To follow up on this, if you do not already know, it is not that
difficult to build the subarray data type by using type contiguous to
define a row/column (row in C, column in Fortran) and then type vector
to define the subarray, building on the contiguous datatype. This is,
of course, for a 2-D array. It is a natural extension to arrays of
higher dimension.
I hope this helps.
Dave.
Jeff Squyres wrote:
> Unfortunately, nothing has changed in this area -- ROMIO does not
> support LAM's SUBARRAY datatype.
>
> On Apr 28, 2005, at 5:10 AM, Heiko Bauke wrote:
>
>> Hi,
>>
>> today I tried to use a subarray data type in MPI-IO. But all I got was
>> the error message:
>>
>> Error: Unsupported datatype passed to ADIOI_Count_contiguous_blocks
>>
>> In the LAM mailing list archives I read in a post from 2003 that the
>> error is originated by ROMIO and it is because LAM treats
>> MPI_Type_darray and MPI_Type_subarray as first-class MPI-2 derived
>> datatypes, whereas ROMIO assumes that these are built on top of MPI-1
>> derived datatypes and this needs to be fixed in ROMIO.
>>
>> At the moment I am using LAM/MPI 7.0.6. Is there any chance that the
>> problems with MPI_Type_darray and MPI_Type_subarray will be fixed in
>> future LAM/MPI versions or Open-MPI?
>>
>>
>> Heiko
>>
>> --
>> -- Wer so tut, als bringe er die Menschen zum Nachdenken, den lieben
>> -- sie. Wer sie wirklich zum Nachdenken bringt, den hassen sie.
>> -- (Aldous Huxley, 1894-1963)
>> -- Heiko Bauke @ http://www.uni-magdeburg.de/bauke
>> _______________________________________________
>> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>>
>
|
