[gecode-users] Gecode/J and time stability

[Guillaume Richaud]

Hi,

I tried to solve some problems with Gecode/J and the time spend to solve 
the same instance seems to be quite unstable. In the example 
QueensJavaPropagator, i removed Distinct constraints. So, the solver 
just had to instantiate each variable to find a solution.
Solving 20 times the same instance:
nb vars=100, min=18ms, max=216ms, av=36ms
nb vars=500, min=371ms, max=3581ms, av=1544ms
nb vars=1000, min=2170ms, max=7333ms, av=6236ms

And i observe a strange phenomena: time variations are similar at each run.
For example with nb vars=600
           Run 1     Run2
1st:     825ms /  761ms
2nd:  1853ms / 1744ms
3rd:   1391ms / 1308ms
4th:   2890ms / 2751ms
5th:   1422ms / 1399ms
6th:   4141ms / 3963ms
7th:     963ms / 945ms
8th:   5073ms / 4775ms
9th:     903ms / 862ms
10th: 5031ms / 4415ms

Do you know what is the cause of the problem ?

Thank,
  Guillaume

ps: I also tried to force garbage collector before solving...

-------------- next part --------------
A non-text attachment was scrubbed...
Name: QueensJavaPropagator.java
Type: java/*
Size: 6640 bytes
Desc: not available
URL: <http://www.ps.uni-sb.de/pipermail/users/attachments/20071102/fc9bd452/attachment.bin>