Author/Authors :
QCD_TARO Collaboration، نويسنده , , K. Akemi، نويسنده , , Ph. deForcrand، نويسنده , , M. Fujisaki، نويسنده , , T. Hashimoto، نويسنده , , H.C. Hege، نويسنده , , S. Hioki، نويسنده , , O. Miyamura، نويسنده , , A. Nakamura، نويسنده , , M. Okuda، نويسنده , , I.O. Stamatescu، نويسنده , , Y. Tago، نويسنده , , T. Takaishi، نويسنده ,
Abstract :
A parallelization study of Quantum Chromodynamics (QCD) simulation on a non-dedicated MIMD computer, Fujitsu AP1000, is reported. The AP1000 is a highly parallel computer with distributed memory. Data communication is fast not only among the nearest neighbor processing units, but also between widely separated ones. We study SU(3) quenched QCD simulation with lattice sizes of 324 and 323 × 48, on the 512-processor AP1000, and report parallelization techniques and physical results for a renormalization group study and hadron spectroscopy. The AP1000 is efficient for these QCD simulations and shows good performance. A bottleneck occurs in the current simulation as data is passed to the front-end processor for storage.
