Automatic Extraction of Pipeline Parallelism for Embedded Software Using Linear Programming

Cordes, Daniel; Heinig, Andreas; Marwedel, Peter; Mallik, Arindam

doi:10.1109/icpads.2011.31

Cited by 18 publications

(7 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Subhlok et al considered a model where every task can be perfectly parallelized on homogeneous processors and provided mapping solutions for opti mizing throughput [9] as well as solutions for latency throughput trade-offs [10] . Recently, Kudlur et al and Cordes et al used integer linear programming to extract data parallelism from streaming pipelines [11], [12]. While they considered replication, they all assumed that the work of replicated stages should be evenly divided and assigned to replicas.…”

Section: Related Workmentioning

confidence: 99%

Adding data parallelism to streaming pipelines for throughput optimization

Agrawal

Buhler

Chamberlain

2013

20th Annual International Conference on High Performance Computing

View full text Add to dashboard Cite

The streaming model is a popular model for writing high-throughput parallel applications. A streaming application is represented by a graph of computation stages that communicate with each other via FIFO channels. In this paper, we consider the problem of mapping streaming pipelines -streaming applications where the graph is a linear chain -onto a set of computing resources in order to maximize its throughput. In a parallel setting, subsets of stages, called components, can be mapped onto different computing resources. The throughput of an application is determined by the throughput of the slowest component.Therefore, if some stage is much slower than others, then it may be useful to replicate the stage's code and divide its workload among two or more replicas in order to increase throughput. However, pipelines may consist of some replicable and some non-replicable stages. In this paper, we address the problem of mapping these partially replicable streaming pipelines onto both homogeneous and heterogeneous platforms so as to maximize throughput.We consider two types of platforms, homogeneous platforms -where all resources are identical, and hetero geneous platforms -where resources may have different speeds. In both cases, we consider two network topologies -unidirectional chain and clique. We provide polynomial time algorithms for mapping partially replicable pipelines onto unidirectional chains for both homogeneous and heterogeneous platforms. For homogeneous platforms, the algorithm for unidirectional chains generalizes to clique topologies. However, for heterogeneous platforms, map ping these pipelines onto clique topologies is NP-complete.We provide heuristics to generate solutions for cliques by applying our chain algorithms to a series of chains sampled from the clique. Our empirical results show that these heuristics rapidly converge to near-optimal solutions.

show abstract

Section: Related Workmentioning

confidence: 99%

Adding data parallelism to streaming pipelines for throughput optimization

Agrawal

Buhler

Chamberlain

2013

20th Annual International Conference on High Performance Computing

View full text Add to dashboard Cite

show abstract

“…One of the possible solutions to deal with the heterogeneity of the underlying platform is to use a model that to consider different execution times for loop iterations of the program. Thus, these execution times are taken into account to select the processing units according to its performance characteristics, as well as, to determine the number of processing units that are used simultaneously . An efficient and economical way of using computational resources is to adopt policies and approaches for deciding the task granularity at run time based on computational resource utilisation constraints, quality of service requirements, and the average task deployment metrics .…”

Section: Future Research Directionsmentioning

confidence: 99%

Survey on the run‐time systems of enterprise application integration platforms focusing on performance

Freire¹,

Frantz²,

Roos‐Frantz³

et al. 2018

Softw Pract Exp

View full text Add to dashboard Cite

Companies are taking advantage of cloud computing to upgrade their business processes. Cloud computing requires interaction with many kinds of applications, so it is necessary to improve the performance of software tools that allow keeping information on all these applications consistent and synchronised. Integration platforms are specialised software tools that provide support to design, implement, run, and monitor integration solutions, which aim to orchestrate a set of applications so as to promote compatibility among their data or to develop new functionality on top of the current ones. The run-time system is the part of the integration platform responsible for running the integration solutions, which makes its performance the uttermost important issue. The contribution of this article is two-fold: a framework and an evaluation of integration platforms. The former is a framework composed of ten properties grouped into two dimensions to evaluate the run-time systems focusing on performance. Using this framework as reference, the second contribution is an evaluation of nine open-source integration platforms, which represent the state-of-the-art, provide support to the integration patterns, and follow the pipes-and-filters architectural style. In addition, as a result of this work, we suggest open research directions that can be explored to improve the performance of the run-time systems and at the same time may be useful to adapt them to the context of cloud computing. KEYWORDScloud computing, enterprise application integration, integration patterns, integration platform, run-time system Companies often need to use their software ecosystems 1,2 to support and improve their business processes. These ecosystems are composed of many applications, usually designed without taking into account their possible integration. The field of study known as Enterprise Integration Applications (EAI) seeks to provide methodologies, techniques, and tools for the design and implementation of integration solutions. 3 In general terms, an integration solution aims to orchestrate a set of applications to keep them synchronised or provide new features that can be built from those already developed. An integration solution is composed of processes that contain the integration logic and ports that encapsulate adapters with communication protocols to connect applications to the integration solution. . 341 language is focused on the elaboration of conceptual models for the integration solution, with a level of abstraction close to the domain of the problem. The development toolkit is a set of tools that allows the implementation of the solution, that is, the transformation of the conceptual model into executable code. The testing environment allows testing individual parts or the entire integration solution, with the aim of identifying and eliminating possible defects in the implementation. The monitoring tool is used to monitor, at run time, the operation of the integration solution and detects errors that may occur during message proce...

show abstract

“…In [21], an ILP formulation is utilized to search a smaller design space and find an appropriate configuration for ASIPs, with the objective of minimizing the system area and satisfying system runtime constraints in pipelined processors. An ILP-based mapping approach is presented in [22] to minimize the most expensive path in a pipeline under the constraints of program dependency and the maximal number of concurrently executed components. In summary, the aforementioned methods do not significantly consider the discussed two factors (cyclic task dependency and communication overhead) in software pipeline.…”

Section: Related Workmentioning

confidence: 99%

Software Pipeline–Based Partitioning Method with Trade-Off between Workload Balance and Communication Optimization

Huang¹,

Xiu²,

Yu³

et al. 2015

ETRI J

View full text Add to dashboard Cite

For a multiprocessor System‐on‐Chip (MPSoC) to achieve high performance via parallelism, we must consider how to partition a given application into different components and map the components onto multiple processors. In this paper, we propose a software pipeline–based partitioning method with cyclic dependent task management and communication optimization. During task partitioning, simultaneously considering computation load balance and communication optimization can cause interference, which leads to performance loss. To address this issue, we formulate their constraints and apply an integer linear programming approach to find an optimal partitioning result — one that requires a trade‐off between these two factors. Experimental results on a reconfigurable MPSoC platform demonstrate the effectiveness of the proposed method, with 20% to 40% performance improvements compared to a traditional software pipeline–based partitioning method.

show abstract

Automatic Extraction of Pipeline Parallelism for Embedded Software Using Linear Programming

Cited by 18 publications

References 20 publications

Adding data parallelism to streaming pipelines for throughput optimization

Adding data parallelism to streaming pipelines for throughput optimization

Survey on the run‐time systems of enterprise application integration platforms focusing on performance

Software Pipeline–Based Partitioning Method with Trade-Off between Workload Balance and Communication Optimization

Contact Info

Product

Resources

About