SciNet: Lessons Learned from Building a Power-efficient Top-20 System and Data Centre

Loken, C.; Grüner, Daniel; Groer, L.; Peltier, W. Richard; Bunn, Neil; Craig, Michael; Henriques, Teresa; Dempsey, J.L.; Yu, Ching-Hsing; Chen, Joseph; Dursi, L. J.; Chong, J. Kenneth; Northrup, Scott; Pinto, Jaime; Knecht, N. S.; Zon, Ramses van

doi:10.1088/1742-6596/256/1/012026

Cited by 437 publications

(329 citation statements)

References 4 publications

Supporting

Mentioning

323

Contrasting

Order By: Relevance

“…Nevertheless, the addition of more fossils and genes may help narrow confidence or credible intervals, and steady progress is being made in the development of better RMC methods, model selection, and fossil assignment validation. These methods, combined with novel sources of time constraints such as those provided by horizontal or endosymbiotic gene transfer (see Shih and Matzke 2013) (Loken et al 2010).…”

Section: Resultsmentioning

confidence: 99%

On the Age of Eukaryotes: Evaluating Evidence from Fossils and Molecular Clocks

Eme

Sharpe

Brown

et al. 2014

Cold Spring Harbor Perspectives in Biology

223

155

View full text Add to dashboard Cite

Our understanding of the phylogenetic relationships among eukaryotic lineages has improved dramatically over the few past decades thanks to the development of sophisticated phylogenetic methods and models of evolution, in combination with the increasing availability of sequence data for a variety of eukaryotic lineages. Concurrently, efforts have been made to infer the age of major evolutionary events along the tree of eukaryotes using fossilcalibrated molecular clock-based methods. Here, we review the progress and pitfalls in estimating the age of the last eukaryotic common ancestor (LECA) and major lineages. After reviewing previous attempts to date deep eukaryote divergences, we present the results of a Bayesian relaxed-molecular clock analysis of a large dataset (159 proteins, 85 taxa) using 19 fossil calibrations. We show that for major eukaryote groups estimated dates of divergence, as well as their credible intervals, are heavily influenced by the relaxed molecular clock models and methods used, and by the nature and treatment of fossil calibrations. Whereas the estimated age of LECA varied widely, ranging from 1007 (943-1102) Ma to 1898 (1655 -2094) Ma, all analyses suggested that the eukaryotic supergroups subsequently diverged rapidly (i.e., within 300 Ma of LECA). The extreme variability of these and previously published analyses preclude definitive conclusions regarding the age of major eukaryote clades at this time. As more reliable fossil data on eukaryotes from the Proterozoic become available and improvements are made in relaxed molecular clock modeling, we may be able to date the age of extant eukaryotes more precisely.

show abstract

Section: Resultsmentioning

confidence: 99%

On the Age of Eukaryotes: Evaluating Evidence from Fossils and Molecular Clocks

Eme

Sharpe

Brown

et al. 2014

Cold Spring Harbor Perspectives in Biology

223

155

View full text Add to dashboard Cite

show abstract

“…The two optimizations are performed on a massively parallel supercomputer [58]. Each optimization function evaluation requires the solution of eight aerostructural solutions: five for the cruise conditions, two for the maneuver conditions, and one for the stability point.…”

Section: Computational Resourcesmentioning

confidence: 99%

Multipoint High-Fidelity Aerostructural Optimization of a Transport Aircraft Configuration

Kenway

Martins

2014

Journal of Aircraft

309

184

View full text Add to dashboard Cite

This paper presents multi-point high-fidelity aerostructural optimizations of a long-range wide-body transonic transport aircraft configuration. The aerostructural analysis employs Euler CFD with a 2 million cell mesh and a structural finite element model with 300 000 degrees of freedom. The coupled adjoint sensitivity method is used to efficiently compute gradients, enabling the use of gradient-based optimization with respect to hundreds of aerodynamic shape and structural sizing variables. The NASA Common Research Model is used as the baseline configuration, together with a wingbox structure that was designed for this study. Two design optimization problems are solved: one where takeoff gross weight is minimized, and another where fuel burn is minimized. Each optimization uses a multi-point formulation with 5 cruise conditions and 2 maneuver conditions. Each of the optimization problems have 476 design variables, including wing planform, airfoil shape, and structural thickness variables. Optimized results are obtained within 36 hours of wall time using 435 processors. The resulting optimal configurations are discussed and analyzed for the aerostructural trade-offs resulting from each objective. The takeoff gross weight minimization results in a 4.2% reduction in takeoff gross weight with a 6.6% fuel burn reduction, while the fuel-burn optimization resulted in an 11.2% fuel burn reduction with no significant change in the takeoff gross weight.

show abstract

“…Figure 2 shows the percent error of our calculation for Λ = 10 6 for various excitation levels. These high-cutoff calculations were performed on SciNet [30]. Computations with lower cutoffs were performed on personal computers.…”

Section: Comparison To Previous Resultsmentioning

confidence: 99%

Component twist method for higher twists in D1-D5 CFT

2017

View full text Add to dashboard Cite

The deformation operator of the D1D5 orbifold CFT, a twist 2 operator, drives the CFT towards the black hole dual and its physics is key to understanding thermalization in the D1D5 system. To further study this deformation, we extend previous work on the effect of twist 2 operators to a method that works for higher orders, in the continuum limit. Our component twist method works by building higher twist operators out of twist 2 operators together with knowledge of Bogoliubov transformations. Consequently, this method sidesteps limitations in Lunin-Mathur technology by avoiding lifts to the covering space. We verify the method by reproducing results obtainable with Lunin-Mathur technology. Going further, our method upholds a previously conjectured scaling law in the continuum limit that applies to any generic configuration of twists. We illustrate this with computations for a new configuration of two twist 2 operators that twists three copies together.

show abstract

SciNet: Lessons Learned from Building a Power-efficient Top-20 System and Data Centre

Cited by 437 publications

References 4 publications

On the Age of Eukaryotes: Evaluating Evidence from Fossils and Molecular Clocks

On the Age of Eukaryotes: Evaluating Evidence from Fossils and Molecular Clocks

Multipoint High-Fidelity Aerostructural Optimization of a Transport Aircraft Configuration

Component twist method for higher twists in D1-D5 CFT

Contact Info

Product

Resources

About