Stochastic modelling of diffusion equations. Implementation and performance of the Mixage3D code on a multinode parallel machine based on the transputer T9000 Original Research Article

This paper presents the implementation of the code Mixage3D on the new parallel machine TN310. Mixage3D code has been already used to make numerical measures on the TNode-Tandem parallel machine. This implementation allowed us to make a comparative study between the parallel machine TNode-Tandem and TN310 and the Convex one. Three transputer configurations were chosen, a matrix 3 × 3 transputers, a matrix 5 × 5 and a configuration of 7 × 7 of transputers. Two cases of calculus domain were used: 21 × 21 × 21 and 105 × 105 × 105 of points. The CPU times reached on TN310 were 5 times better than those on TNode-Tandem.