Speed-up of CFD codes using analytical FE calculations
Today combining supercomputer performances with algorithmic optimizations large linear systems are solved in a very short time. Consequently reduction of FE calculations time becomes crucial. This paper presents how these calculations can be efficiently speed-up by replacing them by analytical ones and by fully vectorizing their assembling. Moreover the simplicity of these calculations enables on efficient parallelization on CRAY YMP.
Unable to display preview. Download preview PDF.
- 1.Chabard, Metivet B., Pot G., Thomas B., An efficient finite element method for the computation of 3D turbulent incompressible flowsGoogle Scholar
- 2.Gregoire J.P., Nitrosso B., Pot G., Conjugate gradient performances enhanced in the C.F.D. code N3S by using a better vectorization of matrix vector product, Proceedings of IMACs 91,.Dublin, July 1991.Google Scholar
- 3.Moulin V., Caruso A., Daubert O., Pot G., Thomas B. Improvement of FE. algorithms implemented in CFD code N3S for turbulent or dilatable flows. Proceedings of 5th Int. Symp. on Refined Flow Modelling and Turbulence Measurements, Paris, September 1993.Google Scholar