Previous: 6. Arithmetic Precision: 64-bit
Up: 6. Arithmetic Precision: 64-bit
Next: 2. Filtered databases:
Figure 21:
PerfectClub: Dimension JANUS 64-bit vs JANUS 32-bit in sampled database
|
Figure 22:
PerfectClub: Dimension Simplex 64-bit vs 32-bit in sampled database
|
Figure 23:
PerfectClub: Dimension Fourier-Motzkin 64-bit vs 32-bit in sampled database
|
Figure 24:
PerfectClub: Dimension Fourier-Motzkin 64-bit vs 32-bit in sampled database
|
We compare the run times between two versions, 32-bit and 64-bit, of
each algorithm. Then we compare the numbers of exceptions between
32-bit and 64-bit versions of these implementations. In
6_fig:PerfectClub_satisfiability_dimension_JV64_JV32_100,
6_fig:PerfectClub_satisfiability_dimension_LS64_LS32_100 and
6_fig:PerfectClub_satisfiability_dimension_FM64_FM32_100, we
can see that the execution time ratios are between and ,
thus the sacrifice in execution time for using (higher
precision) instead of is to be considered.
Table 13:
PerfectClub: Numbers of exceptions in sampled database
|
PerfectClub |
timeout = 2 minutes |
#overflows |
#timeouts |
#operations |
JANUS 64-bit |
5 |
0 |
12668 |
JANUS 32-bit |
8 |
0 |
12668 |
Simplex 64-bit |
9 |
0 |
12668 |
Simplex 32-bit |
59 |
0 |
12668 |
Fourier-Motzkin 64-bit |
0 |
2 |
12668 |
Fourier-Motzkin 32-bit |
1 |
2 |
12668 |
Double Description 64-bit |
2 |
7 |
12668 |
Double Description 32-bit |
18 |
7 |
12668 |
|
6_tab:PerfectClub_satisfiability_64_32_exception_100 shows
that for PerfectClub sampled database, numbers of exceptions are fewer
using the higher precision as expected. We have similar results with
SPEC95.
Previous: 6. Arithmetic Precision: 64-bit
Up: 6. Arithmetic Precision: 64-bit
Next: 2. Filtered databases:
Nguyen Que Duong
2006-09-16