arXiv subjects

Th. Lippert

Publications and source records attributed to Th. Lippert.

79 records · Page 5Linked to original sources

arXiv1995

Hyper-Systolic Processing on APE100/Quadrics: N^2-Loop Computations

We investigate the performance gains from hyper-systolic implementations of n^2-loop problems on the massively parallel computer Quadrics, exploiting its 3-dimensional interprocessor connectivity. For illustration we study the communication aspects of an exact molecular dynamics simulation of n particles with Coulomb (or gravitational) interactions. We compare the interprocessor communication costs of the standard-systolic and the hyper-systolic approaches for various granularities. We predict gain factors as large as 3 on the Q4 and 8 on the QH4 and measure actual performances on these machine configurations. We conclude that it appears feasable to investigate the thermodynamics of a full gravitating n-body problem with O(10000) particles using the new method on a QH4 system.

hep-lat↗

arXiv1995

QCD with dynamical Wilson fermions - first results from SESAM

First results of a recently started simulation of full QCD with two flavours of sea-quarks at a coupling of $β= 5.6$ on a $16^3 \times 32$ lattice are presented. Emphasis is laid on the statistical significance that can be achieved by an ``integrated luminosity'' of 140 TFlop$\times$hrs, for Hybrid Monte Carlo simulations at four intermediate values of $\frac{m_π}{m_ρ}$. The simulation takes place on the Quadrics QH2 at DESY/Zeuthen and DFG/Bielefeld. The performance is optimized by means of BiCGStab and the chronological inversion method of Brower et al. We discuss the systematic errors arising from lack of the molecular dynamic's reversibility on the 32-bit QH2. For plaquette and meson correlators we find integrated autocorrelation times of $< 20$ units of molecular dynamics time and exponential autocorrelation times of about $50$ units. Using these results we perform preliminary measurements of the central potential and $π$ and $ρ$ correlators on independent configurations and obtain first estimates of the lattice spacings at three values of the dynamical hopping parameter.

hep-lat↗

arXiv1995

Hyper-Systolic Parallel Computing

A new class of parallel algorithms is introduced that can achieve a complexity of O(n^3/2) with respect to the interprocessor communication, in the exact computation of systems with pairwise mutual interactions of all elements. Hitherto, conventional methods exhibit a communicational complexity of O(n^2). The amount of computation operations is not altered for the new algorithm which can be formulated as a kind of h-range problem, known from the mathematical field of Additive Number Theory. We will demonstrate the reduction in communicational expense by comparing the standard-systolic algorithm and the new algorithm on the connection machine CM5 and the CRAY T3D. The parallel method can be useful in various scientific and engineering fields like exact n-body dynamics with long range forces, polymer chains, protein folding or signal processing.

hep-lat↗

arXiv1995

Monopoles in Compact U(1) -- Anatomy of the Phase Transition

We present evidence that the existence of a first order phase transition in compact U(1) with Wilson action is not related to monopole loops wrapping around the toroidal lattice, as has been previously suggested. Our analysis is based on the suppression of such loops by `soft boundary conditions' that correspond to an infinitely large chemical potential for the monopoles on the boundary, during the updating process. It is observed that the double peak structure characteristic for the first order phase transition reappears at sufficiently large lattice sizes and separations from the lattice boundary.

hep-lat↗

arXiv1995

First-Order Signals in Compact QED with Monopole Suppressed Boundaries

Pure gauge compact QED on hypercubic lattices is considered with periodically closed monopole currents suppressed. We compute observables on sublattices which are nested around the centre of the lattice in order to locate regions where translation symmetry is approximately recovered. Our Monte Carlo simulations on $24^4$-lattices give indications for a first-order nature of the U(1) phase transition.

hep-lat↗

arXiv1994

Accelerating Wilson Fermion Matrix Inversions by Means of the Stabilized Biconjugate Gradient Algorithm

The stabilized biconjugate gradient algorithm BiCGStab recently presented by van der Vorst is applied to the inversion of the lattice fermion operator in the Wilson formulation of lattice Quantum Chromodynamics. Its computational efficiency is tested in a comparative study against the conjugate gradient and minimal residual methods. Both for quenched gauge configurations at beta= 6.0 and gauge configurations with dynamical fermions at beta=5.4, we find BiCGStab to be superior to the other methods. BiCGStab turns out to be particularly useful in the chiral regime of small quark masses.

hep-lat↗

arXiv1993

Monopole clusters and critical dynamics in four-dimensional U(1)

We investigate monopoles in four-dimensional compact U(1) with Wilson action. We focus our attention on monopole clusters as they can be identified unambiguously contrary to monopole loops. We locate the clusters and determine their properties near the U(1) phase transition. The Coulomb phase is characterized by several small clusters, whereas in the confined phase the small clusters coalesce to one large cluster filling up the whole system. We find that clusters winding around the periodic lattice are absent within both phases and during the transition. However, within the confined phase, we observe periodically closed monopole loops if cooling is applied.

hep-lat↗