PARALLEL PROCESSING OF F.F. PROGRAMS

Abstract

As the new computer architectures emerge using parallel processing methods for more processing power, the need to explore and exploit parallelism in Finite Element Analysis, a numerical method that uses large amount of calculations arises naturally. In this dissertation, a parallel processing of a generalized sequential software PSPSAS (plane stress, plain t.ra.tn a rid axisymmetric) has been implemented. A parallelised program POWER PSPSAS is tried on a MIMD, tightly coupled parallel architecture machine TATA-ELXSI 3220. A parallel programming tool POWER FORTRAN Accelerator TM (PFA) is used in knowing data dependencies of sequential program. After structuring of a program , the PFA directives, assertions have been Inserted at proper locations, breaking data dependencies. The profile listing by pixie-sampling (Basic-block counting) are taken to know savings in CPU cycles & to judge the effectiveness of a program. Also, different scheduling options and chunk options tried in a program to achieve a better load balancing on to available number of processors present in a Silicon Graphics Workstation atNewComputationalFacilities (NCF).