Search

The complexity of modern and upcoming computing architectures poses severe challenges for code developers and application specialists, and forces them to expose the highest possible degree of parallelism, in order to make the best use of the available hardware. The Intel$^{(R)}$ Xeon Phi$^{(TM)}$ of second generation (code-named Knights Landing, henceforth KNL) is the latest many-core system, which implements several interesting hardware features like for example a large number of cores per node (up to 72), the 512 bits-wide vector registers and the high-bandwidth memory. The unique features of KNL make this platform a powerful testbed for modern HPC applications. The performance of codes on KNL is therefore a useful proxy of their readiness for future architectures. In this work we describe the lessons learnt during the optimisation of the widely used codes for computational astrophysics P-Gadget-3, Flash and Echo. Moreover, we present results for the visualisation and analysis tools VisIt and yt. These examples show that modern architectures benefit from code optimisation at different levels, even more than traditional multi-core systems. However, the level of modernisation of typical community codes still needs improvements, for them to fully utilise resources of novel architectures., Comment: 16 pages, 10 figures, 4 tables. To be published in Future Generation of Computer Systems (FGCS), Special Issue on "On The Road to Exascale II: Advances in High Performance Computing and Simulations"

The increasing interest in the usage of Artificial Intelligence techniques (AI) from the research community and industry to tackle "real world" problems, requires High Performance Computing (HPC) resources to efficiently compute and scale complex algorithms across thousands of nodes. Unfortunately, typical data scientists are not familiar with the unique requirements and characteristics of HPC environments. They usually develop their applications with high-level scripting languages or frameworks such as TensorFlow and the installation process often requires connection to external systems to download open source software during the build. HPC environments, on the other hand, are often based on closed source applications that incorporate parallel and distributed computing API's such as MPI and OpenMP, while users have restricted administrator privileges, and face security restrictions such as not allowing access to external systems. In this paper we discuss the issues associated with the deployment of AI frameworks in a secure HPC environment and how we successfully deploy AI frameworks on SuperMUC-NG with Charliecloud., 6 pages, 2 figures, 2019 IEEE High Performance Extreme Computing Conference

Natural language processing (NLP) problems are ubiquitous in classical computing, where they often require significant computational resources to infer sentence meanings. With the appearance of quantum computing hardware and simulators, it is worth developing methods to examine such problems on these platforms. In this manuscript we demonstrate the use of quantum computing models to perform NLP tasks, where we represent corpus meanings, and perform comparisons between sentences of a given structure. We develop a hybrid workflow for representing small and large scale corpus data sets to be encoded, processed, and decoded using a quantum circuit model. In addition, we provide our results showing the efficacy of the method, and release our developed toolkit as an open software suite., For associated code, see https://github.com/ICHEC/QNLP

We describe a strategy for code modernisation of Gadget, a widely used community code for computational astrophysics. The focus of this work is on node-level performance optimisation, targeting current multi/many-core IntelR architectures. We identify and isolate a sample code kernel, which is representative of a typical Smoothed Particle Hydrodynamics (SPH) algorithm. The code modifications include threading parallelism optimisation, change of the data layout into Structure of Arrays (SoA), auto-vectorisation and algorithmic improvements in the particle sorting. We obtain shorter execution time and improved threading scalability both on Intel XeonR ($2.6 \times$ on Ivy Bridge) and Xeon PhiTM ($13.7 \times$ on Knights Corner) systems. First few tests of the optimised code result in $19.1 \times$ faster execution on second generation Xeon Phi (Knights Landing), thus demonstrating the portability of the devised optimisation solutions to upcoming architectures., Comment: 8 pages, 2 columns, 4 figures, accepted as paper at HPCS Proceedings 2017, IEEE XPLORE

A global quantitative picture of the phonon-induced two-electron spin relaxation in GaAs double quantum dots is presented using highly accurate numerical calculations. Wide regimes of interdot coupling, magnetic field magnitude and orientation, and detuning are explored in the presence of a nuclear bath. Most important, the unusually strong magnetic anisotropy of the singlet-triplet relaxation can be controlled by detuning switching the principal anisotropy axes: a protected state becomes unprotected upon detuning, and vice versa. It is also established that nuclear spins can dominate spin relaxation for unpolarized triplets even at high magnetic fields, contrary to common belief. These findings are central to designing quantum dots geometries for spin-based quantum information processing with minimal environmental impact., Comment: 8 pages, 8 figures

The influence of the spin-orbit interactions on the energy spectrum of two-electron laterally coupled quantum dots is investigated. The effective Hamiltonian for a spin qubit pair proposed in F. Baruffa et al., Phys. Rev. Lett. 104, 126401 (2010) is confronted with exact numerical results in single and double quantum dots in zero and finite magnetic field. The anisotropic exchange Hamiltonian is found quantitatively reliable in double dots in general. There are two findings of particular practical importance: i) The model stays valid even for maximal possible interdot coupling (a single dot), due to the absence of a coupling to the nearest excited level, a fact following from the dot symmetry. ii) In a weak coupling regime, the Heitler-London approximation gives quantitatively correct anisotropic exchange parameters even in a finite magnetic field, although this method is known to fail for the isotropic exchange. The small discrepancy between the analytical model (which employes the linear Dresselhaus and Bychkov-Rashba spin-orbit terms) and the numerical data for GaAs quantum dots is found to be mostly due to the cubic Dresselhaus term., Comment: 15 pages, 11 figures

The effects of spin-orbit coupling on the two-electron spectra in lateral coupled quantum dots are investigated analytically and numerically. It is demonstrated that in the absence of magnetic field the exchange interaction is practically unaffected by spin-orbit coupling, for any interdot coupling, boosting prospects for spin-based quantum computing. The anisotropic exchange appears at finite magnetic fields. A numerically accurate effective spin Hamiltonian for modeling spin-orbit-induced two-electron spin dynamics in the presence of magnetic field is proposed., Comment: 4 pages, 3 figures; paper rewritten

Context. The numerical simulation of turbulence and flows in almost ideal astrophysical plasmas with large Reynolds numbers motivates the implementation of magnetohydrodynamical (MHD) computer codes with low resistivity. They need to be computationally efficient and scale well with large numbers of CPU cores, allow obtaining a high grid resolution over large simulation domains, and be easily and modularly extensible, for instance, to new initial and boundary conditions. Aims. Our aims are the implementation, optimization, and verification of a computationally efficient, highly scalable, and easily extensible low-dissipative MHD simulation code for the numerical investigation of the dynamics of astrophysical plasmas with large Reynolds numbers in three dimensions (3D). Methods. The new GOEMHD3 code discretizes the ideal part of the MHD equations using a fast and efficient leap-frog scheme that is second-order accurate in space and time and whose initial and boundary conditions can easily be modified. For the investigation of diffusive and dissipative processes the corresponding terms are discretized by a DuFort-Frankel scheme. To always fulfill the Courant-Friedrichs-Lewy stability criterion, the time step of the code is adapted dynamically. Numerically induced local oscillations are suppressed by explicit, externally controlled diffusion terms. Non-equidistant grids are implemented, which enhance the spatial resolution, where needed. GOEMHD3 is parallelized based on the hybrid MPI-OpenMP programing paradigm, adopting a standard two-dimensional domain-decomposition approach. Results. The ideal part of the equation solver is verified by performing numerical tests of the evolution of the well-understood Kelvin-Helmholtz instability and of Orszag-Tang vortices. The accuracy of solving the (resistive) induction equation is tested by simulating the decay of a cylindrical current column. Furthermore, we show that the computational performance of the code scales very efficiently with the number of processors up to tens of thousands of CPU cores. This excellent scalability of the code was obtained by simulating the 3D evolution of the solar corona above an active region (NOAA AR1249) for which GOEMHD3 revealed the energy distribution in the solar atmosphere in response to the energy influx from the chromosphere through the transition region, taking into account the weak Joule current dissipation and viscosity in the almost dissipationless solar corona.Conclusions. The new massively parallel simulation code GOEMHD3 enables efficient and fast simulations of almost ideal astrophysical plasma flows with large Reynolds numbers well resolved and on huge grids covering large domains. Its abilities are verified by comprehensive set of tests of ideal and weakly dissipative plasma phenomena. The high-resolution (20483 grid points) simulation of a large part of the solar corona above an observed active region proves the excellent parallel scalability of the code up to more than 30 000 processor cores.