### Table 1: Visibility as a function of the type of edges versus location of observer point For a vertical re ex edge, the Zone 0 is the only one where an observer cannot see the vertical edge. Although the re ex edge is visible from the other three zones, for two of these zones the re ex edge produces occlusions. If the observer is located in Zone 3, he/she can see the vertical edge and the edge does not block the observer view to other parts of the environment.

### Tableau (4) shows that this candidate set has been fully ordered by the time that NOVCDOB gets its hands on it. NOVCDOB is therefore unable to favour deletion of the marked consonant b. In other words, if b is in the inventory, NOVCDOB is ranked so low that it cannot cause the cluster in /katba/ to be simplified by deleting the more marked consonant b. So this ranking conforms in its predictions with FCD.

### Table 3: The views required for evaluating other changes

2001

"... In PAGE 18: ...3 Other Changes The third category is a reservoir of change scenarios that cause changes that cannot be classified in one of the other categories. For Sagitta 2000 these scenarios are listed in Table3 . Most of these scenarios cause changes in one of the external components used by Sagitta 2000, and these changes may influence Sagitta 2000 as well.... ..."

Cited by 9

### Table 3 demonstrates that our reduced mathematical formulations do speed up the process of multi-domain clock skew schedule. Especially, in benchmark circuits S3384, S6669, S9234 and S13207, the mathematical solver cannot solve the original formula- tions within twelve hours, but the mathematical solver can solve our reduced formula- tions within twelve hours. However, on the other hand, in benchmark circuits S15850 and S38584, although a lot of redundant binary variables and a lot of redundant con- straints can be pruned, the corresponding problem complexity still cannot be solved within twelve hours.

"... In PAGE 17: ...1 The Minimization of Max_Overlap Here, our objective function is to minimize the value of max_overlap under the con- straint of given clocking domains. Table3 demonstrates the advantage of exploiting the ASAP and ALAP schedule of each register to prune the redundancies. The column Origi- nal Formulation describes the original mathematical formulations proposed by [14].... In PAGE 17: ... We use the notation t/o to denote that the problem complexity cannot be solved within twelve hours. Table3 . The advantage of pruning redundancies for the minimization of max_overlap.... In PAGE 18: ... The column CPU time gives the CPU time (in seconds) spent in both the proce- dure ZONE_BASED_SCHEDULING and the mathematical solver. On the other hand, from Table3 , we can find the results of whole-circuit scheduling, which solves the whole circuit as a single zone. Clearly, in each benchmark circuit, the zone-based scheduling is much faster than the whole-circuit scheduling.... ..."

### Table 1: Some evolutions and which systems support them. Note that none of the other systems directly address instance adaptation. aThis evolution cannot be de ned directly, but can be implemented by replacing the attribute in the new version with generic methods for reading and writing.[3] bSee p. 20 for de nition. cSupport for telescope evolutions is not possible in our database model, as currently presented. (cf., p. 20 for discussion.)

"... In PAGE 10: ... Our proposed scheme, however, will provide a mechanism for managing class evolutions that include dependent attributes. (See Table1 (p.19) for a comparison of the evolution capabilities of various systems.... ..."

### Table 1: Some evolutions and which systems support them. Note that none of the other systems directly address instance adaptation. aThis evolution cannot be de ned directly, but can be implemented by replacing the attribute in the new version with generic methods for reading and writing.[3] bSee p. 20 for de nition. cSupport for telescope evolutions is not possible in our database model, as currently presented. (cf., p. 20 for discussion.)

"... In PAGE 10: ... Our proposed scheme, however, will provide a mechanism for managing class evolutions that include dependent attributes. (See Table1 (p.19) for a comparison of the evolution capabilities of various systems.... ..."

### Table2. Threshold values (and percentage change from estimated value) for the change in predicted model outcome, from partridge exclusion to pheasant and partridge co-existence, for those parameters not directly estimated in this study to which the model outcome was considered to be highly sensitive. In all cases, values were determined with all other model equation parameters kept fixed at their estimated levels. Percentage change cannot be calculated for either apheasant or apartridge since the estimated value of apheasant is 0, and there is no threshold value for apartridge

"... In PAGE 9: ...analyses were also conducted on apheasant, dpheasant, and b, a and b for both host spe- cies, since these were the other quantities not directly estimated in this study to which the model outcome was considered to be highly sensitive. However, the model outcome of partridge exclusion was also relatively robust to changes in these values (see Table2 for a summary of all sensitivity analyses conducted). SPATIAL SEPARATION When the model was run for each host species alone with the parasite, the qualitative outcome for the pheasant was unchanged (remaining in co-existence with the parasite), while that for the partridge chan- ged from exclusion of the host to exclusion of the parasite.... ..."

### Table 1 shows the size of our cell libraries constructed for ACT1 and ACT2 architectures. The new cell matching method consumes 8.2% of the whole CPU time on the av- erage, while Boolean method consumes 91.9% of the total CPU time on the average as shown in the FIgure 3. Currently, we cannot use our method for other FPGA cell architectures whose basic cell has large number of in- put variables, since it takes into account only MUX-based FPGA basic cell whose variable number is not so large as ACT series.

"... In PAGE 2: ... Currently, we cannot use our method for other FPGA cell architectures whose basic cell has large number of in- put variables, since it takes into account only MUX-based FPGA basic cell whose variable number is not so large as ACT series. Table1 . Our library size for FPGA cell match- ing ACT1 ACT2 input number of memory size number of memory size number entries (# of bytes) entries (# of bytes) 8 20160 645120 10080 322560 7 57960 927360 32760 524160 6 59040 472320 39660 317280 5 24960 99840 21240 84960 4 3732 7464 3828 7656 3 175 175 173 173 2 10 10 10 10 1 2 2 2 2 Total 166039 2152291 107753 1256801 0 20 40 60 80 100 5 10 15 20 25 time percentage for cell matching (%) benchmark numbers results by Boolean cell matching results by our cell matching Figure 2.... ..."

### Table 8: P10(x) = xn ? 1 Again the accuracy of the Jenkins/Traub program drastically decreases for relatively small de- grees (n = 50). It cannot be used for degrees n gt; 60 : : :70. On the other hand our method yields the best results with nearly computer accuracy up to the degree n = 10000. The eigenvalue method is slightly worse regarding the accuracy but could be used only for degrees up to about 500. This is 16

1993

"... In PAGE 16: ...2.5 High Order, Well Conditioned Polynomials As a last example we consider high degree polynomials up to n = 10000 for the very well conditioned polynomials P10(x) = xn ? 1 (23) P11(x) = xn + 1: (24) The results are summarized in Table8 and 9 and graphically depicted in Figs. 4 and 5.... ..."

Cited by 2

### Table 4: Contributions to the estimated theoretical errors for s determination. s has been determined from the C-parameter using log-R and R matching schemes and are found to di er by 0.0012. The other matching schemes have not been tried out and so the matching scheme uncertainty cannot be determined for s as determined from C-parameter.

"... In PAGE 5: ... The theoretical error is obtained from an estimate of the hadronization uncertainty and of the errors coming from the uncalculated higher orders in the QCD predictions. The rst part of Table4 shows the variation in the tted value of s due to di erent hadronisation corrections. The hadronisation correction using Jetset has been taken as a reference point.... ..."