### Table 6: N-gram statistics computed on the 5 million word text and the information stored in Markov source models.

1992

"... In PAGE 3: ... Phonotactic constraints Phone, diphone and triphone statistics, computed on the 5 million word original text, are used to provide phonotac- tic constraints. Table6 gives the information stored in the Markov sources (1-gram to 3-gram) estimated from the oc- currence frequencies on the original text in bits/phone[4]. For now only the 1-gram and 2-gram constraints have been... ..."

Cited by 17

### Table 23: Con gurations of all the environment states for Figure 5.

"... In PAGE 31: ... In order to understand a model in its entirety, it is useful to account for information computed and stored at each state. Table23 shows the con guration of each environment state depicted in Figure 5. Technically, this information could be provided by labeling states with appropriate atomic propositions.... ..."

### Table 4: N-gram statistics computed on the 5 million word text and the information stored in Markov source models. Condition Corr. Subs. Del. Ins. Acc.

1992

"... In PAGE 3: ... Phonotactic constraints Phone, diphone and triphone statistics, computed on the 5 million word original text, are used to provide phonotac- tic constraints. Table4 gives the information stored in the Markov sources (1-gram to 3-gram) estimated from the oc- currence frequencies on the original text in bits/phone[3]. For now only the 1-gram and 2-gram constraints have been incorporated in the model.... ..."

Cited by 2

### Table 3: Traditional method data

1994

"... In PAGE 16: ...1 Data for the Traditional Approach The iterative, bit-vector method for computing liveness information requires four sets be computed and and stored at each block in the program: the in, out, use, and def sets. In Table3 the number of bits required (in thousands) per set is shown for each program. For the entire Perfect suite, this works out to about 300K 32-bit words of storage required for all four sets.... ..."

Cited by 3

### Table 3: Traditional method data

1994

"... In PAGE 16: ...1 Data for the Traditional Approach The iterative, bit-vector method for computing liveness information requires four sets be computed and and stored at each block in the program: the in, out, use, and def sets. In Table3 the number of bits required (in thousands) per set is shown for each program. For the entire Perfect suite, this works out to about 300K 32-bit words of storage required for all four sets.... ..."

Cited by 3

### Table 1: Suspension forming activities in the example programs 3.2 Graph size The graphs in Figure 3 illustrate the number of live heap cells throughout the execution of each of the benchmark programs: The Y-axis shows the number of live heap cells, plotted against the number of cells claimed (X-axis). This X-axis plot is e ectively time (see earlier discussion in section 2), as each cell claim marks a signi cant event for the storage manager. Initially, the live graph size is zero, then the graphs grow as intermediate results are demanded, computed and stored, becoming part of the nal answer. When execution ter- minates, the live graph is the program result. Depending on the nature of the intermediate computations and the size and structure of the answer, the graph can take many shapes.

1991

"... In PAGE 7: ... To avoid this we chose to implement all partial applications with binary AP cells. The frequency information that we collect allows us to compute the relative costs of the schemes (see Table1 ). We found that, on average, VAP cells are used in 95% of all suspensions.... ..."

Cited by 5

### Table 1: Comparison of several data structures. \M quot; indicates entities that are explicitly stored, \C quot; means that the information may be calculated in O(1), and \- quot; marks entities that require computations of complexity O(1).

1998

"... In PAGE 8: ... Of course, such attributes may also be added to the directed-edge data structure at the cost of additional memory. Finally, Table1 shows a comparison of the mesh data structures considered in this article. Please note that the individual triangles and shared vertex representation do not provide any immediate neighborhood information which is crucial for many applications.... ..."

Cited by 23

### Table 1: Comparison of several data structures. \M quot; indicates entities that are explicitly stored, \C quot; means that the information may be calculated in O(1), and \- quot; marks entities that require computations of complexity O(1).

1998

"... In PAGE 8: ...ttributes, e.g. colors, surface normals, or texture coordinates [CB97]. Of course, such attributes may also be added to the directed-edge data structure at the cost of additional memory. Finally, Table1 shows a comparison of the mesh data structures considered in this article. Please note that the individual triangles and shared vertex representation do not provide any immediate neighborhood information which is crucial for many applications.... ..."

Cited by 23

### TABLE II COMPUTATIONAL OVERHEADS OF THE MAJOR STEPS OF PRIVACY-PRESERVING WIRELESS COMMUNICATION. *THE OVERHEAD OF THE SERVER DOES NOT INCLUDE THE OVERHEAD OF THE VERIFICATION OF THE CREDIT CARD INFORMATION OR THE OVERHEAD OF SEARCHING AND STORING DATA IN A DATABASE

2004

Cited by 1

### Table 7a shows the percentage of PA-RISC instructions executed in each class. The arithmetic means of the percentage of instructions in each class is calulated to get a quot;typical quot; distribution of instructions by class. For the floating point benchmarks, we see that loads are typically the most common, followed by floating point computation, integer computation, stores, branches, and NOPs, in that order. For the integer benchmarks, we see that loads are typically the most common, followed by integer computation, branches, stores, and NOPs, in that order.

"... In PAGE 8: ...5 8.0 Table7 a: PA-RISC Instruction Classes as Percent of Total Instruction Counts Table 7a shows the percentage of PA-RISC instructions executed in each class. The arithmetic means of the percentage of instructions in each class is calulated to get a quot;typical quot; distribution of instructions by class.... In PAGE 8: ... Table7 b: Geometric Means of Instruction Count Ratios by Instruction Class Table 7b shows the geometic means of the MIPS/PA- RISC and SPARC/PA-RISC instruction count ratios in each instruction class. This is summarized from Tables 1b, 2a, 2b, 3, 4a, 5a, and 6.... In PAGE 8: ...Table 7b: Geometric Means of Instruction Count Ratios by Instruction Class Table7 b shows the geometic means of the MIPS/PA- RISC and SPARC/PA-RISC instruction count ratios in each instruction class. This is summarized from Tables 1b, 2a, 2b, 3, 4a, 5a, and 6.... ..."

Cited by 1