Filters








2,415 Hits in 7.1 sec

Dynamic control flow in large-scale machine learning

Yuan Yu, Peter Hawkins, Michael Isard, Manjunath Kudlur, Rajat Monga, Derek Murray, Xiaoqiang Zheng, Martín Abadi, Paul Barham, Eugene Brevdo, Mike Burrows, Andy Davis (+3 others)
2018 Proceedings of the Thirteenth EuroSys Conference on - EuroSys '18  
Our approach extends the use of dataflow graphs to represent machine learning models, offering several distinctive features.  ...  These applications benefit from the ability to make rapid control-flow decisions across a set of computing devices in a distributed system.  ...  ACKNOWLEDGMENTS We gratefully acknowledge contributions from our colleagues at Google and from members of the wider machine learning community, and the feedback that we have received from them and from  ... 
doi:10.1145/3190508.3190551 dblp:conf/eurosys/YuABBBDDGHHIKMM18 fatcat:5u4gcsi5fba33mv2nyni32h424

Spinning fast iterative data flows

Stephan Ewen, Kostas Tzoumas, Moritz Kaufmann, Volker Markl
2012 Proceedings of the VLDB Endowment  
The extension alleviates for the lack of mutable state in dataflows and allows for exploiting the sparse computational dependencies inherent in many iterative algorithms.  ...  Training a model over a large data corpus frequently requires a dataflow (like MapReduce) for preprocessing the data (e. g., for joining different sources and normalization), a specialized system for the  ...  The abstraction integrates well with the dataflow programming paradigm, can exploit the inherent computational dependencies between data elements, allowing for very efficient execution of many graph and  ... 
doi:10.14778/2350229.2350245 fatcat:mlgvsxdlrngntdm33n2drq6h5m

Spinning Fast Iterative Data Flows [article]

Stephan Ewen, Kostas Tzoumas, Moritz Kaufmann, Volker Markl
2012 arXiv   pre-print
The extension alleviates for the lack of mutable state in dataflows and allows for exploiting the sparse computational dependencies inherent in many iterative algorithms.  ...  In our experiments, the improved dataflow system is highly competitive with specialized systems while maintaining a transparent and unified dataflow abstraction.  ...  The abstraction integrates well with the dataflow programming paradigm, can exploit the inherent computational dependencies between data elements, allowing for very efficient execution of many graph and  ... 
arXiv:1208.0088v1 fatcat:urzipkk2bbbgzeh6qyayphzvoq

GoDEL: A Multidirectional Dataflow Execution Model for Large-Scale Computing

Abhishek Kulkarni, Michael Lang, Andrew Lumsdaine
2011 2011 First Workshop on Data-Flow Execution Models for Extreme Scale Computing  
Implemented with efficiency and programmer productivity as its goals, we describe the syntax and semantics of the GoDEL language and discuss its implementation and runtime.  ...  Dataflow programming models, being declarative in nature, lead to improved programmability at scale by implicitly managing the computation and communication for the application.  ...  This work was partly performed at the Ultrascale Systems Research Center (USRC), a collaboration between Los Alamos National Laboratory and the New Mexico Consortium (NMC).  ... 
doi:10.1109/dfm.2011.12 fatcat:oohxvm4xlndcrp4qkam2wxhi3e

An Empirical Evaluation of Flow Based Programming in the Machine Learning Deployment Context [article]

Andrei Paleyes, Christian Cabrera, Neil D. Lawrence
2022 arXiv   pre-print
Results reveal that FBP is a suitable paradigm for data collection and data science tasks, and is able to simplify data collection and discovery when compared with SOA.  ...  Evaluation is done with respect to different application domains, ML deployment stages, and code quality metrics.  ...  A duality of control flow and data flow for building software was explored by the computer science community before [16, 23, 50] .  ... 
arXiv:2204.12781v1 fatcat:gpb3ktcqbnfffndeeed6gstg3a

Control Flow Versus Data Flow in Distributed Systems Integration: Revival of Flow-Based Programming for the Industrial Internet of Things

Wilhelm Hasselbring, Maik Wojcieszak, Schahram Dustdar, Schahram Dustdar
2021 IEEE Internet Computing  
We discuss coupling in distributed systems integration and reflect the history of business process modeling with respect to data and control flow.  ...  When we consider the application layer of networked infrastructures, data and control flow are important concerns in distributed systems integration.  ...  remote procedure calls, asynchronous messaging, and data streaming support different forms of dataflow and control flow across distributed integrated systems, with various degrees of coupling introduced  ... 
doi:10.1109/mic.2021.3053712 fatcat:cvqlq4eya5chnamrjq46bmrkri

Component reuse methodology for multi-clock Data-Flow parallel embedded Systems

Anne Marie Chana, Patrice Quinton, Steven Derrien
2014 ARIMA  
We discuss the relationship of this approach with multi-clock architecture, latency-insensitive design, multidimensional data-flow systems and stream programming  ...  Flexible parallel components are assembled using a scheduling method which combines the synchronous data-flow principle of balance equations and the polyhedral scheduling technique.  ...  They are based on the same model, the synchronous dataflow graph. Stream programming have been developed to handle systems computing on streams of data.  ... 
doi:10.46298/arima.1979 fatcat:bi6gyfa27jd5vatza752ubttae

The MUSE machine—An architecture for structured data flow computation

David F. Brailsford, R. James Duckworth
1985 New generation computing  
problems with the unconstrained generation of fresh result tokens if a pure data flow approach is adopted.  ...  Although data-driven computation frees the architecture from the constraints of the single program counter, processor and global memory, inherent in the classic von Neumann computer, there can still be  ...  The construction of the hardware was financed by Plessey Office Systems, Beeston, Nottingham to whom we are also grateful, in conjunction with the Science and Engineering Research Council of Great Britain  ... 
doi:10.1007/bf03037068 fatcat:slaonuymtrerbf2fpysmlneraa

SCnC: Efficient Unification of Streaming with Dynamic Task Parallelism

Dragos Sbirlea, Jun Shirako, Ryan Newton, Vivek Sarkar
2011 2011 First Workshop on Data-Flow Execution Models for Extreme Scale Computing  
CnC is a general purpose parallel programming paradigm with a task-parallel look and feel but based on dataflow graph principles. Its expressivity extends to any arbitrary task graph.  ...  Stream processing is a special form of the dataflow execution model that offers extensive opportunities for optimization and automatic parallelization.  ...  A comparison between the task-based dataflow model and streaming is started in [17] , but their work relies on special language and the comparison with data-flow can only be taken as a guideline, as their  ... 
doi:10.1109/dfm.2011.13 fatcat:rljjcq2evrc3zi3qpnvpfoqa3i

A New Compilation Flow for Software-Defined Radio Applications on Heterogeneous MPSoCs

Mickaël Dardaillon, Kevin Marquet, Tanguy Risset, Jérôme Martin, Henri-Pierre Charles
2016 ACM Transactions on Architecture and Code Optimization (TACO)  
Fig. 6 : 6 compute() method and reduced graph expression of the graph ofFig. 4in PaDaF. Fig. 7 : 7 Details of the front end compilation flow with C++ examples for clarity.  ...  Evaluation of the compilation flow is presented in Section 6 and related works are presented in Section 7.  ... 
doi:10.1145/2910583 fatcat:zvgvyrsahrbjlkretgonnbnbk4

TTADF: Power Efficient Dataflow-Based Multicore Co-Design Flow

Ilkka Hautala, Jani Boutellier, Olli Silven
2019 IEEE transactions on computers  
Index Terms-Dataflow, application specific processor, design flow, low power. ! • I. Hautala and Olli Silvén are with  ...  The results of the TTADF co-design-based solutions are compared against previous manually created designs and a recent dataflow-based design flow, showing that TTADF provides very high energy efficiency  ...  ACKNOWLEDGMENTS The work was partially funded by the Academy of Finland project 309693 UNICODE and Tauno Tönning Foundation.  ... 
doi:10.1109/tc.2019.2937867 fatcat:bkhi52rqkbhivdvtffztgp2424

Couillard: Parallel programming via coarse-grained Data-flow Compilation

Leandro A.J. Marzulo, Tiago A.O. Alves, Felipe M.G. França, Vítor Santos Costa
2014 Parallel Computing  
More specifically, for an application that follows the wavefront method, running with big inputs, Trebuchet achieved up to 4.7% speedup over Intel Ò TBB novel flow-graph approach and up to 44% over OpenMP  ...  be run in parallel and connect them in a data-flow graph.  ...  Acknowledgments We thank FAPERJ, CAPES and the Euro-Brazilian Windows consortium for the financial support given to the authors of this work.  ... 
doi:10.1016/j.parco.2014.10.002 fatcat:dqqzlned35gd3glu43kdl4m4oe

A Data-Flow Soft-Core Processor for Accelerating Scientific Calculation on FPGAs

Lorenzo Verdoscia, Roberto Giorgi
2016 Mathematical Problems in Engineering  
Data-flow instructions serve to describe a program and to dynamically change the context of a data-flow program graph inside the accelerator, on-the-fly.  ...  Our proposed design aims at combining the performance of a fine-grained data-flow architecture with the flexibility of reconfiguration, without requiring a partial reconfiguration or new bit-stream for  ...  processor architecture whose resources are always sufficient to compute any one dot product) with the related context management and data-flow accelerator.  ... 
doi:10.1155/2016/3190234 fatcat:xb7gwn4zinh2va3kir4ckjzxgi

Compositionality in synchronous data flow

Stavros Tripakis, Dai Bui, Marc Geilen, Bert Rodiers, Edward A. Lee
2013 ACM Transactions on Embedded Computing Systems  
Motivated by the need for incremental and modular code generation from hierarchical SDF models, we introduce in this paper DSSF profiles.  ...  DSSF (Deterministic SDF with Shared FIFOs) forms a compositional abstraction of composite actors that can be used for modular compilation.  ...  Acknowledgments We would like to thank Jörn Janneck and Maarten Wiggers for their valuable input.  ... 
doi:10.1145/2442116.2442133 fatcat:fhmrjoiymbdgbkyi643hxjqv3q

Compositionality in Synchronous Data Flow

Stavros Tripakis, Dai Bui, Bert Rodiers, Edward A. Lee
2010 Proceedings of the 1st ACM/IEEE International Conference on Cyber-Physical Systems - ICCPS '10  
Motivated by the need for incremental and modular code generation from hierarchical SDF models, we introduce in this paper DSSF profiles.  ...  DSSF (Deterministic SDF with Shared FIFOs) forms a compositional abstraction of composite actors that can be used for modular compilation.  ...  Acknowledgments We would like to thank Jörn Janneck and Maarten Wiggers for their valuable input.  ... 
doi:10.1145/1795194.1795223 dblp:conf/iccps/TripakisBRL10 fatcat:juprc6dlqzhphekw4nxcd3kg3m
« Previous Showing results 1 — 15 out of 2,415 results