### Table 7: Average latency to adeliver a message using the TokenFD atomic broadcast algorithm in the two-location wide area network model.

### Table 1: Results for the 4 widely used data sets Data Set g g method HA MA Rand FM Jaccard

"... In PAGE 12: ... We use the K-means and PAM procedure in Splus 6. Table1 lists the results obtained by the CLUES algorithm with the CH index and the Silhouette index. Table 1 also lists the results obtained by PAM and K-means.... In PAGE 12: ... Table 1 lists the results obtained by the CLUES algorithm with the CH index and the Silhouette index. Table1 also lists the results obtained by PAM and K-means. The second column represents the true number of clusters.... ..."

Cited by 1

### Table 1: Results for the 4 widely used data sets Data Set g g method HA MA Rand FM Jaccard

"... In PAGE 12: ... We use the K-means and PAM procedure in Splus 6. Table1 lists the results obtained by the CLUES algorithm with the CH index and the Silhouette index. Table 1 also lists the results obtained by PAM and K-means.... In PAGE 12: ... Table 1 lists the results obtained by the CLUES algorithm with the CH index and the Silhouette index. Table1 also lists the results obtained by PAM and K-means. The second column represents the true number of clusters.... ..."

### Table 1. Comparison of results between grids with and without diagonals. New results

1994

"... In PAGE 2: ... For two-dimensional n n meshes without diagonals 1-1 problems have been studied for more than twenty years. The so far fastest solutions for 1-1 problems and for h-h problems with small h 9 are summarized in Table1 . In that table we also present our new results on grids with diagonals and compare them with those for grids without diagonals.... ..."

Cited by 11

### Table 1: Processor configurations. 4-wide 8-wide

"... In PAGE 4: ... by performing cycle by cycle instruction-level simulation, including execution down any speculative path until a branch misprediction is detected. Table1 lists the architectural parameters for the 4-wide and 8-wide superscalar processors. In Wattch, the CAM cell of the evaluated designs was based on the CAM model in [1].... ..."

### Table 2: Estimation of prototype wide FOV HMD Advantage of wide FOV

"... In PAGE 4: ... We used the images captured on a moving car on this experiment. Table2 shows the result of comparison... ..."

### Table 5: Response time vs request arrival rate In both cases, the LS algorithm has the best performance in a wide range of request arrival rates. The simulation result shows that the time-slice algorithm (especially the LS algorithm) performs better than the FCFS algorithm and the SJF algorithm under a 37

1997

"... In PAGE 36: ...he maximum throughput of the tape subsystem is 4.72 requests/hour. Here, we consider a tape subsystem with a higher performance tape drive. The average response time of the FCFS, SJF, RR, and LS algorithms are tabulated in Table5 . Again, those... ..."

Cited by 10

### Table 2: Round Trip Delay Times (in ms.) Table 3: Throughput (in KB/s)

1992

Cited by 1

### Table 4: CostSendi,j, CostWaiti,j, CostOrderj and Tutoredi,j in the three-location wide area network model, using the TokenFD atomic broadcast algorithm.

"... In PAGE 16: ...TokenFD atomic broadcast Table4 present the CostSendi,j, CostWaiti,j, CostOrderj and Tutoredi,j matrices in a system with the TokenFD atomic broadcast algorithm. The average latency of atomic broadcast is then easily derived following the results presented in Section A.... ..."

### Table 1. These equations de ne the core of a portable endgame algorithm. By modifying the factorizations, code suitable for execution on a wide range of high-performance architectures can be derived. Fix a basis fcig64 i=1 of C, and de ne

1996

"... In PAGE 2: ...7 giganodes was near the maximum that the target architecture could store in RAM. Two main results are reported here: Table1 on page 18 gives equations de ning the dynamic programming solution to chess endgames. Using the techniques described in this paper, the factor- izations can be modi ed to produce e cient code for most current parallel and vector architectures.... In PAGE 3: ... Section 5 develops a generalized version of the formalism of Section 4, and describes the chess endgame algorithm in terms of this formalism. Section 6 presents equations de ning the dynamic programming solution to chess endgames ( Table1 ). Section 6.... In PAGE 14: ... The presentation is intended to illustrate the parallel code development methodology used to describe parallelization of the chess endgame algorithm, in Section 6.1 and Table1 . The exposition of the chess material, however, does not depend on any of the results here.... In PAGE 17: ...ection 6.1 gives the fundamental factorization. Section 6.2 describes the modi cation of the equations of Table1 to exploit symmetry. Section 6.... In PAGE 18: ... If v 2 B then Lv is the projection of v onto the subspace of B generated by basis elements that are not holes. Table1 de nes the piece-unmove operators. Figure 2 illustrates the compu- tation of the integrand in the expression for XR;1 in Table 1.... In PAGE 18: ... These equations are vectorizable as well [Smitley 1991]. The vectorized im- plementation of Table1 by Burton Wendro et al. has supported this claim [Wendro et al.... In PAGE 20: ... When pieces other than the k are moved, the induced motion in the hyper- board remains within the wedge. Thus, the induced functions X0 p;s: B0 7! B0 have the same form as Table1 when s 1. However, when the k is moved outside its fundamental region, the resulting position must be transformed so that the k is in its fundamental region.... In PAGE 29: ... The CM-2 six-piece code required approximately 1200 seconds for initialization and between 111 and 172 seconds to compute Ki+1 from Ki. Exact timings depend on S (for instance, as is clear from Table1 , XQ;s is slower than either XR;s or XB;s) as well as run-time settable factorization choices and load on the front end. Per-node time per endgame (time to solve the endgame divided by number of nodes in the state-space) is faster by a factor of approximately 6000 than timings of di erent endgames reported using classical techniques [van den Herik and Herschberg 1985b; Thompson 1986; Nefkens 1985; van den Herik et al.... In PAGE 30: ... Thus, although per-node timing comparisons based on radically di erently sized state-spaces are not very meaningful, the large per-node timing di erential of the current program compared to classical programs does tend to support the hypothesis that the techniques reported here lend themselves to e cient parallel implementation. The only program with per-node time of comparable speed to the author apos;s CM-200 implementation is the vectorized implementation of Table1 by Burton Wendro et al. [1993], although this implementation currently solves only a single four-piece endgame.... In PAGE 30: ... [1993], although this implementation currently solves only a single four-piece endgame. The CM-200 source code implementing Table1 is currently available from ftp://ftp.... ..."

Cited by 5