Searcharxiv⌕ Search

arXiv subjects

Peter Boyle

Publications and source records attributed to Peter Boyle.

At least 55 records · Page 3Linked to original sources
arXiv2017

Accelerating HPC codes on Intel(R) Omni-Path Architecture networks: From particle physics to Machine Learning

We discuss practical methods to ensure near wirespeed performance from clusters with either one or two Intel(R) Omni-Path host fabric interfaces (HFI) per node, and Intel(R) Xeon Phi(TM) 72xx (Knight's Landing) processors, and using the Linux operating system. The study evaluates the performance improvements achievable and the required programming approaches in two distinct example problems: firstly in Cartesian communicator halo exchange problems, appropriate for structured grid PDE solvers that arise in quantum chromodynamics simulations of particle physics, and secondly in gradient reduction appropriate to synchronous stochastic gradient descent for machine learning. As an example, we accelerate a published Baidu Research reduction code and obtain a factor of ten speedup over the original code using the techniques discussed in this paper. This displays how a factor of ten speedup in strongly scaled distributed machine learning could be achieved when synchronous stochastic gradient descent is massively parallelised with a fixed mini-batch size. We find a significant improvement in performance robustness when memory is obtained using carefully allocated 2MB "huge" virtual memory pages, implying that either non-standard allocation routines should be used for communication buffers. These can be accessed via a LD\_PRELOAD override in the manner suggested by libhugetlbfs. We make use of a the Intel(R) MPI 2019 library "Technology Preview" and underlying software to enable thread concurrency throughout the communication software stake via multiple PSM2 endpoints per process and use of multiple independent MPI communicators. When using a single MPI process per node, we find that this greatly accelerates delivered bandwidth in many core Intel(R) Xeon Phi processors.

cs.DC↗
arXiv2017

BSM Kaon Mixing at the Physical Point

We present a progress update on the calculation of beyond the standard model (BSM) kaon mixing matrix elements at the physical point. Simulations are performed using 2+1 flavour domain wall lattice QCD with the Iwasaki gauge action at 3 lattice spacings and with pion masses ranging from 430 MeV to the physical pion mass.

hep-lat↗
arXiv2017

Testing algorithms for critical slowing down

We present the preliminary tests on two modifications of the Hybrid Monte Carlo (HMC) algorithm. Both algorithms are designed to travel much farther in the Hamiltonian phase space for each trajectory and reduce the autocorrelations among physical observables thus tackling the critical slowing down towards the continuum limit. We present a comparison of costs of the new algorithms with the standard HMC evolution for pure gauge fields, studying the autocorrelation times for various quantities including the topological charge.

hep-lat↗
arXiv2017

Isospin Breaking Corrections to the HVP with Domain Wall Fermions

We present results for the QED and strong isospin breaking corrections to the hadronic vacuum polarization using $N_f=2+1$ Domain Wall fermions. QED is included in an electro-quenched setup using two different methods, a stochastic and a perturbative approach. Results and statistical errors from both methods are directly compared with each other.

hep-lat↗
arXiv2017

BSM Kaon Mixing at the Physical Point

We present new preliminary results for bag parameters and ratios of the BSM kaon mixing operators measured at the physical point. The results are obtained from simulations of domain wall fermion QCD with 2+1 flavours with an Iwasaki gauge, and now include pion and kaon masses very close to the physical point. We compare these results to our collaboration's previous results obtained from heavier simulated quarks.

hep-lat↗
arXiv2016

Electromagnetic Corrections to Meson Masses and the HVP

We present an exploratory study of the electromagnetic corrections to meson masses and the hadronic vacuum polarization using $N_f=2+1$ Domain Wall fermions. These corrections are estimated with two different approaches, a stochastic approach using $U(1)$ gauge configurations for the photon fields, and a perturbative approach through a QED perturbative expansion of the QCD+QED path integral. We compare results and statistical errors from both methods.

hep-lat↗
arXiv2016

Charm Physics with Domain Wall Fermions and Physical Pion Masses

We present RBC/UKQCD's charm project using $N_f=2+1$ flavour ensembles with inverse lattice spacings in the range $1.73-2.77\,\mathrm{GeV}$ and two physical pion mass ensembles. Domain wall fermions are used for the light as well as the charm quarks. We discuss our strategy for the extraction of the decay constants $f_D$ and $f_{D_s}$ and their extrapolation to the continuum limit, physical pion masses and the physical heavy quark mass. Our preliminary results are $f_D=208.7(2.8)\,\mathrm{MeV}$ and $f_{D_s}=246.4(1.9)\,\mathrm{MeV}$ where the quoted error is statistical only. We outline our current approach to extend the reach in the heavy quark mass and present preliminary results.

hep-lat↗
arXiv2016

Hierarchically deflated conjugate residual

We present a progress report on a new class of multigrid solver algorithm suitable for the solution of 5d chiral fermions such as Domain Wall fermions and the Continued Fraction overlap. Unlike HDCG \cite{Boyle:2014rwa}, the algorithm works directly on a nearest neighbour fine operator. The fine operator used is Hermitian indefinite, for example $Γ_5 D_{dwf}$, and convergence is achieved with an indefinite matrix solver such as outer iteration based on conjugate residual. As a result coarse space representations of the operator remain nearest neighbour, giving an 8 point stencil rather than the 81 point stencil used in HDCG. It is hoped this may make it viable to recalculate the matrix elements of the little Dirac operator in an HMC evolution.

hep-lat↗
arXiv2016

A massive momentum-subtraction scheme

A new renormalization scheme is defined for fermion bilinears in QCD at non vanishing quark masses. This new scheme, denoted RI/mSMOM, preserves the benefits of the nonexceptional momenta introduced in the RI/SMOM scheme, and allows a definition of renormalized composite fields away from the chiral limit. Some properties of the scheme are investigated by performing explicit one-loop computation in dimensional regularization.

hep-lat↗
arXiv2016

A massive momentum-subtraction scheme

We introduce a new massive renormalization scheme, denoted mSMOM, as a modification of the existing RI/SMOM scheme. We use SMOM for defining renormalized fermion bilinears in QCD at non-vanishing fermion mass. This scheme has properties similar to those of the SMOM scheme, such as the use of non-exceptional symmetric momenta, while in contrast to SMOM, it defines the renormalized fields away from the chiral limit. Here we discuss some of the properties of mSMOM, and present non-perturbative arguments for deriving some renormalization constants. The results of a 1-loop calculation in dimensional regularization are briefly summarised to illustrate some properties of the scheme.

hep-lat↗
arXiv2016

An exploratory study of heavy domain wall fermions on the lattice

We report on an exploratory study of domain wall fermions (DWF) as a lattice regularisation for heavy quarks. Within the framework of quenched QCD with the tree-level improved Symanzik gauge action we identify the DWF parameters which minimise discretisation effects. We find the corresponding effective 4$d$ overlap operator to be exponentially local, independent of the quark mass. We determine a maximum bare heavy quark mass of $am_h\approx 0.4$, below which the approximate chiral symmetry and O(a)-improvement of DWF are sustained. This threshold appears to be largely independent of the lattice spacing. Based on these findings, we carried out a detailed scaling study for the heavy-strange meson dispersion relation and decay constant on four ensembles with lattice spacings in the range $2.0-5.7\,\mathrm{GeV}$. We observe very mild $a^2$ scaling towards the continuum limit. Our findings establish a sound basis for heavy DWF in dynamical simulations of lattice QCD with relevance to Standard Model phenomenology.

hep-lat↗
arXiv2015

Grid: A next generation data parallel C++ QCD library

In this proceedings we discuss the motivation, implementation details, and performance of a new physics code base called Grid. It is intended to be more performant, more general, but similar in spirit to QDP++\cite{QDP}. Our approach is to engineer the basic type system to be consistently fast, rather than bolt on a few optimised routines, and we are attempt to write all our optimised routines directly in the Grid framework. It is hoped this will deliver best known practice performance across the next generation of supercomputers, which will provide programming challenges to traditional scalar codes. We illustrate the programming patterns used to implement our goals, and advances in productivity that have been enabled by using new features in C++11.

hep-lat↗
arXiv2015

Domain Wall Charm Physics with Physical Pion Masses: Decay constants, bag and $ξ$ parameters

We provide an overview of RBC/UKQCD's charm project on 2+1 flavour physical pion mass ensembles using Möbius Domain Wall Fermions for the light as well as for the charm quark. We discuss the analysis strategy in detail and present results at the different stages of the analysis for $D$ and $D_s$ decay constants as well as the bag and $ξ$ parameters. We also discuss future approaches to extend the reach in the heavy quark mass.

hep-lat↗
arXiv2015

$N_f=2+1+1$ renormalisation of four-quark operators

When several four-quark operators are allowed to mix through renormalisation, this can considerably amplify the problems coming from perturbative truncation and discretisation effects. In this work we investigate whether our previous $B_K$ strategy can conveniently be generalised to a wider set of operators, corresponding to the so-called "SUSY $B_K$" basis of four-quark operators. We show that the discretisation effects, when plotted as a function of $ap$, are surprisingly independent of the lattice spacing. They appear reasonably under control up to very large energy scales. This allows us to discuss the effect of varying the intermediate scale on which the perturbative matching is done, and therefore the prospects of future high-precision studies with a Rome-Southampton renormalisation.

hep-lat↗
arXiv2015

Towards the physical point hadronic vacuum polarisation from Moebius DWF

We present steps towards the computation of the leading-order hadronic contribution to the muon anomalous magnetic moment on RBC/UKQCD physical point DWF ensembles. We discuss several methods for controlling and reducing uncertainties associated to the determination of the HVP form factor.

hep-lat↗
arXiv2015

Charm physics with Moebius Domain Wall Fermions

We present results showing that Domain Wall fermions are a suitable discretisation for the simulation of heavy quarks. This is done by a continuum scaling study of charm quarks in a Möbius Domain Wall formalism using a quenched set-up. We find that discretisation effects remain well controlled by the choice of Domain Wall parameters preparing the ground work for the ongoing dynamical $2+1f$ charm program of RBC/UKQCD.

hep-lat↗
arXiv2014

NPR step-scaling across the charm threshold

Matching Non-Perturbative Renormalisation on the lattice and perturbative renormalisation would benefit from higher matching scales, which are needed for observables entering their permile era such as BK. In this work we lay down a strategy, within the Rome-Southampton framework, to push this scale higher across the charm threshold, and apply it to an exploration of the BK running from 3 GeV to 9 GeV. This is done on Nf = 2+1+1 ensembles generated by the RBC-UKQCD collaboration, and features a close study of the discretisation effects.

hep-lat↗