Clustered Software Queue for Efficient Pipelined Multithreading

Zhang, Yuanming; Ootsu, Kanemitsu; Yokota, Takashi; Baba, Takanobu

doi:10.1109/pdcat.2009.24

Cited by 6 publications

(11 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The four cores share one L3 cache. In order to further reduce the inter-core overheads, we adopted a clustered SW queue technique [14] in pipeline. When the queue size is 64*64, the inter-core communication latency is about 20 cycles.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Automatic Thread Decomposition for Pipelined Multithreading

Zhang

Ootsu

Yokota

et al. 2010

2010 IEEE 16th International Conference on Parallel and Distributed Systems

Self Cite

View full text Add to dashboard Cite

An appropriate automatic thread decomposition approach is critical for pipelined multithreading (PMT) to maximize pipeline performance with balanced thread size on target multi-core processor. This paper presents an automatic thread decomposition approach, which maps the pipeline thread decomposition problem onto a graph-theoretic framework to construct an optimized DAG with minimal bottleneck node size and balanced node size under constrained core number. In this approach, control dependence is treated as special data dependence and then an effective mechanism is proposed to remove redundant control dependences. A heuristic decomposition algorithm is given to generate an optimized pipeline. The algorithm has been evaluated on a commodity multi-core processor, and experimental results show that it has achieved speedup ranging from 113% to 174% on several SPEC CPU 2000 benchmark programs.

show abstract

Section: Discussionmentioning

confidence: 99%

“…Communication overheads have great impact on the pipeline performance [14]. To minimize communication overheads, we should avoid any unnecessary inter-thread communication.…”

Section: Minimum Communication Overheadsmentioning

confidence: 99%

Automatic Thread Decomposition for Pipelined Multithreading

Zhang

Ootsu

Yokota

et al. 2010

2010 IEEE 16th International Conference on Parallel and Distributed Systems

Self Cite

View full text Add to dashboard Cite

show abstract

“…Clustered software queue: The clustered software queue [20] -or CSQ -delays the production of data to minimize data sharing and use a bit flip to notify when the communication buffer is full or empty. Despite an algorithm similar to BatchQueue, two differences remain between them.…”

Section: Related Workmentioning

confidence: 99%

An Improvement of OpenMP Pipeline Parallelism with the BatchQueue Algorithm

Preud'Homme

Sopena

Thomas

et al. 2012

2012 IEEE 18th International Conference on Parallel and Distributed Systems

View full text Add to dashboard Cite

In the context of multicore programming, pipeline parallelism is a solution to easily transform a sequential program into a parallel one without requiring a whole rewriting of the code. The OpenMP stream-computing extension presented by Pop and Cohen proposes an extension of OpenMP to handle pipeline parallelism. However, their communication algorithm relies on Multiple-producer-Multiple-Consumer queues, while pipelined applications mostly deal with linear chains of communication, i.e., with only a single producer and a single consumer.To improve the performance of the OpenMP streamextension, we propose to add a more specialized SingleProducer-Single-Consumer communication algorithm called BatchQueue and to select it for one-to-one communication. Our evaluation shows that BatchQueue is then able to improve the throughput up to a factor 2 on an 8-core machine both for example application and real applications. Our study shows therefore that using specialized and efficient communication algorithms can have a significant impact on the overall performance of pipelined applications.

show abstract

“…Nevertheless, according to [10], multi-threading can be used to minimize the waste of processing cycles in a single core when many requests are made to access external memory. And this being the case, throughput can be increased.…”

Section: Definition Of the Experimentsmentioning

confidence: 99%

“…Using multiple cores, different stages of a pipeline can run concurrently. If each stage is assigned to a different core, the packet budget is effectively multiplied by the number of cores [10].…”

Section: Definition Of the Experimentsmentioning

confidence: 99%

How different messaging semantics can affect SCA applications performances: a benchmark comparison

Bernier

Latour

Zapata

2011

Analog Integr Circ Sig Process

View full text Add to dashboard Cite

Software Communications Architecture (SCA) compliant radios typically contain a large number of software components. Some software components provide access to hardware devices while others perform signal processing. By interacting with each other, the software components implement a radio communications standard. To interact, the software components use a middleware called Common Object Request Broker Architecture (CORBA). Using CORBA, each interaction is carried out as an exchange of messages between two components. CORBA supports two main types of messaging: one-way and twoway. This article explores the differences between the two types of messaging and provides performance metrics. The article also describes design approaches that can be used to avoid common pitfalls associated with the use of both types of messaging.

show abstract

Clustered Software Queue for Efficient Pipelined Multithreading

Cited by 6 publications

References 21 publications

Automatic Thread Decomposition for Pipelined Multithreading

Automatic Thread Decomposition for Pipelined Multithreading

An Improvement of OpenMP Pipeline Parallelism with the BatchQueue Algorithm

How different messaging semantics can affect SCA applications performances: a benchmark comparison

Contact Info

Product

Resources

About