7 Hits in 4.8 sec

Software challenges in extreme scale systems

Vivek Sarkar, William Harrod, Allan E Snavely
2009 Journal of Physics, Conference Series  
Carlson is a member of the research staff at the IDA Center for Computing Sciences where, since 1990, his focus has been on applications and system tools for large-scale parallel and distributed computers  ...  them, and scaling multiple chips to complete systems, for a range of real system applications, from highly scalable deep space exploration to trans-petaflops level supercomputing.  ...  These numbers are derived from the following: -For all applications, bisection bandwidth should be no less than current 3-D topologies scaled to a petaflop/s performance rates.  ... 
doi:10.1088/1742-6596/180/1/012045 fatcat:iukutry2dvbitfdh6ng7kgz564

Mesoscopic simulations at the physics-chemistry-biology interface [article]

Massimo Bernaschi, Simone Melchionna, Sauro Succi
2019 arXiv   pre-print
We discuss the Lattice Boltzmann-Particle Dynamics (LBPD) multiscale paradigm for the simulation of complex states of flowing matter at the interface between Physics, Chemistry and Biology.  ...  on Tsubame (2012), up to a world-record (to the best of our knowledge) of 20 PetaFlops/s (sustained performance) for protein crowding on Titan (2013).  ...  In fact, scaling up the size L by a factor 10 would take to the order of Exaflops, still feasible on present-day leading-edge Petaflops/s computers.  ... 
arXiv:1905.02261v1 fatcat:k4zbtaqgmnfkvn2ioxnybg26ba

Pathway to the Square Kilometre Array - The German White Paper - [article]

F. Aharonian, T. G. Arshakian, B. Allen, R. Banerjee, R. Beck, W. Becker, D. J. Bomans, D. Breitschwerdt, M. Brüggen, A. Brunthaler, B. Catinella, D. Champion (+57 others)
2013 arXiv   pre-print
At present the energy consumption of a Petaflops/s system is about a few megawatts. For example, the power consumption of the 1 Petaflop/s Blue Gene system JUGENE is about 2 megawatt (MW).  ...  The leading system of the June 2011 Top 500 list has a power consumption of 10 MW for a peak performance of 8 Petaflops/s.  ... 
arXiv:1301.4124v1 fatcat:2gqq3xjnjbcxhnf36w3au7imvq

Machine Learning Parallelism Could Be Adaptive, Composable and Automated

Hao Zhang
We examine them and show that they significantly boost the efficiency or scalability of ML training on clusters 2-10x in their applicable scenarios.  ...  Particularly, ML scale-up is usually underestimated in terms of the amount of knowledge and time required to map from an appropriate distribution strategy to the model.  ...  This figure measures the computational cost using the metric Petaflop/s-days, which consists of performing 10 15 neural net operations per second for one day, or a total of about 10 20 operations.  ... 
doi:10.1184/r1/14402450 fatcat:be5w3hpokjcvplwhzqjnkz5re4

The Frontiers of Nuclear Science, A Long Range Plan [article]

The DOE/NSF Nuclear Science Advisory Committee
2008 arXiv   pre-print
Petaflop/s,.by.FY2009 unite.these.facilities,.'s.SciDAC.Initiative  ...  EXO is a 136 Xe project supported in part by the DOE Office of High Energy Physics and NSF.  ...  ReCommeNDAtioN iV The experiments at the Relativistic Heavy ion Collider have discovered a new state of matter at extreme temperature and density-a quark-gluon plasma that exhibits unexpected, almost perfect  ... 
arXiv:0809.3137v1 fatcat:jr7exfxxizatdevyno7ew5jck4

Distributed and multiscale computing for scientific applications

Mohamed Ben Belgacem, Bastien Chopard, Nabil Abdennadher
The federated European resources have delivered a computing performance that excessed 2 PetaFlops/s.  ...  Currently, the Tianhe-2 supercomputer located at the Sun Yat-sen University in China delivers ∼33. 86 PetaFLOPS/s and is considered as the world's fastest supercomputer. • HPC cluster: a HPC cluster  ...  Coupling of a PRACE machine (superMUC [136] supercomputer) with an EGI cluster has been also be done showing that the MMSF can be also handle supercomputers machines.  ... 
doi:10.13097/archive-ouverte/unige:48195 fatcat:3353amihkvattkhyr677dgb7du

Capturing the impact of external interference on HPC application performance

Aamer Shah
In the Top500 list, which maintains a record of the worlds most powerful supercomputers, the maximum achieved computation performance has increased from 59.7 gigaflop/s in 1993 1 to 415.5 petaflop/s in  ...  Such a network architecture provides high bandwidth at low latency among neighboring compute nodes, and allows for scalability at lower cost.  ... 
doi:10.25534/tuprints-00011774 fatcat:mylkgwaqbjai7ltbhcgldruutu