Výsledky vyhledávání - "Latency hiding"

Scalable unsupervised ML: Latency hiding in distributed sparse tensor decomposition

Autor: Nabil Abubaker, M. Ozan Karsavuran, Cevdet Aykanat

Publikováno v: IEEE Transactions on Parallel and Distributed Systems (TPDS)

Latency overhead in distributed-memory parallel CPD-ALS scales with the number of processors, limiting the scalability of computing CPD of large irregularly sparse tensors. This overhead comes in the form of sparse reduce and expand operations perfor

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::794440acf76a2d96032e39a6e2963551
https://hdl.handle.net/11693/111701

Zobrazit plný text záznamu

Akademický článek

Tento výsledek nelze pro nepřihlášené uživatele zobrazit.
K zobrazení výsledku je třeba se přihlásit.

Conference

Tento výsledek nelze pro nepřihlášené uživatele zobrazit.
K zobrazení výsledku je třeba se přihlásit.

Analyzing the Effect of Local Rounding Error Propagation on the Maximal Attainable Accuracy of the Pipelined Conjugate Gradient Method

Autor: Siegfried Cools, Wim Vanroose, Emrullah Fatih Yetkin, Emmanuel Agullo, Luc Giraud

Publikováno v: SIAM Journal on Matrix Analysis and Applications
SIAM Journal on Matrix Analysis and Applications, Society for Industrial and Applied Mathematics, 2018, 39 (1), pp.426-450. 〈10.1137/17M1117872〉
SIAM Journal on Matrix Analysis and Applications, 2018, 39 (1), pp.426-450. ⟨10.1137/17M1117872⟩
SIAM Journal on Matrix Analysis and Applications, Society for Industrial and Applied Mathematics, 2018, 39 (1), pp.426-450. ⟨10.1137/17M1117872⟩
SIAM journal on matrix analysis and applications

Pipelined Krylov subspace methods typically offer improved strong scaling on parallel HPC hardware compared to standard Krylov subspace methods for large and sparse linear systems. In pipelined methods the traditional synchronization bottleneck is mi

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::e71acd21eea071a4298cfa257938b903
https://hal.inria.fr/hal-01753411/document

Zobrazit plný text záznamu

Optimal additional data layers amount determining for interconnect latency hiding scheme

Autor: A. B. Novikov, G. I. Evtushenko

Publikováno v: Vestnik Voronežskogo Gosudarstvennogo Universiteta Inženernyh Tehnologij, Vol 79, Iss 1, Pp 95-98 (2017)

The key component of parallel computing efficiency is the structure of data exchange between computing nodes. It is necessary to reduce the latency of data exchange to improve the efficiency of parallel computing. A B+2R algorithm for overlapping del

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::1cad87fc0f5b6711c1a5278e16c7a5fb
https://doi.org/10.20914/2310-1202-2017-1-95-98

Zobrazit plný text záznamu

Analysis of rounding error accumulation in Conjugate Gradients to improve the maximal attainable accuracy of pipelined CG

Autor: Cools, Siegfried, Yetkin, Emrullah Fatih, Agullo, Emmanuel, Giraud, Luc, Vanroose, Wim

Publikováno v: [Research Report] RR-8849, Inria Bordeaux Sud-Ouest. 2016

Pipelined Krylov solvers typically offer better scalability in the strong scaling limit compared to standard Krylov methods. The synchronization bottleneck is mitigated by overlapping time-consuming global communications with useful computations in t

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=dedup_wf_001::7efd8242b5cbc5a60ac9038efc04e4f7
https://hal.inria.fr/hal-01262716

Zobrazit plný text záznamu

On rounding error resilience, maximal attainable accuracy and parallel performance of the pipelined Conjugate Gradients method for large-scale linear systems in PETSc

Autor: Luc Giraud, Siegfried Cools, Wim Vanroose, Emrullah Fatih Yetkin, Emmanuel Agullo

Publikováno v: Proceedings of the Exascale Applications and Software Conference, ACM, Stockholm, Sweden, April 2016
EASC
EASC '16 Proceedings of the Exascale Applications and Software Conference 2016
EASC 2016-Exascale Applications and Software Conference
EASC 2016-Exascale Applications and Software Conference, Apr 2016, Stockholm, Sweden. pp.1-10, ⟨10.1145/2938615.2938621⟩

International audience; Pipelined Krylov solvers typically display better strong scaling compared to standard Krylov methods for large linear systems. The synchronization bottleneck is mitigated by overlapping time-consuming global communications wit

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::fea2058f34b26db5a028e7042f5a5c17
https://hdl.handle.net/10067/1359470151162165141

Zobrazit plný text záznamu

Vyhledávací nástroje:

Upřesnit hledání