predicting impossible timetable

[liquid]

Consider following case:
Data consist of about 270 activities. There are about 120 max activities placed during the first minute of computation which goes down to about 60 in next minutes. This amount stays unchanged during pretty long time (at least 2 times the average time needed for solving similar set).
The question is: does it prove given data set to lead to impossible timetable or should one wait much longer for solving?

[Liviu Lalescu]

I think it is impossible, but I advise you to wait as much as possible.

There is a trick you could use, for more advanced usage: under GNU/Linux like, run FET from console. Under Windows, you need to compile FET enabling console output (ask me for details). Then, run FET from console. There are some messages (going fast, but you can read what you need, or you can redirect the console output to a file and read from there). There is one: "Warning - explored removal: id==a, time==b, times==c". You are interested in c (times). Normally, c should be at most 3 (higher for more difficult files, maybe). If it goes very high, higher than say 10 or 30 or even 100, then these are strong hints that the timetable is impossible.

You could even modify FET, so that if c (times) is too high (higher than a defined x), to stop. Ask me for details.

[liquid]

Yes, times grows slowly. After 4 mins, it reached max value of 8. I stopped generation. I'd better try to change some constraints.

[ekawaluya]

In my experience, the most easily, without the constraints generated time tables. Without constraint in just seconds. Then enter the constraints one by one, according to the priorities.