"... In PAGE 6: ... 3.2 Results Table1 shows three values for the computed savings on a series of parameterised models. The first column lists the model parameter (size), the second is the sum of all fire call times on saturated nodes, the third eliminates the unfeasible calls due to the condition tsat + f lt; tbf , while the fourth shows the realistic savings for the first-call-first strategy.... ..."

"... In PAGE 11: ...e. BCET , WCETNP , and WCETP , together with the experimental collected results, denoted as Experimental values, are listed for each of the examples in Table5 . To give a quantita- tive idea about the accuracy of the theoretical approach in which processor pipelining architecture is taken into account in comparison with the case where no pipelining is considered in the calcula- tion, the differences with reference to real system measurements are presented.... In PAGE 12: ... In other words, similar to the instruction prefetching, the efficiency of the pipelining analysis is dependent upon the instructions involved as well as the program context. ulululululululululululululululululululululululululululululululululululululululululululululululululululululululul Codes NIFPIi (bytes) ri /wi (bytes) MINIETi (clks) ulululululululululululululululululululululululululululululululululululululululululululululululululululululululul ulululululululululululululululululululululululululululululululululululululululululululululululululululululululul MOV dx, 003ch 3 0/0 4 ADD cl, 01h 3 0/0 4 OR dl, cl 2 0/0 3 MOV dl, 0ch 2 0/0 3 NEG cx 2 0/0 3 XOR dl, 01h 3 0/0 4 CMP dx, cx 2 0/0 3 DEC dx 2 0/0 3 INC cx 2 0/0 3 SUB dl, cl 2 0/0 3 XCHG dx, cx 2 0/0 4 TEST dl, cl 2 0/0 3 PUSH dx 1 0/0 14 POP dx 1 0/0 14 ulululululululululululululululululululululululululululululululululululululululululululululululululululululululul br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br br Table 1: Instruction Timing Information for Example 1 Obviously, according to Table5 , the pipelining approach provides a remarkable improve- ment in estimation accuracy over the conventional non-pipelining approach with reference to the real system measurement. It can be seen that the difference of the computed execution time by the pipelining analysis is 9.... In PAGE 14: ...Table5 shows that the difference of 20.... ..."

"... In PAGE 11: ...e. BCET , WCETNP , and WCETP , together with the experimental collected results, denoted as Experimental values, are listed for each of the examples in Table5 . To give a quantita- tive idea about the accuracy of the theoretical approach in which processor pipelining architecture is taken into account in comparison with the case where no pipelining is considered in the calcula- tion, the differences with reference to real system measurements are presented.... In PAGE 12: ... In other words, similar to the instruction prefetching, the efficiency of the pipelining analysis is dependent upon the instructions involved as well as the program context. a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0 Codes NIFPIi (bytes) ri /wi (bytes) MINIETi (clks) a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0 a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0 MOV dx, 003ch 3 0/0 4 ADD cl, 01h 3 0/0 4 OR dl, cl 2 0/0 3 MOV dl, 0ch 2 0/0 3 NEG cx 2 0/0 3 XOR dl, 01h 3 0/0 4 CMP dx, cx 2 0/0 3 DEC dx 2 0/0 3 INC cx 2 0/0 3 SUB dl, cl 2 0/0 3 XCHG dx, cx 2 0/0 4 TEST dl, cl 2 0/0 3 PUSH dx 1 0/0 14 POP dx 1 0/0 14 a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0a2a0a1a0a1a0 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 a3 Table 1: Instruction Timing Information for Example 1 Obviously, according to Table5 , the pipelining approach provides a remarkable improve- ment in estimation accuracy over the conventional non-pipelining approach with reference to the real system measurement. It can be seen that the difference of the computed execution time by the pipelining analysis is 9.... In PAGE 14: ...Example 4 shows the case where both categories of instructions are included. Table5 shows that the difference of 20.... ..."

"... In PAGE 4: ...s 11.31. The execution time is given in the third column. No ef- fort has been directed to optimize algorithms in terms of execution speed. The theoretical computational complexity for both methods is given in Table3 , where C6 by C6 is the original image size, C8 is the number of nodes at the finest level, CA by CA is the search range at the finest mesh level, C1 is the number of iterations at the hexagonal matching procedure, CB is the search step size at the finest mesh level, and C4 is the number of levels. This table provides approxi- mate number of addition operations involved in the motion estima- tion stage.... In PAGE 4: ... This table assumes a regular mesh as opposed to content-based mesh to make the calculations tractable. From the results in Table ?? and the comparisons in Table3 , we can claim that the proposed method performs equally or slightly better in terms of PSNR and runs significantly faster. The latter is a direct consequence of pyra- mid image representation.... ..."

"... In PAGE 24: ... With the values of the other constants as shown, the thermal expansion of the LaRC-SI layers is almost completely irrelevant to the output of the model. The values that our model predicted are summarized in Table2 . As can be noted, the values vary only slightly from the measured values.... ..."