### Table 9: Summary of the decomposition of a real-valued function.

"... In PAGE 5: ... This produces the representative columns of Table 8. The composition of this decomposition is shown as h in Table9 . Note that the function represented by the decomposition (i.... ..."

### Table 6: Real function eld over Fp key exchange | NUCOMP/composition.

"... In PAGE 12: ... The cor- responding key exchange protocol in the principal class [9] is very similar to that in real quadratic number elds; each communication partner has to perform two binary exponentiations of principal ideals and maintain the correspond- ing distances. We have also implemented this protocol, and for each nite eld and genus pair in Table6 and Table 7, we have performed a number of key exchanges using random eld discriminants of the given genus and random ex- ponents bounded by qg: As in the imaginary function eld case, we expect each communication partner to perform 2 log2 qg NUDUPL or ideal squaring oper- ations and half as many NUCOMP or ideal multiplication operations per key exchange. We performed 4000 key exchanges using both NUCOMP and compo- sition for g 5; 2000 for 5 lt; g 10; 1000 for 10 lt; g 15; and 500 for g gt; 15: The ratio of the total time for all key exchanges using NUCOMP over the total time using ideal multiplication is given for each genus/ eld pair.... ..."

### Table 7: Real function eld over GF(2n) key exchange | NUCOMP/composition.

"... In PAGE 12: ... The cor- responding key exchange protocol in the principal class [9] is very similar to that in real quadratic number elds; each communication partner has to perform two binary exponentiations of principal ideals and maintain the correspond- ing distances. We have also implemented this protocol, and for each nite eld and genus pair in Table 6 and Table7 , we have performed a number of key exchanges using random eld discriminants of the given genus and random ex- ponents bounded by qg: As in the imaginary function eld case, we expect each communication partner to perform 2 log2 qg NUDUPL or ideal squaring oper- ations and half as many NUCOMP or ideal multiplication operations per key exchange. We performed 4000 key exchanges using both NUCOMP and compo- sition for g 5; 2000 for 5 lt; g 10; 1000 for 10 lt; g 15; and 500 for g gt; 15: The ratio of the total time for all key exchanges using NUCOMP over the total time using ideal multiplication is given for each genus/ eld pair.... ..."

### Table 3: Real estate usage by functional unit type.

"... In PAGE 5: ... Real estate:Transistor level analysis of published work provided the approximations for each functional unit type. This model is presented in Table3 . #28Since the FPMul unit is used iteratively for division, the FPDiv unit does not consume any die space and is not mentioned in the table.... ..."

### Table 10: Random samples of a function with real-valued inputs.

"... In PAGE 8: ...n exhaustive table. We are assuming, instead, that we are given a nite partial table3. We will again use an example to identify challenges posed by real-valued variables and explore some approaches to those challenges. Table10 is a set of samples from a function of the form f : [0; 1]2 ! [0; 1], de ned as: f(x; y) = ( y if x 0:5 1 ? y if x gt; 0:5 If we chose x as a column variable and y a row variable then the resulting partition matrix is shown in Table 11. This table demonstrates a signi cant change from the familiar partition matrices for functions with binary inputs | each column and each row only has one entry.... ..."

### Table 3: Special functions dedicated to the texturing process. Establishment of the texture properties according to the real-valued function.

2001

### Table 2: Real Calculation and Accepted Domain for Mathematical Functions Routine Name Real Calculation Domain AINT [-minreal, maxreal]

"... In PAGE 22: ...ttp://www.eecs.lehigh.edu/~mschulte/compiler/report2.html Table 1 shows relative error bounds for each interval version of the mathematical functions. Table2 shows the domain over which the relative error bounds are valid. If a function takes an interval as an input in this domain, then it calculates the result interval within the relative errror bounds.... ..."