Abstract—The goal of this study is to adapt the multiscale fluid solver EULAG [Prusa et al., Computers & Fluids, vol 37, 2008] to the future GPU-based high-performance computing platforms. The EULAG model has a proven record of successful applications in a range of environmental fluid dynamics, and excellent efficiency and scalability on conventional supercomputer architectures. Currently, the model is being implemented as a new dynamical core of COSMO (Consortium for Small-scale Modeling) weather prediction framework. The EULAG code combines features of a stencil and point wise computations. Its communication scheme consists of both halo exchange subroutines and global reduction functions. Within the project, two main modules of EULAG, namely the multidimensional positive definite advection transport algorithm, MPDATA, and the variational generalized conjugate residual, GCR, elliptic pressure solver are analyzed and optimized. Relevant techniques have been chosen and applied to accelerate code execution on modern GPU architectures: stencil decomposition, block decomposition (with weighting analysis between computation and communication), reduction of intercache communication by partitioning of cores into independent teams, cache reusing and vectorization.
Testing and validation of the new GPU implementation have been carried out based on modeling decaying turbulence of a homogeneous incompressible fluid in a triply-periodic cube. Simulations performed using the standard version of EULAG and its new GPU implementation give similar solutions. Preliminary results of the parallel performance of the new implementation show a promising increase in terms of computational efficiency.

Index Terms—Anelastic model, EULAG, hybrid architectures, parallel computing.

B. Rosa and D. K. Wójcik are with the Institute of Meteorology and Water Management - National Research Institute, Podleśna 61 Street, 01-673 Warsaw, Poland (e-mail: bogdan.rosa@imgw.pl).
P. K. Smolarkiewicz is with the European Centre for Medium-Range Weather Forecasts, Reading, RG2 9AX, UK.
M. Ciżnicki is with the Poznan Supercomputing and Networking Center, Noskowskiego 10 Street, 61-704 Poznań, Poland.
K. A. Rojek and R. Wyrzykowski are with the Czestochowa University of Technology, Dabrowskiego 69 Street, 42-201 Czestochowa, Poland.

Cite: Bogdan Rosa, Miłosz Ciżnicki, Krzysztof A. Rojek, Damian K. Wójcik, Piotr K. Smolarkiewicz, and Roman Wyrzykowski, "Porting Multiscale Fluid Model EULAG to Modern Heterogeneous Architectures," International Journal of Applied Physics and Mathematics vol. 4, no. 3, pp. 188-195, 2014.