PREPARING FOR EXASCALE: ORNL Leadership Computing Application Requirements and Strategy

Joubert, Wayne; Kothe, Douglas B.; Nam, Hai Ah

doi:10.2172/974637

Cited by 7 publications

(11 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In all cases, the behavior of l for the presented range of p and N is well approximated by Equation (9). It is interesting that the resulting fit in the values of A varies slightly around the value A = 0.30 (1) and the values of D are practically the same, D = 7.7(1), for all investigated networks except A-k1, for which D is much smaller, D = 6.1 (2).…”

Section: Dependence On the Local Connectivitymentioning

confidence: 70%

“…Before saturation, the width grows as w 2 (t) ∼ t 2β , where β = 0.326 (5). In the steady state, the width is stationary and depends on the system size w 2 ∞ ∼ N 2α , α = 0.49 (1). The two values of the exponents α and β are close to those of the KPZ universality class [10], α = 1/2 and β = 1/3.…”

Section: Basic Conservative Pdes Schemementioning

confidence: 81%

“…Values of u 0 , i.e., the average speed on the time profile for the network with p = 0, are presented in Table IV. Another interesting observation is that u 0 scales with k as u 0 ∝ k −0.84 (1) . The data collapse shown in Figure 19 can therefore be treated in the rescaled variables u k 0.84 and (p/k).…”

Section: Dependence On the Local Connectivitymentioning

confidence: 96%

“…Progress in computation hardware in the last decade has been mainly in the direction of multicore/distributed systems. It is a big challenge to use modern hardware effectively, and creating a single program able to orchestrate a huge number of nodes and cores is not trivial [1,2].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Synchronization of conservative parallel discrete event simulations on a small-world network

Ziganurova

Shchur

2018

Phys. Rev. E

View full text Add to dashboard Cite

We examine the question of the influence of sparse long-range communications on the synchronization in parallel discrete event simulations. We build a model of the evolution of local virtual times in a conservative algorithm including several choices of local links. All network realizations belong to the small-world network class. We find that synchronization depends on the average shortest path of the network. The time profile dynamics are similar to the surface profile growth, which helps to analyze synchronization effects using a statistical physics approach. Without long-range links of the nodes, the model belongs to the universality class of the Kardar-Parisi-Zhang equation for surface growth. We find that the critical exponents depend logarithmically on the fraction of long-range links. We present the results of simulations and discuss our observations.

show abstract

Section: Dependence On the Local Connectivitymentioning

confidence: 70%

Section: Basic Conservative Pdes Schemementioning

confidence: 81%

Section: Dependence On the Local Connectivitymentioning

confidence: 96%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Synchronization of conservative parallel discrete event simulations on a small-world network

Ziganurova

Shchur

2018

Phys. Rev. E

View full text Add to dashboard Cite

show abstract

“…Under UNEDF, considerable effort has been put into optimizing the NUCCOR-CCSD(T) V2.0 doubles kernel for the Jaguar Cray XT5 system, Oak Ridge Leadership Computing Facility's (OLCF) flagship leadership computing resource. 31) To continue to achieve breakthrough discoveries in low-energy nuclear physics using CC, we need to push investigations to larger medium-mass nuclei in larger model spaces. This requires continued emphasis on scaling the CC algorithms and performance optimization.…”

Section: §1 Introductionmentioning

confidence: 99%

Computational Aspects of Nuclear Coupled-Cluster Theory

Hagen

Nam

2012

Prog. Theor. Phys. Suppl.

Self Cite

View full text Add to dashboard Cite

We discuss computational aspects of the spherical coupled-cluster method specific to the nuclear many-body problem. Using chiral nucleon-nucleon interaction at next-to-next-tonext-to leading order (N 3 LO) with cutoff Λ = 500 MeV, we present coupled-cluster results for the ground state of 40 Ca. Scaling and performance studies are presented together with challenges we meet with when extending the coupled-cluster effort to nuclei mass hundred and beyond. §1. IntroductionThe low-energy nuclear many-body problem is a challenging undertaking, however, in the last decade there has been significant progress in first principle calculations of nuclei. 1)-9) With the advance of chiral effective field theory 10)-15) which allows for a systematic and consistent derivation of the nuclear forces rooted in lowenergy Quantum Chromo Dynamics (QCD), and with the development of advanced many-body methods and high performance computing facilities, first principle computations of medium mass and neutron-rich nuclei at the extremes of the nuclear chart are now within reach. The computational cost involved in these calculations grows rapidly as one moves beyond the lightest towards the medium mass region and beyond, and it has been crucial to implement techniques that scale gently with system size and code development that leverages the benefits of modern architecture at high-performance computing facilities.The coupled-cluster (CC) method is an optimal approach to medium mass and neutron-rich nuclei as it is an ideal compromise between accuracy on the one hand and computational cost on the other. Coupled-cluster theory was introduced in nuclear physics in the late 1950s by Coester and Kümmel 16), 17) and was shortly thereafter introduced in quantum chemistry byČížek. 18), 19) Coupled-cluster theory has now been established as the "gold standard" for first principle computations in quantum chemistry (see Ref. 20) for a recent review). Only in the last decade has coupled-cluster re-emerged in the nuclear physics community and has established itself as a state-of-the-art approach to structure of medium mass and neutron-rich nuclei. 21)-28)The nuclear coupled-cluster code suite developed at Oak Ridge National Laboratory, called NUCCOR (Nuclear Coupled-Cluster -Oak Ridge) has been further

show abstract

Accelerated application development: The ORNL Titan experience

Joubert

Archibald

Berrill

et al. 2015

Computers & Electrical Engineering

Self Cite

View full text Add to dashboard Cite

The use of computational accelerators such as NVIDIA GPUs and Intel Xeon Phi processors is now widespread in the high performance computing community, with many applications delivering impressive performance gains. However, programming these systems for high performance, performance portability and software maintainability has been a challenge. In this paper we discuss experiences porting applications to the Titan system. Titan, which began planning in 2009 and was deployed for general use in 2013, was the first multi-petaflop system based on accelerator hardware. To ready applications for accelerated computing, a preparedness effort was undertaken prior to delivery of Titan. In this paper we report experiences and lessons learned from this process and describe how users are currently making use of computational accelerators on Titan.

show abstract

PREPARING FOR EXASCALE: ORNL Leadership Computing Application Requirements and Strategy

Cited by 7 publications

References 8 publications

Synchronization of conservative parallel discrete event simulations on a small-world network

Synchronization of conservative parallel discrete event simulations on a small-world network

Computational Aspects of Nuclear Coupled-Cluster Theory

Accelerated application development: The ORNL Titan experience

Contact Info

Product

Resources

About