Filters








50,079 Hits in 4.4 sec

Asynchronous updates in large parallel systems

Albert G. Greenberg, S. Shenker, Alexander L. Stolyar
1996 Proceedings of the 1996 ACM SIGMETRICS international conference on Measurement and modeling of computer systems - SIGMETRICS '96  
The key issue for this asynchronous updating technique is: how fast do the local times make progress in the large system limit?  ...  This asynchronous updating appears to allow the simulation to achieve a high degree of parallelism, with very low overhead in processor synchronization.  ...  We feel these results reveal the key structure of the asynchronous updating simulation technique, and demonstrates its scalability y to large sys- tems.  ... 
doi:10.1145/233013.233028 dblp:conf/sigmetrics/GreenbergSS96 fatcat:4b4gvhso2ver7ewy25ts5a5ebm

Asynchronous updates in large parallel systems

Albert G. Greenberg, S. Shenker, Alexander L. Stolyar
1996 Performance Evaluation Review  
The key issue for this asynchronous updating technique is: how fast do the local times make progress in the large system limit?  ...  This asynchronous updating appears to allow the simulation to achieve a high degree of parallelism, with very low overhead in processor synchronization.  ...  We feel these results reveal the key structure of the asynchronous updating simulation technique, and demonstrates its scalability y to large sys- tems.  ... 
doi:10.1145/233008.233028 fatcat:64ov3ii5ybg4zdwogdalnwzc7a

Effect of asynchronicity on the universal behaviour of coupled map lattices [article]

Neelima Gupte, T.M. Janaki, Sudeshna Sinha
2002 arXiv   pre-print
We investigate the spatiotemporal dynamics of coupled circle map lattices, evolving under synchronous (parallel) updating on one hand and asynchronous (random) updating rules on the other.  ...  By marked contrast numerical evidence here strongly indicates that asynchronous evolution eliminates most of the attractor states arising from special sets of initial conditions in synchronous systems,  ...  In that sense asynchronous updating yields more generic and robust phenomena in extended systems.  ... 
arXiv:nlin/0205020v1 fatcat:br5vmfyygnfijjdgwk5zrbhpbm

Processing Big Data Graphs on Memory-Restricted Systems

Harshvardhan, Nancy M. Amato, Lawrence Rauchweger
2014 Proceedings of the 23rd international conference on Parallel architectures and compilation - PACT '14  
In this work, we propose a novel RAM-Disk hybrid approach to graph processing that can scale well from a single shared-memory node to large distributed-memory systems.  ...  We show that without modifying the algorithms, this approach can scale from small memory-constrained systems (such as tablets) to large-scale distributed machines with 16, 000+ cores.  ...  An implementation of our approach in the stapl Graph Library [2] allows us to process large graphs on systems ranging from small-scale systems such as off-the-shelf PCs or Android tablets, to large high-end  ... 
doi:10.1145/2628071.2671429 dblp:conf/IEEEpact/HarshvardhanAR14 fatcat:jgs6z7ospjbhtdgtyjmhqp2f34

Balancing the Communication Load of Asynchronously Parallelized Machine Learning Algorithms [article]

Janis Keuper, Franz-Josef Pfreundt
2015 arXiv   pre-print
In the context of large scale learning, as utilized by many Big Data applications, efficient parallelization of SGD is in the focus of active research.  ...  Asynchronous Stochastic Gradient Descent (ASGD) outperforms other, mostly MapReduce based, parallel algorithms solving large scale machine learning problems.  ...  In [8] , we introduced a lock-free parallelization method for the computation of stochastic gradient optimization of large scale machine learning algorithms, which is based on the asynchronous communication  ... 
arXiv:1510.01155v1 fatcat:y3eu66h5ircmbk7s3lujszeiwi

Frog: Asynchronous Graph Processing on GPU with Hybrid Coloring Model

Xuanhua Shi, Xuan Luo, Junling Liang, Peng Zhao, Sheng Di, Bingsheng He, Hai Jin
2018 IEEE Transactions on Knowledge and Data Engineering  
In this work, we mainly answer the four questions: (1) how to partition the vertices in a sparse graph with maximized parallelism, (2) how to process large-scale graphs which are out of GPU memory, (3)  ...  We find that majority of vertices (about 80%) are colored with only a few colors, such that they can be read and updated in a very high degree of parallelism without violating the sequential consistency  ...  In contrast to synchronous execution models, which update all values simultaneously by using values from previous time step, asynchronous systems update them using the most recent values.  ... 
doi:10.1109/tkde.2017.2745562 fatcat:f5db7c5f5nesvn7qmnvxtbo46q

A Stabilizing Control Algorithm for Asynchronous Parallel Quadratic Programming via Dual Decomposition [article]

Kooktae Lee
2019 arXiv   pre-print
We aim to mitigate this synchronization penalty in PQP problems by implementing asynchronous updates of the dual variable.  ...  In general, distributed and parallel optimization requires synchronization of data at each iteration step due to the interdependency of data.  ...  Dual Decomposition with Asynchronous update In order to alleviate this synchronization penalty, we consider asynchronous updates of dual variable y.  ... 
arXiv:1911.11012v1 fatcat:6qmmedhbjvhp7o4gps4tfw74le

Tracking Mean Field Dynamics by Synchronous Computations of Recurrent Multilayer Perceptrons [chapter]

Jiann-Ming Wu, Jung-Chao Ban, Chun-Chang Wu
2014 Computational and Numerical Simulations  
This approach decomposes a large system to several sparsely connected small systems, updates mean activations inside each small system synchronously and updates decomposed systems sequentially.  ...  Synchronous parallel updating is explored for emulating asynchronous updating (9) for tracking mean field dynamics.  ... 
doi:10.5772/57217 fatcat:jhkiysvwozf2hmorsywejy6o6a

A Block-Asynchronous Relaxation Method for Graphics Processing Units

Hartwig Anzt, Stanimire Tomov, Jack Dongarra, Vincent Heuveline
2012 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops & PhD Forum  
For this purpose, we developed a set of asynchronous iteration algorithms in CUDA and compared them with a parallel implementation of synchronous relaxation methods on CPU-based systems.  ...  Due to the explosion of parallelism in todays architecture designs, the significance and the need for asynchronous methods, as the ones described in this work, is expected to grow.  ...  While for small iteration numbers and problem sizes we have a factor of around 5, it rises over 10 for large systems and high total iteration numbers.  ... 
doi:10.1109/ipdpsw.2012.11 dblp:conf/ipps/AnztTDH12 fatcat:twbwjkclsfetdmq3buyz5quwre

A Parallel Strategy for Convolutional Neural Network Based on Heterogeneous Cluster for Mobile Information System

Jilin Zhang, Junfeng Xiao, Jian Wan, Jianhua Yang, Yongjian Ren, Huayou Si, Li Zhou, Hangdi Tu
2017 Mobile Information Systems  
The strategy is based on both synchronous parallel and asynchronous parallel approaches; the model training process can reduce the dependence on the heterogeneous architecture in the premise of ensuring  ...  More and more machine learning methods have been used in the field of mobile information systems to study user behavior and classify usage patterns, especially convolutional neural network.  ...  Parallel Method. In the heterogeneous system, data parallel was chosen to train the model in general.  ... 
doi:10.1155/2017/3824765 fatcat:7m5ml5wdprga3fqyk4pk6fhlwy

Optimization of asynchronous graph processing on GPU with hybrid coloring model

Xuanhua Shi, Junling Liang, Sheng Di, Bingsheng He, Hai Jin, Lu Lu, Zhixiang Wang, Xuan Luo, Jianlong Zhong
2015 Proceedings of the 20th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming - PPoPP 2015  
We find that majority of vertices (about 80%) are colored with only a few colors, such that they can be read and updated in a very high degree of parallelism without violating the sequential consistency  ...  Many parallel graph algorithms adopt the asynchronous computing model to accelerate the iterative convergence.  ...  We partition the graph based on the colors assigned to vertices, and process/update the vertices in parallel.  ... 
doi:10.1145/2688500.2688542 dblp:conf/ppopp/ShiLDHJLWLZ15 fatcat:33u3utczobgp3fbcvolj6seqja

Block-asynchronous Multigrid Smoothers for GPU-accelerated Systems

Hartwig Anzt, Stanimire Tomov, Mark Gates, Jack Dongarra, Vincent Heuveline
2012 Procedia Computer Science  
The hardware target for the new algorithms is top-of-the-line, highly parallel hybrid architectures -multicore-based systems enhanced with GPGPUs.  ...  This paper explores the need for asynchronous iteration algorithms as smoothers in multigrid methods.  ...  It allows a large freedom in the update order and the number of updates per component, while every component update uses the latest available values for the other components.  ... 
doi:10.1016/j.procs.2012.04.002 fatcat:kred4zhysbdpheqydrak7fsdw4

Optimization of asynchronous graph processing on GPU with hybrid coloring model

Xuanhua Shi, Junling Liang, Sheng Di, Bingsheng He, Hai Jin, Lu Lu, Zhixiang Wang, Xuan Luo, Jianlong Zhong
2015 SIGPLAN notices  
We find that majority of vertices (about 80%) are colored with only a few colors, such that they can be read and updated in a very high degree of parallelism without violating the sequential consistency  ...  Many parallel graph algorithms adopt the asynchronous computing model to accelerate the iterative convergence.  ...  We partition the graph based on the colors assigned to vertices, and process/update the vertices in parallel.  ... 
doi:10.1145/2858788.2688542 fatcat:5y54whzap5dx3kdh3we57gqqiq

Parallel asynchronous particle swarm optimization

Byung-Il Koh, Alan D. George, Raphael T. Haftka, Benjamin J. Fregly
2006 International Journal for Numerical Methods in Engineering  
In this study, we introduce a parallel asynchronous PSO (PAPSO) algorithm to enhance computational efficiency.  ...  20 processor system).  ...  We thank Jaco Schutte for providing the parallel synchronous PSO algorithm used for the evaluations.  ... 
doi:10.1002/nme.1646 pmid:17224972 pmcid:PMC1769316 fatcat:ko2egtkddjhnxhj74oyhsrpona

Iterative Sparse Triangular Solves for Preconditioning [chapter]

Hartwig Anzt, Edmond Chow, Jack Dongarra
2015 Lecture Notes in Computer Science  
We demonstrate the performance gains that this approach can have on GPUs in the context of solving sparse linear systems with a preconditioned Krylov subspace method.  ...  We also illustrate the effect of using asynchronous iterations.  ...  Therefore, the asynchronous method converges in the asymptotic sense for any triangular system [11] .  ... 
doi:10.1007/978-3-662-48096-0_50 fatcat:k4p5ysqnhza3djruia4kvsc5j4
« Previous Showing results 1 — 15 out of 50,079 results