GPGPU

GPGPU, or the general-purpose computing on graphics processing units, is a quickly developing, emerging technology that aims at utilizing the computing power of modern graphics cards in numerical simulations of typical engineering problems.

In our team, we focus on two issues:

  1. How to use the GPUs effectively in simulations of computational fluid dynamics?
  2. How to implement efficiently some fundamental operations, like sparse matrix-dense vector multiplication on the massively parallel architecture of GPUs?

We tackle these problems in collaboration with industry. In particular, we collaborated with Vratis in the development of their SpeedIT Flow™ and SpeedIT Flow CL™, the first commercial CFD solvers working entirely on the GPUs using CUDA or OpenCL technology. Moreover, our PhD student, Jakub Poła, took part in the development of AMD’s library of basic sparse linear algebra functions, clSPARSE.

Representative publications

  • J. L. Greathouse, K. Knox, J. Poła, K. Varaganti, M. Daga,
    clSPARSE: A Vendor-Optimized Open-Source Sparse BLAS Library
    IWOCL ’16 Proceedings of the 4th International Workshop on OpenCL, Vienna, Austria (2016)
    doi: 10.1145/2909437.2909442
    (conference presentation)
  • Z. Koza, M. Matyka, S. Szkoda, L. Miroslaw,
    Compressed Multirow Storage Format for Sparse Matrices on Graphics Processing Units,
    SIAM J. Sci. Comput. 36-2, C219-C239 (2014)
  • Z. Koza, M. Matyka, L. Miroslaw, J. Poła,
    Sparse Matrix-Vector Product,
    [in:] Numerical Computations with GPUs, Springer-Verlag, Berlin (2014)
  • T. Tomczak, K. Zadarnowska, Z. Koza, M. Matyka, L. Miroslaw,
    Acceleration of iterative Navier-Stokes solvers on graphics processing units,
    Int. J. Comp. Fluid Dyn. 27:4-5, 201-209 (2013)
  • S. Szkoda, Z. Koza, M. Tykierko,
    Accelerating cellular automata simulations using AVX and CUDA,
    arxiv.org/abs/1208.2428 (2012)
  • Z. Malecha, Ł. Miroslaw, T. Tomczak, Z. Koza, M. Matyka, W. Tarnawski, D. Szczerba
    GPU-based simulation of 3D blood flow in abdominal aorta using OpenFOAM
    Arch. Mech. 63, 137 – 161 (2011)

Courses taught

Staff engaged