Implementation of Computer Graphics and Image-Processing Applications on a GPU Cluster – running

The objective of the project is to efficiently implement computer graphics and image-processing algorithms that can hardly run on a single Graphics Processing Unit (GPU) due to their high computational and storage costs. Although the conventional GPU has been originally developed for traditional computer graphics applications, nowadays, it is used more and more as a general purpose coprocessor in various applications such as signal processing (Fourier transform, continuous reconstruction of discrete signals), image processing (resampling, compression, segmentation, registration, tomography reconstruction), physical simulation (flow simulation, modeling multiple scattering in inhomogeneous materials), or solving linear equation systems and differential equations. A GPU implementation is worthwhile to study in those applications that require SIMD (Single Instruction Multiple Data) processing. If dynamic data structures are required, the CPU-based computational capacity of the supercomputer can also be exploited, but sharing different subtasks between the CPU and GPU can be reasonable as well.

Project owner:
Tóth Márton József (Irányítástechnika és Informatika Tanszék)
Web address:
Irányítástechnika és Informatika Tanszék (VIK-IIT)