Assignment 1
Part A
clingo
clingo tsp.lp
| size |
ans |
time |
| 5 |
38 |
0.001 |
| 10 |
48 |
0.050 |
| 15 |
29 |
0.351 |
| 20 |
unknown |
too long to solve |
time grows exponentialy.
swi-prolog
| size |
ans |
time |
| 5 |
38 |
very fast |
| 10 |
48 |
10+s |
| 15 |
unknow |
too long to solve |
| 20 |
unknown |
too long to solve |
I didn't find a way to calculate runtime in swi-prolog. But it is clearly to see that it is way slower than clingo.
Google OR-Tools
AMPL
Part B
clingo
clingo task.lp
| size |
processors |
deadline |
result |
time |
| 10 |
3 |
20 |
yes |
0.031 |
| 20 |
6 |
30 |
no |
4.498 |
| 20 |
8 |
30 |
yes |
0.233 |
| 30 |
6 |
60 |
unknown |
too long to solve |
| 30 |
8 |
60 |
yes |
2.961 |
| 30 |
8 |
80 |
yes |
6.701 |
| 30 |
8 |
90 |
yes |
9.245 |
With either too high or too low constrain, the runtime grows exponentially. If the constrain is too low, it must iterate through all answers to find one correct result. If it is too high, then it may waste too much time fill the first processor.
Google OR-Tools
Part C
clingo
clingo cut.lp
| size |
ans |
time |
| 5 |
5 |
0.678 |
| 10 |
9 |
0.866 |
| 15 |
13 |
1.122 |
| 20 |
17 |
1.902 |
| 25 |
21 |
6.379 |
| 30 |
25 |
34.466 |
One important factor of the runtime is the max number of cut to check. If we set max = 10 with size = 30, we get the time of 0.914. However, it is impossible to get the max size before running, I believe this time would be a great representation of the runtime of clingo.
AMPL
