[Gate-users] Jobs killed "memory exceeded"?

Wed Sep 17 10:05:39 CEST 2014

Hello,

     Did Valgrind identified a memory leak ? I just want to stress that 
one difficulty with optical photons is that a huge amount of them can be 
produced during the tracking of one track. There is an option with 
optical photon which allows the production to be limited to a given 
amount, N. When this amount is reached, the track "t" is suspended, and 
the N optical photons are tracked, to clear them from memory, and then 
the tracking of "t" restarts, etc. I just want to be sure you are aware 
of this mechanism, sorry for time wasting if you were already aware.

Cheers,
Marc

On 09/16/2014 11:01 PM, Dávid Völgyes wrote:
> Hi David,
>
> Yes, I just checked a minute ago, and all three of the patches are
> incorporated into the develop branch.
> So I would modify my advice to this one:
> Amy could try the latest development code, maybe it solves the memory
> leak problem.
> (This is actually better idea than just applying our patches, because
> there is at least another one memory leak fix in the development
> branch. (I did not check all the commits, so there may be more.))
>
> In general, if somebody can create a short test GATE macro (the
> shorter the better) for any bug, and he/she can share it, then it is
> very useful for the actual code development.
>
> Best,
>    David
>
>
> On Tue, Sep 16, 2014 at 9:12 PM, David Sarrut
> <David.Sarrut at creatis.insa-lyon.fr> wrote:
>> Hello,
>>
>> David : correct me if I am wrong, but those patches are already included in
>> the develop branch of the git version, right ?
>>
>> thanks,
>> David
>>
>> On Tue, Sep 16, 2014 at 1:43 PM, Dávid Völgyes <david.volgyes at gmail.com>
>> wrote:
>>> Hi Amy,
>>>
>>> Hopefully, Sebastiens suggestion will help you, but if not, you can try
>>> the followings:
>>>
>>> Gergely Patay and me sent some source code patches in June to the mailing
>>> list, they could help.
>>> The only drawback is that you have to compile GATE from source.
>>>
>>> Most of my fixes are related to memory leaks, but actually most of them
>>> are ITK related, and I do not see any alfa/beta source specific
>>> modification. Anyway, you can try these patches, it will not harm. If you do
>>> so, please let us (the mailing list) know the result (helped or did not).
>>>
>>> Applying the patches is actually quite easy:
>>>
>>> First you have to download the patch files (see below), and uncompress
>>> them with gunzip.
>>>
>>> After that you have to do everything exactly like the GATE install manual
>>> says, except this part:
>>>
>>> - when you get a copy from the source:
>>> git clone http://git.opengatecollaboration.org/git/opengate-public.git
>>>
>>> - then you have to enter to the GATE source directory:
>>> cd opengate-public
>>>
>>> - and you have to apply the patches:
>>> git apply  cppfix_dvolgyes.patch geant4-geometry.patch interfile.patch
>>>
>>> (the actual filenames can contain path-parts, like
>>> ../patches/cppfix_dvolgyes.patch, it depends on your actual directory
>>> layout.)
>>>
>>> After that you should follow the manual. (cmake, compile, install, etc.)
>>>
>>> As a last hope without actually debugging the code, you can try different
>>> geant/root/gate/clhep versions,
>>> maybe it comes from a recently introduced or recently fixed bug. Of
>>> course, this takes a lot of time, and there is no guarantee for the success.
>>>
>>> (The original e-mail what contained the patches is quoted below my
>>> signature.)
>>>
>>> Good luck!
>>>
>>> Cheers,
>>>    David
>>>
>>> On Wed, Jun 18, 2014 at 10:54 AM, Patay Gergely <patay.gergely at mediso.hu>
>>> wrote:
>>>> Dear Gaters,
>>>>
>>>> here are a few patches against the upstream (ie. master branch) Gate
>>>> source
>>>> code. Since I am not a member of the collaboration, I cannot directly
>>>> commit
>>>> them into the git repository.
>>>>
>>>> The first patch was written by Dávid Völgyes (formerly at Mediso and at
>>>> COMPET). These are C++ bugfixes (unchecked ranges, memory corruptions).
>>>> [1]
>>>>
>>>> https://dl.dropboxusercontent.com/u/43097994/gate7/cppfix_dvolgyes.patch.gz
>>>>
>>>> The second patch contains Gate wrappers to all the missing Geant4 solid
>>>> objects from
>>>>
>>>> http://geant4.web.cern.ch/geant4/UserDocumentation/UsersGuides/ForApplicationDeveloper/html/ch04.html
>>>> .
>>>> With this patch the following extra shapes will be available: cut tube,
>>>> elliptical cone, generic trapezoid, hyperbolic tube, parallelepiped,
>>>> paraboloid tube, tetrahedron, torus, twisted box, twisted trapezoid,
>>>> generic
>>>> twisted trapezoid and twisted tube segment.
>>>> [2]
>>>>
>>>> https://dl.dropboxusercontent.com/u/43097994/gate7/geant4-geometry.patch.gz
>>>>
>>>> The third patch is a heavy modification of the Interfile reader. Besides
>>>> some code cleaning and refactoring, it solves the following issues:
>>>> - if the key 'name of data file' is present in the header file then it
>>>> won't
>>>> guess the name
>>>> - the 'number of bytes per pixel' key (1,2,4,8...) is implemented
>>>> - the 'data offset in bytes' key is implemented
>>>> - the 'number format' key now accepts 'signed integer', 'unsigned
>>>> integer',
>>>> 'short float', 'long float' and (for convenience) 'float', as seen in
>>>> the
>>>> Interfile 3.3 standard.
>>>> The interfile reader thus accepts signed or unsigned 8, 16, 32 or 64 bit
>>>> integer as well as 32 (float) or 64 bit (double) floating point data
>>>> from
>>>> little or big endian binary file.
>>>> [3]
>>>> https://dl.dropboxusercontent.com/u/43097994/gate7/interfile.patch.gz
>>>>
>>>> cheers,
>>>>   Gergely
>>>>
>>>> ---
>>>> Gergely Patay
>>>> physicist
>>>> Mediso Medical Imaging Systems Ltd.
>>>> Hungary
>>>
>>>
>>> On Tue, Sep 16, 2014 at 11:40 AM, JAN Sebastien <sebastien.jan at cea.fr>
>>> wrote:
>>>>
>>>>
>>>> Hi Amy,
>>>>
>>>> As explained by Marc, this is a memory leak…probably du to the fact that
>>>> the energy of optical photons which are generated is out of the range
>>>> defined in the xml files (material & surface). Try to correct that point and
>>>> fell free to tell us if the problem is solved.
>>>>
>>>> Cheers
>>>>
>>>> Sebastien
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> De : gate-users-bounces at lists.opengatecollaboration.org
>>>> [mailto:gate-users-bounces at lists.opengatecollaboration.org] De la part de
>>>> Marc Chamberland
>>>> Envoyé : lundi 15 septembre 2014 16:06
>>>> À : Amy Meldrum
>>>> Cc : Gate Users
>>>> Objet : Re: [Gate-users] Jobs killed "memory exceeded"?
>>>>
>>>>
>>>>
>>>> Hi Amy,
>>>>
>>>>
>>>>
>>>> Perhaps there is a memory leak? If there is, it's gonna be very annoying
>>>> for you to troubleshoot if you have no C++ experience.
>>>>
>>>>
>>>>
>>>> I'd suggest you look at using Valgrind to detect memory leaks. You can
>>>> find Valgrind here: http://valgrind.org
>>>>
>>>>
>>>>
>>>> The Quick Start page should help you run it. Feel free to post the output
>>>> to the mailing list or send it to me and I'll have a quick look.
>>>>
>>>>
>>>>
>>>> Good luck!
>>>>
>>>> Marc
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> __________________________
>>>>
>>>>
>>>>
>>>> Marc Chamberland, MSc
>>>>
>>>> PhD candidate
>>>> Department of Physics
>>>> Carleton University
>>>> Ottawa (ON)
>>>>
>>>>
>>>>
>>>> On Sep 14, 2014, at 8:55, Amy Meldrum <ameldru at g.clemson.edu> wrote:
>>>>
>>>>
>>>>
>>>> Gate users,
>>>>
>>>>
>>>>
>>>> I have been running simulations using optical photons that run to
>>>> completion when I am using a beta source. However, when I change the source
>>>> to an alpha source, suddenly all of my simulations only run for a few
>>>> minutes before being killed. It seems like the memory usage skyrockets
>>>> beyond some limit and get killed.
>>>>
>>>>
>>>>
>>>> Why would changing only the source do this? Is there something I can do
>>>> to stop this from happening? The simulation runs fine for a few particles,
>>>> but I am not sure why it would behave like this.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> Please help. Thank you,
>>>>
>>>>
>>>>
>>>> Amy
>>>>
>>>> _______________________________________________
>>>> Gate-users mailing list
>>>> Gate-users at lists.opengatecollaboration.org
>>>> http://lists.opengatecollaboration.org/mailman/listinfo/gate-users
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Gate-users mailing list
>>>> Gate-users at lists.opengatecollaboration.org
>>>> http://lists.opengatecollaboration.org/mailman/listinfo/gate-users
>>>
>>>
>>> _______________________________________________
>>> Gate-users mailing list
>>> Gate-users at lists.opengatecollaboration.org
>>> http://lists.opengatecollaboration.org/mailman/listinfo/gate-users
>>
>>
>>
>> --
>> David Sarrut, Phd
>> Directeur de recherche CNRS
>> CREATIS, UMR CNRS 5220, Inserm U 1044
>> Centre de lutte contre le cancer Léon Bérard
>> 28 rue Laënnec, 69373 Lyon cedex 08
>> Tel : 04 78 78 51 51 / 06 74 72 05 42
>> http://www.creatis.insa-lyon.fr/~dsarrut
>> _________________________________
>>   "2 + 2 = 5,  for extremely large values of 2"
>> _________________________________
> _______________________________________________
> Gate-users mailing list
> Gate-users at lists.opengatecollaboration.org
> http://lists.opengatecollaboration.org/mailman/listinfo/gate-users