Keep in mind that MPI does not guarantee that files written with MPI-2 IO
are not necessarily readable outside of the MPI-2 IO.
It might be safest to write a serial MPI application to read/postprocess
your data. Specifically, just because a program calls MPI_INIT /
MPI_FINALIZE doesn't mean that it *has* to be run in parallel.
On 8/5/06 1:30 PM, "Christopher G. Roberts" <croberts_at_[hidden]> wrote:
>
> Thanks Pierre. The direct access format does work, you just determine the
> offset and use a DO loop to read the data into an array.
>
> Chris
>
>> Dear Christopher,
>>
>> I guess you are producing a raw binary file, which is lacking the
>> structure of an unformatted fortran file. Using a fortran interface to the
>> MPI I/O library is not expected to change the internal file structure.
>>
>> If your buffers are of equal size, you should be able to read the file in
>> Fortran as a binary direct file. Be aware that the unit for expressing the
>> size of direct access buffers is compiler-dependant.
>>
>> If not, all you need is to write a few C routines for opening, making
>> offsets, and read buffers of specified length.
>>
>> In case my analysis is correct, I attach a C template I am using for
>> manipulating in fortran large raw files produced either by acquisition
>> hardware devices or by ab-initio quantum calculations. Be aware that this
>> template might not be 64-bit clean. This template was contributed by
>> Francoise Roch at the SCCI, cf.
>> http://www.obs.ujf-grenoble.fr/osug/content/blogcategory/83/63/
>>
>> Pierre.
>>
>>
>> Christopher G. Roberts wrote:
>>> Dear Sirs:
>>>
>>>
>>> I have successfully implemented a parallel write to file using the
>>> Fortran subroutine, MPI_FILE_WRITE() and using the MPI_FILE_SET_VIEW to
>>> displace the file pointers for each processor. The file output contains
>>> all integers and is unformmatted. The issue is that I cannot open the
>>> file in a typical Fortran program because the file does not contain the
>>> "headers" and "tails" values common to a Fortran unformatted file. The
>>> worst part is I have all my analysis codes written in F90.
>>>
>>> Is there any way I can overcome this problem or do I have to write a C
>>> function to open the data file, read it to a 1D array, pass it back to
>>> Fortran by reference, and unpack it in column major order?
>>>
>>> If you need me to send a schematic of this issue, please let me know.
>>>
>>> Thanks for your time,
>>>
>>> Chris Roberts
>>>
>>> _______________________________________________ This list is archived at
>>> http://www.lam-mpi.org/MailArchives/lam/
>>>
>>
>>
>> -- Soutenez le mouvement SAUVONS LA RECHERCHE : National :
>> http://recherche-en-danger.apinc.org/ Grenoble :
>> http://recherchegrenoble.free.fr et http://etatsg.free.fr
>>
>> _/_/_/_/ _/ _/ Dr. Pierre VALIRON _/ _/ _/ _/
>> Laboratoire d'Astrophysique _/ _/ _/ _/ Observatoire de
>> Grenoble / UJF _/_/_/_/ _/ _/ BP 53 F-38041 Grenoble Cedex 9
>> (France) _/ _/ _/ http://www-laog.obs.ujf-grenoble.fr _/
>> _/ _/ mail: Pierre.Valiron_at_[hidden] _/ _/ _/
>> Phone: +33 4 7651 4787 Fax: +33 4 7644 8821 _/ _/_/
>>
>>
>> _______________________________________________ This list is archived at
>> http://www.lam-mpi.org/MailArchives/lam/
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
--
Jeff Squyres
Server Virtualization Business Unit
Cisco Systems
|
