arXiv subjects

Erik Schnetter

Publications and source records attributed to Erik Schnetter.

At least 55 records · Page 3Linked to original sources

arXiv2017

General Relativistic Three-Dimensional Multi-Group Neutrino Radiation-Hydrodynamics Simulations of Core-Collapse Supernovae

We report on a set of long-term general-relativistic three-dimensional (3D) multi-group (energy-dependent) neutrino-radiation hydrodynamics simulations of core-collapse supernovae. We employ a full 3D two-moment scheme with the local M1 closure, three neutrino species, and 12 energy groups per species. With this, we follow the post-core-bounce evolution of the core of a nonrotating $27$-$M_\odot$ progenitor in full unconstrained 3D and in octant symmetry for $\gtrsim$$ 380\,\mathrm{ms}$. We find the development of an asymmetric runaway explosion in our unconstrained simulation. We test the resolution dependence of our results and, in agreement with previous work, find that low resolution artificially aids explosion and leads to an earlier runaway expansion of the shock. At low resolution, the octant and full 3D dynamics are qualitatively very similar, but at high resolution, only the full 3D simulation exhibits the onset of explosion.

astro-ph.HE↗

arXiv2017

Coarse graining flow of spin foam intertwiners

Simplicity constraints play a crucial role in the construction of spin foam models, yet their effective behaviour on larger scales is scarcely explored. In this article we introduce intertwiner and spin net models for the quantum group $\text{SU}(2)_k \times \text{SU}(2)_k$, which implement the simplicity constraints analogous to 4D Euclidean spin foam models, namely the Barrett-Crane (BC) and the Engle-Pereira-Rovelli-Livine/Freidel-Krasnov (EPRL/FK) model. These models are numerically coarse grained via tensor network renormalization, allowing us to trace the flow of simplicity constraints to larger scales. In order to perform these simulations we have substantially adapted tensor network algorithms, which we discuss in detail. The BC and the EPRL/FK model behave very differently under coarse graining: While the unique BC intertwiner model is a fixed point and therefore constitutes a 2D topological phase, BC spin net models flow away from the initial simplicity constraints and converge to several different topological phases. Most of these phases correspond to decoupling spin foam vertices, however we find also a new phase in which this is not the case, and in which a non-trivial version of the simplicity constraints holds. The coarse graining flow of the BC spin net models indicates furthermore that the phase transitions are not of second order. The EPRL/FK model by contrast reveals a far more intricate and complex dynamics. We observe an immediate flow away from the original simplicity constraints, however, with the truncation employed here, the models generically do not converge to a fixed point. The results show that the imposition of simplicity constraints can indeed lead to interesting and complex dynamics. Thus we will need to further develop coarse graining tools to efficiently study the large scale behaviour of spin foam models, in particular for the EPRL/FK model.

gr-qc↗

arXiv2016

pocl: A Performance-Portable OpenCL Implementation

OpenCL is a standard for parallel programming of heterogeneous systems. The benefits of a common programming standard are clear; multiple vendors can provide support for application descriptions written according to the standard, thus reducing the program porting effort. While the standard brings the obvious benefits of platform portability, the performance portability aspects are largely left to the programmer. The situation is made worse due to multiple proprietary vendor implementations with different characteristics, and, thus, required optimization strategies. In this paper, we propose an OpenCL implementation that is both portable and performance portable. At its core is a kernel compiler that can be used to exploit the data parallelism of OpenCL programs on multiple platforms with different parallel hardware styles. The kernel compiler is modularized to perform target-independent parallel region formation separately from the target-specific parallel mapping of the regions to enable support for various styles of fine-grained parallel resources such as subword SIMD extensions, SIMD datapaths and static multi-issue. Unlike previous similar techniques that work on the source level, the parallel region formation retains the information of the data parallelism using the LLVM IR and its metadata infrastructure. This data can be exploited by the later generic compiler passes for efficient parallelization. The proposed open source implementation of OpenCL is also platform portable, enabling OpenCL on a wide range of architectures, both already commercialized and on those that are still under research. The paper describes how the portability of the implementation is achieved. Our results show that most of the benchmarked applications when compiled using pocl were faster or close to as fast as the best proprietary OpenCL implementation for the platform at hand.

cs.DC↗

arXiv2016

An Operator-Based Local Discontinuous Galerkin Method Compatible With the BSSN Formulation of the Einstein Equations

Discontinuous Galerkin Finite Element (DGFE) methods offer a mathematically beautiful, computationally efficient, and efficiently parallelizable way to solve hyperbolic partial differential equations. These properties make them highly desirable for numerical calculations in relativistic astrophysics and many other fields. The BSSN formulation of the Einstein equations has repeatedly demonstrated its robustness. The formulation is not only stable but allows for puncture-type evolutions of black hole systems. To-date no one has been able to solve the full (3+1)-dimensional BSSN equations using DGFE methods. This is partly because DGFE discretization often occurs at the level of the equations, not the derivative operator, and partly because DGFE methods are traditionally formulated for manifestly flux-conservative systems. By discretizing the derivative operator, we generalize a particular flavor of DGFE methods, Local DG methods, to solve arbitrary second-order hyperbolic equations. Because we discretize at the level of the derivative operator, our method can be interpreted as either a DGFE method or as a finite differences stencil with non-constant coefficients.

gr-qc↗

arXiv2016

A Survey of High Level Frameworks in Block-Structured Adaptive Mesh Refinement Packages

Over the last decade block-structured adaptive mesh refinement (SAMR) has found increasing use in large, publicly available codes and frameworks. SAMR frameworks have evolved along different paths. Some have stayed focused on specific domain areas, others have pursued a more general functionality, providing the building blocks for a larger variety of applications. In this survey paper we examine a representative set of SAMR packages and SAMR-based codes that have been in existence for half a decade or more, have a reasonably sized and active user base outside of their home institutions, and are publicly available. The set consists of a mix of SAMR packages and application codes that cover a broad range of scientific domains. We look at their high-level frameworks, and their approach to dealing with the advent of radical changes in hardware architecture. The codes included in this survey are BoxLib, Cactus, Chombo, Enzo, FLASH, and Uintah.

cs.DC↗

arXiv2016

Neutrino-Driven Convection in Core-Collapse Supernovae: High-Resolution Simulations

We present results from high-resolution semi-global simulations of neutrino-driven convection in core-collapse supernovae. We employ an idealized setup with parametrized neutrino heating/cooling and nuclear dissociation at the shock front. We study the internal dynamics of neutrino-driven convection and its role in re-distributing energy and momentum through the gain region. We find that even if buoyant plumes are able to locally transfer heat up to the shock, convection is not able to create a net positive energy flux and overcome the downwards transport of energy from the accretion flow. Turbulent convection does, however, provide a significant effective pressure support to the accretion flow as it favors the accumulation of energy, mass and momentum in the gain region. We derive an approximate equation that is able to explain and predict the shock evolution in terms of integrals of quantities such as the turbulent pressure in the gain region or the effects of non-radial motion of the fluid. We use this relation as a way to quantify the role of turbulence in the dynamics of the accretion shock. Finally, we investigate the effects of grid resolution, which we change by a factor 20 between the lowest and highest resolution. Our results show that the shallow slopes of the turbulent kinetic energy spectra reported in previous studies are a numerical artefact. Kolmogorov scaling is progressively recovered as the resolution is increased.

astro-ph.HE↗

arXiv2015

Ab initio holography

We apply the quantum renormalization group to construct a holographic dual for the U(N) vector model for complex bosons defined on a lattice. The bulk geometry becomes dynamical as the hopping amplitudes which determine connectivity of space are promoted to quantum variables. In the large N limit, the full bulk equations of motion for the dynamical hopping fields are numerically solved for finite systems. From finite size scaling, we show that different phases exhibit distinct geometric features in the bulk. In the insulating phase, the space gets fragmented into isolated islands deep inside the bulk, exhibiting ultra-locality. In the superfluid phase, the bulk exhibits a horizon beyond which the geometry becomes non-local. Right at the horizon, the hopping fields decay with a universal power-law in coordinate distance between sites, while they decay in slower power-laws with continuously varying exponents inside the horizon. At the critical point, the bulk exhibits a local geometry whose characteristic length scale diverges asymptotically in the IR limit.

hep-th↗

arXiv2014

Chemora: A PDE Solving Framework for Modern HPC Architectures

Modern HPC architectures consist of heterogeneous multi-core, many-node systems with deep memory hierarchies. Modern applications employ ever more advanced discretisation methods to study multi-physics problems. Developing such applications that explore cutting-edge physics on cutting-edge HPC systems has become a complex task that requires significant HPC knowledge and experience. Unfortunately, this combined knowledge is currently out of reach for all but a few groups of application developers. Chemora is a framework for solving systems of Partial Differential Equations (PDEs) that targets modern HPC architectures. Chemora is based on Cactus, which sees prominent usage in the computational relativistic astrophysics community. In Chemora, PDEs are expressed either in a high-level \LaTeX-like language or in Mathematica. Discretisation stencils are defined separately from equations, and can include Finite Differences, Discontinuous Galerkin Finite Elements (DGFE), Adaptive Mesh Refinement (AMR), and multi-block systems. We use Chemora in the Einstein Toolkit to implement the Einstein Equations on CPUs and on accelerators, and study astrophysical systems such as black hole binaries, neutron stars, and core-collapse supernovae.

cs.MS↗

arXiv2013

Error-analysis and comparison to analytical models of numerical waveforms produced by the NRAR Collaboration

The Numerical-Relativity-Analytical-Relativity (NRAR) collaboration is a joint effort between members of the numerical relativity, analytical relativity and gravitational-wave data analysis communities. The goal of the NRAR collaboration is to produce numerical-relativity simulations of compact binaries and use them to develop accurate analytical templates for the LIGO/Virgo Collaboration to use in detecting gravitational-wave signals and extracting astrophysical information from them. We describe the results of the first stage of the NRAR project, which focused on producing an initial set of numerical waveforms from binary black holes with moderate mass ratios and spins, as well as one non-spinning binary configuration which has a mass ratio of 10. All of the numerical waveforms are analysed in a uniform and consistent manner, with numerical errors evaluated using an analysis code created by members of the NRAR collaboration. We compare previously-calibrated, non-precessing analytical waveforms, notably the effective-one-body (EOB) and phenomenological template families, to the newly-produced numerical waveforms. We find that when the binary's total mass is ~100-200 solar masses, current EOB and phenomenological models of spinning, non-precessing binary waveforms have overlaps above 99% (for advanced LIGO) with all of the non-precessing-binary numerical waveforms with mass ratios <= 4, when maximizing over binary parameters. This implies that the loss of event rate due to modelling error is below 3%. Moreover, the non-spinning EOB waveforms previously calibrated to five non-spinning waveforms with mass ratio smaller than 6 have overlaps above 99.7% with the numerical waveform with a mass ratio of 10, without even maximizing on the binary parameters.

gr-qc↗

arXiv2013

Cactus: Issues for Sustainable Simulation Software

The Cactus Framework is an open-source, modular, portable programming environment for the collaborative development and deployment of scientific applications using high-performance computing. Its roots reach back to 1996 at the National Center for Supercomputer Applications and the Albert Einstein Institute in Germany, where its development jumpstarted. Since then, the Cactus framework has witnessed major changes in hardware infrastructure as well as its own community. This paper describes its endurance through these past changes and, drawing upon lessons from its past, also discusses future

cs.CE↗

arXiv2013

Performance and Optimization Abstractions for Large Scale Heterogeneous Systems in the Cactus/Chemora Framework

We describe a set of lower-level abstractions to improve performance on modern large scale heterogeneous systems. These provide portable access to system- and hardware-dependent features, automatically apply dynamic optimizations at run time, and target stencil-based codes used in finite differencing, finite volume, or block-structured adaptive mesh refinement codes. These abstractions include a novel data structure to manage refinement information for block-structured adaptive mesh refinement, an iterator mechanism to efficiently traverse multi-dimensional arrays in stencil-based codes, and a portable API and implementation for explicit SIMD vectorization. These abstractions can either be employed manually, or be targeted by automated code generation, or be used via support libraries by compilers during code generation. The implementations described below are available in the Cactus framework, and are used e.g. in the Einstein Toolkit for relativistic astrophysics simulations.

cs.DC↗

arXiv2013

From Physics Model to Results: An Optimizing Framework for Cross-Architecture Code Generation

Starting from a high-level problem description in terms of partial differential equations using abstract tensor notation, the Chemora framework discretizes, optimizes, and generates complete high performance codes for a wide range of compute architectures. Chemora extends the capabilities of Cactus, facilitating the usage of large-scale CPU/GPU systems in an efficient manner for complex applications, without low-level code tuning. Chemora achieves parallelism through MPI and multi-threading, combining OpenMP and CUDA. Optimizations include high-level code transformations, efficient loop traversal strategies, dynamically selected data and instruction cache usage strategies, and JIT compilation of GPU code tailored to the problem characteristics. The discretization is based on higher-order finite differences on multi-block domains. Chemora's capabilities are demonstrated by simulations of black hole collisions. This problem provides an acid test of the framework, as the Einstein equations contain hundreds of variables and thousands of terms.

physics.comp-ph↗

arXiv2013

Coarse graining of spin net models: dynamics of intertwiners

Spin foams are models of quantum gravity and therefore quantum space time. A key open issue is to determine the possible continuum phases of these models. Progress on this issue has been prohibited by the complexity of the full four--dimensional models. We consider here simplified analogue models, so called spin nets, that retain the main dynamical ingredient of spin foams, the simplicity constraints. For a certain class of these spin net models we determine the phase diagram and therefore the continuum phases via a coarse graining procedure based on tensor network renormalization. This procedure will also reveal an unexpected fixed point, which turns out to define a new triangulation invariant vertex model.

gr-qc↗

arXiv2013

The Transient Gravitational-Wave Sky

Interferometric detectors will very soon give us an unprecedented view of the gravitational-wave sky, and in particular of the explosive and transient Universe. Now is the time to challenge our theoretical understanding of short-duration gravitational-wave signatures from cataclysmic events, their connection to more traditional electromagnetic and particle astrophysics, and the data analysis techniques that will make the observations a reality. This paper summarizes the state of the art, future science opportunities, and current challenges in understanding gravitational-wave transients.

gr-qc↗

arXiv2013

GRHydro: A new open source general-relativistic magnetohydrodynamics code for the Einstein Toolkit

We present the new general-relativistic magnetohydrodynamics (GRMHD) capabilities of the Einstein Toolkit, an open-source community-driven numerical relativity and computational relativistic astrophysics code. The GRMHD extension of the Toolkit builds upon previous releases and implements the evolution of relativistic magnetised fluids in the ideal MHD limit in fully dynamical spacetimes using the same shock-capturing techniques previously applied to hydrodynamical evolution. In order to maintain the divergence-free character of the magnetic field, the code implements both hyperbolic divergence cleaning and constrained transport schemes. We present test results for a number of MHD tests in Minkowski and curved spacetimes. Minkowski tests include aligned and oblique planar shocks, cylindrical explosions, magnetic rotors, Alfvén waves and advected loops, as well as a set of tests designed to study the response of the divergence cleaning scheme to numerically generated monopoles. We study the code's performance in curved spacetimes with spherical accretion onto a black hole on a fixed background spacetime and in fully dynamical spacetimes by evolutions of a magnetised polytropic neutron star and of the collapse of a magnetised stellar core. Our results agree well with exact solutions where these are available and we demonstrate convergence. All code and input files used to generate the results are available on http://einsteintoolkit.org. This makes our work fully reproducible and provides new users with an introduction to applications of the code.

gr-qc↗

arXiv2012

A New Monte Carlo Method for Time-Dependent Neutrino Radiation Transport

Monte Carlo approaches to radiation transport have several attractive properties such as simplicity of implementation, high accuracy, and good parallel scaling. Moreover, Monte Carlo methods can handle complicated geometries and are relatively easy to extend to multiple spatial dimensions, which makes them potentially interesting in modeling complex multi-dimensional astrophysical phenomena such as core-collapse supernovae. The aim of this paper is to explore Monte Carlo methods for modeling neutrino transport in core-collapse supernovae. We generalize the Implicit Monte Carlo photon transport scheme of Fleck & Cummings and gray discrete-diffusion scheme of Densmore et al. to energy-, time-, and velocity-dependent neutrino transport. Using our 1D spherically-symmetric implementation, we show that, similar to the photon transport case, the implicit scheme enables significantly larger timesteps compared with explicit time discretization, without sacrificing accuracy, while the discrete-diffusion method leads to significant speed-ups at high optical depth. Our results suggest that a combination of spectral, velocity-dependent, Implicit Monte Carlo and discrete-diffusion Monte Carlo methods represents a robust approach for use in neutrino transport calculations in core-collapse supernovae. Our velocity-dependent scheme can easily be adapted to photon transport.

astro-ph.SR↗

arXiv2012

Numerical simulations with a first order BSSN formulation of Einstein's field equations

We present a new fully first order strongly hyperbolic representation of the BSSN formulation of Einstein's equations with optional constraint damping terms. We describe the characteristic fields of the system, discuss its hyperbolicity properties, and present two numerical implementations and simulations: one using finite differences, adaptive mesh refinement and in particular binary black holes, and another one using the discontinuous Galerkin method in spherical symmetry. The results of this paper constitute a first step in an effort to combine the robustness of BSSN evolutions with very high accuracy numerical techniques, such as spectral collocation multi-domain or discontinuous Galerkin methods.

gr-qc↗

arXiv2012

A Massive Data Parallel Computational Framework for Petascale/Exascale Hybrid Computer Systems

Heterogeneous systems are becoming more common on High Performance Computing (HPC) systems. Even using tools like CUDA and OpenCL it is a non-trivial task to obtain optimal performance on the GPU. Approaches to simplifying this task include Merge (a library based framework for heterogeneous multi-core systems), Zippy (a framework for parallel execution of codes on multiple GPUs), BSGP (a new programming language for general purpose computation on the GPU) and CUDA-lite (an enhancement to CUDA that transforms code based on annotations). In addition, efforts are underway to improve compiler tools for automatic parallelization and optimization of affine loop nests for GPUs and for automatic translation of OpenMP parallelized codes to CUDA. In this paper we present an alternative approach: a new computational framework for the development of massively data parallel scientific codes applications suitable for use on such petascale/exascale hybrid systems built upon the highly scalable Cactus framework. As the first non-trivial demonstration of its usefulness, we successfully developed a new 3D CFD code that achieves improved performance.

cs.DC↗