Modelling of the parallel resolution of a numerical problem on a locally distributed computing system
The authors investigate the behaviour of a parallel computing system, consisting of a number of independent processors, which execute in parallel the numerical solution of a partial differential equation. As a measure of system they consider the speed gain obtained per iteration step on the parallel processors with respect to the one-processor execution of the equivalent task. For the performance evaluation of the parallel system they use a probabilistic model of the computation scheme, based upon several main assumptions concerning computation and communication times. They provide the exact solution of the model in the case of two processors and then develop a heuristic method allowing the approximate solution of the model in the case of three or more parallel processors. 8 references.
- Research Organization:
- Univ. of Paris, Orsay, France
- OSTI ID:
- 5134943
- Report Number(s):
- CONF-8208209-
- Journal Information:
- Perform. Eval. Rev.; (United States), Vol. 4; Conference: ACM sigmetrics conference on measurement and modeling of computer systems, Seattle, WA, USA, 30 Aug 1982
- Country of Publication:
- United States
- Language:
- English
Similar Records
Mapping unstructured grid computations to massively parallel computers. Ph. D. Thesis - Rensselaer Polytechnic Inst. , Feb. 1992
Parallel sparse matrix computations in iterative solvers on distributed memory machines
Related Subjects
COMPUTERS
PARALLEL PROCESSING
PERFORMANCE
PARTIAL DIFFERENTIAL EQUATIONS
COMPUTER CALCULATIONS
DISTRIBUTED DATA PROCESSING
NUMERICAL SOLUTION
DATA PROCESSING
DIFFERENTIAL EQUATIONS
EQUATIONS
PROCESSING
PROGRAMMING
990200* - Mathematics & Computers