A Capabilities-Aware Framework for Using Computational Accelerators in Data-Intensive Computing

Multicore computational accelerators such as GPUs are now commodity components for highperformance computing at scale. While such accelerators have been studied in some detail as stand-alone computational engines, their integration in large-scale distributed systems raises new challenges and trade-offs. In this paper, we present an exploration of resource management alternatives for building asymmetric accelerator-based distributed systems. We present these alternatives in the context of a capabilities-aware framework for data-intensive computing, which uses an enhanced implementation of the MapReduce programming model for accelerator-based clusters, compared to the state of the art. The framework can transparently utilize heterogeneous accelerators for deriving high performance with low programming effort. Our work is the first to compare heterogeneous types of accelerators, GPUs and a Cell processors, in the same environment and the first to explore the trade-offs between compute-efficient and control-efficient accelerators on data-intensive systems. Our investigation shows that our framework scales well with the number of different compute nodes. Furthermore, it runs simultaneously on two different types of accelerators, successfully adapts to the resource capabilities, and performs 26.9% better on average than a static execution approach.

Local DNA damage by proton microbeam irradiation induces poly(ADP-ribose) synthesis in mammalian cells

Cellular recovery from ionizing radiation (IR)-induced damage involves poly(ADP-ribose) polymerase (PARP-1 and PARP-2) activity, resulting in the induction of a signalling network responsible for the maintenance of genomic integrity. In the present work, a charged particle microbeam delivering 3.2 MeV protons from a Van de Graaff accelerator has been used to locally irradiate mammalian cells. We show the immediate response of PARPs to local irradiation, concomitant with the recruitment of ATM and Rad51 at sites of DNA damage, both proteins being involved in DNA strand break repair. We found a co-localization but no connection between two DNA damage-dependent post-translational modifications, namely poly(ADP-ribosyl)ation of nuclear proteins and phosphorylation of histone H2AX. Both of them, however, should be considered and used as bona fide immediate sensitive markers of IR damage in living cells. This technique thus provides a powerful approach aimed at understanding the interactions between the signals originating from sites of DNA damage and the subsequent activation of DNA strand break repair mechanisms.

Nonlinear electrostatic excitations of charged dust in degenerate ultra-dense quantum dusty plasmas

The linear and nonlinear properties of low-frequency electrostatic excitations of charged dust particles (or defects) in a dense collisionless, unmagnetized Thomas-Fermi plasma are investigated. A fully ionized three-component model plasma consisting of electrons, ions, and negatively charged massive dust grains is considered. Electrons and ions are assumed to be in a degenerate quantum state, obeying the Thomas-Fermi density distribution, whereas the inertial dust component is described by a set of classical fluid equations. Considering large-amplitude stationary profile travelling-waves in a moving reference frame, the fluid evolution equations are reduced to a pseudo-energy-balance equation, involving a Sagdeev-type potential function. The analysis describes the dynamics of supersonic dust-acoustic solitary waves in Thomas-Fermi plasmas, and provides exact predictions for their dynamical characteristics, whose dependence on relevant parameters (namely, the ion-to-electron Fermi temperature ratio, and the dust concentration) is investigated. An alternative route is also adopted, by assuming weakly varying small-amplitude disturbances off equilibrium, and then adopting a multiscale perturbation technique to derive a Korteweg–de Vries equation for the electrostatic potential, and finally solving in terms for electric potential pulses (electrostatic solitons). A critical comparison between the two methods reveals that they agree exactly in the small-amplitude, weakly superacoustic limit. The dust concentration (Havnes) parameter h = Zd0nd0/ne0 affects the propagation characteristics by modifying the phase speed, as well as the electron/ion Fermi temperatures. Our results aim at elucidating the characteristics of electrostatic excitations in dust-contaminated dense plasmas, e.g., in metallic electronic devices, and also arguably in supernova environments, where charged dust defects may occur in the quantum plasma regime.

Relative biological effectiveness (RBE) and out-of-field cell survival responses to passive scattering and pencil beam scanning proton beam deliveries

The relative biological effectiveness (RBE) of passive scattered (PS) and pencil beam scanned (PBS) proton beam delivery techniques for uniform beam configurations was determined by clonogenic survival. The radiobiological impact of modulated beam configurations on cell survival occurring in- or out-of-field for both delivery techniques was determined with intercellular communication intact or physically inhibited. Cell survival responses were compared to those observed using a 6 MV photon beam produced with a linear accelerator. DU-145 cells showed no significant difference in survival response to proton beams delivered by PS and PBS or 6 MV photons taking into account a RBE of 1.1 for protons at the centre of the spread out Bragg peak. Significant out-of-field effects similar to those observed for 6 MV photons were observed for both PS and PBS proton deliveries with cell survival decreasing to 50-60% survival for scattered doses of 0.05 and 0.03 Gy for passive scattered and pencil beam scanned beams respectively. The observed out-of-field responses were shown to be dependent on intercellular communication between the in-and out-of-field cell populations. These data demonstrate, for the first time, a similar RBE between passive and actively scanned proton beams and confirm that out-of-field effects may be important determinants of cell survival following exposure to modulated photon and proton fields

Shape corrections to exchange-correlation potentials by gradient-regulated seamless connection of model potentials for inner and outer region

Shape corrections to the standard approximate Kohn-Sham exchange-correlation (xc) potentials are considered with the aim to improve the excitation energies (especially for higher excitations) calculated with time-dependent density functional perturbation theory. A scheme of gradient-regulated connection (GRAC) of inner to outer parts of a model potential is developed. Asymptotic corrections based either on the potential of Fermi and Amaldi or van Leeuwen and Baerends (LB) are seamlessly connected to the (shifted) xc potential of Becke and Perdew (BP) with the GRAC procedure, and are employed to calculate the vertical excitation energies of the prototype molecules N-2, CO, CH2O, C2H4, C5NH5, C6H6, Li-2, Na-2, K-2. The results are compared with those of the alternative interpolation scheme of Tozer and Handy as well as with the results of the potential obtained with the statistical averaging of (model) orbital potentials. Various asymptotically corrected potentials produce high quality excitation energies, which in quite a few cases approach the benchmark accuracy of 0.1 eV for the electronic spectra. Based on these results, the potential BP-GRAC-LB is proposed for molecular response calculations, which is a smooth potential and a genuine "local" density functional with an analytical representation. (C) 2001 American Institute of Physics.

Cycloidal magnetic order in the compound IrMnSi

A new compound, IrMnSi, has been synthesized, and its crystal structure and magnetic properties have been investigated by means of neutron powder diffraction, magnetization measurements, and first-principles theory. The crystal structure is found to be of the TiNiSi type (ordered Co2P, space group Pnma). The Mn-projected electronic states are situated at the Fermi level, giving rise to metallic binding, whereas a certain degree of covalent character is observed for the chemical bond between the It and Si atoms. A cycloidal, i.e., noncollinear, magnetic structure was observed below 460 K, with the propagation vector q=[0,0,0.4530(5)] at 10 K. The magnetism is dominated by large moments on the Mn sites, 3.8 mu(B)/atom from neutron diffraction. First-principles theory reproduces the propagation vector of the experimental magnetic structure as well as the angles between the Mn moments. The calculations further result in a magnetic moment of 3.21 mu(B) for the Mn atoms, whereas the Ir and Si moments are negligible, in agreement with observations. A calculation that more directly incorporates electron-electron interactions improves the agreement between the theoretical and experimental magnetic moments. A band mechanism is suggested to explain the observed magnetic order.

Bright Laser-Driven Neutron Source Based on the Relativistic Transparency of Solids

Neutrons are unique particles to probe samples in many ?elds of research ranging from biology to material sciences to engineering and security applications. Access to bright, pulsed sources is currently
limited to large accelerator facilities and there has been a growing need for compact sources over the recent years. Short pulse laser driven neutron sources could be a compact and relatively cheap way to
produce neutrons with energies in excess of 10 MeV. For more than a decade experiments have tried to obtain neutron numbers suf?cient for applications. Our recent experiments demonstrated an ion acceleration mechanism based on the concept of relativistic transparency. Using this new mechanism, we produced an intense beam of high energy (up to 170 MeV) deuterons directed into a Be converter to
produce a forward peaked neutron ?ux with a record yield, on the order of 1010 n=sr. We present results comparing the two acceleration mechanisms and the ?rst short pulse laser generated neutron radiograph.

Accelerating code on multi-cores with FastFlow

FastFlow is a programming framework specifically targeting cache-coherent shared-memory multi-cores. It is implemented as a stack of C++ template libraries built on top of lock-free (and memory fence free) synchronization mechanisms. Its philosophy is to combine programmability with performance. In this paper a new FastFlow programming methodology aimed at supporting parallelization of existing sequential code via offloading onto a dynamically created software accelerator is presented. The new methodology has been validated using a set of simple micro-benchmarks and some real applications. © 2011 Springer-Verlag.

Deployment on GPUs of an application in computational atomic physics

This paper describes the deployment on GPUs of PROP, a program of the 2DRMP suite which models electron collisions with H-like atoms and ions. Because performance on GPUs is better in single precision than in double precision, the numerical stability of the PROP program in single precision has been studied. The numerical quality of PROP results computed in single precision and their impact on the next program of the 2DRMP suite has been analyzed. Successive versions of the PROP program on GPUs have been developed in order to improve its performance. Particular attention has been paid to the optimization of data transfers and of linear algebra operations. Performance obtained on several architectures (including NVIDIA Fermi) are presented.

A filamentation instability for streaming cosmic rays

We demonstrate that cosmic rays form filamentary structures in the precursors of supernova remnant shocks due to their self-generated magnetic fields. The cosmic ray filamentation results in the growth of a long-wavelength instability, and naturally couples the rapid non-linear amplification on small scales to larger length-scales. Hybrid magnetohydrodynamics-particle simulations are performed to confirm the effect. The resulting large-scale magnetic field may facilitate the scattering of high-energy cosmic rays as required to accelerate protons beyond the knee in the cosmic ray spectrum at supernova remnant shocks. Filamentation far upstream of the shock may also assist in the escape of cosmic rays from the accelerator.

GIANT LOBES OF CENTAURUS A RADIO GALAXY OBSERVED WITH THE SUZAKU X-RAY SATELLITE

We report on Suzaku observations of selected regions within the southern giant lobe of the radio galaxy Centaurus A. In our analysis we focus on distinct X-ray features detected with the X-ray Imaging Spectrometer within the range 0.5-10 keV, some of which are likely associated with fine structure of the lobe revealed by recent high-quality radio intensity and polarization maps. With the available photon statistics, we find that the spectral properties of the detected X-ray features are equally consistent with thermal emission from hot gas with temperatures kT > 1 keV, or with a power-law radiation continuum characterized by photon indices Gamma similar to 2.0 +/- 0.5. However, the plasma parameters implied by these different models favor a synchrotron origin for the analyzed X-ray spots, indicating that a very efficient acceleration of electrons up to greater than or similar to 10 TeV energies is taking place within the giant structure of Centaurus A, albeit only in isolated and compact regions associated with extended and highly polarized radio filaments. We also present a detailed analysis of the diffuse X-ray emission filling the whole field of view of the instrument, resulting in a tentative detection of a soft excess component best fitted by a thermal model with a temperature of kT similar to 0.5 keV. The exact origin of the observed excess remains uncertain, although energetic considerations point to thermal gas filling the bulk of the volume of the lobe and mixed with the non-thermal plasma, rather than to the alternative scenario involving a condensation of the hot intergalactic medium around the edges of the expanding radio structure. If correct, this would be the first detection of the thermal content of the extended lobes of a radio galaxy in X-rays. The corresponding number density of the thermal gas in such a case is n(g) similar to 10(-4) cm(-3), while its pressure appears to be in almost exact equipartition with the volume-averaged non-thermal pressure provided by the radio-emitting electrons and the lobes' magnetic field. A prominent large-scale fluctuation of the Galactic foreground emission, resulting in excess foreground X-ray emission aligned with the lobe, cannot be ruled out. Although tentative, our findings potentially imply that the structure of the extended lobes in active galaxies is likely to be highly inhomogeneous and non-uniform, with magnetic reconnection and turbulent acceleration processes continuously converting magnetic energy to internal energy of the plasma particles, leading to possibly significant spatial and temporal variations in the plasma beta parameter around the volume-averaged equilibrium condition beta similar to 1.

Diffusive shock acceleration at laser-driven shocks: studying cosmic-ray accelerators in the laboratory

The non-thermal particle spectra responsible for the emission from many astrophysical systems are thought to originate from shocks via a first order Fermi process otherwise known as diffusive shock acceleration. The same mechanism is also widely believed to be responsible for the production of high energy cosmic rays. With the growing interest in collisionless shock physics in laser produced plasmas, the possibility of reproducing and detecting shock acceleration in controlled laboratory experiments should be considered. The various experimental constraints that must be satisfied are reviewed. It is demonstrated that several currently operating laser facilities may fulfil the necessary criteria to confirm the occurrence of diffusive shock acceleration of electrons at laser produced shocks. Successful reproduction of Fermi acceleration in the laboratory could open a range of possibilities, providing insight into the complex plasma processes that occur near astrophysical sources of cosmic rays.

RADIATIVE SIGNATURES OF RELATIVISTIC SHOCKS

Particle-in-cell simulations of relativistic, weakly magnetized collisionless shocks show that particles can gain energy by repeatedly crossing the shock front. This requires scattering off self-generated small length-scale magnetic fluctuations. The radiative signature of this first-order Fermi acceleration mechanism is important for models of both the prompt and afterglow emission in gamma-ray bursts and depends on the strength parameter a = lambda e/delta B/mc(2) of the fluctuations (lambda is the length scale and vertical bar delta B vertical bar is the magnitude of the fluctuations). For electrons (and positrons), acceleration saturates when the radiative losses produced by the scattering cannot be compensated by the energy gained on crossing the shock. We show that this sets an upper limit on both the electron Lorentz factor gamma <10(6) (n/1 cm(-3))(-1/6)(-1/6) and on the energy of the photons radiated during the scattering process h omega(max) <40Max(a, 1)(n/1 cm(-3))(1/6)(-1/6) eV, where n is the number density of the plasma and (gamma) over bar is the thermal Lorentz factor of the downstream plasma, provided a <a(crit) similar to 10(6). This rules out "jitter" radiation on self-excited fluctuations with a <I as a source of gamma rays, although high-energy photons might still be produced when the jitter photons are upscattered in an analog of the synchrotron self-Compton process. In fluctuations with a > 1, radiation is generated by the standard synchrotron mechanism, and the maximum photon energy rises linearly with a, until saturating at 70 MeV, when a = a(crit).