicle routing problem. A repository of artefacts is developed through system initiation, analys...
详细信息
Information on control and dataflows within a program is often required when building tools that support or investigate software development and maintenance activities. However, existing source code analysis tools fo...
详细信息
暂无评论