[vtkusers] For people at kitware : ReleaseDataFlagOn question
Naveen Kumar Polapally
naveenpkumar at hotmail.com
Thu Nov 8 21:12:42 EST 2001
hi all,
As i said earlier i am using mpi over clan, and there is clan optimization
due to which it holds the memory and tries to allocate any future memory
requests from this held memory. To change this you have to undefine the flag
in mpidirectory/mpid/via/dreg.h
/*
#define LAZY_MEM_UNREGISTER
*/
and recompile and then compile vtk with this mpilibraries.
After which release data feature works as expected.
>From: "Naveen Kumar Polapally" <naveenpkumar at hotmail.com>
>To: berkgeveci at hotmail.com, vtkusers at public.kitware.com
>Subject: Re: [vtkusers] For people at kitware : ReleaseDataFlagOn question
>Date: Thu, 08 Nov 2001 06:23:50
>
>I am using mpich-1.2.0 over clan on cluster running Redhat7.1.
>
>
>>From: "Berk Geveci" <berkgeveci at hotmail.com>
>>To: naveenpkumar at hotmail.com, vtkusers at public.kitware.com
>>Subject: Re: [vtkusers] For people at kitware : ReleaseDataFlagOn question
>>Date: Wed, 07 Nov 2001 17:49:15 -0500
>>
>>
>>Which MPI implementation is this ? On which platform ?
>>
>>-Berk
>>
>>
>>>From: "Naveen Kumar Polapally" <naveenpkumar at hotmail.com>
>>>To: vtkusers at public.kitware.com
>>>Subject: [vtkusers] For people at kitware : ReleaseDataFlagOn question
>>>Date: Wed, 07 Nov 2001 22:34:27
>>>
>>>hi all,
>>>For some time now i have been trying to use ReleaseData feature in my
>>>parallel program but couldn't do so. Now i found out that the problem is
>>>not
>>>with VTK but it is with MPI. It seems we cannot release the memory
>>>allocated
>>>after MPI::Init. I have a simple program here you can check the memory
>>>consumption on your machine .
>>>
>>>I think some thing should be done about it because the release data
>>>feature
>>>is most useful in case of large datasets and parallel programs.
>>>
>>>
>>>#include <iostream.h>
>>>#include "mpi++.h"
>>>#include <unistd.h>
>>>#include <stdlib.h>
>>>
>>>int
>>>main(int argc, char *argv[])
>>>{
>>> MPI::Init(argc, argv);
>>>
>>> int rank = MPI::COMM_WORLD.Get_rank();
>>> int size = MPI::COMM_WORLD.Get_size();
>>>
>>> int *Ptr=new int[10000000];
>>>
>>> for(int i=0;i<10000000;i++) Ptr[i]=0;
>>> cout<<" allocated "<<endl;
>>> sleep(15);
>>>
>>> delete Ptr;
>>> cout<<" just deleted "<<endl;
>>> sleep(15);
>>>
>>> cout << "Hello World! I am " << rank << " of " << size << endl;
>>>
>>> MPI::Finalize();
>>>}
>>>
>>>
>>>Thanks,
>>>naveen.
>>>
>>>
>>>
>>>
>>>>From: "Naveen Kumar Polapally" <naveenpkumar at hotmail.com>
>>>>To: vtkusers at public.kitware.com
>>>>Subject: Re: [vtkusers] ReleaseDataFlagOn question
>>>>Date: Tue, 06 Nov 2001 22:37:59
>>>>
>>>>
>>>>hi,
>>>>
>>>>For some days now i have been complaining that ReleaseData feature is
>>>>not
>>>>working for graphical pipeline while it is working for imaging pipeline.
>>>>No
>>>>wonder no one answered, because i was wrong.
>>>>
>>>>
>>>>I am rephrasing my question,for a pipeline shown below the release data
>>>>machanism is working. But when i parallelize it it does not work.
>>>>
>>>>vtkImageReader *reader = vtkImageReader::New();
>>>>reader->SetDataByteOrderToBigEndian();
>>>>reader->SetDataExtent(0, 511, 0,511,Mainzmin,Mainzmax);
>>>>reader->SetFilePrefix("/home/devel/vtk/vtkdata/vfem/slice");
>>>>reader->ReleaseDataFlagOn();
>>>>reader->SetDataSpacing(1, 1, 1);
>>>>
>>>>
>>>>vtkSynchronizedTemplates3D *iso = vtkSynchronizedTemplates3D::New()
>>>>iso->SetInput(reader->GetOutput());
>>>>iso->SetValue(0,600);
>>>>iso->Update();
>>>>
>>>>I narrowed my problem to the method shown below.
>>>>
>>>>vtkShortArray::~vtkShortArray()
>>>>{
>>>> if ((this->Array) && (!this->SaveUserArray))
>>>> {
>>>> delete [] this->Array;
>>>> }
>>>> delete [] this->Tuple;
>>>>}
>>>>
>>>>Both the parallel and sequential executions call this method and both of
>>>>them delete the this->Array. This is the array which holds all the data
>>>>of
>>>>the output of vtkImageReader. But only in case of sequential program i
>>>>see
>>>>a reduction of memory usage, while in case of parallel program i don't
>>>>see
>>>>any change in memory usage. I am unable to explain this as the memory
>>>>gets
>>>>allocated in vtkShortArray::Allocate method into Array when you delete
>>>>it,
>>>>it has to free the memory.
>>>>
>>>>
>>>>Any help is appreciated.
>>>>
>>>>
>>>>
>>>>thanks,
>>>>naveen.
>>>>
>>>>_________________________________________________________________
>>>>Get your FREE download of MSN Explorer at
>>>>http://explorer.msn.com/intl.asp
>>>>
>>>>_______________________________________________
>>>>This is the private VTK discussion list.
>>>>Please keep messages on-topic. Check the FAQ at:
>>>><http://public.kitware.com/cgi-bin/vtkfaq>
>>>>Follow this link to subscribe/unsubscribe:
>>>>http://public.kitware.com/mailman/listinfo/vtkusers
>>>
>>>
>>>_________________________________________________________________
>>>Get your FREE download of MSN Explorer at
>>>http://explorer.msn.com/intl.asp
>>>
>>>_______________________________________________
>>>This is the private VTK discussion list.
>>>Please keep messages on-topic. Check the FAQ at:
>>><http://public.kitware.com/cgi-bin/vtkfaq>
>>>Follow this link to subscribe/unsubscribe:
>>>http://public.kitware.com/mailman/listinfo/vtkusers
>>
>
>
>_________________________________________________________________
>Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp
>
>_______________________________________________
>This is the private VTK discussion list.
>Please keep messages on-topic. Check the FAQ at:
><http://public.kitware.com/cgi-bin/vtkfaq>
>Follow this link to subscribe/unsubscribe:
>http://public.kitware.com/mailman/listinfo/vtkusers
_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp
More information about the vtkusers
mailing list