Graphical processing units and scientific applications

This special issue of the International Journal of High Performance Computing Applications collects extended versions of the best three papers presented at the International Workshop on GPUs and Scientific Applications (GPUScA 2010) held in Vienna in September 2010, in conjunction with PACT 2010 – the Annual International Conference on Parallel Architectures and Compilation Techniques. GPUs are cost-effective platforms for computationally intensive applications, providing tremendous peak performance. However, it is a major challenge to deliver the intrinsic performance of such architectures to end applications. The workshop addressed programming approaches and key techniques to leverage the computing power of GPUs. The paper Fast GPU perspective grid construction and triangle tracing for exhaustive ray tracing of highly coherent rays from Lancelot Perrotte and Guillaume Saupin, CEA-LIST (France), addresses the problem of computing, storing and sorting, at an interactive rate, all of the intersections between millions of triangles (a 3D scene) and millions of rays starting from the same point. This paper focuses on the fast GPU construction of a grid in projective space referencing the triangles of a 3D scene. It introduces a fast GPU algorithm which is used to build a grid of the rays constituting the scene, in the same projective space. This ray-based grid is computed during the initialization of the scene, which allows higher performance to be achieved, and the construction of a triangle-based grid in distinct passes for very large scenes, without having to manage memory transfers between CPUs and GPUs. This algorithm works the same way for both static and dynamic scenes, allowing interactive processing of complex and dynamic scenes to be achieved. These optimizations are used to speed up the geometrical computations used in the nuclear field to evaluate the impact of radiative sources on an operator. These geometrical computations are similar to those of traditional ray tracing, except that only highly coherent rays are thrown. The paper A framework for GPU accelerated deformable object modeling from Aria Shahingohar and Roy Eagleson, University of Western Ontario (Canada), describes a framework that uses multi-core CPUs and GPUs found in personal computers to accelerate the computations needed for a class of deformable object modeling algorithms. In recent years there has been growing interest in using deformable objects in computer applications such as animation, video games, garment CAD, and surgical simulation. Deformable object modeling is quite computationally expensive. However, since most of the related calculations can be parallelized, the authors have developed a framework that utilizes NVIDIA’s CUDA technology to accelerate a set of deformable object modeling algorithms by transferring their core computations to the GPU. Their results show that frame rates can be improved more than 20 times using GPU compared with using a multi-core CPU. In addition, they have developed a method called ‘Local Shape Matching’ which is an extension of the ‘Shape Matching’ method. Using this new method they have achieved fast and robust simulations. The paper Combining lattice Boltzmann and discrete element methods on a graphics processor from Andreas Monitzer, University of Applied Sciences Technikum Wien (Austria), deals with an original GPU-based implementation of the lattice Boltzmann method. It allows the simulation of fluids using basic arithmetic operations with a linear complexity, as is demonstrated in the paper. Additionally, the discrete element method can also be adapted to the new model. After outlining the method themselves and the integration of these two into a single simulation, this article shows a way to implement it on graphics cards using the CUDA platform.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.