174 Hits in 7.3 sec

Beyond the Memory Wall: A Case for Memory-centric HPC System for Deep Learning [article]

Youngeun Kwon, Minsoo Rhu
2019 arXiv   pre-print
We propose a memory-centric deep learning system that can transparently expand the memory capacity available to the accelerators while also providing fast inter-device communication for parallel training  ...  As the models and the datasets to train deep learning (DL) models scale, system architects are faced with new challenges, one of which is the memory capacity bottleneck, where the limited physical memory  ...  MEMORY-CENTRIC HPC SYSTEM ARCHITECTURE FOR DEEP LEARNING In this paper, we propose a new architectural solution for future HPC systems optimized for deep learning.  ... 
arXiv:1902.06468v1 fatcat:f5xgnihbizgchk5ds5qcof6ysy

Will supercomputers be super-data and super-AI machines?

Yutong Lu, Depei Qian, Haohuan Fu, Wenguang Chen
2018 Communications of the ACM  
Over the past several decades, China has put significant effort into improving its own HPC through a series of key projects under its national research and development program.  ...  -have developed a series of domestic supercomputing systems, including Dawning 4000A (2005, 11.2TFlops); Tianhe-1A (2011, 4.7PFlops, number one in the TOP500); Sunway BlueLight (2011, 1PFlops); Tianhe-  ...  Deep-learning applications.  ... 
doi:10.1145/3239556 fatcat:tmxojh6syzft5efobo4cbesufy

Eurolab-4-HPC Long-Term Vision on High-Performance Computing [article]

Theo Ungerer, Paul Carpenter
2018 arXiv   pre-print
The objective of the Eurolab-4-HPC vision is to provide a long-term roadmap from 2023 to 2030 for High-Performance Computing (HPC).  ...  Radical changes in computing are foreseen for the next decade.  ...  Deep learning is a machine learning technique inspired by the neural learning process of the human brain.  ... 
arXiv:1807.04521v1 fatcat:5neetrgubjhnvcajcktpkohrzq

ETP4HPC's Strategic Research Agenda for High-Performance Computing in Europe 4 [article]

Michael Malms, Marcin Ostasz, Maike Gilliot, Pascale Bernier-Bruna, Laurent Cargemel, Estela Suarez, Herbert Cornelius, Marc Duranton, Benny Koren, Pascale Rosse-Laurent, María S. Pérez-Hernández, Manolis Marazakis (+11 others)
2020 Zenodo  
This new concept well reflects the main trend of this SRA – it is not only about developing HPC technology in order to build competitive European HPC systems but also about making our HPC solutions work  ...  It continues the tradition of a structured approach to the identification of key research objectives.  ...  Artificial Intelligence related workloads for HPC: tored on Edge systems as part of a global process allowing for • Acceleration of computation for IA: current Deep-Learning continuous  ... 
doi:10.5281/zenodo.4605343 fatcat:lcsgbea5dzgdfmj5dkw6pr7vni

Rethinking High Performance Computing Platforms

Ole Weidner, Malcolm Atkinson, Adam Barker, Rosa Filgueira Vicente
2016 Proceedings of the ACM International Workshop on Data-Intensive Distributed Computing - DIDC '16  
These requirements are not met by the current production HPC platform models and policies. This results in a loss of opportunity, productivity and innovation for new computational methods and tools.  ...  In this paper we critically discuss the dominant HPC platform model and describe the challenges it creates for second generation applications because of its asymmetric resource view, interfaces and software  ...  In the second scenario, users "learn" from the first scenario and define the application wall-time limit very pessimistically.  ... 
doi:10.1145/2912152.2912155 dblp:conf/hpdc/WeidnerABV16 fatcat:4ujp2ojn7zcevggbdxim2sguyi

Rethinking High Performance Computing Platforms: Challenges, Opportunities and Recommendations [article]

Ole Weidner, Malcolm Atkinson, Adam Barker, Rosa Filgueira
2017 arXiv   pre-print
These requirements are not met by the current production HPC platform models and policies. This results in a loss of opportunity, productivity and innovation for new computational methods and tools.  ...  In this paper we critically discuss the dominant HPC platform model and describe the challenges it creates for second generation applications because of its asymmetric resource view, interfaces and software  ...  In the second scenario, users "learn" from the first scenario and define the application wall-time limit very pessimistically.  ... 
arXiv:1702.05513v2 fatcat:is7d2uvbyvbljp6xslrlkjgj6u

Parallelizing Training of Deep Generative Models on Massive Scientific Datasets [article]

Sam Ade Jacobs, Brian Van Essen, David Hysom, Jae-Seung Yeom, Tim Moon, Rushil Anirudh, Jayaraman J. Thiagaranjan, Shusen Liu, Peer-Timo Bremer, Jim Gaffney, Tom Benson, Peter Robinson (+2 others)
2019 arXiv   pre-print
We present a novel tournament method to train traditional as well as generative adversarial networks built on LBANN, a scalable deep learning framework optimized for HPC systems.  ...  Training deep neural networks on large scientific data is a challenging task that requires enormous compute power, especially if no pre-trained models exist to initialize the process.  ...  ACKNOWLEDGMENT This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344 (LLNL-CONF-677443).  ... 
arXiv:1910.02270v1 fatcat:ib26rk5qcrbw7mz3taymmudgrq

Serverless End Game: Disaggregation enabling Transparency [article]

Pedro García-López, Aleksander Slominski, Simon Shillaker, Michael Behrendt, Barnard Metzler
2020 arXiv   pre-print
Can we close the curtains of distributed systems complexity for the majority of users?  ...  For many years, the distributed systems community has struggled to smooth the transition from local to remote computing.  ...  ACKNOWLEDGMENTS is work has been partially supported by the EU Horizon2020 programme under grant agreement No 825184.  ... 
arXiv:2006.01251v1 fatcat:sklmosqtvjcppf44rr77okd6oa

Recent Advances in Computer Architecture: The Opportunities and Challenges for Provenance

Nikilesh Balakrishnan, Thomas Bytheway, Lucian Carata, Oliver R. A. Chick, James Snee, Sherif Akoush, Ripduman Sohan, Margo I. Seltzer, Andy Hopper
2015 Workshop on the Theory and Practice of Provenance  
In recent years several hardware and systems fields have made advances in technology that open new opportunities and challenges for provenance systems.  ...  In this paper we look at such technologies and discuss the implications they have for provenance.  ...  This data can be used to reason about system behaviour, for example, recent work shows that memory heat maps can be used to learn about normal system behaviour and deviations in behaviour can be detected  ... 
dblp:conf/tapp/BalakrishnanBCC15 fatcat:wqt2vu6qdfe7tjqsjhmzkbrt3a

D5.2: Market and Technology Watch Report Year2

Aris Sotiropoulos
2019 Zenodo  
A new benchmark related to IO performance (IO500) initiated in 2017 is gaining momentum with 63 entries while a proposal for a new benchmark on large scale deep learning called Deep500 have been presented  ...  It is thus the continuation of a well-established effort to carry out an assessment of the HPC market based on market surveys, supercomputing conferences, and exchanges between vendors and experts involved  ...  , Mellanox and others) to provide a complete solution including OS and tuned drivers ready for HPC [121].  ... 
doi:10.5281/zenodo.6805970 fatcat:oqhe5lwwizbf3aputurg4o5byq

Prototype of application and infrastructure performance models - Final version

Alfio Lazzaro, Karthee Sivalingam, Nina Mujkanovic, Indika Kumara, Piero Fraternali, Rocio Nahime Torres, Giovanni Quattrocchi, Kamil Tokmakov, Ralf Schneider, Paul Mundt
2020 Zenodo  
This deliverable reports on the progress of the SODALITE project in developing a model for application and infrastructure performance.  ...  As part of this M12 deliverable, this report describes the prototype model developed and also reports on the baseline performance of use case applications.  ...  As neither of these cases is frequently run (there is no online learning carried out), processing can be batched -lending itself well to the increased computational capabilities of HPC systems.  ... 
doi:10.5281/zenodo.3822241 fatcat:yf3uwy2a3nhbljsph3b3egbtea

Towards HPC and Big Data Analytics Convergence: Design and Experimental Evaluation of a HPDA Framework for eScience at Scale

Donatello Elia, Sandro Fiore, Giovanni Aloisio
2021 IEEE Access  
ACKNOWLEDGMENT The authors kindly acknowledge PRACE for awarding access to MareNostrum4 at Barcelona Supercomputing Center (BSC), Spain, as well as the support by the BSC team.  ...  Moreover, the authors would like to acknowledge Antonio Aloisio for his editing and proofreading work on this paper.  ...  ; • including learning techniques to further extend the provided HPDA framework towards machine and deep learning as well as AI support to tackle intelligencedriven data-centric scenarios.  ... 
doi:10.1109/access.2021.3079139 fatcat:zej3qjtcrvbgpnhe7a3ijv4f5e

Enabling particle applications for exascale computing platforms

Susan M Mniszewski, James Belak, Jean-Luc Fattebert, Christian FA Negre, Stuart R Slattery, Adetokunbo A Adedoyin, Robert F Bird, Choongseok Chang, Guangye Chen, Stéphane Ethier, Shane Fogerty, Salman Habib (+12 others)
2021 The international journal of high performance computing applications  
The Exascale Computing Project (ECP) is invested in co-design to assure that key applications are ready for exascale computing.  ...  Libraries are modular instantiations that multiple applications can utilize or be built upon; CoPA has developed the Cabana particle library, PROGRESS/BML libraries for QMD, and the SWFFT and fftMPI parallel  ...  Acknowledgements This work was performed as part of the Co-design Center for Particle Applications, supported by the Exascale Computing Project This paper describes objective technical results and analysis  ... 
doi:10.1177/10943420211022829 fatcat:7goiighkmjgjnosvzjubskjplq

Research trend of large-scale supercomputers and applications from the TOP500 and Gordon Bell Prize

Weimin Zheng
2020 Science China Information Sciences  
In highperformance computing domain, there are two famous awards: The TOP500 list for the fastest 500 supercomputers in the world and the Gordon Bell Prize for the best HPC (high-performance computing)  ...  The first trend we observe is that heterogeneous architectures are widely used in the construction of supercomputing systems.  ...  He is now the editor in chief of the Journal Big Data. Bell Prize: exascale deep learning for climate analytics  ... 
doi:10.1007/s11432-020-2861-0 fatcat:73afvlh5wneq3oqkj2nlnkv3jy

Status and progress of China SKA Regional Centre prototype [article]

Tao An, Xiaocong Wu, Baoqiang Lao, Shaoguang Guo, Zhijun Xu, Weijia Lv, Yingkang Zhang, Zhongli Zhang
2022 arXiv   pre-print
In this model, the SKAO will be supported by a global network of SKA Regional Centres (SRCs) distributed around the world in its member countries to build an end-to-end science data system that will provide  ...  The computing resources needed to process, distribute, curate and use the vast amount of data that will be generated by the SKA telescopes are too large for the SKAO to manage on its own.  ...  Only a hybrid heterogeneous architecture system can effectively perform the SKA data processing tasks. • Data-centric rather than compute-centric.  ... 
arXiv:2206.13022v2 fatcat:edxf6cocxvd4to3yeinmvnyk7q
« Previous Showing results 1 — 15 out of 174 results