### Table 3. Summary of 2-way direction join algorithms sweep line range search data

2001

"... In PAGE 9: ... If the evaluation needs only 1-dimensional range searches, B-trees are su cient for the temporary data structure. Column 4 of Table3 shows the data structures used in the evaluation of a direction join. Theorem 3.... ..."

Cited by 3

### Table 1: Comparison between constant thinning and geometric thinning algorithms.

"... In PAGE 7: ... The best result is in bold text, and the worst result is in italic text. In the Table1 , the simulations of CT and GT are compared. In these experiments, we had 1024 processors and = 1:2.... In PAGE 7: ....05 and 1.1. The value of tmax is 1.5. The Figure 1 describes partly the results shown in Table1 . A logarithmic scale (for h) is used in the gure.... ..."

### Table 2: Performance of parallel thinning

1997

"... In PAGE 17: ... For the image of size 128 128, the number of iterations required by the thinning algorithm is 15. The thinning time and other performance data of the algorithm for various numbers of the processors are tabularized in Table2 . The \e ciency quot; measure is to re ect the e ectiveness of using more processors to solve the same problem.... In PAGE 19: ... There- fore, we consider the saving due to remapping of only a few percents in the overall thinning time in both cases satisfactory. In comparison with the inter-processor communication cost which accounts also for a few percents (see Table2... ..."

Cited by 5

### Table 2: Performance of parallel thinning

1997

"... In PAGE 20: ... For the image of size 128 128, the number of iterations required Figure 7: The image pattern and the thinning result by the thinning algorithm is 15. The thinning time and other performance data of the algorithm for various numbers of the processors are tabularized in Table2 . The \e ciency quot; Table 2: Performance of parallel thinning... In PAGE 23: ... Therefore, we consider the saving due to remapping of only a few percents in the overall thinning time in both cases satisfactory. In comparison with the inter-processor communi- cation cost which accounts also for a few percents (see Table2 ), this saving is signi cant. 6 APPLICATION 3: PARALLEL UNSTRUCTURED GRID PARTITIONING In computational uid dynamics, physical domains of interest are tessellated using struc- tured, block structured or unstructured grids.... ..."

Cited by 5

### Table 3.1 correct value in the sweeping with ordering i = 1 : I; j = 1 : J since they all depend on their left and down neighbors. After four sweepings are nished a single communication among them will get the correct solution in the whole domain. In general if the charac- teristics are straight lines, sweepings with di erent orderings are almost independent of each other and the parallel sweeping algorithm should be as e cient as the the original fast sweeping algorithm. However if the characteristics are curved then di erent orderings implemented sequentially may propagate information faster on a curved characteristics than di erent orderings implemented in parallel.

### Table 4 Comparison of ECA, Strobe, SWEEP, and PVM

"... In PAGE 63: ...63 A comparison of the features of popular algorithms, such as Strobe, C-Strobe, SWEEP, Nested-SWEEP, Posse [19], and PVM is shown in Table4 inspired by [1]. As we can see PVM inherits all the advantages from its ancestors, e.... ..."

### Table 3 Sweeps taken for matrices of various sizes

in Abstract

2001

"... In PAGE 7: ... This is an evidence for the importance of sorting in block Jacobi-like methods. Some experimental results for matrices of different sizes are given in Table3 . We can see from this table that it takes a few more sweeps to converge for a matrix with complex eigenvalues than for a matrix with all real ARTICLE IN PRESS Table 1 Sweeps and lower block triangular norms for 40 C2 40 matrices obtained by using the block algorithm without sorting Sweep Lower block triangular norm Matrix 1 Matrix 2 Matrix 3 0 1.... ..."

### Table 1: Thin-client computing platforms.

2003

"... In PAGE 2: ...are 1.2 10.d. In this paper, we also refer to Citrix Metaframe by Citrix ICA (Independent Computing Architecture), which is the re- mote display protocol used by the Citrix platform. As summarized in Table1 , these platforms span a range of differences in the encod- ing of display primitives, policies for updating the client display, algorithms for compressing screen updates, caching mechanisms, supported display color depth, and transport protocol used. To evaluate their performance, we designed an experimental network testbed and various experiments to exercise each of the thin-client platforms on single-user web browsing workloads.... In PAGE 9: ... 4.1 TCP Retransmission Frequency As shown in Table1 , all of the thin clients were based on TCP except for SunRay, which uses UDP. In addition, the fat clients use HTTP which is built on top of TCP.... ..."

Cited by 2

### Table 2. Thinning to 22 points with f2.

2000

"... In PAGE 10: ... In order to make one concrete example, we let c = 0:2, n = 22. Table2 re ects the numerical results, and Figures 9,10 show the two subsets Y = X22 selected by the method AT1 (left) and NAT (right) along with the graphs of their corresponding least squares... In PAGE 10: ... Thinning to 22 points with f2. Table2 indicates that small subsets of X output by an adaptive thin- ning algorithm can serve as good sets of centres for approximating the data (x; f(x)) : x 2 X, by a sum of translates of ` to the chosen centres, and that these approximations are superior to the piecewise linear interpolants on these subsets. Acknowledgments.... ..."

Cited by 1

### Table 2: Another comparison between constant thinning and geometric thinning algo- rithms. Here, = 1:1.

"... In PAGE 7: ... A logarithmic scale (for h) is used in the gure. In Table2 , we have another CT versus GT comparison. The di erence to Table 1 is... In PAGE 8: ... (In the O(h) time proof in [10], a very safe value h0 gt; 100 ln p was used.) The results of Table2 are partly illustrated in Figure 2. In Table 3, there are results of another kind of comparison between the CT and GT algorithms.... ..."