[vtkusers] Memory leak vtkEnsightGoldReader in Python
Bill Lorensen
bill.lorensen at gmail.com
Wed Mar 5 12:39:03 EST 2014
I don't see any leaks via valgrind. Maybe Berk's discovery is the
culprit for the memory growth.
On Wed, Mar 5, 2014 at 12:33 PM, Berk Geveci <berk.geveci at kitware.com> wrote:
> I went over the EnSight Gold reader carefully and there is only one
> small memory issue. Instead of a clearing a collection object every
> timestep, it keeps adding items to it. This accumulates to something
> small (7.5MB over 1000 timesteps in my example). It doesn't show up as
> a leak because the object is properly deleted at the end. Try the
> attached patch. With this change, I can see that the memory footprint
> of your Python example stays flat. I used Instruments on the Mac for
> this. It shows me all memory allocations of an applications.
>
> Best,
> -berk
>
>
> On Wed, Mar 5, 2014 at 9:32 AM, Berk Geveci <berk.geveci at kitware.com> wrote:
>> By the way, I have been looking into this as well. So far, I found no
>> memory issues either.
>>
>> PS: The time stepping for the EnSight reader is tested. I recently
>> added further coverage for it also.
>>
>> -berk
>>
>> On Wed, Mar 5, 2014 at 4:00 AM, Bernhard Righolt <b.w.righolt at tudelft.nl> wrote:
>>> Dear Bill,
>>>
>>> With your C++ script, I am indeed facing the same issue (although the python
>>> code runs significantly quicker for my case).
>>>
>>> I shared a few files, and a script to create more data here:
>>> https://www.dropbox.com/s/ndg6mwam2xdfmwq/data.tar.gz My experience is that
>>> with dissimilar data, the memory leak seems to be larger, but with this data
>>> (which just copies the time steps), it is still present,
>>>
>>> Best regards,
>>> Bernhard
>>>
>>>
>>> On 4 March 2014 21:27, Bill Lorensen <bill.lorensen at gmail.com> wrote:
>>>>
>>>> Attached is my C++ program and CMakeLists.txt file. But I don't think
>>>> this is a python issue...
>>>>
>>>> On Tue, Mar 4, 2014 at 3:10 PM, Bernhard Righolt <b.w.righolt at tudelft.nl>
>>>> wrote:
>>>> > Dear Bill,
>>>> >
>>>> > I am a C++ speaker, so I don't mind to test your code on my current
>>>> > large
>>>> > dataset. Can you maybe share your code, such that I can test it? I've
>>>> > never
>>>> > used VTK within C++, if it solves the issues I am experiencing with
>>>> > Python,
>>>> > I am very willing to switch. I hope it is not too complicated to link
>>>> > the
>>>> > libraries correctly.
>>>> >
>>>> > Bernhard
>>>> >
>>>> >
>>>> >
>>>> > On 4 March 2014 21:05, Bill Lorensen <bill.lorensen at gmail.com> wrote:
>>>> >>
>>>> >> Bernhard,
>>>> >>
>>>> >> Thanks for your patience. Since I'm not a python person, I converted
>>>> >> your program to C++. I ran valgrind and found no memory leaks.
>>>> >>
>>>> >> Perhaps the small dataset you provided does not exercise the code that
>>>> >> is leaking.
>>>> >>
>>>> >> Can you possible place a dataset that you know has leaksat some
>>>> >> accessible
>>>> >> site?
>>>> >>
>>>> >> I will say, that I don't think any of the current ensight tests access
>>>> >> multiple time steps. Your data and test script will be a good starting
>>>> >> point for such a test.
>>>> >>
>>>> >> Bill
>>>> >>
>>>> >> On Tue, Mar 4, 2014 at 12:08 AM, Bernhard Righolt
>>>> >> <b.w.righolt at tudelft.nl> wrote:
>>>> >> > Therr is a python script inside the archive which is the same script
>>>> >> > that I
>>>> >> > used on the large case.
>>>> >> >
>>>> >> > Bernhard
>>>> >> >
>>>> >> >
>>>> >> > On Mar 4, 2014 12:22 AM, "Bill Lorensen" <bill.lorensen at gmail.com>
>>>> >> > wrote:
>>>> >> >>
>>>> >> >> This is great. Can you send a complete python script that
>>>> >> >> illustrates
>>>> >> >> the
>>>> >> >> problem.?
>>>> >> >>
>>>> >> >> On Mar 3, 2014 2:47 PM, "Bernhard Righolt" <b.w.righolt at tudelft.nl>
>>>> >> >> wrote:
>>>> >> >> >
>>>> >> >> > Hi Bill,
>>>> >> >> >
>>>> >> >> > I added a very small sample data set. I am not sure if the memory
>>>> >> >> > problem I am encountering with my big data-set is measurable for
>>>> >> >> > this
>>>> >> >> > one.
>>>> >> >> >
>>>> >> >> > Anyhow, I would like to repeat that I am using Python 2.6 and VTK
>>>> >> >> > 5.10,
>>>> >> >> >
>>>> >> >> > Best regards,
>>>> >> >> > Bernhard
>>>> >> >> >
>>>> >> >> >
>>>> >> >> > On 3 March 2014 19:25, Bill Lorensen <bill.lorensen at gmail.com>
>>>> >> >> > wrote:
>>>> >> >> >>
>>>> >> >> >> The ensight readers could use more testing. Can you provide a
>>>> >> >> >> "small"
>>>> >> >> >> dataset and program that cause memory leaks?
>>>> >> >> >>
>>>> >> >> >> Bill
>>>> >> >> >>
>>>> >> >> >>
>>>> >> >> >> On Mon, Mar 3, 2014 at 10:38 AM, Bernhard Righolt
>>>> >> >> >> <b.w.righolt at tudelft.nl> wrote:
>>>> >> >> >> > Dear all,
>>>> >> >> >> >
>>>> >> >> >> > A typical case file, looks as follows (ASCII).
>>>> >> >> >> >
>>>> >> >> >> > $ head -n 20 case.case
>>>> >> >> >> > FORMAT
>>>> >> >> >> > type: ensight gold
>>>> >> >> >> >
>>>> >> >> >> > GEOMETRY
>>>> >> >> >> > model: 1 sliceCentre.000.mesh
>>>> >> >> >> >
>>>> >> >> >> > VARIABLE
>>>> >> >> >> > vector per node: 1 U sliceCentre.*****.U
>>>> >> >> >> >
>>>> >> >> >> > TIME
>>>> >> >> >> > time set: 1
>>>> >> >> >> > number of steps: 11550
>>>> >> >> >> > filename start number: 1
>>>> >> >> >> > filename increment: 1
>>>> >> >> >> > time values:
>>>> >> >> >> > 0.002
>>>> >> >> >> > 0.004
>>>> >> >> >> > 0.006
>>>> >> >> >> > 0.008
>>>> >> >> >> > 0.01
>>>> >> >> >> > $
>>>> >> >> >> >
>>>> >> >> >> > The mesh (snippet only), is as follows:
>>>> >> >> >> >
>>>> >> >> >> > $ head -n 10 sliceCentre.000.mesh
>>>> >> >> >> > Ensight Geometry File
>>>> >> >> >> > =====================
>>>> >> >> >> > node id assign
>>>> >> >> >> > element id assign
>>>> >> >> >> > part
>>>> >> >> >> > 1
>>>> >> >> >> > sliceCentre.000.mesh
>>>> >> >> >> > coordinates
>>>> >> >> >> > 299028
>>>> >> >> >> > 0.00000e+00
>>>> >> >> >> > $ sed -n -e 897091,897098p sliceCentre.000.mesh
>>>> >> >> >> > 0.00000e+00
>>>> >> >> >> > 0.00000e+00
>>>> >> >> >> > 0.00000e+00
>>>> >> >> >> > tria3
>>>> >> >> >> > 595264
>>>> >> >> >> > 162106 162714 162799
>>>> >> >> >> > 162799 162226 162106
>>>> >> >> >> > 162714 162106 162021
>>>> >> >> >> >
>>>> >> >> >> >
>>>> >> >> >> > And the corresponding datafile
>>>> >> >> >> > $ head -n 10 sliceCentre.00001.U
>>>> >> >> >> > vector
>>>> >> >> >> > part
>>>> >> >> >> > 1
>>>> >> >> >> > coordinates
>>>> >> >> >> > 5.33425e-04
>>>> >> >> >> > 2.69515e-03
>>>> >> >> >> > -1.65164e-03
>>>> >> >> >> > 5.33425e-04
>>>> >> >> >> > 1.39069e-04
>>>> >> >> >> > 2.69515e-03
>>>> >> >> >> >
>>>> >> >> >> >
>>>> >> >> >> > However, I doubt if something is wrong with my datafiles
>>>> >> >> >> > itself,
>>>> >> >> >> > because I
>>>> >> >> >> > can easily read them in Paraview for example, the only
>>>> >> >> >> > difference
>>>> >> >> >> > is
>>>> >> >> >> > the
>>>> >> >> >> > large amount of data-files, e.g (124GB in this case)
>>>> >> >> >> >
>>>> >> >> >> > Best regards,
>>>> >> >> >> > Bernhard
>>>> >> >> >> >
>>>> >> >> >> >
>>>> >> >> >> > On 27 February 2014 22:28, Berk Geveci
>>>> >> >> >> > <berk.geveci at kitware.com>
>>>> >> >> >> > wrote:
>>>> >> >> >> >>
>>>> >> >> >> >> Nothing strikes me as obviously wrong in your script. Can you
>>>> >> >> >> >> send
>>>> >> >> >> >> me
>>>> >> >> >> >> your case file? I'll take a look.
>>>> >> >> >> >>
>>>> >> >> >> >> -berk
>>>> >> >> >> >>
>>>> >> >> >> >> On Thu, Feb 27, 2014 at 3:20 AM, Bernhard Righolt
>>>> >> >> >> >> <b.w.righolt at tudelft.nl> wrote:
>>>> >> >> >> >> > Dear all,
>>>> >> >> >> >> >
>>>> >> >> >> >> > Currently I am using a Python script to process information
>>>> >> >> >> >> > stored
>>>> >> >> >> >> > in
>>>> >> >> >> >> > the
>>>> >> >> >> >> > EnSight Gold format. My Python (2.6) scipt uses VTK (5.10.0)
>>>> >> >> >> >> > to
>>>> >> >> >> >> > process
>>>> >> >> >> >> > the
>>>> >> >> >> >> > file, where I used the vtkEnSightGoldReader for reading the
>>>> >> >> >> >> > data,
>>>> >> >> >> >> > and
>>>> >> >> >> >> > loop
>>>> >> >> >> >> > over time steps.
>>>> >> >> >> >> >
>>>> >> >> >> >> > In principle this works fine for smaller datasets, however,
>>>> >> >> >> >> > for
>>>> >> >> >> >> > large
>>>> >> >> >> >> > datasets (i.e. 11k files of 12MB), I see the memory usage
>>>> >> >> >> >> > (recorded via
>>>> >> >> >> >> > top)
>>>> >> >> >> >> > increasing with time while the process is running. This
>>>> >> >> >> >> > filling
>>>> >> >> >> >> > of
>>>> >> >> >> >> > the
>>>> >> >> >> >> > memory goes slow, but in some cases problems are inevitable.
>>>> >> >> >> >> > Apparently,
>>>> >> >> >> >> > I
>>>> >> >> >> >> > am doing some things wrong in my script with respect to
>>>> >> >> >> >> > memory
>>>> >> >> >> >> > management.
>>>> >> >> >> >> >
>>>> >> >> >> >> > I am using the following script (stripped obviously, but
>>>> >> >> >> >> > still
>>>> >> >> >> >> > showing
>>>> >> >> >> >> > this
>>>> >> >> >> >> > issue)
>>>> >> >> >> >> >
>>>> >> >> >> >> > import vtk
>>>> >> >> >> >> >
>>>> >> >> >> >> > reader = vtk.vtkEnSightGoldReader()
>>>> >> >> >> >> >
>>>> >> >> >> >> > reader.SetCaseFileName("case.case")
>>>> >> >> >> >> > reader.Update()
>>>> >> >> >> >> >
>>>> >> >> >> >> > # Get time values
>>>> >> >> >> >> > timeset=reader.GetTimeSets()
>>>> >> >> >> >> > time=timeset.GetItem(0)
>>>> >> >> >> >> > timesteps=time.GetSize()
>>>> >> >> >> >> >
>>>> >> >> >> >> > #reader.ReleaseDataFlagOn()
>>>> >> >> >> >> >
>>>> >> >> >> >> > for j in range(timesteps):
>>>> >> >> >> >> > curTime=time.GetTuple(j)[0]
>>>> >> >> >> >> > print curTime
>>>> >> >> >> >> > reader.SetTimeValue(curTime)
>>>> >> >> >> >> > reader.Update()
>>>> >> >> >> >> >
>>>> >> >> >> >> > #reader.RemoveAllInputs()
>>>> >> >> >> >> >
>>>> >> >> >> >> >
>>>> >> >> >> >> > I tried a few things as you can see
>>>> >> >> >> >> > - reader.ReleaseDataFlagOn(), without success
>>>> >> >> >> >> > - reader.RemoveAllInputs(), did not get it to run
>>>> >> >> >> >> > - running gc.collect at the end of the loop, following
>>>> >> >> >> >> >
>>>> >> >> >> >> >
>>>> >> >> >> >> >
>>>> >> >> >> >> > http://vtk.1045678.n5.nabble.com/VTK-memory-management-td5715661.html
>>>> >> >> >> >> > - Using DeepCopy as suggested here:
>>>> >> >> >> >> >
>>>> >> >> >> >> >
>>>> >> >> >> >> >
>>>> >> >> >> >> >
>>>> >> >> >> >> > http://vtk.1045678.n5.nabble.com/Memory-leak-in-VTK-Python-module-td1234002.html
>>>> >> >> >> >> > However, I am already experiencing this without a call to
>>>> >> >> >> >> > GetOutput
>>>> >> >> >> >> >
>>>> >> >> >> >> >
>>>> >> >> >> >> > My question is, how can I properly unload/replace the data
>>>> >> >> >> >> > that
>>>> >> >> >> >> > is
>>>> >> >> >> >> > stored in
>>>> >> >> >> >> > the memory, instead of using more memory continuously?
>>>> >> >> >> >> >
>>>> >> >> >> >> > Best regards,
>>>> >> >> >> >> > Bernhard
>>>> >> >> >> >> >
>>>> >> >> >> >> > _______________________________________________
>>>> >> >> >> >> > 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
>>>> >> >> >> >> >
>>>> >> >> >> >> > Follow this link to subscribe/unsubscribe:
>>>> >> >> >> >> > http://www.vtk.org/mailman/listinfo/vtkusers
>>>> >> >> >> >> >
>>>> >> >> >> >
>>>> >> >> >> >
>>>> >> >> >> >
>>>> >> >> >> > _______________________________________________
>>>> >> >> >> > 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
>>>> >> >> >> >
>>>> >> >> >> > Follow this link to subscribe/unsubscribe:
>>>> >> >> >> > http://www.vtk.org/mailman/listinfo/vtkusers
>>>> >> >> >> >
>>>> >> >> >>
>>>> >> >> >>
>>>> >> >> >>
>>>> >> >> >> --
>>>> >> >> >> Unpaid intern in BillsBasement at noware dot com
>>>> >> >> >
>>>> >> >> >
>>>> >>
>>>> >>
>>>> >>
>>>> >> --
>>>> >> Unpaid intern in BillsBasement at noware dot com
>>>> >
>>>> >
>>>>
>>>>
>>>>
>>>> --
>>>> Unpaid intern in BillsBasement at noware dot com
>>>
>>>
--
Unpaid intern in BillsBasement at noware dot com
More information about the vtkusers
mailing list