### Table 2: Statistics concerning the GPU-based visibility algorithm.

2005

"... In PAGE 10: ...3). Table2 summarizes the statistics of the GPU visibility algorithm. The rst column shows the name and number of polygons used in the test (i.... ..."

Cited by 5

### Table 2: Statistics concerning the GPU-based visibility algorithm.

2005

"... In PAGE 10: ...3). Table2 summarizes the statistics of the GPU visibility algorithm. The rst column shows the occluder name and speci es the number of polygons used to model it (i.... ..."

Cited by 5

### Table 1. Performance of GPU-based 2D vector eld visualization in frames per second on an ATI Radeon 9800 XT GPU.

2007

"... In PAGE 16: ... Figure 6 (d) is based on a smaller sampling distance, which results in short streaks. Table1 shows performance measurements for di erent GPU-based 2D vec-... ..."

Cited by 2

### Table 3: Framerates per second for the GPU-based ray caster employ- ing either the space-optimized or the performance-optimized frag- ment program.

2004

"... In PAGE 7: ... 6.2 Rendering Table3 presents the timings that our GPU-based ray caster achieved for the images presented in this paper acquired for a 400 400 viewport. Sphere, Ell (volume), and Ell (iso) correspond to the left, middle, and right image of Figure 1 respectively.... In PAGE 8: ... Theoretically, if the early depth-test prohibits further computation on nished rays, the same or even lower performance should be ex- pected for tile-based rendering. The timings in Table3 , therefore, were acquired with the viewport split into 16 16 tiles, which we experienced as optimal with respect to the overhead introduced for the tile handling. Texture bind operations turned out not to be the bottleneck for our implementation, even if they are not avoided by simultaneously binding the current render target as texture (see Section 5.... ..."

Cited by 12

### Table 4: Performance measures in frames per sec- ond for GPU-based construction and rendering of vector plots.

in GPU-PIV

### Table 1: Performance of GPU-based 2D dye advection in frames per second on ATI X800 XT.

"... In PAGE 8: ... In all example images, dye placement was based on user interaction the user virtually painted newly released dye into the ow by using mouse interaction. Performance numbers are documented in Table1 . Rendering speed depends linearly on the number of texels, as shown in the comparison of different viewport sizes.... ..."

### Table 2. Relighting benchmark for both envi- ronment lighting and local lighting, itemized by the illumination sampling speed and the GPU-based relighting speed.

"... In PAGE 7: ... Relighting. The relighting performance is summarized in Table2 . For all test scenes, we maintain interactive relight- ing speed under arbitrary illumination, viewpoint, and ma- terial changes.... ..."

### Table 1: Performance of all three GPU-based 2D advection tech- niques in frames per second measured on an NVIDIA GeForce 6800 GT graphics board.

2005

Cited by 1

### Table 1 Combinatorial optimization problems and their geometric equivalents

"... In PAGE 10: ... Similarly, if we color the graph with a minimum number of colors, then this is equivalent with dividing the collection of lines into a minimum number of subcollections such that each subcollection contains no parallel pairs of lines. Table1 gives an overview of problems in graph theory and their geometric equivalent. Since in this paper we focus on parallel line grouping, we propose two combi- natorial algorithms that can be used to partition a graph of parallel pairs into subgraphs which are or which resemble cliques.... ..."

### Table 1. An example of the permutation problem

2003

"... In PAGE 1: ... In this paper we focus speci cally on the case where the feature values have been permuted in a random manner. Table1 shows a simple example of this type of permutation problem. We would like to be able to learn the joint probability density of the original data on the left given only the permuted data and knowl- edge of the type of permutations that may have been Table 1.... In PAGE 1: ...pplied to the data (e.g., cyclic shifts). Two questions naturally arise: (a) how hard is this type of learning problem in general? and (b) what kinds of algorithms can we use to solve this problem in practice? In considering the rst problem, our intuition tells us that the \more di erent quot; the features in the original (unpermuted) table are then the \easier quot; the unscram- bling problem may be. For example, in Table1 , the distributions of each individual feature in the table on the left appear quite di erent from each other, so that one hopes that given enough data one could eventu- ally recover a model for the original data given only permuted data. In Section 2 we make this notion of learnability precise by introducing the notion of a Bayes-optimal permutation error rate.... In PAGE 4: ... E? C is proportional to the overlap of the individual fea- ture densities p (~xi) in the space S. For example, for the data on the left in Table1 we would expect the overlap of the 4 densities, as re ected by E? C, to be quite small. Furthermore, we would expect intuitively that the permutation error rate E? P should also be low in this case, and more generally that it should be re- lated to E? C in some manner.... ..."

Cited by 3