arXiv subjects

Simon Heybrock

Publications and source records attributed to Simon Heybrock.

12 recordsLinked to original sources

arXiv2023

Octet baryon isovector charges from $N_f = 2 + 1$ lattice QCD

We determine the axial, scalar and tensor isovector charges of the nucleon, sigma and cascade baryons as well as the difference between the up and down quark masses, $m_u-m_d$. We employ gauge ensembles with $N_f=2+1$ non-perturbatively improved Wilson fermions at six values of the lattice spacing in the range $a\approx (0.039 - 0.098) \,$fm, generated by the Coordinated Lattice Simulations (CLS) effort. The pion mass $M_\pi$ ranges from around $430 \, $MeV down to a near physical value of $130 \, $MeV and the linear spatial lattice extent $L$ varies from $6.5\,M_{\pi}^{-1}$ to $3.0\,M_{\pi}^{-1}$, where $L M_\pi \geq 4$ for the majority of the ensembles. This allows us to perform a controlled interpolation/extrapolation of the charges to the physical mass point in the infinite volume and continuum limit. Investigating SU(3) flavour symmetry, we find moderate symmetry breaking effects for the axial charges at the physical quark mass point, while no significant effects are found for the other charges within current uncertainties.

hep-lat↗

arXiv2020

Scipp: Scientific data handling with labeled multi-dimensional arrays for C++ and Python

Scipp is heavily inspired by the Python library xarray. It enriches raw NumPy-like multi-dimensional arrays of data by adding named dimensions and associated coordinates. Multiple arrays are combined into datasets. On top of this, scipp introduces (i) implicit handling of physical units, (ii) implicit propagation of uncertainties, (iii) support for histograms, i.e., bin-edge coordinate axes, which exceed the data's dimension extent by one, and (iv) support for event data. In conjunction these features enable a more natural and more concise user experience. The combination of named dimensions, coordinates, and units helps to drastically reduce the risk for programming errors. The core of scipp is written in C++ to open opportunities for performance improvements that a Python-based solution would not allow for. On top of the C++ core, scipp's Python components provide functionality for plotting and content representations, e.g., for use in Jupyter Notebooks. While none of scipp's concepts in isolation is novel per-se, we are not aware of any project combining all of these aspects in a single coherent software package.

cs.MS↗

arXiv2017

Baryonic and mesonic 3-point functions with open spin indices

We have implemented a new way of computing three-point correlation functions. It is based on a factorization of the entire correlation function into two parts which are evaluated with open spin- (and to some extent flavor-) indices. This allows us to estimate the two contributions simultaneously for many different initial and final states and momenta, with little computational overhead. We explain this factorization as well as its efficient implementation in a new library which has been written to provide the necessary functionality on modern parallel architectures and on CPUs, including Intel's Xeon Phi series.

hep-lat↗

arXiv2016

Multiple right-hand-side setup for the DD-αAMG

The setup cost of a modern solver such as DD-αAMG (Wuppertal Multigrid) is a significant contribution to the total time spent on solving the Dirac equation, and in HMC it can even be dominant. We present an improved implementation of this algorithm with modified computation order in the setup procedure. By processing multiple right-hand sides simultaneously we can alleviate many of the performance issues of the default single right-hand-side setup. The main improvements are as follows: By combining multiple right-hand sides the message size for off-chip communication is larger, which leads to better utilization of the network bandwidth. Many matrix-vector products are replaced by matrix-matrix products, leading to better cache reuse. The synchronization overhead inflicted by on-chip parallelization (threading), which is becoming crucial on many-core architectures such as the Intel Xeon Phi, is effectively reduced. In the parts implemented so far, we observe a speedup of roughly 3x compared to the optimized version of the single right-hand-side setup on realistic lattices.

hep-lat↗

arXiv2015

Adaptive algebraic multigrid on SIMD architectures

We present details of our implementation of the Wuppertal adaptive algebraic multigrid code DD-$α$AMG on SIMD architectures, with particular emphasis on the Intel Xeon Phi processor (KNC) used in QPACE 2. As a smoother, the algorithm uses a domain-decomposition-based solver code previously developed for the KNC in Regensburg. We optimized the remaining parts of the multigrid code and conclude that it is a very good target for SIMD architectures. Some of the remaining bottlenecks can be eliminated by vectorizing over multiple test vectors in the setup, which is discussed in the contribution of Daniel Richtmann.

physics.comp-ph↗

arXiv2015

QPACE 2 and Domain Decomposition on the Intel Xeon Phi

We give an overview of QPACE 2, which is a custom-designed supercomputer based on Intel Xeon Phi processors, developed in a collaboration of Regensburg University and Eurotech. We give some general recommendations for how to write high-performance code for the Xeon Phi and then discuss our implementation of a domain-decomposition-based solver and present a number of benchmarks.

cs.DC↗

arXiv2014

Lattice QCD with Domain Decomposition on Intel Xeon Phi Co-Processors

The gap between the cost of moving data and the cost of computing continues to grow, making it ever harder to design iterative solvers on extreme-scale architectures. This problem can be alleviated by alternative algorithms that reduce the amount of data movement. We investigate this in the context of Lattice Quantum Chromodynamics and implement such an alternative solver algorithm, based on domain decomposition, on Intel Xeon Phi co-processor (KNC) clusters. We demonstrate close-to-linear on-chip scaling to all 60 cores of the KNC. With a mix of single- and half-precision the domain-decomposition method sustains 400-500 Gflop/s per chip. Compared to an optimized KNC implementation of a standard solver [1], our full multi-node domain-decomposition solver strong-scales to more nodes and reduces the time-to-solution by a factor of 5.

hep-lat↗

arXiv2011

A nested Krylov subspace method to compute the sign function of large complex matrices

We present an acceleration of the well-established Krylov-Ritz methods to compute the sign function of large complex matrices, as needed in lattice QCD simulations involving the overlap Dirac operator at both zero and nonzero baryon density. Krylov-Ritz methods approximate the sign function using a projection on a Krylov subspace. To achieve a high accuracy this subspace must be taken quite large, which makes the method too costly. The new idea is to make a further projection on an even smaller, nested Krylov subspace. If additionally an intermediate preconditioning step is applied, this projection can be performed without affecting the accuracy of the approximation, and a substantial gain in efficiency is achieved for both Hermitian and non-Hermitian matrices. The numerical efficiency of the method is demonstrated on lattice configurations of sizes ranging from 4^4 to 10^4, and the new results are compared with those obtained with rational approximation methods.

hep-lat↗

arXiv2010

Double-pass variants for multi-shift BiCGstab(ell)

In analogy to Neuberger's double-pass algorithm for the Conjugate Gradient inversion with multi-shifts we introduce a double-pass variant for BiCGstab(ell). One possible application is the overlap operator of QCD at non-zero chemical potential, where the kernel of the sign function is non-Hermitian. The sign function can be replaced by a partial fraction expansion, requiring multi-shift inversions. We compare the performance of the new method with other available algorithms, namely partial fraction expansions with restarted FOM inversions and the Krylov-Ritz method using nested Krylov subspaces.

hep-lat↗

arXiv2010

Short-recurrence Krylov subspace methods for the overlap Dirac operator at nonzero chemical potential

The overlap operator in lattice QCD requires the computation of the sign function of a matrix, which is non-Hermitian in the presence of a quark chemical potential. In previous work we introduced an Arnoldi-based Krylov subspace approximation, which uses long recurrences. Even after the deflation of critical eigenvalues, the low efficiency of the method restricts its application to small lattices. Here we propose new short-recurrence methods which strongly enhance the efficiency of the computational method. Using rational approximations to the sign function we introduce two variants, based on the restarted Arnoldi process and on the two-sided Lanczos method, respectively, which become very efficient when combined with multishift solvers. Alternatively, in the variant based on the two-sided Lanczos method the sign function can be evaluated directly. We present numerical results which compare the efficiencies of a restarted Arnoldi-based method and the direct two-sided Lanczos approximation for various lattice sizes. We also show that our new methods gain substantially when combined with deflation.

hep-lat↗

arXiv2009

A nested Krylov subspace method for the overlap operator

We present a novel method to compute the overlap Dirac operator at zero and nonzero quark chemical potential. To approximate the sign function of large, sparse matrices, standard methods project the operator on a much smaller Krylov subspace, on which the matrix function is computed exactly. However, for large lattices this subspace can still be too large for an efficient calculation of the sign function. The idea of the new method is to nest Krylov subspace approximations by making a further projection on an even smaller subspace, which is then small enough to compute the sign function efficiently, and this without any noticeable loss of numerical accuracy. We demonstrate the efficiency of the method both on Hermitian and non-Hermitian matrices.

hep-lat↗

arXiv2009

Krylov subspace methods and the sign function: multishifts and deflation in the non-Hermitian case

Rational approximations of the matrix sign function lead to multishift methods. For non-Hermitian matrices long recurrences can cause storage problems, which can be circumvented with restarts. Together with deflation we obtain efficient iterative methods, as we show in numerical experiments for the overlap Dirac operator at non-vanishing quark chemical potential for lattices up to size 10^4.

hep-lat↗