### Table 3. In this case, the time budget based on ! Lw(1) satisfies the liveness and given latency/throughput constraints. Thus, we can conclude that j = 1 and the best solution may exists between ! Lw(0) and !

"... In PAGE 9: ...468125 16.698500 Table3 : Lw(0) and Lw(1) for Each Path Lw(0)[msec.] Lw(1)[msec.... ..."

### Table 1. Family Living Arrangements of Children, by Whether Married is Married or Cohabiting: Children Under Age 17 and Living With Their Mother, 1987-88 NSFH

"... In PAGE 13: ... This excludes less than 5 percent of all children and for the sake of ease of expression, we will not repeat this constraint when describing the living arrangements of all children. In Table1 , we see that about 70 percent of all children are living with both biological parents, about 10 percent are in stepfamilies, and 20 percent are in single-mother families. Cohabitation is relatively unimportant for biological families, but clearly has a major impact on the definitions of step and mother-only families.... In PAGE 13: ...S. among the variables presented in Table1 . Consequently, it is important to note that logit analyses (not presented here) indicate... ..."

### Table 2. Detailed data of live range splitting.

1998

"... In PAGE 8: ...2% in optimistic coalesc- ing with the help of live range splitting, which is ana- lyzed in detail in Table 2. Among those violating chunks in Table 1, the rst column in Table2 shows the number of chunks that have been the candidate of live range splitting and the number of nodes in those chunks5. A candidate chunk belongs to one of the following three categories after live range splitting: 5It should be noted that if the set of these candidate chunks, the set of potentially-spilled chunks, and the set of violating chunks are denoted by S, P, and V , respectively, the following relationship always holds: S P V .... In PAGE 8: ... Partially-spilled/colored, meaning that some splits are spilled while the rest are colored. Table2 shows the number of fully-spilled, fully-colored, and partially-spilled/colored chunks, and the number of nodes in those chunks6. The ratio of successfully colored nodes to all split candidate nodes is an average of 52.... In PAGE 8: ...andidate chunks is an average of 86.8%. This is the reason why optimistic coloring spills less compared to aggressive coalescing. One thing to note is that the number of actually- spilled chunks by aggressive coalescing in Table 1 is not equal to the number of split candidate chunks (which are, in fact, actual-spill candidate chunks) by opti- mistic coalescing in Table2 . The reason is that com- pared to spilling in aggressive coalescing, splitting in optimistic coalescing imposes more constraints on sub- sequent coloring in the selection phase, making more chunks below the stack be actual-spill candidates.... ..."

Cited by 21

### Table II Constraints limiting employees apos; total investment and choices between the four investment vehicles Constraints imposed on the investment choices offered to the current and former employees of France Telecom. The most severe constraint on investor behavior is presented by the rule that no more than 1/4 of annual salary can be invested into the long-lived assets. We find only 169 individuals in the data for whom the FF 9,000 constraint on the Multiplix investment binds, but estimate the 1/4 annual salary constraint to be binding for 8,375 individuals. Only 265 individuals requested the maximum amount of shares of FF 823,200.

### Table 1: Protocol family constraint summary

"... In PAGE 8: ... The constraints are not affected by whether the failure model for clients is crash or Byzantine. A summary of the constraints for the protocol family is presented in Table1 . Proofs that the constraints provide the safety and liveness guarantees indicated, as well as precise definitions of those guarantees, are presented in [15].... In PAGE 21: ... Increasing m improves the space-efficiency of protocol members. To increase m, one must also increase QC and N (constraints on m, QC and N are listed in Table1 ). Figure 6 shows the effect of increasing m (along with QC and N) on mean write response time for the synchronous repairable protocol member.... ..."

### Table 1: Protocol family constraint summary

2003

"... In PAGE 8: ... The constraints are not affected by whether the failure model for clients is crash or Byzantine. A summary of the constraints for the protocol family is presented in Table1 . Proofs that the constraints provide the safety and liveness guarantees indicated, as well as precise definitions of those guarantees, are presented in [15].... In PAGE 21: ... Increasing m improves the space-efficiency of protocol members. To increase m, one must also increase QC and N (constraints on m, QC and N are listed in Table1 ). Figure 6 shows the effect of increasing m (along with QC and N) on mean write response time for the synchronous repairable protocol member.... ..."

### Table 1.The basic strong constraint types of RTSTDs and some compound constraint. (a) Simultaneity constraint (b) Precedence constraint (c) Leads-tot constraint

2001

"... In PAGE 7: ... Constraint arcs impose extra constraints on the occurrence of events. The strong simultaneity constraint c1 from Table1 (a) con nes the two events e and f to happen simultaneously and thus maps to automaton Ac1 = true :e ^ :f ^ :PE (e ^ f) _ PE In contrast, the weak counterpart of above simultaneity constraint maps to automaton Ac1w = (e 6 = f) ^ violationi :e ^ :f ^ :violationi ^ :PE (e ^ f ^ :violationi) _ PE true which does not impose any constraints, but signals constraint violation through the distinguished event violationi. The violationi events from all automata implementing weak constraints, where the index i distinguishes the events used in the di erent au- tomata, are later collected and or-ed together to form the preemption event PE.... In PAGE 7: ... Weak variants of the other constraint types are derived analogously; we will therefore not elaborate on them in the remainder. The strong precedence constraint from Table1 (b) requires that event f may not occur before e, which is enforced by automaton Ac2 = true :e ^ :f ^ :PE e _ PE . Note that f need not happen, even if e has already been observed.... In PAGE 7: ... Note that f need not happen, even if e has already been observed. A strong timed leads-to constraint, as in Table1 (c), is implemented by automaton Ac3 = true :e ^ :f ^ :PE e ^ :f ^ :PE c := 0 f _ PE (f _ PE) ^ c lt; t :f ^ :PE , where c is a clock that is local to this automaton. Being motivated by the needs of synchronous system design, we slightly deviate from the usual interpretation of... In PAGE 8: ... Note that in case t = 1, which denotes an unbounded liveness constraint, proper behaviour is enforced by B uchi acceptance and the middle state being non-accepting, causing f to eventually happen. Finally, a strong con ict-t constraint, as in Table1 (d), asks for e and f to keep a temporal distance of at least t time units, which is implemented by Ac4 = true :e ^ :f ^ :PE PE c := 0 c := 0 :e ^ f ^ :PE e ^ :f ^ :PE c lt; t ^ :f ^ :PE c lt; t ^ :e ^ :PE c t _ PE c t _ PE Now it remains to issue a preemption event PE whenever some weak constraint has been violated. Therefore, we \or together quot; the violation signals violation1; violation2; : : : of all weak constraints actually occurring in the timing diagram by APE = PE , (violation1 _ violation2 _ ::: _ violationk) The semantics of a complete timing diagram body is the conjunction of the seman- tics of its parts, yet projected to the behaviour on the ports only, thus eliminating all event names.... ..."

Cited by 4

### Table 1.The basic strong constraint types of RTSTDs and some compound constraint. (a) Simultaneity constraint (b) Precedence constraint (c) Leads-tot constraint

2001

"... In PAGE 7: ... Constraint arcs impose extra constraints on the occurrence of events. The strong simultaneity constraint c1 from Table1 (a) con nes the two events e and f to happen simultaneously and thus maps to automaton Ac1 = true :e ^ :f ^ :PE (e ^ f) _ PE In contrast, the weak counterpart of above simultaneity constraint maps to automaton Ac1w = (e 6 = f) ^ violationi :e ^ :f ^ :violationi ^ :PE (e ^ f ^ :violationi) _ PE true which does not impose any constraints, but signals constraint violation through the distinguished event violationi. The violationi events from all automata implementing weak constraints, where the index i distinguishes the events used in the di erent au- tomata, are later collected and or-ed together to form the preemption event PE.... In PAGE 7: ... Weak variants of the other constraint types are derived analogously; we will therefore not elaborate on them in the remainder. The strong precedence constraint from Table1 (b) requires that event f may not occur before e, which is enforced by automaton Ac2 = true :e ^ :f ^ :PE e _ PE . Note that f need not happen, even if e has already been observed.... In PAGE 7: ... Note that f need not happen, even if e has already been observed. A strong timed leads-to constraint, as in Table1 (c), is implemented by automaton Ac3 = (:f ^ :PE) _ c t true :e ^ :f ^ :PE e ^ :f ^ :PE c := 0 f _ PE (f _ PE) ^ c lt; t , where c is a clock that is local to this automaton. Being motivated by the needs of synchronous system design, we slightly deviate from the usual interpretation of... In PAGE 8: ... Note that in case t = 1, which denotes an unbounded liveness constraint, proper behaviour is enforced by B uchi acceptance and the middle state being non-accepting, causing f to eventually happen. Finally, a strong con ict-t constraint, as in Table1 (d), asks for e and f to keep a temporal distance of at least t time units, which is implemented by Ac4 = true :e ^ :f ^ :PE PE c := 0 c := 0 :e ^ f ^ :PE e ^ :f ^ :PE c lt; t ^ :f ^ :PE c lt; t ^ :e ^ :PE c t _ PE c t _ PE Now it remains to issue a preemption event PE whenever some weak constraint has been violated. Therefore, we \or together quot; the violation signals violation1; violation2; : : : of all weak constraints actually occurring in the timing diagram by APE = PE , (violation1 _ violation2 _ ::: _ violationk) The semantics of a complete timing diagram body is the conjunction of the seman- tics of its parts, yet projected to the behaviour on the ports only, thus eliminating all event names.... ..."

Cited by 4

### Table 2. The number of hypotheses generated per frame is 30 and the time taken per iteration is 49.5 ms (i.e. 20.2 fps). Figure 3 illustrates the recognition process on a typical test- ing clip. The proposed algorithm can also be implemented to work on a live video feed from camera at the mentioned frame rate. If more constraints (such as average duration) are added, the accuracy can be boosted to around 90%.

"... In PAGE 4: ... Table2 : Classification result on segmenting and recognising continuous gestures. Notation used (see [8]): Accuracy = (N-I-D-S)/N, N is the total number of gestures appeared in the testing set (30 video clips), I is the number of insertions, D is the number of deletions, S is the number of substitutions and C is the number of correctly recognised gestures.... ..."

Cited by 1

### Table 1: Minimizing integral versus total MaxLive. For comparison purposes, we also investigated a model minimizing integral MaxLive. This integer programming model di ers from the model of Figure 4 in that the equa- tions labeled Lifetimes and MaxLive are replaced by an equation de ning the number of FIFO bu ers needed to 35

1995

"... In PAGE 6: ... Algorithm 1 is directly applicable to this integer programming model as well. The complexity of these models, in number of variables and fundamental constraints, is shown in Table1 for a loop with N operations, an initiation interval of II, ereg regis- ter dependence edges, esched scheduling dependence edges, and a machine with m kinds of machine resources. The ma- jor complexity increase, when minimizing MaxLive instead of integral MaxLive, is that N II variables are needed to keep track of the register requirements of each operation in each row of the MRT instead of N variables for each op- eration over the entire MRT.... ..."

Cited by 31