Overview of the QCDSP and QCDOC computers

Boyle, P. A.; Chen, D.; Christ, Norman H.; Clark, M. A.; Cohen, Saul D.; Cristian, C.; Dong, Zhihua; Gara, Alan; Joó, Bálint; Jung, Chulwoo; Kim, C.; Levkova, L.; Liao, X.; Liu, G.; Mawhinney, Robert D.; Ohta, Shigemi; Petrov, K.; Wettig, Tilo; Yamaguchi, Akinobu

doi:10.1147/rd.492.0351

Cited by 63 publications

(46 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The calculations reported here were performed on the QCDOC computers [65][66][67][68] at Columbia University, Edinburgh University, and at Brookhaven National Laboratory (BNL), and Argonne Leadership Class Facility (ALCF) BlueGene/P resources at Argonne National Laboratory (ANL).…”

Section: Acknowledgmentsmentioning

confidence: 99%

Continuum limit ofBKfrom2+1flavor domain wall QCD

et al. 2011

Self Cite

View full text Add to dashboard Cite

2We determine the neutral kaon mixing matrix element B K in the continuum limit with 2+1 flavors of domain wall fermions, using the Iwasaki gauge action at two different lattice spacings. These lattice fermions have near exact chiral symmetry and therefore avoid artificial lattice operator mixing.We introduce a significant improvement to the conventional NPR method in which the bare matrix elements are renormalized non-perturbatively in the RI-MOM scheme and are then converted into the MS scheme using continuum perturbation theory. In addition to RI-MOM, we introduce and implement four non-exceptional intermediate momentum schemes that suppress infrared non-perturbative uncertainties in the renormalization procedure. We compute the conversion factors relating the matrix elements in this family of RI-SMOM schemes and MS at one-loop order.Comparison of the results obtained using these different intermediate schemes allows for a more reliable estimate of the unknown higher-order contributions and hence for a correspondingly more robust estimate of the systematic error. We also apply a recently proposed approach in which twisted boundary conditions are used to control the Symanzik expansion for off-shell vertex functions leading to a better control of the renormalization in the continuum limit.We control chiral extrapolation errors by considering both the NLO SU(2) chiral effective theory, and an analytic mass expansion. We obtain B MS K (3 GeV) = 0.529(5) stat (15) χ (2) FV (11) NPR . This corresponds toB RGI K = 0.749(7) stat (21) χ (3) FV (15) NPR . Adding all sources of error in quadrature we obtainB RGI K = 0.749(27) combined , with an overall combined error of 3.6%.3

show abstract

Section: Acknowledgmentsmentioning

confidence: 99%

Continuum limit ofBKfrom2+1flavor domain wall QCD

et al. 2011

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, design, production and deployment of such cost-efficient machines becomes more challenging with prices per GFlops for commodity systems going down and technology becoming more complex. For example, the latest generation of custom machines, apeNEXT [1] and QCDOC [2], was based on custom designed processors. Given the costs and the risks involved in an ASIC design, this has become less of an option today.…”

Section: Introductionmentioning

confidence: 99%

QPACE -- a QCD parallel computer based on cell processors

Pleiter¹,

Maurer²

2010

Proceedings of the XXVII International Symposium on Lattice Field Theory — PoS(LAT2009)

View full text Add to dashboard Cite

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.

show abstract

“…At the time, physicist Richard Feynman and his son Carl were working on the network of the Connection Machine and optimizing LQCD algorithms. This was followed by the development of "QCD on a chip," or QCDOC computers, by a group from Columbia University (Boyle et al, 2005). A collaborator from the Columbia group went on to join IBM and designed the closely related commercial project, the IBM BlueGene/L computer (Chromodynamics, 2015).…”

Section: Massively Parallel Computing -Lattice Gauge Calculationsmentioning

confidence: 99%