Implementation, Scheduling, and Adaptation of Partial Expansion Graphs on Multicore Platforms

Zaki, George; Plishker, William; Bhattacharyya, Shuvra S.; Fruth, Frank

doi:10.1007/s11265-016-1107-8

Cited by 7 publications

(10 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [21], the optimal vectorization of an SDFG is achieved by multiplying the rates of the original graph by integers resulting in less invocation of the actors of the SDFG. Partial Expansion Graphs (sPEGs) [24] formulation provides a framework in which the vectorization of actors is integrated efficiently for multiprocessor scheduling context. Zaki et al use Particle Swarm Optimization (PSO) to find and adjust the amount of expansion, or vectorization, of the actors of the graph.…”

Section: Avoiding Graph Expansionmentioning

confidence: 99%

Numerical Representation of Directed Acyclic Graphs for Efficient Dataflow Embedded Resource Allocation

Arrestier

Desnos

Juárez

et al. 2019

ACM Trans. Embed. Comput. Syst.

View full text Add to dashboard Cite

Stream processing applications running on Heterogeneous Multi-Processor Systems on Chips (sHMPSoCs) require efficient resource allocation and management, both at compile-time and at runtime. To cope with modern adaptive applications whose behavior can not be exhaustively predicted at compile-time, runtime managers must be able to take resource allocation decisions on-the-fly, with a minimum overhead on application performance. Resource allocation algorithms often rely on an internal modeling of an application. Directed Acyclic Graphs (sDAGs) are the most commonly used models for capturing control and data dependencies between tasks. DAGs are notably often used as an intermediate representation for deploying applications modeled with a dataflow Model of Computation (MoC) on HMPSoCs. Building such intermediate representation at runtime for massively parallel applications is costly both in terms of computation and memory overhead. In this paper, an intermediate representation of DAGs for resource allocation is presented. This new representation shows improved performance for run-time analysis of dataflow graphs with less overhead in both computation time and memory footprint. The performances of the proposed representation are evaluated on a set of computer vision and machine learning applications.

show abstract

Section: Avoiding Graph Expansionmentioning

confidence: 99%

Numerical Representation of Directed Acyclic Graphs for Efficient Dataflow Embedded Resource Allocation

Arrestier

Desnos

Juárez

et al. 2019

ACM Trans. Embed. Comput. Syst.

View full text Add to dashboard Cite

show abstract

“…This operation artificially reduces the repetition vector size, or decreases the values held by the repetition vector. This method has been employed for SDF graphs under real-time constraints [25], and also under data-driven scheduling of Partial Expansion Graphs (PEG) [24]. The StreamIT benchmark has also been successfully transformed into coarse SDF graphs for the RAW architecture [7], by an unfolding technique using actor fusion and fission.…”

Section: On the Clustering Of Dataflow Graphsmentioning

confidence: 99%

Modeling Nested for Loops with Explicit Parallelism in Synchronous DataFlow Graphs

Honorat

Desnos

Pelcat

et al. 2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

A common problem when developing signal processing applications is to expose and exploit parallelism in order to improve both throughput and latency. Many programming paradigms and models have been introduced to serve this purpose, such as the Synchronous DataFlow (SDF) Model of Computation (MoC). SDF is used especially to model signal processing applications. However, the main difficulty when using SDF is to choose an appropriate granularity of the application representation, for example when translating imperative functions into SDF actors. In this paper, we propose a method to model the parallelism of perfectly nested for loops with any bounds and explicit parallelism, using SDF. This method makes it possible to easily adapt the granularity of the expressed parallelism, thanks to the introduced concept of SDF iterators. The usage of SDF iterators is then demonstrated on the Scale Invariant Feature Transform (SIFT) image processing application.

show abstract

“…Partial expansion graphs (PEG) [4,5,10] are proposed to address these limitations. This approach allows the designer or design tool to control the degree to which each actor is expanded rather than expanding actors based on their SDF repetition counts.…”

Section: Introductionmentioning

confidence: 99%

“…Problem statement: Existing PEG solutions are not well adapted to achieve efficient scheduling on many-core architectures. They involve the use of shared buffers [10] or the addition of split/join actors [5]. The first solution requires the implementation of a buffer manager to coordinate data production and consumption of expanded actors on shared buffers.…”

Section: Introductionmentioning

confidence: 99%

Toward Efficient Many-core Scheduling of Partial Expansion Graphs

Tran

Bhattacharyya

Talpin

et al. 2018

Proceedings of the 21st International Workshop on Software and Compilers for Embedded Systems

Self Cite

View full text Add to dashboard Cite

Transformation of synchronous data flow graphs (SDF) into equivalent homogeneous SDF representations has been extensively applied as a pre-processing stage when mapping signal processing algorithms onto parallel platforms. While this transformation helps fully expose task and data parallelism, it also presents several limitations such as an exponential increase in the number of actors and excessive communication overhead. Partial expansion graphs were introduced to address these limitations for multi-core platforms. However, existing solutions are not well-suited to achieve efficient scheduling on many-core architectures. In this article, we develop a new approach that employs cyclo-static data flow techniques to provide a simple but efficient method of coordinating the data production and consumption in the expanded graphs. We demonstrate the advantage of our approach through experiments on real application models.

show abstract

Implementation, Scheduling, and Adaptation of Partial Expansion Graphs on Multicore Platforms

Cited by 7 publications

References 21 publications

Numerical Representation of Directed Acyclic Graphs for Efficient Dataflow Embedded Resource Allocation

Numerical Representation of Directed Acyclic Graphs for Efficient Dataflow Embedded Resource Allocation

Modeling Nested for Loops with Explicit Parallelism in Synchronous DataFlow Graphs

Toward Efficient Many-core Scheduling of Partial Expansion Graphs

Contact Info

Product

Resources

About