Distributed Green's function Monte Carlo calculations.
Item
-
Title
-
Distributed Green's function Monte Carlo calculations.
-
Identifier
-
AAI9510644
-
identifier
-
9510644
-
Creator
-
Chen, Jian.
-
Contributor
-
Adviser: David Arnow
-
Date
-
1994
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Computer Science
-
Abstract
-
A useful and interesting problem is to solve a linear partial differential equation using the Monte Carlo method. For example, the only general method for obtaining exact solutions to the Schrodinger equation is the Green's function Monte Carlo method. The fermion Green's function Monte Carlo requires computations on multi processors since it demands on memory and CPU power. Parallel simulations, like distributed Monte Carlo, have their unique characteristics which do not exist in sequential Monte Carlo computations. For instance, an distributed estimator is needed. Load balancing has to be taken into account to achieve better computational performance. There is no guarantee that correlations will not occur in such a distributed computation if subsequences of a single pseudorandom sequence is used by individual worker processes. Therefore, implementing a distributed random number generator system to support these simulations is essential. A procedure that integrates all above was developed to carry out distributed Green's function Monte Carlo calculations. A distributed message passing system, DP, that is suitable for large scale scientific computations is employed to carry out data transmission and process synchronization of the distributed computation.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
