[vtkusers] memory leak in vtkStructuredPoints or am I doing something wrong?

[David Lonie]

Hi Robert,

I took a look at your test case (thanks, btw -- these help a lot!) and
while I can reproduce the results, it appears that linux is just caching
the memory for the process, not leaking it.

I modified your script to do 50 cycles of load/clears, with this result:

./vtkbug

Press 1 for zebra, 2 for phallucia


1


Data loaded: 352 MB in use.


Data cleared: 248 MB in use.


Data loaded: 378 MB in use.


Data cleared: 378 MB in use.


Data loaded: 352 MB in use.


Data cleared: 352 MB in use.


Data loaded: 352 MB in use.


Data cleared: 352 MB in use.
[snipped 45 load/clears]
Data loaded: 378 MB in use.
Data cleared: 378 MB in use.

While the memory is still marked as reserved by the process, it doesn't
grow, and the 'leaked' memory is reused by later allocations.

I also verified that the dataset destructors are being called at the
appropriate times, and valgrind confirms that no memory is leaking at exit.

I did a quick google search for clearing linux memory caches, and it
appears that it may be possible to instruct linux to drop the cache if this
is causing problems, but I suspect the operating system will reclaim cached
memory when it starts running out. Unless you start seeing memory
allocations fail or excessive swapping, this can likely be ignored.

HTH,
Dave

On Tue, Jun 21, 2016 at 2:36 AM, RobertS <spir.robert at gmail.com> wrote:

> Hello all,
> I'm doing visualization of series of 3D volumes stored in VTK files. First
> I
> load all vtk files to vector of vtkStructuredPoints using this simple code
> inside cycle
>
> vtkSmartPointer<vtkStructuredPointsReader> reader =
> vtkSmartPointer<vtkStructuredPointsReader>::New();
> reader->SetFileName(filename);
> reader->Update();
> ptsCopy = vtkSmartPointer<vtkStructuredPoints>::New();
> ptsCopy->DeepCopy(reader->GetOutput());
> loadedPoints.push_back(ptsCopy);
> reader->CloseVTKFile();
>
> then when I want to change the dataset I just call loadedPoints.clear() and
> start loading other vtk files.
> Now on windows, everything goes as expected, by calling .clear() memory get
> freed. This is not the case in linux, here the memory is stil occupied and
> since I'm working with very large datasets (>12GB) this is a serious
> problem.
> Now another interesting thing is, that with some datasets the memory gets
> cleared even in linux, but with some others it doesn't...
> I'm using VTK 7.0 compiled from source, vs2015 in windows, gcc6.1 in linux
>
> I made a small project that only loads the vtk files to vector and then
> clears it, displaying amount of used memory in between. I can reproduce the
> memory leak with this code in linux, in windows it clears the memory.
> http://tesla.math.sk/vtkbug.7z
> Here you can download two datasets, zebra is causing the leak, phallusia is
> not
> http://tesla.math.sk/zebra.7z
> http://tesla.math.sk/phallusia.7z
>
> Thanks for any help
>
> Robert Spir PhD
> Department of Mathematics
> Slovak University of Technology
> Radlinskeho 11
> 813 68 Bratislava
> Slovakia
>
>
>
>
>
> --
> View this message in context:
> http://vtk.1045678.n5.nabble.com/memory-leak-in-vtkStructuredPoints-or-am-I-doing-something-wrong-tp5738805.html
> Sent from the VTK - Users mailing list archive at Nabble.com.
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the VTK FAQ at:
> http://www.vtk.org/Wiki/VTK_FAQ
>
> Search the list archives at: http://markmail.org/search/?q=vtkusers
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/vtkusers
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtkusers/attachments/20160621/816defba/attachment.html>