15.01.2013 Views

U. Glaeser

U. Glaeser

U. Glaeser

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Results Using Probabilistic Technique<br />

The signal probability and activity for the primary inputs are both specified to be 0.5. The load capacitance<br />

at the output nodes is specified to a unit capacitance and the internal node capacitances are specified to<br />

one-half unit capacitance. The maximum number of inputs allowed for each partition level is 10 inputs.<br />

The simulations are run on SPARCstation 5, and the results are shown in Table 20.2. Power dissipation<br />

due to the internal node ranges from 9.38% to 22.4% of the overall power consumption. Hence, the<br />

internal nodes power is a significant portion of the total power consumption. The result is given in term<br />

of power measure φ (switching activity × fanouts).<br />

Results Using Statistical Technique<br />

Similar to the probabilistic technique, the signal probability and activity for the primary inputs are<br />

specified to be 0.5. The sample period of each simulation run is specified to be 100 unit clock cycles.<br />

The relative error is specified to be 30%, and the minimum threshold is specified to be 3%. In the<br />

simulation, 5 V power supply is used. The threshold voltages for both PMOS and NMOS devices are<br />

specified to be 1 V. The results are tabulated in Table 20.3.<br />

The experiment shows that the percentage of internal nodes power consumption to overall power<br />

consumption ranges from 7.75% to 18.59%. The result of the same simulation with the charge-sharing<br />

option being switched off is shown in Table 20.4.<br />

The computation time is faster by up to 10% for certain cases when the simulation is run with the<br />

charge-sharing option switched off. The percentage of internal power to overall power only changes by<br />

0.1% to 0.2% when charge sharing is not taken into consideration, so neglecting the charge-sharing effect<br />

among the internal node capacitances will not affect the overall result significantly.<br />

© 2002 by CRC Press LLC<br />

TABLE 20.1 The ISCAS-85 Benchmark Circuits<br />

Circuit No. of Inputs No. of Outputs No. of Nodes No. of Levels<br />

C1355 41 32 514 23<br />

C17 5 2 6 3<br />

C1908 33 25 880 40<br />

C2670 233 140 1161 32<br />

C3540 50 22 1667 47<br />

C432 36 7 160 17<br />

C499 41 32 202 11<br />

C5315 178 123 2290 49<br />

C6288 32 32 2416 124<br />

C7552 207 108 3466 43<br />

C880 60 26 357 23<br />

TABLE 20.2 Results of Probabilistic Technique<br />

Circuit<br />

CPU Time<br />

SPARCstation 5 (seconds)<br />

Internal Nodes<br />

Power Measure<br />

Total<br />

Power Measure (φ)<br />

C1355 9.82 40.9 237.7<br />

C17 0.03 0.51 3.77<br />

C1908 72.92 32.3 343.97<br />

C2670 34.36 55.4 505.7<br />

C3540 86.14 63.1 590.3<br />

C432 29.68 8.92 72.96<br />

C499 12.98 26.6 118.79<br />

C5315 398.74 118.9 1106.3<br />

C6288 1786.44 228.9 1300.1<br />

C7552 709.71 172.3 1516.8<br />

C880 6.58 22.3 162.0

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!