Stefan Solbrig

QPACE: Quantum Chromodynamics Parallel Computing on the Cell Broadband Engine

Application-driven computers for lattice gauge theory simulations have often been based on system-on-chip designs, but the development costs can be prohibitive for academic project budgets. An alternative approach uses compute nodes based on a commercial processor tightly coupled to a custom-designed network processor. Preliminary analysis shows that this solution offers good performance, but it also entails several challenges, including those arising from the processors multicore structure and from implementing the network processor on a field-programmable gate array.

Center vortices and Dirac eigenmodes in SU(2) lattice gauge theory

Abstract We study the interplay between Dirac eigenmodes and center vortices in SU(2) lattice gauge theory. In particular, we focus on vortex-removed configurations and compare them to an ensemble of configurations with random changes of the link variables. We show that removing the vortices destroys all zero modes and the near zero modes are no longer coupled to topological structures. The Dirac spectrum for vortex-removed configurations in many respects resembles a free spectrum thus leading to a vanishing chiral condensate. Configurations with random changes leave the topological features of the Dirac eigensystem intact. We finally show that smooth center vortex configurations give rise to zero modes and topological near zero modes.

Quantitative comparison of filtering methods in lattice QCD

Abstract.We systematically compare filtering methods used to extract topological excitations (like instantons, calorons, monopoles and vortices) from lattice gauge configurations, namely APE-smearing and spectral decompositions based on lattice Dirac and Laplace operators. Each of these techniques introduces ambiguities, which can invalidate the interpretation of the results. We show, however, that all these methods, when handled with care, reveal very similar topological structures. Hence, these common structures are free of ambiguities and faithfully represent infrared degrees of freedom in the QCD vacuum. As an application we discuss an interesting power law for the clusters of filtered topological charge.

QPACE - a QCD parallel computer based on Cell processors

QPACE is a novel parallel computer which has been developed to be primarily used for lattice QCD simulations. The compute power is provided by the IBM PowerXCell 8i processor, an enhanced version of the Cell processor that is used in the Playstation 3. The QPACE nodes are interconnected by a custom, application optimized 3-dimensional torus network implemented on an FPGA. To achieve the very high packaging density of 26 TFlops per rack a new water cooling concept has been developed and successfully realized. In this paper we give an overview of the architecture and highlight some important technical details of the system. Furthermore, we provide initial performance results and report on the installation of 8 QPACE racks providing an aggregate peak performance of 200 TFlops.

QCD on the Cell Broadband Engine

We evaluate IBMs Enhanced Cell Broadband Engine (BE) as a possible building block of a new generation of lattice QCD machines. The Enhanced Cell BE will provide full support of double-precision floating-point arithmetics, including IEEE-compliant rounding. We have developed a performance model and applied it to relevant lattice QCD kernels. The performance estimates are supported by micro- and application-benchmarks that have been obtained on currently available Cell BE-based computers, such as IBM QS20 blades and PlayStation 3. The results are encouraging and show that this processor is an interesting option for lattice QCD applications. For a massively parallel machine on the basis of the Cell BE, an application-optimized network needs to be developed.

Dependence of Dirac eigenmodes on boundary conditions for SU(2) lattice gauge theory

Abstract We analyze zero modes of the Dirac operator for SU(2) lattice gauge theory. We find that the zero modes are strongly localized in all 4 directions. The position of these lumps depends on the boundary conditions we use for the Dirac operator. We compare periodic boundary conditions and anti-periodic boundary conditions and find that the position of the zero modes jumps for about one third of the configurations.

iDataCool: HPC with Hot-Water Cooling and Energy Reuse

iDataCool is an HPC architecture jointly developed by the University of Regensburg and the IBM Research and Development Lab Boblingen. It is based on IBM’s iDataPlex platform, whose air-cooling solution was replaced by a custom water-cooling solution that allows for cooling water temperatures of 70°C/158°F. The system is coupled to an adsorption chiller by InvenSor that operates efficiently at these temperatures. Thus a significant portion of the energy spent on HPC can be recovered in the form of chilled water, which can then be used to cool other parts of the computing center. We describe the architecture of iDataCool and present benchmarks of the cooling performance and the energy (reuse) efficiency.

Status of the QPACE Project

We give an overview of the QPACE project, which is pursuing the development of a massively parallel, scalable supercomputer for LQCD. The machine is a three-dimensional torus of identical processing nodes, based on the PowerXCell 8i processor. The nodes are connected by an FPGAbased, application-optimized network processor attached to the PowerXCell 8i processor. We present a performance analysis of lattice QCD codes on QPACE and corresponding hardware benchmarks.

Remnant index theorem and low-lying eigenmodes for twisted mass fermions

Abstract We analyze the low-lying spectrum and eigenmodes of lattice Dirac operators with a twisted mass term. The twist term expels the eigenvalues from a strip in the complex plane and all eigenmodes obtain a non-vanishing matrix element with γ 5 . For a twisted Ginsparg–Wilson operator the spectrum is located on two arcs in the complex plane. Modes due to non-trivial topological charge of the underlying gauge field have their eigenvalues at the edges of these arcs and obey a remnant index theorem. For configurations in the confined phase we find that the twist mainly affects the zero modes, while the bulk of the spectrum is essentially unchanged.

Loss-function learning for digital tissue deconvolution

The gene expression profile of a tissue averages the expression profiles of all cells in this tissue. Digital tissue deconvolution (DTD) addresses the following inverse problem: Given the expression profile y of a tissue, what is the cellular composition c of that tissue? If X is a matrix whose columns are reference profiles of individual cell types, the composition c can be computed by minimizing \(\mathcal {L}(y-Xc)\) for a given loss function \(\mathcal {L}\). Current methods use predefined all-purpose loss functions. They successfully quantify the dominating cells of a tissue, while often falling short in detecting small cell populations.