"... In PAGE 10: ... As in the unoptimized case, the fraction of reads can be predicted from the port utilizations. If A, B, and C represent the utilizations (fractions of cycles that the port is busy) of their respective ports and x is the fraction of reads, then the following relationship should hold: x = A + B A + B + C Table2 gives the predicted and computed x for the benchmarks.... ..."

"... In PAGE 11: ...ttf. The empirical results for the entropy-based adaptive coding methods are shown in Table3 . The... ..."

"... In PAGE 7: ...ttf. The empirical results for the entropy-based adaptive coding methods are shown in Table2 . The results show that if enhanced by using the linear and nonlinear SLWE schemes, a gain of nearly 9% and 10% (respectively) is obtained with respect to the traditional entropy-based encoding scheme.... ..."

"... In PAGE 4: ... Although compilation time could not be accurately measured due to limitations in the imple- mentation of the research framework, compilation seemed to be faster when partial inlining was used, perhaps due to reduced code growth. Code growth The first column of Table1 shows the percentage change in code growth for the benchmarks comparing re- gion formation using demand-driven full inlining, and the use of partial inlining in addition to full inlining, calculated using Equation 1. With the exception of clinpack, the inclu- sion of partial inlining controlled code growth better than using only full inlining; code growth was reduced by from 1% to 6%.... In PAGE 4: ... With clinpack, how- ever, an additional instance of inlining was performed that had not been performed in the full inlining version before no other procedures were eligible for inlining. Runtime performance The second column of Table1 shows the percentage change in runtime performance for the benchmarks over the procedure-basedstrategy, calculated using Equation 1. Per- formance increase was demonstrated for all benchmarks, ranging from 0.... ..."

"... In PAGE 9: ... For this problem ten databases were used from the Irvine repository and the Star/Galaxy database. Table1 gives the percent of correctly separated points as well as CPU times using an average of ten SLA runs on the smooth misclassi cation minimization problem (20). These quantities are compared with those of a parametric minimization method (PMM) applied to an LPEC associated with the misclassi cation minimization [18, 1].... In PAGE 9: ... These quantities are compared with those of a parametric minimization method (PMM) applied to an LPEC associated with the misclassi cation minimization [18, 1]. Table1 shows that the much simpler SLA algorithm obtained a separation that was almost as good as the parametric method for solving the LPEC at considerably less computing cost. Each problem was solved using no more than a maximum average of 7 LPs over ten runs.... ..."

"... In PAGE 8: ... In fact, we tried to implement classes and methods similar to the ones typically used by programmers in terms of the names, parameters, and form of use (see example in section 4). Table1 presents the most important methods provided. 4 Practical Example of Implementation The example presented in this section has resulted from the study meant to demonstrate and evaluate the usefulness of the transactions programming approach proposed.... ..."