The main object of this paper is to demonstrate the feasibility of coupling parallel codes in the framework of large scale scientific computing. The second aim is to provide a numerical tool to solve 3D frontal polymerization in liquid problems where the multiple physical scales require intensive computation on MIMD architecture. We extend our 2D modelization of such phenomena by coupling a 3D reaction diffusion system well known in solid combustion with the 3D Navier-Stokes equations written in Boussinesq approximation. We develop two separate parallel codes for each physical model, each code running on its own topological network of processors with its own data structures. The non-blocking communications, manage the interaction terms of the two physical models on each code, are performed through a Portable Inter Program Communication Library (PIPCL). This communication library computes array-based communication schedules creating a one-to-one implicit mapping between each data representation. The data distribution of each code is hidden to the others, allowing an easy parallel program inter-operability on heterogeneous computers.
ASJC Scopus subject areas
- Theoretical Computer Science
- Hardware and Architecture
- Computer Networks and Communications
- Computer Graphics and Computer-Aided Design
- Artificial Intelligence