TABLE OF CONTENTS Parallel computers provide opportunities to simulate larger problems and longer timescales, with enhanced detail and complexity, as in some of the following examples :
Crash scenarios can be modeled, including detailed descriptions of both occupants and passenger compartments. Typical applications use distributed-memory parallel computers with as many as 32 processors.
Turbulence simulations greatly benefit from hundreds and even thousands of processors. The calculations can be extended to include complex theoretical models for turbulence. Likewise a fully nonlinear simulation of turbulence on parallel computers can model extended length and time scales.
Material behavior characterized by a statistical description of flaws can be simulated with systems with billions of degrees of freedom.
The cavitation of the fluid column illustrated in Figure 2.1 of Chapter 2 runs in two hours on 64 processors of a Linux cluster parallel computer using the ParaDyn program. The equivalent run on a slower single-processor computer would require about 256 hours for a single simulation of the fluid column! At present ( 2006 ) it takes too long to run a set of problems to study the behavior of such systems without access to a parallel computer. There are applications in many fields that have benefitted from the increased capacity and speeds afforded by parallel computers.
The two most popular parallel computer architectures are: (i) shared-memory computers, and (ii) distributed-memory computers. A shared-memory computer consists of many processors sharing a common memory with relatively expensive hardware connecting the processors to the full shared...
