"... In PAGE 11: ...1, Ratio cost of a core node of type 1 divided by the fiber unitary cost: f1/F = 20, Ratio cost of a core node of type 2 divided by the fiber unitary cost: f2/F = 50, Ratio cost of a core node of type 3 divided by the fiber unitary cost: f3/F = 100, Edge node capacity in the 10 edge nodes network, traffic matrices A and B: Cj = 1000 Gbit/s, Edge node capacity in the 34 edge nodes network, traffic matrix A: Cj = 2000 Gbit/s, Edge node capacity in the 34 edge nodes network, traffic matrix B: Cj = 2800 Gbit/s. The results are presented in Table1 where we portray the total cost of the design for each case and traffic matrix as well as the percentage of the core, fiber and delay costs that compose the solution. The solution time and the optimality gap are given at the end.... ..."

"... In PAGE 7: ...Final Revision IBM Systems Journal 7 while nodes are milestones. Different activities and milestones are described in Table1 . The final product delivered approximately 64 thousand lines of C++ code, the port required over 8 person- years of effort, and took 16 months to complete.... In PAGE 10: ... The granularity of the task schedules was typically from 1 to 4 weeks and involved from 1 to 3 software professionals. Examples of tasks are design-level reviews, design of a small component, and unit test of a component (see Table1 ). From the raw data we can calculate a number of task parameters, such as the planned task duration, the actual task duration, early task starts, late task starts, and so on.... ..."

"... In PAGE 6: ... Each CTG is different in the number of nodes, the number of edges, the number of allocated PEs, and the number of branching nodes. The second column of Table1 summarizes the characteristics of the 12 CTGs used for the experiments. For the task assignment step, we used the GA-based task assignment algorithm [7] to assign each task in a CTG to a PE.... In PAGE 6: ... For the condition-aware task scheduling algorithm, we experimented two versions, one without the profile information and the other with the profile information. Table1 shows the experi- mental results for twelve CTGs. The third, fourth and fifth columns show the normalized energy consumption by task schedules gener- ated from three task scheduling algorithms, the condition-unaware algorithm (denoted by CU), the condition-aware algorithm without the profile information (denoted by CAno profile) and the condition- aware algorithm with the profile information (denoted by CAprofile).... In PAGE 6: ... With the profile information, the condition-aware algo- rithm further reduced the energy consumption by 50% on average. Table1 also shows that the energy efficiency of the condition- aware algorithms varies significantly depending on the character- istic of a CTG. For example, the energy consumption in ctg5 is reduced by 73%.... ..."

"... In PAGE 15: ... The result of this renaming is shown in Table 3 where the variable msg in line number 46 is renamed twice, because of the two instantiations of process medium in line number 39. The classification of each variable occurrence in the flowgraph of the ABP specification as a def, c-use, or p-use is given in Table4 . Note that the p-nodes 21 and 64, representing recursive instantiations of processes quot;safemesser quot; and quot;safeacker quot; respectively (line numbers 71 and 104), do not contain any uses and defs since the values of the variables in these instantiations have not been changed (see case e), Section 2.... ..."

"... In PAGE 2: ... The same holds with the dataset we processed in our experiments. As can be seen in the statistics of section 3 ( Table1 ) we have only 0.27 edges per node in our graph.... In PAGE 4: ...The first three columns of Table1 contain information on the nodes of our graph (unique permalinks with content, all permalinks, news URLs). The other three columns provide information on the edges of the graph (links to permalinks with content, links to all permalinks, links to news URLs).... ..."

"... In PAGE 5: ... There is no port assignment in both [13] and [5]. The benchmarks shown in Table2 are from [25]. The examples used in this study are data-dominated behavioral descriptions with predominantly arithmetic operations that are commonly encountered in signal and image processing applications.... In PAGE 5: ... The initial scheduling and functional unit binding solutions are generated by the LOPASS high-level synthesis system that targets low power designs [5]. The minimum register number shown in Table2 is obtained by a left-edge [26] register binding algorithm, which is optimal for a comparability graph generated from non-hierarchical designs, i.... ..."