Solving Kinetic Equations on GPU's
Abstract
The main aim of the present paper is to translate efficiently a semi-regular method of solution of the full non-linear Boltzmann equation into a parallel code to be executed on a GPU. The efficiency of the algorithm is assessed by solving the classical two-dimensional driven cavity flow. It is shown that it is possible to cut down the computing time of the sequential codes of two order of magnitudes. This paper is organized as follows. Sections 2 and 3 are devoted to a concise description of the mathematical model and the adopted numerical method. In Section 4 the key aspects of the GPU hardware architecture and CUDATM programming model are briefly described and implementation details are provided. Sections 5 is devoted to the description of the test problem and the discussion of the results. Concluding remarks are presented in Section 6.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2011
- Accession Number
- ADA587866
Entities
People
- Aldo Frezzotti
- Gian Pietro Ghiroldi
- Livio Gibelli
Organizations
- Polytechnic University of Milan