"... In PAGE 24: ... Table 9 report respectively the odds ratios of the interacted variables for non LCC (D_LCC = 0 and hence, ORk = Exp[ k]) and for LCC (D_LCC = 1 and hence, ORk = Exp[ k + k]). It is straightforward to show that if Xk is a non-interacted variable, its odds ratio will be: ORk = Exp[ k] Table8 reports the odds ratios for non-interacted variables. Finally, the odds ratio of the dummy variable D_LCC can be written as: ORk = Exp[ 0 + Pm k=1 kXk + Pn k=1 kXk] Exp[Pn k=1 kXk] which yields: ORD_LCC = Exp quot; 0 + m X k=1 kXk # The interpretation of the odds ratio for the dummy variable D_LCC is somehow trickier, as it also comprises the m interacted variables.... ..."

"... In PAGE 3: ... We also benchmarked some standard non- interactive applications which are closer to re- alistic (interactive) applications than the micro- benchmarks. The results of these tests have been summarized in Table2 . When taking a closer look at these numbers one notices that the slowdown for most benchmarks is lower than a factor of 2.... ..."

"... In PAGE 4: ... Note that the protein pairs without overlapping domains in AP matrix are not included in the test data in the measurement. Table1 shows the sensitivities and specificities of each test group depending on the ratios of interacting and non-interacting set of protein pairs. The data in each test group is divided further into two subgroups; one group is the test set of protein pairs which has a matching PIP value in PIP distributions and the other group is the test set of protein pairs without matching PIP value in PIP distribution.... In PAGE 4: ... The data in each test group is divided further into two subgroups; one group is the test set of protein pairs which has a matching PIP value in PIP distributions and the other group is the test set of protein pairs without matching PIP value in PIP distribution. As shown in Table1 , very high sensitivities and specificities were achieved for the test groups with matching PIP values, whereas moderate sensitivities and specificities were achieved for the test groups without matching PIP values. In the test, it was revealed that protein pairs with common domains in AP matrix are amenable to have matching PIP values in the PIP distributions.... ..."

"... In PAGE 10: ...Table 3a: Variations of Input Data Streams for Testing Utilities (these were used for the non-interactive utility programs) ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul Input Streams for Interactive Utilities ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul # Character Types NULL character Input stream size (no. of strings) ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul 1 printable+nonprintable bu 10 2 printable+nonprintable bu 100 3 printable+nonprintable bu 1000 ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul 4 printable bu 10 5 printable bu 100 6 printable bu 1000 ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul 7 printable+nonprintable 10 8 printable+nonprintable 100 9 printable+nonprintable 1000 ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul 10 printable 10 11 printable 100 12 printable 1000 ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul ulululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br Table3 b: Variations of Input Data Streams for Testing Utilities (these were used for the interactive utility programs)... ..."

"... In PAGE 9: ...Table 3a: Variations of Input Data Streams for Testing Utilities (these were used for the non-interactive utility programs) ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul Input Streams for Interactive Utilities ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul # Character Types NULL character Input stream size (no. of strings) ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul 1 printable+nonprintable YES 10 2 printable+nonprintable YES 100 3 printable+nonprintable YES 1000 ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul 4 printable YES 10 5 printable YES 100 6 printable YES 1000 ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul 7 printable+nonprintable 10 8 printable+nonprintable 100 9 printable+nonprintable 1000 ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul 10 printable 10 11 printable 100 12 printable 1000 ululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul ulululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululululul br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br Table3 b: Variations of Input Data Streams for Testing Utilities (these were used for the interactive utility programs)... ..."

"... In PAGE 4: ...1 Non-interactive design transfer model We first performed TAM optimization using the design flow in Figure 2. In Table1 , we compare the testing times (in clock cycles) and CPU times (in seconds) of the proposed hierarchical TAM op- timization method with those of the corresponding flat methods in [5, 13]. Though hierarchical TAM optimization was performed based on both methods in [5, 13] for each SOC, results are pre- sented here for only one of the two methods for each SOC due to insufficient space.... In PAGE 4: ... The percentage change in test- ing time A1CC using hierarchical TAM optimization is calculated as CC CWCXCTD6 A0CC CUD0CPD8 CC CUD0CPD8 A2 BDBCBC. In Table1 (a), we present results for p22810. The hierarchi- cal TAM optimization flow was based on the ILP and enumera- tion method of [5].... In PAGE 4: ... While the testing times obtained are higher than those obtained by (unrealistically) assuming that the SOC hierarchy can be flattened, the results ob- tained here are more realistic for hierarchical TAM design. Note from Table1 (a) that the testing time for the hierarchical method levels out at 366260 cycles at CF BP BGBC. A total top-level TAM width of 40 is therefore an effective choice for this SOC.... In PAGE 4: ... A total top-level TAM width of 40 is therefore an effective choice for this SOC. In Table1 (b), we present results for p34392. Hierarchical TAM optimization was performed based on the heuristic method of [13].... In PAGE 4: ... In the hierarchical optimization flow, the SOC is parti- tioned per mega-core and the ILP model runs significantly faster. In Table1 (c), we present results for p93791. Hierarchical TAM optimization was performed using the heuristic method [13].... In PAGE 4: ... A TAM width of 16 was supplied to each mega-core prior to system- level TAM optimization. In Table1 (d), we present results for a586710. Hierarchical TAM optimization was performed using ILP and enumeration [5].... In PAGE 5: ... In Table 2, we compare the testing times and CPU times for hierarchical TAM optimization with those for the corresponding flat methods in [5, 13]. We do not list the testing times and CPU times of the flat methods, since these are already listed in Table1 . We also compare the testing times for the in- teractive design transfer model with the testing times for the non- interactive model.... ..."