"... In PAGE 4: ... The initial layer file which serves as an initial solution to our SA implementation contains the layer assignment for the wires in the original layout. Table1 shows the effectiveness of the SA algorithm. The second column gives the number of wire segments in a circuit.... ..."

"... In PAGE 7: ...Table2 y) is erroneously eliminated by traditional-DEE during the pruning stage, confirming that traditional-DEE is not provably accurate with backbone flexibility. 5.... In PAGE 7: ... The significance of BD lies in its ability to generate lower- energy conformations than traditional-DEE. As Table2 shows for the GrsA redesign, even after performing backbone minimization, the energy of the fixed-backbone rigid-rotamer GMEC is significantly higher (by more than 4 kcal/mol) than the energy of the flexible-backbone rigid-rotamer GMEC. Thus, for models that incorporate backbone flexibility, the BD criterion should be used instead of traditional-DEE when accuracy is preferred over speed.... In PAGE 7: ... From that initial conformation, flexible-backbone minimiza- tion and side-chain dihedral minimization resulted in virtually equal energies. In the GC121 redesign, differences of almost 10 kcal/mol were observed between the energies resulting from side-chain and backbone minimization ( Table2 ). Experiments on a set of surface residues of GC121 [a description of this system can be found in (Georgiev et al.... In PAGE 8: ... Hence, in order to further improve the accuracy of the model, both minimization types could be incorporated simultaneously. A somewhat surprising result ( Table2 ) is that even very small changes in the backbone can result in a significant improvement in the conformational energy. Since the ligand in the GrsA system is also allowed to rotate, translate, and flex, however, it was not certain to what extent the energy improvement was due to protein backbone movements.... ..."

"... In PAGE 6: ... Energy-per-division after voltage scaling. 5 Comparison between Radix-4, 8 and 16 Table5 summarizes the energy dissipation and area for the implementations of radix-4, 8 and 16. It was not possi- ble to implement dual voltage with our cell library so that entry d-v is an estimate of a possible implementation.... In PAGE 7: ... In fact, voltage scaling is more effective for simpler datapaths, where the ratio delay MSBs delay LSBs gt; 2. The data from Table 3 and Table5 are plotted in Fig- ure 11. The target is to obtain an implementation that is as close as possible to the origin: minimum energy and la- tency.... ..."

"... In PAGE 6: ... Energy-per-division after voltage scaling. 5 Comparison between Radix-4, 8 and 16 Table5 summarizes the energy dissipation and area for the implementations of radix-4, 8 and 16. It was not possi- ble to implement dual voltage with our cell library so that entry d-v is an estimate of a possible implementation.... In PAGE 7: ... In fact, voltage scaling is more effective for simpler datapaths, where the ratio delay MSBs delay LSBs a1 a21 . The data from Table 3 and Table5 are plotted in Fig- ure 11. The target is to obtain an implementation that is as close as possible to the origin: minimum energy and la- tency.... ..."

"... In PAGE 5: ... set freq specifies the value of the next CPU or cache frequencies. Rules learned from ST in Table 2 are shown in Table3 . Note that the number of rules is very small because of the simplified system setting chosen in our hypothetical example.... ..."