### Table 6.6 Estimated cost of mentor work hours Component Monthly Estimated Cost ($) Steve Bell (est. $30/hr, 7.5 hrs/wk) 900/month Matthew Cheselka (est. $15/hr, 7.5 hrs/wk) 450/month

1997

### Table 3.1: Cost Estimates for Light Pulser System Peter Eckstein, Martin Kocian, Klaus Schubert, Rainer Schwierz, Bernhard Spaan Dept. of Physics and Astronomy, University of Edinburgh,Edinburgh, UK Roland Bernet, Philip Clark, Steve Playfer Dept. of Physics, University of Liverpool, Liverpool, UK John Fry, Steve McMahon, Sean Colloby University of Bergen, Bergen, Norway Gerald Eigen Currently, the manpower devoted to the system includes 17 people, (6 Faculty, 7 PhD-Physicists and 4 PhD-Students), several of whom will be working full time on the light pulser:

### Table I: Information Categories in CWIS [Adapted from Wiggins, 1995, pp. 518, 519] Type of Information Some examples General information about the university Brief history, photographs of campus, campus maps, parking information, visitor information (main sights, places to eat) Academics Catalogs, calendars, enrolment procedures, academic advising, closed class list; noncredit courses; adult and night classes; enrolments Housing Residence halls, local housing options Alumni Information How to join association, event

### Table 1: Rms analysis errors at selected time instances for all experiments. results show that it is worthwhile to concentrate on estimating model error statistics before expending a great deal of computational e ort on dynamically propagating error covariances. Acknowledgements I would like to thank Steve Cohn, Dave Parrish, and especially Ricardo Todling for making their shallow-water code available to me. Most computations were performed on the CRAY C90 of the NASA Center for Computational Sciences at Goddard Space Flight Center. Many discussions with Steve Cohn and Ricardo Todling have helped me to develop the ideas presented in this paper. This research was supported by the NASA Earth Observing System Interdisciplinary Project on Data Assimilation, led by Richard Rood. I am grateful to my employers at Delft Hydraulics for their constructive and exible attitude regarding this research, and for permitting the sometimes complicated arrangements that have been required for me to be able to complete it. References

"... In PAGE 16: ... 6.1 Experimental design The implementation of the forecast model used for our experiments is described in TC94 (Section 2 and Table1 ). The evolution of the actual atmosphere is simulated by applying a random perturbation to the forecast model at each model timestep; i.... In PAGE 21: ...even under faulty assumptions about model error characteristics. See also Table1 , which summarizes the sample rms analysis errors attained in this and all other experiments. estimated Gaussian spectrum (experiment ASKF-I) estimated Gaussian spectrum (experiment ASKF-II) estimated power-law spectrum (experiment ASKF-III) Figure 5: Model error spectra, Experiments ASKF-I{ASKF-III.... In PAGE 25: ... It is easy to concoct examples either way, although it is not clear how realistic any such examples would be. Table1 summarizes the bottom-line performance obtained in each experiment. Listed are the sample rms analysis errors at 2, 4, and 6 days.... ..."

### Table 2: Percentage of connections violating timing constraints after detailed routing completion. 5 Conclusions We have presented a timing-driven router for FPGAs with segments of various lengths. The router is based on the hierarchical strategy and suited for the special properties of FPGA routing architectures. Experimental results show that our router is very e ective in reducing the number of connections violating timing constraints. Acknowledgments The authors would like to thank Steve Brown and Baharam Fallah of University of Toronto for providing us with the benchmark circuits, Nick Haruyama for helpful discussions, and Cherng-Shiuan Wang for implementing the detailed router.

"... In PAGE 10: ...nd loose cases for timing constraints. If the delay bound B(ti) was less than dmin1, then set B(ti) = dmin. Each circuit was routed by the algorithm and the percentage of source-sink pairs violating the delay bounds was computed. The results are shown in the column \Timing-driven quot; of Table2 . For the purpose of comparison, we also routed the circuits by the same routing algorithm, with the cost function for the linear assignment to minimize the wire length and the delay bound distributions/redistributions being turned o , i.... In PAGE 10: ...o minimize the wire length and the delay bound distributions/redistributions being turned o , i.e., C(3) ij in Equation (4) was set according to the cost function illustrated in Figure 8(b); this leads to a non-timing-driven routing approach [19, 25, 30]. The results are given in the column \Non-timing-driven quot; of Table2 . For all the circuits, the timing-driven routing algorithm substantially reduced the percentage of connections violating the delay bounds.... ..."

### Table 2: Experimental results. the timing constraints during the routing. It should be noted that the rates of satisfying delay bounds in the experiments depend on the generations of the delay bounds. If the delay bounds generated are too stringent, it is very likely that not all the net delay bounds can be satis ed. 6 Conclusions We present in this paper a timing-driven global routing algorithm for symmetrical array based FPGAs. The algorithm performs routing with the objective to satisfy timing constraints by utiliz- ing various segments e ciently. The preliminary experimental results show that the algorithm is promising. Acknowledgements We would like to thank Steve Brown and Baharam Fallah of University of Toronto for providing us with the benchmark circuits, and Nick Haruyama of AT amp;T Bell Laboratories for the helpful discussions.

"... In PAGE 15: ... Each circuit is routed by the algorithm and the percentage of source-sink pairs which meet the delay bounds is computed. The results are shown in the second column of Table2 . For comparison, we also route the circuits by the same routing algorithm, but with cost function for the linear assignment to minimize the wire length and the delay bound distributions/redistributions being turned o .... In PAGE 15: ... For comparison, we also route the circuits by the same routing algorithm, but with cost function for the linear assignment to minimize the wire length and the delay bound distributions/redistributions being turned o . The results are shown in the third column of Table2 . For all the circuits, the timing-driven routing algorithm achieves better results for satisfying the net delay bounds.... In PAGE 15: ... For all the circuits, the timing-driven routing algorithm achieves better results for satisfying the net delay bounds. The improvements are shown in the last column of Table2 . Up to 19% of improvement is obtained.... ..."

### Table 1. Benchmark Parameters The parallelization strategy implemented in DMMD up to now is a particle decomposition algorithm. In this method particles are distributed uniformly over the processors where they remain resident during the simulation. Each processor thus stores coordinates velocities and forces of the particles without replication. In the force loop Newton apos;s 3rd law is applied and communication of the coordinates and the forces between all processors is realized by two counter-rotating systolic loops.The performance is compared with results for the benchmark program of Steve Plimpton (PD2). However, this comparison should not be considered as decisive, since some algorithmical details are very much di erent implemented as it is in DMMD. The main di erence is the use of a replicated data algorithm, i.e. each PE stores coordinates and forces of all particles in the system, which leads to a very much smaller communication overhead when small systems are simulated on many PEs, but also limits the system size to a very much smaller number of particles

### Table 4. The ve largest primes found during the GFN search

1999

"... In PAGE 7: ... During the extension of the search, Steve Scott discovered the largest known GFN prime 48594216 + 1. Table4 is a list of the 5 largest primes that were found. 5.... ..."

Cited by 3

### Table 4: Higher-Order Economic Losses

"... In PAGE 5: ...oss at a single site is $3.6 million. In addition to environmental losses, other higher-order eco- nomic losses are possible in a large New Madrid earthquake. Table4 lists some of these higher-order losses and the pa- rameters that influence their levels (Eguchi et al., 1993; Wiggins, 1994).... ..."

### Table 2. Evaluation results for the digital camera ontology in Figure 2

"... In PAGE 10: ... We collected a set of 137 digital camera reviews in natural language text from the Web3 with over 57,000 words4. Table2 depicts the results of our extraction 3 http://www.steves-digicams.... ..."