An Implementation of the Codelet Model

Suettlerlein, Joshua; Zuckerman, Stéphane; Gao, Guang R.

doi:10.1007/978-3-642-40047-6_63

Cited by 42 publications

(24 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In particular, the DAGuE runtime [Bosilca et al 2012], the University of Delaware codelet model [Suettlerlein et al 2013] and the related SWARM environment [ETI International 2014], and the T* runtime of OpenStream [Pop and Cohen 2013] are all "feed-forward" or "argument fetch" data-driven models. In such runtimes, tasks notify their successors upon completion.…”

Section: Related Workmentioning

confidence: 99%

Compiler/Runtime Framework for Dynamic Dataflow Parallelization of Tiled Programs

Kong

Pop

Pouchet

et al. 2015

ACM Trans. Archit. Code Optim.

View full text Add to dashboard Cite

Task-parallel languages are increasingly popular. Many of them provide expressive mechanisms for intertask synchronization. For example, OpenMP 4.0 will integrate data-driven execution semantics derived from the StarSs research language. Compared to the more restrictive data-parallel and fork-join concurrency models, the advanced features being introduced into task-parallel models in turn enable improved scalability through load balancing, memory latency hiding, mitigation of the pressure on memory bandwidth, and, as a side effect, reduced power consumption.In this article, we develop a systematic approach to compile loop nests into concurrent, dynamically constructed graphs of dependent tasks. We propose a simple and effective heuristic that selects the most profitable parallelization idiom for every dependence type and communication pattern. This heuristic enables the extraction of interband parallelism (cross-barrier parallelism) in a number of numerical computations that range from linear algebra to structured grids and image processing. The proposed static analysis and code generation alleviates the burden of a full-blown dependence resolver to track the readiness of tasks at runtime. We evaluate our approach and algorithms in the PPCG compiler, targeting OpenStream, a representative dataflow task-parallel language with explicit intertask dependences and a lightweight runtime. Experimental results demonstrate the effectiveness of the approach.

show abstract

Section: Related Workmentioning

confidence: 99%

Compiler/Runtime Framework for Dynamic Dataflow Parallelization of Tiled Programs

Kong

Pop

Pouchet

et al. 2015

ACM Trans. Archit. Code Optim.

View full text Add to dashboard Cite

show abstract

“…Project Runnemede [6,21,25] introduces codelets, which are similar to i-lets and are supported directly by the operating system [11]. Codelets are small self-contained units of computation with run-to-completion semantics assumed by default.…”

Section: Related Workmentioning

confidence: 99%

Cutting out the middleman: OS-level support for x10 activities

Mohr¹,

Buchwald²,

Zwinkau³

et al. 2015

Proceedings of the ACM SIGPLAN Workshop on X10

View full text Add to dashboard Cite

In the X10 language, computations are modeled as lightweight threads called activities. Since most operating systems only offer relatively heavyweight kernel-level threads, the X10 runtime system implements a user-space scheduler to map activities to operating-system threads in a many-to-one fashion. This approach can lead to suboptimal scheduling decisions or synchronization overhead. In this paper, we present an alternative X10 runtime system that targets OctoPOS, an operating system designed from the ground up for highly parallel workloads on PGAS architectures. OctoPOS offers an unconventional execution model based on i-lets, lightweight self-contained units of computation with (mostly) runto-completion semantics that can be dispatched very efficiently. We are able to do a 1-to-1 mapping of X10 activities to i-lets, which results in a slim runtime system, avoiding the need for user-level scheduling and its costs. We perform microbenchmarks on a prototype many-core hardware architecture and show that our system needs fewer than 2000 clock cycles to spawn local and remote activities.

show abstract

“…This hybrid approach can exploit the principle of computing and data locality exhibited by massive multithreaded applications to confine the traffic mostly inside the rings for better traffic management [6,25] (not shown as it is out of the scope of this paper). Also, using efficient run-time systems (e.g., the Codelet model [49]) can offer the opportunity to exploit such locality easily. Our proposed architecture provides a customised router architecture which processes the traffic and also bypasses it when necessary for better latency improvement.…”

Section: Introductionmentioning

confidence: 99%

Ring-mesh: a scalable and high-performance approach for manycore accelerators

Mazumdar

Scionti

2019

J Supercomput

View full text Add to dashboard Cite

There are increasing number of works addressing the design challenges of fast, scalable solutions for the growing number of new type of applications. Recently, many of the solutions aimed at improving processing element capabilities to speed up the execution of machine learning application domain. However, only a few works focused on the interconnection subsystem as a potential source of performance improvement. Wrapping many cores together offer excellent parallelism, but it brings other challenges (e.g., adequate interconnections). Scalable, power-aware interconnects are required to support such a growing number of processing elements, as well as modern applications. In this paper, we propose a scalable and energy efficient Network-on-Chip architecture fusing the advantages of rings as well as the 2D-mesh without using any bridge router to provide high-performance. A dynamic adaptation mechanism allows to better adapt to the application requirements. Simulation results show efficient power consumption (up to 141.3% saving for connecting 1024 cores), 2x (on average) throughput growth with better scalability (up to 1024 processing elements) compared to popular 2D-mesh while tested in multiple statistical traffic pattern scenarios.

show abstract

An Implementation of the Codelet Model

Cited by 42 publications

References 18 publications

Compiler/Runtime Framework for Dynamic Dataflow Parallelization of Tiled Programs

Compiler/Runtime Framework for Dynamic Dataflow Parallelization of Tiled Programs

Cutting out the middleman: OS-level support for x10 activities

Ring-mesh: a scalable and high-performance approach for manycore accelerators

Contact Info

Product

Resources

About