Re: [netcdf-hdf] using POSIX calls when configured for MPI-IO

  • To: Robert Latham <robl@xxxxxxxxxxx>
  • Subject: Re: [netcdf-hdf] using POSIX calls when configured for MPI-IO
  • From: Quincey Koziol <koziol@xxxxxxxxxxxx>
  • Date: Sat, 15 Sep 2007 19:03:06 -0500
Hi Rob,

On Sep 14, 2007, at 6:16 PM, Robert Latham wrote:


Bet you're going to get tired of me soon!

        No, I'm happy to help you out, keep 'em coming... :-)

I'd really like to see all I/O calls using MPI-IO instead of posix
fread (as in nc_check_for_hdf5).   The HDF5 guys might recognize that
I've bugged them about this last year :>

ROMIO can take a file system prefix to override automatic file system
detection.  Such a prefix could be "testfs:" for our tracing file
system driver (does no I/O, but logs which processes wrote how much)
or it could be "pvfs2:" if we want to acess this userspace file system
and do not happen to have the kernel component running.

So for example, in nc_check_for_hdf5, instead of using fread, maybe
use MPI_File_open and MPI_File_read when examining the magic number,
and USE_PARALLEL_MPIIO is defined.  Likewise, nc4_create_file,
and nc_get_file_version use POSIX calls.

I think you might want the "MPI-POSIX" driver in HDF5 (look at H5Pset_fapl_mpiposix), but I don't know if the netCDF-4 library is exporting a way to use that in netCDF-4 applications. Hmm, but re- reading your note again, I'm not certain. Can you take a look at the MPI-POSIX file driver and see if that's what you want?


Attachment: smime.p7s
Description: S/MIME cryptographic signature