"... In PAGE 20: ... The address bits for the ROM block are timely incremented by a 4-bit counter as shown in Figure 8. Table3 shows the algorithm dataflow. In the first cycle, the field element a, whose multiplicative inverse is required, is written into the BRAM.... ..."

"... In PAGE 3: ... 2. This architecture highlights the dynamic nature of DW by incorporating five flows as shown in Table1 (Lee et al., 2001).... ..."

"... In PAGE 10: ... The programs used for this comparison include a recursive Fibonacci program, Matrix Multiply6, and Livermore Kernel 5. Table1 shows the number of the cycles for various data sizes. Table 1: Comparative analysis of Scheduled Dataflow with MIPS.... In PAGE 12: ... We use the Matrix Multiply program for this purpose. For the experiments in Table1 , we created 10 Scheduled Dataflow threads for the Matrix Multiply (innermost loop) application. For both SDF and MIPS, each thread computed 5 innermost loop iteration (i.... ..."

"... In PAGE 19: ... We selected four structural faults from our classification (Figure 4) to be seeded: Omission of Clause, Omission of Conditional, Trans- position Between Regions, and Transposition Within Region. The second column in Table2 , (labeled Total Faults), gives the total number of faults seeded. The next four columns give the numbers of faults of each type that were seeded.... In PAGE 20: ... Therefore, for this study, we manually generated all tests for both Mothra and Combat. The last two columns in Table2 give the total number of faults detected by dataflow testing and mutation testing.... In PAGE 22: ... 5.2 Analysis of Results Table2 shows that dataflow testing detected seventy-two of the seventy-four faults seeded, whereas mutation testing detected all faults seeded. The two faults that were not detected by dataflow testing were, somewhat surprisingly, trivial faults.... In PAGE 23: ... The two subject programs for which the fault-detecting test sets were poorly dataflow adequate were mid and tritype. An inspection of the fault types seeded in the third through sixth columns of Table2 reveals that the faults seeded for mid and tritype are not well distributed: there are no Transposition Between Regions or Transposition Within Region type of faults. Thus, there may be a relation between types of faults seeded and dataflow adequacy of test sets that expose these faults.... In PAGE 23: ... Thus, there may be a relation between types of faults seeded and dataflow adequacy of test sets that expose these faults. The dataflow adequacy scores for newton show that the fault-detecting test set was only about 80% all-uses adequate, although Table2 shows that newton had a good distribution of seeded faults. This anomaly might be due to the specific test sets that were selected for newton.... ..."

"... In PAGE 19: ... We selected four structural faults from our classi cation (Figure 4) to be seeded: Omission of Clause, Omission of Conditional, Trans- position Between Regions, and Transposition Within Region. The second column in Table2 , (labeled Total Faults), gives the total number of faults seeded. The next four columns give the numbers of faults of each type that were seeded.... In PAGE 20: ... Therefore, for this study, we manually generated all tests for both Mothra and Combat. The last two columns in Table2 give the total number of faults detected by dataflow testing and mutation testing.... In PAGE 22: ... 5.2 Analysis of Results Table2 shows that dataflow testing detected seventy-two of the seventy-four faults seeded, whereas mutation testing detected all faults seeded. The two faults that were not detected by dataflow testing were, somewhat surprisingly, trivial faults.... In PAGE 23: ... The two subject programs for which the fault-detecting test sets were poorly dataflow adequate were mid and tritype. An inspection of the fault types seeded in the third through sixth columns of Table2 reveals that the faults seeded for mid and tritype are not well distributed: there are no Transposition Between Regions or Transposition Within Region type of faults. Thus, there may be a relation between types of faults seeded and dataflow adequacy of test sets that expose these faults.... In PAGE 23: ... Thus, there may be a relation between types of faults seeded and dataflow adequacy of test sets that expose these faults. The dataflow adequacy scores for newton show that the fault-detecting test set was only about 80% all-uses adequate, although Table2 shows that newton had a good distribution of seeded faults. This anomaly might be due to the speci c test sets that were selected for newton.... ..."

"... In PAGE 5: ... 5 Examples In this section we present a number of examples illustrating the performance of the retiming heuristics proposed in the paper. We started by considering specific datapath bindings for the three char- acteristic loops shown in Table1 . Then, the algorithm was applied to solve Problem 2, i.... ..."

"... In PAGE 5: ... 5 Examples In this section we present a number of examples illustrating the performance of the retiming heuristics proposed in the paper. We started by considering specific datapath bindings for the three char- acteristic loops shown in Table1 . Then, the algorithm was applied to solve Problem 2, i.... ..."

"... In PAGE 12: ...Table1... ..."