An investigation of the applicability of parallel computation to demanding chemical engineering problems. (Volumes I and II)
Scientists and engineers have a large number of problems that are too large or complex to run on anything less than a supercomputer. There is general agreement that massively parallel computers are the supercomputers of the future. The question for engineers is to what extent they can make use of today's parallel computers. This is a nontrivial concern because program development time on parallel platforms tends to be slow and machine dependent. This thesis centers on the general approaches one can use when trying to develop code for a parallel architecture. The goal of this thesis was to determine the viability of parallel computing as a tool capable of extending the scope of problems typical of those found in chemical engineering. This was accomplished through the examination of parallelization approaches for a number of problems of importance in chemical engineering. The problem investigations utilized both MIMD and SIMD parallel architectures. The MIMD machine was a BBN TC2000 and the SIMD machine was a Thinking Machine CM-2. The key lessons resulting from the work that has been done on the two architectures is summarized by the following points: (1) The attainment of good parallel performance for all problems investigated indicates that large scale engineering applications have nontrivial degrees of parallelism. (2) Performance depends critically on the algorithm and the optimal algorithm is in turn dependent on the architecture. (3) The current state of software development tools and the fact that algorithm performance is dependent on the hardware means that the burden of achieving good parallel performance is on the programmer. (4) The MIMD paradigm was more versatile than the SIMD paradigm, and therefore, the more useful parallel architecture when the applications were considered collectively. (5) Although the primary purpose of parallel computing is speed, it is speed that will enable the next level of sophistication in engineering modelling.
- Research Organization:
- Delaware Univ., Newark, DE (United States)
- OSTI ID:
- 5709600
- Resource Relation:
- Other Information: Thesis (Ph.D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
3D gyrokinetic particle-in-cell simulation of fusion plasma microturbulence on parallel computers
Solution of open region electromagnetic scattering problems on hypercube multiprocessors
Related Subjects
99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE
FLUID FLOW
COMPUTERIZED SIMULATION
SUPERCOMPUTERS
ALGORITHMS
PARALLEL PROCESSING
CHEMICAL ENGINEERING
COMPUTER ARCHITECTURE
COMPUTER CODES
COMPUTERS
DIGITAL COMPUTERS
ENGINEERING
MATHEMATICAL LOGIC
PROGRAMMING
SIMULATION
420400* - Engineering- Heat Transfer & Fluid Flow
990200 - Mathematics & Computers