### Table 1. Set of structurally feasible paths for example. PROGRAM SEGMENTS

"... In PAGE 7: ... Once the feasible subtrees are found and weights are assigned, devising a good test structure devolves to finding a balanced covering subset tree. For the program in Figure 2 there are nine feasible subtrees; these are enumerated in Table1 . The subtree is indicated simply by noting the program sequences that belong to it; the column just to the right gives the weights associated with that subtree.... ..."

### Table 3. Path feasibility determination results.

2006

"... In PAGE 7: ...3 Path Feasibility Determination and Path Pref erence Comparison To verify our path feasibility determination techniques, we obtained an initial feasibility graph using withdrawal ob- servation, chose arbitrary paths on the graph starting from the origin AS and ending in a route collector, and applied the nailed-path algorithm to determine which of the paths were feasible. Examples are in Table3 . The Prefix col- umn shows the prefix we tested and thus whether the test was performed in the IPv6 or IPv4 network.... ..."

Cited by 1

### Table 1: Experimental results of scheduling algorithm complexity given by the number of inspected vertices and CPU time (mean value over flfty randomly generated set of input data). B amp;B1 - All feasible solutions. B amp;B2 - Critical Path Bounding, B amp;B3 - Remaining Processing Time Bounding, B amp;B4 - All methods together, ILP - Integer linear programming.

"... In PAGE 7: ... There- fore, comparison of number of processed vertices allows to examine results of both methods. Total number of processed vertices of ILP in Table1 is given by value of variable total nodes declared in lpkit.h of LP SOLVE tool.... In PAGE 7: ... The scheduling problems were generated at random manner so that the number of forward edges was 3/2 of the number of nodes and there was flxed number of backward edges (5,10 or 20). When all bounding methods are combined together (B amp;B4 in Table1 ) the number of inspected states is approximately 81% of all feasible solutions for 8 nodes (with 20 backward edges) and it is 1.1% of all feasible states for 16 nodes.... ..."

### Table 2: Feasible path generation phase

in Exact

"... In PAGE 6: ...its/packet, with d set to 1. Index of the source node is 0. Distance matrix (D) is non-symmetric as shown in Table 1. The computing steps of the optimal DC-MCLP algorithm are shown in Table2 for the feasible path generation phase and Table 3 for matching and link capacity allocation phase. The optimal policy in Table 2 shows the node index with the Table 1: Distance matrix 01234 003154 110534 254021 331303 452410 46 least cost in the previous stage to connect node j.... In PAGE 6: ... The computing steps of the optimal DC-MCLP algorithm are shown in Table 2 for the feasible path generation phase and Table 3 for matching and link capacity allocation phase. The optimal policy in Table2 shows the node index with the Table 1: Distance matrix 01234 003154 110534 254021 331303 452410 46 least cost in the previous stage to connect node j. Numbers in fgin Table 3 represent Rm.... ..."

### TABLE I EXPERIMENTAL RESULTS OF SCHEDULING ALGORITHM COMPLEXITY GIVEN BY CPU TIME (MEAN VALUE OVER FIVE HUNDRED RANDOMLY GENERATED SET OF INPUT DATA) AND THE NUMBER OF INSPECTED VERTICES IN SEARCH TREE. B amp;B0 - ALL FEASIBLE SOLUTIONS, B amp;B1 - CRITICAL PATH BOUNDING, B amp;B2 - REMAINING PROCESSING TIME BOUNDING, B amp;B3 - ALL METHODS TOGETHER, ILP - INTEGER LINEAR PROGRAMMING.

### Table 5. interested in finding the dynamic slice of X at S3. Further assume that the value of A was 15 and therefore S1 is executed and S2 is not executed before arriving at S3. In other words, the use of X at S3 receives value of X defined at S1. By switching the outcome of predicate P2, we determine that a different value of X (the one defined at S2) reaches S3. As a result in our method it is assumed that an implicit dependence between P2 and S3 has been exposed. However, it seems that if P1 evaluates to true, P2 cannot evalu- ate to true. By forcing P2 to evaluate to true we may introduce a spurious implicit dependence. Our argument is that we cannot completely exclude the possi- bility of P1 or P2 being the error. In other words, even though the path is infeasible in the faulty program, it may be feasible in the correct version of the program.

"... In PAGE 9: ... One concern arises from the brute force predicate switching, which is about the feasibility of the switched path. Con- sider the code shown in Table5 (a). Let us assume that we are... In PAGE 9: ... In particular, it may miss an implicit dependence. Now let us consider another example in Table5 (b), in which our method fails to expose an implicit dependence. Let us assume that the value of A computed at statement S2 is 5, and as a result P1 evaluates to false and P2 is not executed.... ..."

### Table 3 Feasibility results for experiment set one

in Experimental Evaluation of Two-Dimensional Media Scaling Techniques for Internet Video Conferencing

1997

"... In PAGE 6: ...able 2 Summary of recent success finite state machine states and transitions...........................................23 Table3 Feasibility results for experiment set one.... In PAGE 41: ... These results conformed to naive expectations. These results are summarized in Table3 . During the peak period of the day, conferencing over this path was almost impossible, while early or... ..."

Cited by 1

### Table 3 Feasibility results for experiment set one

in CONTENTS

1997

"... In PAGE 6: ...able 2 Summary of recent success finite state machine states and transitions...........................................23 Table3 Feasibility results for experiment set one.... In PAGE 41: ... These results conformed to naive expectations. These results are summarized in Table3 . During the peak period of the day, conferencing over this path was almost impossible, while early or... ..."

### Table 4: Comparison of Lagrangian, linear programming, and best known lower bounds for the Pro- Gen/max instances [57]. All figures are averaged over the 1059 feasible instances. Figures in parenthesis denote the corresponding maxima. Unless stated differently, all computation times are in seconds.

2000

"... In PAGE 22: ... Hence, we have also applied the Lagrangian approach to the ProGen/max benchmark set by Schwindt [57], where such maximal time lags exist. The corresponding results are shown in Table4 , again in terms of deviation from the critical path lower bound (Dev. CP).... ..."

Cited by 17

### Table 3. The path pro le of the restructured program.

"... In PAGE 9: ... To show that we have (in some sense) captured the essence of the original program, we path pro led the new program. The path pro le of the new program is shown in Table3 , with paths sorted in ascending order of frequency; it is very similar to the original pro le (Table 2) with some minor di erences due to the restructuring. Summary A well known folk theorem in computer science is that any program can be transformed into a semantically equivalent program consisting of a single recursive function.... ..."