A bst 1: actParticle-in-cell (PIC) is a simulation method widely used in many important scientific areas; such as plasma physics, semiconductor device physics, global climate modeling, and galaxy dynamics. In general, these simulations are extremely computationally intensive and, therefore, very time consuming even on supercomputers. We consider a PIC algorithm which simulates the behavior of charged particles in an electromagnetic field. This study is performed in order to explore parallel processing issues, such as relationships between speedup and problem partitioning schemes, problem size and time duration of each iteration for PIC method on different multiprocessors. A new partitioning scheme, hybrid partitioning, is introduced.Hybrid partitioning has evolved out of two general approaches to PIC problem decomposition on multiprocessors, partitioning particles and partitioning the space. We chose the shared memory multiprocessor environment for analyzing our parallel (distributed computing) algorithms. Two different BBN Butterfly machines (GPlOOO and TC2OOO) were employed as testbeds.

No abstract

No abstract