[alcf-discuss] Parallel hdf5 I/O of strided data

[John R. Cary]

Hi Rob,

First, Thanks for setting up this list.  I wonder whether it
makes sense to have Reply-To list turned
on for this?  I usually just hit reply, and then the list does not get
this.

Rob Latham wrote:
> On Wed, May 06, 2009 at 09:47:31AM -0700, John R. Cary wrote:
>   
>> Is anyone else writing out strided data
>> using parallel HDF5 on Intrepid?  We had this working
>> at one time, before the driver change, but
>> since then we also changed how we do I/O.
>> Now we get garbage out.
>>     
>
> Hi John.  You're going to have to be a lot more specific. 
> - Which file system are you writing to?
>   

/gpfs1/cary

> - what, if any, HDF5 property list tunables are you setting
>   
On file open:

  MPI_Comm comm = MPI_COMM_WORLD;
  MPI_Info info = MPI_INFO_NULL;
  hid_t plistId = H5Pcreate(H5P_FILE_ACCESS);
  H5Pset_fapl_mpio(plistId, comm, info);
  hid_t res = H5Fcreate(filename.c_str(), H5F_ACC_TRUNC, H5P_DEFAULT,
        plistId);



> - what version of HDF5 are you using?
>   


Using HDF5 with:
    HDF5_INC                      :  
-I/gpfs/software/linux-sles10-ppc64/apps/V1R3M0/hdf5-1.8.0/include
    HDF5_LIBDIR                   : 
/gpfs/software/linux-sles10-ppc64/apps/V1R3M0/hdf5-1.8.0/lib
    HDF5_LIBS                     :  
-L/gpfs/software/linux-sles10-ppc64/apps/V1R3M0/hdf5-1.8.0/lib -lhdf5
    H5DIFF                        :
    H5LS                          :
    HDF5 VERSION                  : 1.8.0


> As far as I know, it's entirely possible and not uncommon to get
> correct strided parallel HDF5 writes.  But, the FLASH gang is getting
> corrupted data when setting certain parameters.  I'd like to hear if
> you're doing what they are doing, or if you've found something new.

Thx...John