Hi Juha,
Juha Heinanen wrote on 12/22/2015 09:18 AM:
> i made some tests regarding sems memory consumption when executing
> simple dsm scripts.
>
> the first script plays an audio from db. the second makes two python
> tests
>
> py(in_period('Weekday.Hr08-Hr16') == 1);
> py(in_period('Always') == 1);
>
> and then plays an audio from file.
>
> i make 100 calls to each application at rate about 20 calls per sec and
> after each 100 calls check how much memory sems is using.
...
>
> should i be worried about this? how to find out if memory is actually
> leaking?
There is the STL pool allocator, and then with python there's the
python memory allocation, both of which can make it look like memory
leaks which aren't really. I'd perform long-time tests, and after the
last calls I would also wait a few minutes (transactions/timers
expiring etc). If you're doing tests with really many calls
(millions), at some point the memory should stay stable. If not,
there's definitely a leak (such as what Alex reported).
Also, running under valgrind sometimes shows where the leaks can be found.
Stefan
_______________________________________________
Sems mailing list
[email protected]
http://lists.iptel.org/mailman/listinfo/sems
