Christian Trefftz

Computation of Voronoi diagrams using a graphics processing unit

A parallel algorithm to compute a discrete approximation to the Voronoi diagram is presented. The algorithm, which executes in single instruction multiple data (SIMD) mode, was implemented on a high-end graphics processing unit (GPU) using NVIDIApsilas compute unified device architecture (CUDA) development environment. The performance of the resulting code is investigated and presented, and a mathematical model is developed that predicts the performance of the algorithm.

Parallelizing AES on multicores and GPUs

The AES block cipher cryptographic algorithm is widely used and it is resource intensive. An existing sequential open source implementation of the algorithm was parallelized on multi-core microprocessors and GPUs. Performance results are presented.

Memory-efficient implementation of a graphics processor-based cluster detection algorithm for large spatial databases

Numerous approaches have been proposed for detecting clusters, groups of data in spatial databases. Of these, the algorithm known as Density Based Spatial Clustering of Applications with Noise (DBSCAN) is a recent approach which has proven efficient for larger databases. Graphical Processing Units (GPUs), used originally to aid in the processing of high intensity graphics, have been found to be highly effective as general purpose parallel computing platforms. In this project, a GPU-based DBSCAN program has been implemented: the enhancement in this program allows for better memory scalability for use with very large databases. Algorithm performance, as compared to the original sequential program and to an initial GPU implementation, is investigated and analyzed.

Parallel algorithms to find the Voronoi diagram and the order-k Voronoi diagram

The Voronoi diagram is a classical problem in the area of computational geometry. Given a plane and a set of n seed points, the objective is to divide the plane into tiles (or subsets of points), such that the set of points in a tile are closer to a particular seed point than to any other seed. This problem has become increasingly important in the area of geographic information systems (GIS) as Voronoi diagrams are used in GIS for computing zonal statistics. Current or timely data for GIS is acquired via remotely-sensed devices providing data in raster(regular or gridded) format. This paper describes parallel algorithms to find the Voronoi Diagram, the furthest site Voronoi diagram and the order-k Voronoi diagram. Prototype implementations in Parallaxes are outlined.

Using XML in a compiler course

In this paper we describe how XML can be introduced into a compiler construction course. We make the case that XML and compilers have much in common, and that introducing XML into a compiler course makes sense. We then goes on to demonstrate how XML was used in two recent compiler courses. Finally, we discuss the tradeoffs of using an XML-based project rather than a traditional programming-language project.

Parallel implementations of FGMRES for solving large, sparse non-symmetric linear systems

Abstract The Flexible Generalized Minimal Residual method (FGMRES) is an attractive iterative solver for non-symmetric systems of linear equations. This paper presents several methods for parallelizing FGMRES for a variety of archi- tectures including multi-core CPU, Graphics Processing Units (GPU), and multi-GPU systems. The parallel imple- mentations utilize OpenMP and CUDA kernels, and are organized according to thread scope. The linear systems employed in this study correspond to the discrete analogues of realistic three-dimensional convection-diffusion problems, and range in size to nearly 107 linear equations. All of the parallel implementations, particularly the novel hybrid approach, show a significant speedup over the sequential version. Theoretical insight and perfor- mance data is provided to allow informed decisions as to the most effective parallelization method for a given architecture.

Enumerating Communities for a Deeper Understanding of Community Finding

Often new insights and advancements are made by a detailed study of the problem and the solution space. The area of community finding has had many algorithms proposed recently, but to our knowledge there have not been any detailed studies of the solution space. In this paper, we present two algorithms for enumerating and unranking the possible valid community assignments for a network. To demonstrate the value of our algorithms, we also present some interesting insights gained by examining the solution space of some small networks.

Community finding within the community set space

Community finding algorithms strive to find communities that have a higher connectivity within the communities than between them. Recently a framework called the community set space was introduced which provided a way to measure the quality of community sets. We present a new community finding algorithm, CHI, designed to minimize the violations defined by this framework. It will be shown that the CHI algorithm has similarities to kmeans. It is flexible and fast and can also be tuned to find certain types of communities. It is optimized for the community set framework and results so that it performs better than other algorithms within that framework.

Parallelizing a heuristic for the Maximum Clique Problem on GPUs and clusters of workstations

A heuristic for the Maximum Clique Problem was parallelized on a Graphical Processing Unit and a cluster of workstations. The heuristic is based on a formulation based on neural networks. Performance results are reported.

An Investigation on Testing of Parallelized Code with OpenMP

Testing is a crucial element of software development. As OpenMP becomes more widely used, a relevant question for developers is: How will programs, that have been parallelized with OpenMP, be tested for correctness? OpenMP programs are concurrent programs and as such all the risks of concurrent programming are present as well. This paper presents some observations about testing parallelized loops using OpenMP.