Re: Eclipse gets slower and slower...

Thanks to both of you for your quick replies!

Joachim Schimpf writes:
| 
| Do you also see a significant growth in the size of your
| eclipse process?
| 

It goes up from 9050K (initially) to 9267K (after six requests).
This is what 'top' says in the 'SIZE' column.

| 
| > 
| > The program uses assert/retract. But it is made sure that the database
| > is restored upon each request.
| 
| How are you doing the cleanup exactly?
| 
As in

	retract_all(given_rule(_,_,_,_,_))

There are about 15 dynamic predicates like given_rule, all of them get
facts asserted, and some of them have about up to 50 facts asserted.

What I am using heavily is an array:

	make_array(db_branch_nodes(200000)),

Typically,  the first 4000 entries are filled, and
assignments to it happen all the time.
 
Could that cause the problem?

| > P.S: I tried the profiler. But execution stopped with
| > 'Profiling time alarm' . If you could help me getting the profiler
| > run, that would certainly help. I'm running linux 2.4.18 on a pentium
| > III PC.
| 
| I don't know why your profiler doesn't work. Do you use
| an Eclipse "runtime" version?
| 

No.

But, anyway, I can work around by using cputime/1 to time
interesting fragments.

I came up with one more experiment: the assert/retract happen only
during an initialisation phase when a request comes. 
I can iterate the subsequent phase, say, 50 times and see if
this phase becomes slower (but this phase uses the array). 
If it does not become slower, assert/retract is the problem. I'd say.

But if you have further advice right now, it would be appreciated.

Cheers,

	Peter