Re: Root Mpi

Hello,

I am glad you guys are interested in this work. I implemented TOOMPI to help me parallelize some Geant4 simulations where ROOT objects containing data were exchanged. As you can see the code is very simple and only works in the compiled version.

I used OOMPI on top of LAM-MPI to do the job but the latter has been replaced by OpenMPI and the OOMPI does not work with OpenMPI. However the original libraries (TOOMPI+OOMPI+LAM-MPI) can still be successfully compiled and used nowadays in the most modern systems and the latest ROOT versions, so I did not change nor update anything. But if I need to update it BOOST and OpenMPI libraries are the way to go, and I think you guys are in the right track.

The trick to serialize ROOT objects, maybe you already know, is shown here in this thread by Fons and Phillipe: http://root.cern.ch/phpBB3/viewtopic.php?f=3&t=268&start=0

Cheers, Luis

Luís Augusto Perles, PhD
Radiation Physics Department
University of Texas
MD Anderson Cancer Center
Houston, TX, USA

On 12/13/2010 07:32 AM, Omar Andrés Zapata Mesa wrote:
> Thanks Aaron!
> I see the code and I am getting some new ideas to implements TMessage
> Serialization.
>
> Thanks.
>
> El 13 de diciembre de 2010 02:55, Aaron Dominguez
> <aarondominguez_at_me.com <mailto:aarondominguez_at_me.com>> escribió:
>
> Here are a couple more links on his work which I found in some old
> e-mails. I am also cc'ing him on this thread since maybe he has
> something more recent.
>
> The code:
> http://dropbox.unl.edu/uploads/20101227/f43846b8d36c61af/TOOMPI-0.4.4.tar.gz
>
> A talk he gave:
> http://resource.npl.co.uk/docs/science_technology/ionising%20radiation/clubs_groups/monte_carlo/2004/mcneg/oompi_slides.pdf
>
>
> Aaron
>
> On Dec 13, 2010, at 8:28 AM, Aaron Dominguez wrote:
>
>> I remember that a student implemented MPI in ROOT a few years
>> back. You might want to take a look at how far he
>> got: http://www.osl.iu.edu/research/oompi/
>>
>> Aaron
>>
>> On Dec 12, 2010, at 9:13 PM, Omar Andrés Zapata Mesa wrote:
>>
>>> Hi Ivan.
>>> Thats a Great Question.
>>> let me a time to prepare a good comparison, but at the moment I
>>> can tell you that proof is a power-full tools to grid computing,
>>> but the idea of rootmpi is give support to Message Passing
>>> Interface using MPI
>>> standards http://www.mpi-forum.org/docs/mpi-11-html/mpi-report.html,
>>> that let you not just peer-to-peer communication, it let you
>>> send Broadcast, Gather,Scatter, Reduce and All-toAll messaging;
>>> you can create a messages groups and it have a powefull
>>> framework to work with synchronous and asynchronous process.
>>> Maybe a good idea could be to integrate in some way proof and
>>> rootmpi to run process with job schedulers like condor o glite,
>>> now implemented in proof.
>>>
>>> Thanks!!
>>>
>>>
>>> El 12 de diciembre de 2010 14:18, Ivan Pulido
>>> <mefistofeles87_at_gmail.com <mailto:mefistofeles87_at_gmail.com>>
>>> escribió:
>>>
>>>
>>>
>>> 2010/12/12 Omar Andrés Zapata Mesa
>>> <andresete.chaos_at_gmail.com <mailto:andresete.chaos_at_gmail.com>>
>>>
>>> Hi Routers.
>>> We are writing a experimental code to give support to
>>> root in message passing.
>>> Please visit our
>>> website https://gfifdev.udea.edu.co/plugins/mediawiki/wiki/rootmpi/index.php/Main_Page
>>> and test the code just for linux in the moment.
>>> This code is just an experimental prototype but it works
>>> fine.
>>> Please feel free to suggestions
>>>
>>>
>>> Hello, this seems like a good effort to extend ROOT's
>>> capabilities, congratulations.
>>>
>>> I only have a couple of questions, just to make sure, not
>>> sure if you have tried PROOF but, how is ROOT-mpi different
>>> from PROOF and how would a ROOT user should choose between
>>> these two?
>>>
>>> Thanks.
>>>
>>>
>>> --
>>> Omar Andres Zapata Mesa
>>> Head Developer Phenomenology of Fundamental Interactions
>>> Group (Gfif)
>>> http://gfif.udea.edu.co <http://gfif.udea.edu.co/>
>>> Division of computer science Gfif Developers (Gfif Dev)
>>> http://gfifdev.udea.edu.co <http://gfifdev.udea.edu.co/>
>>> Systems Engineering Student
>>> Universidad de Antioquia At Medellin - Colombia
>>> Usuario Linux #490962
>>>
>>>
>>>
>>>
>>> --
>>> Ivan Pulido
>>> Estudiante de Física
>>> Universidad Nacional de Colombia
>>>
>>>
>>>
>>>
>>> --
>>> Omar Andres Zapata Mesa
>>> Head Developer Phenomenology of Fundamental Interactions
>>> Group (Gfif)
>>> http://gfif.udea.edu.co <http://gfif.udea.edu.co/>
>>> Division of computer science Gfif Developers (Gfif Dev)
>>> http://gfifdev.udea.edu.co <http://gfifdev.udea.edu.co/>
>>> Systems Engineering Student
>>> Universidad de Antioquia At Medellin - Colombia
>>> Usuario Linux #490962
>>>
>>
>
>
>
>
> --
> Omar Andres Zapata Mesa
> Head Developer Phenomenology of Fundamental Interactions Group (Gfif)
> http://gfif.udea.edu.co
> Division of computer science Gfif Developers (Gfif Dev)
> http://gfifdev.udea.edu.co <http://gfifdev.udea.edu.co/>
> Systems Engineering Student
> Universidad de Antioquia At Medellin - Colombia
> Usuario Linux #490962
>
Received on Mon Dec 13 2010 - 18:20:26 CET