Figure 1: With the eight seeds (highlighted and shown in different colors) in a 64×64 grid on the leftmost picture, the progress of each round of the jump flooding algorithm is shown in the other six pictures, with the rightmost being the computed Voronoi diagram. AbstractThis paper studies jump flooding as an algorithmic paradigm in the general purpose computation with GPU. As an example application of jump flooding, the paper discusses a constant time algorithm on GPU to compute an approximation to the Voronoi diagram of a given set of seeds in a 2D grid. The errors due to the differences between the approximation and the actual Voronoi diagram are hardly noticeable to the naked eye in all our experiments. The same approach can also compute in constant time an approximation to the distance transform of a set of seeds in a 2D grid. In practice, such constant time algorithm is useful to many interactive applications involving, for example, rendering and image processing. Besides the experimental evidences, this paper also confirms quantitatively the effectiveness of jump flooding by analyzing the occurrences of errors. The analysis is a showcase of insights to the jump flooding paradigm, and may be of independent interests to other applications of jump flooding.

We propose a Parallel Banding Algorithm (PBA) on the GPU to compute the exact Euclidean Distance Transform (EDT) for a binary image in 2D and higher dimensions. Partitioning the image into small bands to process and then merging them concurrently, PBA computes the exact EDT with optimal linear total work, high level of parallelism and a good memory access pattern. This work is the first attempt to exploit the enormous power of the GPU in computing the exact EDT, while prior works are only on approximation. Compared to these other algorithms in our experiments, our exact algorithm is still a few times faster in 2D and 3D for most input sizes. We illustrate the use of our algorithm in applications such as computing the Euclidean skeleton using the integer medial axis transform, performing morphological operations of 3D volumetric data, and constructing 2D weighted centroidal Voronoi diagrams.

No abstract

Data trees provide a standard abstraction of XML documents with data values: they are trees whose nodes, in addition to the usual labels, can carry labels from an infinite alphabet (data). Therefore, one is interested in decidable formalisms for reasoning about data trees. While some are knownsuch as the two-variable logic -they tend to be of very high complexity, and most decidability proofs are highly nontrivial. We are therefore interested in reasonable complexity formalisms as well as better techniques for proving decidability.Here we show that many decidable formalisms for data trees are subsumed -fully or partially -by the power of tree automata together with set constraints and linear constraints on cardinalities of various sets of data values. All these constraints can be translated into instances of integer linear programming, giving us an NP bound on the complexity of the reasoning tasks. We prove that this bound, as well as the key encoding technique, remain very robust, and allow the addition of features such as counting of paths and patterns, and even a concise encoding of constraints, without increasing the complexity. We also relate our results to several reasoning tasks over XML documents, such as satisfiability of schemas and data dependencies and satisfiability of the twovariable logic.

No abstract

No abstract

This paper presents a novel approach to compute, for a given point set S in R 2 , its Delaunay triangulation T (S). Though prior work mentions the possibility of using the graphics processing unit (GPU) to compute Delaunay triangulations, no known implementation and performance have been reported. Our work uncovers various challenges in the use of GPU for such a purpose. In practice, our approach exploits the GPU to assist in the computation of a triangulation T ′ of S that is a good approximation to T (S). From that, the approach employs the CPU to transform T ′ to T (S). As a major part of the total work is done by the GPU with parallel computing capability, it is a fast and practical approach, particularly for a large number of points (millions with the current state-of-the-art GPU). For such cases, our current implementation can run up to 53% faster on a Core2 Duo machine when compared to Triangle, the well-known fastest Delaunay triangulation implementation.

No abstract