InsertionSortCollider memory leaks

Seems, it is not quite proper way to load particles.

Please, check an attached script with PyRunner and close the bug if it solves the problem for you.?field.comment=Seems, it is not quite proper way to load particles.

Please, check an attached script with PyRunner and close the bug if it solves the problem for you.

Anton, whatever Yade lets you do is a proper way of using it. Why is it not "proper"?

Still leaks.

Anton

Valgrind reports about problems with InsertionSortCollider::handleBoundInversion. All memory leak from this function, as seems. May be something is wrong with creating of the interactions for the new particles?

Valgrind report:
--------------------------------------------------------------------------------
  n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
--------------------------------------------------------------------------------
 47 74,791,217,242 854,819,752 735,410,607 119,409,145 0
86.03% (735,410,607B) (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
->33.37% (285,212,808B) 0x192CC889: InsertionSortCollider::handleBoundInversion(int, int, InteractionContainer*, Scene*) (in /home/sega/YADE/lib/yade-trunk/plugins/libplugins.so)
| ->33.37% (285,212,808B) 0x192CE9AD: InsertionSortCollider::action() (in /home/sega/YADE/lib/yade-trunk/plugins/libplugins.so)
| ->33.37% (285,212,808B) 0xFB86C02: Scene::moveToNextTimeStep() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->33.37% (285,212,808B) 0xFB8796A: SimulationFlow::singleAction() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->33.37% (285,212,808B) 0xFB80B3F: ThreadWorker::callSingleAction() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->33.37% (285,212,808B) 0xFB844F6: ThreadRunner::call() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->33.37% (285,212,808B) 0xFB84572: ThreadRunner::run() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->33.37% (285,212,808B) 0x8B31A52: boost::function0<void>::operator()() const (in /usr/lib64/libboost_python.so.1.39.0)
| ->33.37% (285,212,808B) 0x7B2500E: thread_proxy (in /usr/lib64/libboost_thread.so.1.39.0)
| ->33.37% (285,212,808B) 0x51B665B: start_thread (pthread_create.c:297)
| ->33.37% (285,212,808B) 0x5AF8ECB: clone (clone.S:112)
|
->12.20% (104,301,400B) 0xFB80F4C: InteractionContainer::insert(boost::shared_ptr<Interaction> const&) (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->12.20% (104,301,400B) 0x192CC8B2: InsertionSortCollider::handleBoundInversion(int, int, InteractionContainer*, Scene*) (in /home/sega/YADE/lib/yade-trunk/plugins/libplugins.so)
| ->12.20% (104,301,400B) 0x192CE9AD: InsertionSortCollider::action() (in /home/sega/YADE/lib/yade-trunk/plugins/libplugins.so)
| ->12.20% (104,301,400B) 0xFB86C02: Scene::moveToNextTimeStep() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->12.20% (104,301,400B) 0xFB8796A: SimulationFlow::singleAction() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->12.20% (104,301,400B) 0xFB80B3F: ThreadWorker::callSingleAction() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->12.20% (104,301,400B) 0xFB844F6: ThreadRunner::call() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->12.20% (104,301,400B) 0xFB84572: ThreadRunner::run() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
| ->12.20% (104,301,400B) 0x8B31A52: boost::function0<void>::operator()() const (in /usr/lib64/libboost_python.so.1.39.0)
| ...

I have discovered, that the number of iterations is getting larger and
larger during leak.
Also a simulation freezes.

Anton

On Fri, Feb 18, 2011 at 2:31 PM, Sergei Dorofeenko <email address hidden> wrote:
> Valgrind reports about problems with
> InsertionSortCollider::handleBoundInversion. All memory leak from this
> function, as seems. May be something  is wrong with creating of the
> interactions  for the new particles?
>
> Valgrind report:
> --------------------------------------------------------------------------------
>  n        time(i)         total(B)   useful-heap(B) extra-heap(B)    stacks(B)
> --------------------------------------------------------------------------------
>  47 74,791,217,242      854,819,752      735,410,607   119,409,145            0
> 86.03% (735,410,607B) (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
> ->33.37% (285,212,808B) 0x192CC889: InsertionSortCollider::handleBoundInversion(int, int, InteractionContainer*, Scene*) (in /home/sega/YADE/lib/yade-trunk/plugins/libplugins.so)
> | ->33.37% (285,212,808B) 0x192CE9AD: InsertionSortCollider::action() (in /home/sega/YADE/lib/yade-trunk/plugins/libplugins.so)
> |   ->33.37% (285,212,808B) 0xFB86C02: Scene::moveToNextTimeStep() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
> |     ->33.37% (285,212,808B) 0xFB8796A: SimulationFlow::singleAction() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
> |       ->33.37% (285,212,808B) 0xFB80B3F: ThreadWorker::callSingleAction() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
> |         ->33.37% (285,212,808B) 0xFB844F6: ThreadRunner::call() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
> |           ->33.37% (285,212,808B) 0xFB84572: ThreadRunner::run() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
> |             ->33.37% (285,212,808B) 0x8B31A52: boost::function0<void>::operator()() const (in /usr/lib64/libboost_python.so.1.39.0)
> |               ->33.37% (285,212,808B) 0x7B2500E: thread_proxy (in /usr/lib64/libboost_thread.so.1.39.0)
> |                 ->33.37% (285,212,808B) 0x51B665B: start_thread (pthread_create.c:297)
> |                   ->33.37% (285,212,808B) 0x5AF8ECB: clone (clone.S:112)
> |
> ->12.20% (104,301,400B) 0xFB80F4C: InteractionContainer::insert(boost::shared_ptr<Interaction> const&) (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
> | ->12.20% (104,301,400B) 0x192CC8B2: InsertionSortCollider::handleBoundInversion(int, int, InteractionContainer*, Scene*) (in /home/sega/YADE/lib/yade-trunk/plugins/libplugins.so)
> |   ->12.20% (104,301,400B) 0x192CE9AD: InsertionSortCollider::action() (in /home/sega/YADE/lib/yade-trunk/plugins/libplugins.so)
> |     ->12.20% (104,301,400B) 0xFB86C02: Scene::moveToNextTimeStep() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
> |       ->12.20% (104,301,400B) 0xFB8796A: SimulationFlow::singleAction() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
> |         ->12.20% (104,301,400B) 0xFB80B3F: ThreadWorker::callSingleAction() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
> |           ->12.20% (104,301,400B) 0xFB844F6: ThreadRunner::call() (in /home/sega/YADE/lib/yade-trunk/lib/libcore.so)
> |             ->1...

Hm... If I add one particle and run 100 iters, then add 10 particles and run 1 iter I have body 8 with wrong bound AABB. Seems, it is a cause of a very large amount of a ghost interacitons.

Cause is that velocityBins->bodyBins is resized (in InsertionSortCollider.cpp, line 216) _after_ having been used in BoundDispatcher::action, line 47. Therefore, for added particles used a garbage value of sweep.

Vaclav, could not you to check it? I need this feature, but I do not want to modify this code, because I do not know it in detail. Thanks!

Fix commited in bzr2749. Is it a good solution?