Super and Parallel Computers and Their Impact on Civil Engineering

Kamat, Manohar P.; ,

Super and Parallel Computers and Their Impact on Civil Engineering

by Manohar P. Kamat, (M.ASCE), Assoc. Prof.; Virginia Polytechnic Inst. and State Univ., Dept. of Engrg. Sci. & Mathematics, Blacksburg, VA 24061,

American Society of Civil Engineers, New York, NY
978-0-87262-551-8 (ISBN-13) | 0-87262-551-6 (ISBN-10), 1986, Soft Cover, Pg. 53

See all papers/chapter

Conference information: A Session at Structures Congress | New Orleans, Louisiana, United States | September 15-18, 1986

Out of Print: Not available at ASCE Bookstore.

Document Type: Book - Proceedings

Abstract:

This publication contains four papers on the use of super and parallel computers in the practice of civil engineering. The first paper shows how large nonlinear systems of equations, such as the equilibrium equations in structural mechanics, are particularly suitable to solution on distributed computing systems. The second paper examines advances in computer hardware that have resulted in phenomenal performance ratings of supercomputers and the newer class of machines referred to as mini-supercomputers, which offer high performance at lower prices. The third paper reviews the classification and general architecture of supercomputers including microprocessors, vector and array processors. The paper then discusses their capabilities, and current and projected limitations for solving engineering problems. The paper concludes that even though the gains achieved on today's supercomputers are substantial, they are often suboptimal. The last paper describes the Custom Architectured Parallel Processing System (CAPPS) for the solution of computationally intensive problems. A unique feature of the CAPPS system is that all data transfers among the computational units take place concurrently with the execution of instructions. This eliminates the problem of performance degradation due to data transfers, which plagues many parallel computers. Advances in super and parallel computations have and will continue to be rapid. Accordingly, an assessment of the state-of-the-art, such as the present one, will have to be made frequently and at regular intervals.

Return to search