The Particle-in-cell (PIC) model, which has been used extensively on supercomputers to simulate fusion and space plasma phenomena, is difficult to develop for the Massively Parallel Processor (MPP) because the model requires indirect indexing in computing electric fields. To overcome the difficulties, a parallel PIC algorithm is developed for MPP by mapping particles in a cell randomly to a row of processors. Because of this mapping, the algorithm needs only the nearest neighbor communication to sort particles and to collect charge density for each cell. From the cell charge density, this algorithm then calculates electric fields at the cell by Fast Fourier Transform. The developed PIC code has a speed comparable to that of the vectorized PIC code on CRAY X-MP. The results from simulating the plasma instabilities of a cold electron beam in a hot electron background are presented.
Synchronization techniques are proposed f o r algorithms which spawn processes remotely o n loosely coupled processors based o n run-time characteristics. T h e performance of the proposed synchronization schemesare measured o n the iPSC/2 and S N A P -1 multiprocessors and their implementation cost is discussed. Results show that processes created dynamically throughout a distributed system can be synchronized at comparable overhead and cost to that required f o r fixedlocation process creation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.