Charles H. Still

On the dominant and subdominant behavior of stimulated Raman and Brillouin scattering driven by nonuniform laser beams

In a number of experiments, stimulated Brillouin (SBS) or Raman backscattering (SRS) has been observed to be much more vigorous than the other although the expectations based on linear gain exponents are that they should both be reflecting large amounts of incident light. Multidimensional fluid simulations of the growth and saturation of these two instabilities driven by a nonuniform incident laser beam are presented. On the fast time scale, the nonlinear saturation occurs via an anomalous damping inspired by fundamental studies of Langmuir turbulence [D. F. DuBois et al., Bull. Am. Phys. Soc. 41, 1531 (1996)] and acoustic wave turbulence [B. I. Cohen et al., Phys. Plasmas 4, 956 (1997)]. Over a longer time scale, SRS and SBS are limited by quasilinear processes such as flows induced by the transfer of momentum from the light to the plasma and ion temperature increases caused by a loss of light energy in SBS. The simulations show a reduction of the SBS reflectivity under conditions of strong SRS reflectiv...

Exploring Traditional and Emerging Parallel Programming Models Using a Proxy Application

Parallel machines are becoming more complex with increasing core counts and more heterogeneous architectures. However, the commonly used parallel programming models, C/C++ with MPI and/or OpenMP, make it difficult to write source code that is easily tuned for many targets. Newer language approaches attempt to ease this burden by providing optimization features such as automatic load balancing, overlap of computation and communication, message-driven execution, and implicit data layout optimizations. In this paper, we compare several implementations of LULESH, a proxy application for shock hydrodynamics, to determine strengths and weaknesses of different programming models for parallel computation. We focus on four traditional (OpenMP, MPI, MPI+OpenMP, CUDA) and four emerging (Chapel, Charm++, Liszt, Loci) programming models. In evaluating these models, we focus on programmer productivity, performance and ease of applying optimizations.

Filamentation and forward Brillouin scatter of entire smoothed and aberrated laser beams

Laser–plasma interactions are sensitive to both the fine-scale speckle and the larger scale envelope intensity of the beam. For some time, simulations have been done on volumes taken from part of the laser beam cross-section, and the results from multiple simulations extrapolated to predict the behavior of the entire beam. However, such extrapolation could very well miss effects of the larger scale structure on the fine-scale. The only definitive method is to simulate the entire beam. These very large calculations have not been feasible until recently, but they are now possible on massively parallel computers. Whole beam simulations show the dramatic difference in the propagation and break up of smoothed and aberrated beams.

Stimulated Raman scatter analyses of experiments conducted at the National Ignition Facility a)

Recent experiments conducted at the National Ignition Facility achieved two main goals: providing radiation drive and symmetry suitable for subsequent ignition experiments. Of the many diagnostics fielded, one provided a time-resolved wavelength spectrum of light reflected from the target by stimulated Raman scatter (SRS). SRS occurs when incident light reflects off self-generated electron plasma waves. Analyses indicate that synthetic SRS spectra better match those of experiments when an atomic physics model with greater emissivity is utilized in target modeling, along with less inhibited electron transport (higher flux, with, ideally, nonlocal electron transport). With these models, SRS occurs in a target region where nearest-neighbor quads of laser beams significantly overlap the diagnosed quad. This increases SRS gain at lower density (lower wavelength), a feature consistent with experimental results. Inclusion of this effect of multiple quads sharing a reflected SRS light wave has resulted in improve...

Mapping applications with collectives over sub-communicators on torus networks

The placement of tasks in a parallel application on specific nodes of a supercomputer can significantly impact performance. Traditionally, this task mapping has focused on reducing the distance between communicating tasks on the physical network. This minimizes the number of hops that point-to-point messages travel and thus reduces link sharing between messages and contention. However, for applications that use collectives over sub-communicators, this heuristic may not be optimal. Many collectives can benefit from an increase in bandwidth even at the cost of an increase in hop count, especially when sending large messages. For example, placing communicating tasks in a cube configuration rather than a plane or a line on a torus network increases the number of possible paths messages might take. This increases the available bandwidth which can lead to significant performance gains. We have developed Rubik, a tool that provides a simple and intuitive interface to create a wide variety of mappings for structured communication patterns. Rubik supports a number of elementary operations such as splits, tilts, or shifts, that can be combined into a large number of unique patterns. Each operation can be applied to disjoint groups of processes involved in collectives to increase the effective bandwidth. We demonstrate the use of Rubik for improving performance of two parallel codes, pF3D and Qbox, which use collectives over sub-communicators.

Analyses of laser-plasma interactions in National Ignition Facility ignition targets

A capability to analyze laser-plasma interactions (LPI) for ignition targets to be fielded at the National Ignition Facility has been developed and exercised. LPI in these targets may cause direct energy loss (backscatter) or energy redirection (beam spray, deflection, and energy transfer). These analyses range from analyzing the gain exponents for backscatter and beam spray to performing massively parallel, three-dimensional simulations of laser beam propagation in the most promising candidate ignition target designs. In the former assessment, ignition designs are iterated to reduce the gain exponent values. In the latter, beam propagation simulations are performed to analyze the reflectivity and beam transmission of speckled laser beams in the computed plasma profiles of the ignition targets. In current ignition designs, laser reflectivity is calculated to be well below 10%.

Electron and ion kinetic effects on non-linearly driven electron plasma and ion acoustic waves

Fully non-linear kinetic simulations of electron plasma and ion acoustic waves (IAWs) have been carried out with a new multi-species, parallelized Vlasov code. The numerical implementation of the Vlasov model and the methods used to compute the wave frequency are described in detail. For the first time, the nonlinear frequency of IAWs, combining the contributions from electron and ion kinetic effects and from harmonic generation, has been calculated and compared to Vlasov results. Excellent agreement of theory with simulation results is shown at all amplitudes, harmonic generation being an essential component at large amplitudes. For IAWs, the positive frequency shift from trapped electrons is confirmed and is dominant for the effective electron-to-ion temperature ratio, Z Te/Ti ≳ 10 with Z as the charge state. Furthermore, numerical results demonstrate unambiguously the dependence [R. L. Dewar, Phys. Fluids 15, 712 (1972)] of the kinetic shifts on details of the distribution of the trapped particles, whi...

Concurrent source iteration in the solution of three-dimensional, multigroup discrete ordinates neutron transport equations

A strategy for implementing source iteration on massively parallel computers for use in solving multigroup discrete ordinates neutron transport equations on three-dimensional Cartesian grids is proposed and analyzed. Based on an analysis of the memory requirement and floating-point complexity of the formal matrix-vector multiplication effected by a single source iteration, a data decomposition and communication strategy is presented that is designed to achieve good scalability with respect to all phase-space variables, i.e., neutron position, energy, and direction. A performance model is developed to analyze the scalability properties of the algorithm and to provide computational and heuristic strategies for determining a data decomposition that minimizes wall clock execution time. Numerical results are presented to demonstrate the performance of a specific implementation of this approach on a 1,024-node nCUBE/2.

Stimulated Raman and Brillouin scattering of polarization-smoothed and temporally smoothed laser beams

Control of filamentation and stimulated Raman and Brillouin scattering is shown to be possible by use of both spatial and temporal smoothing schemes. The spatial smoothing is accomplished by the use of phase plates [Y. Kato and K. Mima, Appl. Phys. 329, 186 (1982)] and polarization smoothing [Lefebvre et al., Phys. Plasmas 5, 2701 (1998)] in which the plasma is irradiated with two orthogonally polarized, uncorrelated speckle patterns. The temporal smoothing considered here is smoothing by spectral dispersion [Skupsky et al., J. Appl. Phys. 66, 3456 (1989)] in which the speckle pattern changes on the laser coherence time scale. At the high instability gains relevant to laser fusion experiments, the effect of smoothing must include the competition among all three instabilities.

The Multicomputer Toolbox approach to concurrent BLAS and LACS

The authors describe many of the issues involved in general-purpose concurrent basic linear algebra subprograms (concurrent BLAS or CBLAS) and discuss data-distribution independence, while further generalizing data distributions. They comment on the utility of linear algebra communication subprograms (LACS). They also describe an algorithm for dense matrix-matrix multiplication and also discuss matrix-vector multiplication issues. With regard to communication, they conclude that there is limited leverage in LACS per se as a stand-alone message-passing standard, and propose that needed capabilities instead be integrated in a general, application-level message passing standard, focusing attention on CBLAS and large-scale application needs. Most of the proposed LACS features are similar to existing or needed general-purpose primitives anyway. All of the ideas discussed have been implemented or are under current development within the Multicomputer Toolbox open software system.<<ETX>>